article
stringlengths 360
417k
| abstract
stringlengths 8
2.14k
|
|---|---|
[SECTION: Purpose] Organizations of various sectors are not giving proper attention toward the importance of social context in highly challenging working environment for employees. The social context in highly challenging work settings plays a vital role in shaping employees' proficiencies and behaviors. Social context at work captures the interpersonal interactions that develop as employees perform their jobs, roles and tasks (Chou, 2015; Grant and Parker, 2009). The relational perspective of work design focuses on how the social aspects of a job combined with job roles and/or tasks advance an employee's understanding of the job and how this understanding can impact organizational outcomes (Grant and Parker, 2009). In this regard, social exchange theory is a very useful approach for integrating relational or social work context variables into JCM. The JCM posits that five core job characteristics make jobs intrinsically motivating and satisfying and encourage the achievement of high performance (Hackman and Oldham, 1980). Despite the popularity and continuous use of JCM in job enrichment literature, the model has received multiple criticisms from other scholars (Grant et al., 2010). Two main critiques of the model stand out. First, although the JCM advocates that "Growth Need Strength" (GNS), a trait associated with the need for achievement, moderates the relationship between job characteristics and outcome variables, empirical research has not been conclusive about the moderating effects of GNS (Johns, 2006). Second, the commonly used job design measure in the model, known as the Job Diagnostic Survey (JDS) (Hackman and Oldham, 1980) focused on narrow set of motivational job characteristics which may be less applicable in current work contexts (Parker et al., 2001). If scholars continue to ignore broader range of work design literature, the JCM model runs the risk of being deficient (Morgeson and Humphrey, 2006). More recently, studies in the theory and research of work design have emphasized the importance of the relational aspects of job design (Morgeson and Humphrey, 2006). More specifically, they have suggested that there is a need to explore the role of social support in modifying the effects of job characteristics on outcomes. The research on social networks points out that interpersonal relationships often enhance employees' motivations, opportunities and resources (Frazier and Tupper, 2016; Hongvichit, 2015). Economically constrained research communities require a careful assessment of responsibilities and options for researchers and research subjects alike (Rafique, 2015). Very few research studies are available for managers and practitioners in Pakistan at job design and its affirmative relatedness to enhance the performance of employees in various sectors of Pakistan. Thus, managers in Pakistan are facing hindrances in finding the ways to improve the quality of job design through focusing on various factors which ultimately boost the performance of employees (Ali and Zia-ur-Rehman, 2014). Drawing upon social exchange theory and related literature, in this paper we explore how supervisor support, an important aspect of social context, can act as a moderator in JCM. We propose that social support is a critical factor that will enhance the five core job characteristics for individuals in organizations. Strong social support should encourage employees to take up the challenges associated with high scope jobs and lead to improved productivity. The JCM framework of Hackman and Oldham (1975) includes the five job characteristics that influence work-attitudes and behaviors: skills variety, task identity, task significance, autonomy and feedback. They also suggested that GNS, a trait associated with internal energy achievement orientation, acts as moderator in the job characteristics/outcomes relationship. Although JCM continues to enjoy strong support among scholars for the proposed direct relationship between core job characteristics and outcomes, there is mixed evidence confirming the moderator effects of GNS or other related variables such as personality (e.g. Xie and Johns, 1995). We will first briefly review the research on the relationship between job characteristics and job performance and then discuss the proposed moderating effects of supervisor support in shaping this relationship. Relationship of job characteristics with job performance Standard job performance states the degree to which a management employee accomplish organizationally prescribed work role prospects (Katz and Kahn, 1966). In order to get efficiency at workplace, inventive job performance considered as critical element for organizations (Kanter, 1988). Different research studies conducted in the field of Psychology and Organizational Behavior propose that individual job performance gets affected by characteristics of the work environment, which may in turn interact, with employees' personal characteristics (Hyatt and Prawitt, 2001). The Job Characteristics Model (Hackman and Oldham, 1976) reinforced the value of job enrichment that generates positive effects on work outcomes such as job performance, job satisfaction, motivation and work effectiveness. Job performance is one of the most important individual behaviors that concern organizations and the relationship between job characteristics and job performance has caught the attention of many researchers seeking to enhance scholarly and practical understanding (Schleicher et al., 2004). Furthermore, previous research was mainly based on data collected in the developed Western countries, and there is still a need to test and analyze the job characteristics in different cultural settings to test the efficacy and thus generalizability of JCM. According to Parker et al. (2001), untraditional cultural settings could affect a study's findings. With this need in mind, our study was conducted in Pakistan, a country very different from Western countries in terms of socio-political, cultural and legal aspects. Designing jobs and roles appropriately are imperative in elevating the performance of employees, which is addressed through job characteristic model. So this motivates us to test that how job design can enhance the performance of the employees (Ali and Zia-ur-Rehman, 2014). Designing positions with enriched job characteristics may help with the development of effective performance in a challenging organizational environment. Enriched jobs often promote the creativity that leads to high job performance (Elsbach and Hargadon, 2006). The paper is based on a questionnaire survey. The data were subjected to t-test and correlation analysis using SPSS software. Thus, that was an empirical study. In quintessence, enriched and complex jobs are associated with positive attitudinal/behavioral outcomes. Therefore, we suggest the following hypothesis: H1. Job scope will be positively related to job performance. The moderating role of supervisor support Meta-analytic studies suggest that a substantial amount of variance in job performance is not explained by job characteristics. For example, Fried and Ferris (1987) indicated that the five dimensions of job characteristics are all weakly related to job performance. In a more recent meta-analysis, however, Humphrey and colleagues (2007) found a strong relationship between job characteristics and job performance. This variation between scholars in the observed strength of the relationship between core job characteristics and job performance may be indicative that moderators are at work (Parker and Wall, 1998). In this inquiry, we explore the specific moderating role of supervisor support. Social exchange theory asserts that employees making contributions to the organization in the form of effort and time expect that these investments will be reciprocated by the organization. One way to reciprocate is the provision of support through manager to the employees which they feel that supervisors value their contributions and care about their well-being (Chou, 2015; O'Donnell et al., 2012). Social support is positively linked with different work outcomes (Morgeson and Humphrey, 2006), and also acts as an important buffer against job-related demands (Thomas and Wright, 2013). According to recent findings, employees who receive career-related suggestions and psychological support from their supervisors reciprocate by stretching their performance and going that extra mile (Sun et al., 2014). For employees to be able to cope with a challenging work environment characterized by a need for continuous change and innovation, appropriate supervisor support and guidance is crucial so that employees can accomplish challenging goals with high levels of performance (Carter et al., 2013). Thus, supervisor support in a challenging work environment operates as a pivotal catalyst to improve the in-role performance of employees (Rooney et al., 2008). Supervisor support depends on the interpersonal skills of supervisors and in a challenging job environment it displays in terms of credibility and a deep concern for their subordinates' needs (Hongvichit, 2015). In challenging job situations, strong social bonds between management and employees reflects that supervisors, through their support, are able to adjust individual job conditions in order to facilitate better in-role performance. Therefore, we suggest the following hypothesis: H2. Supervisor support moderates the relationship of job scope with job performance such that the relationship will be stronger when supervisor support is high. Data collection and sample This study was conducted in Pakistan, where people attitude toward research is not serious. Thus, for the collection of quality data it was decided that those sectors should be focused from where the chance of collection of quality data would be higher. Different sectors were chosen to look at the variability of responses regarding job design in them. Various professions were represented in the sample, with respondents belonging to all designations levels ranging from entry level office work to top management for checking the impact of supervisor support in challenging work environment at broader level. The questionnaires were distributed by hand and through mail as well. Data were collected through survey distributed to eight different organizations of five miscellaneous sectors to capture maximum variations. The business activities and operations of the organizations include the banks/financial institutions, the educational sector, the health sector, water and power development authority and Pakistan International Airlines. Reminders were given to the respondents of that concerned organization to fill out questionnaires. Total of 640 distributed questionnaires yielded 426 returns for a response rate of 66 percent. After removing incomplete questionnaires and ones with missing peer-reports, 328 complete useable pairs of responses were available for analyses. To avoid problems associated with common method bias and social desirability, we collected the data from two different sources: self and supervisor-reports. While job scope characteristics, supervisor support and all demographic data were measured with self-reports and job performance data were collected through supervisor-reports. The self and supervisor report versions of the survey received a similar code to match survey responses. A cover letter accompanied all questionnaires that explained the study's scope, the voluntary nature of the survey, and assured participants of strict confidentiality for the responses. Completed responses were personally collected by the first author. Measures In Pakistan, English is the language of instruction in all higher education institutes and it is the official language of correspondence in companies. In general, white collar workers in Pakistan are proficient in English according to previous workplace research conducted in Pakistan (e.g. Raja and Johns, 2010). Therefore, it was decided to distribute the questionnaires in English. Job scope Job scope was measured with the self-reported measures from the JDS (Hackman and Oldham, 1976). Sample items with anchors of 1= very little and 7= very much included for all five characteristics. The reliability of overall job scope was 0.86. In-role performance Job performance was measured using supervisor-reports based on the seven-item measure developed by Williams and Anderson (1991). Five point Likert scale used to measure the responses of respondents. The job performance had an excellent Cronbach's a coefficient of 0.89. Supervisor support We measured supervisor support using the nine-item scale developed by Greenhaus et al. (1990). Sample items with anchors of 1= strongly disagree and 5= strongly agree. The reliability of this measure was 0.83. Control variables Information about demographic factors were collected through self-reports. One-way analysis of variance (ANOVA) revealed significant differences between organizations and age in performance. Therefore, we controlled for the effects of organizations and age in analyses. Confirmatory factor analyses The factor structure of the variables was checked through confirmatory factor analysis (CFA) using AMOS 22 with maximum likelihood estimation. CFA was conducted for each measure in order to examine the model fit and the uniqueness of the measures. CFA analysis ran on a five-factor model of job scope, which resulted in a very good model fit (kh2=194, df= 99, RMSEA=0.04, CFI=0.96, GFI=0.94, AGFI=0.90). The results of the CFA for the one-factor model of performance revealed that the one-factor model produced satisfactory fit (kh2=159, df=27, CFI=0.79, GFI=0.94, AGFI=0.92, RMSEA=0.07). The results of standardized factor loadings indicated that all the items loaded above 0.30. The results of the CFA for the one-factor model of self-reported supervisor support revealed that the one-factor model yielded better fit (kh2=87, df=12, CFI=0.86, GFI=0.94, AGFI=0.90, RMSEA=0.03). The results of standardized factor loadings indicated that all the items loaded above 0.30. Results Table I displays the mean values, standard deviations and correlation between the variables. The highest correlation obtained in the current study was between job scope and in-role performance (r=0.46, p<0.001). With respect to the associations between job scope, supervisor support, job scope correlated significantly with supervisor support (r=0.34, p<0.01). We used simple multiple regression analysis to test H1. Results of this analysis are displayed in Table II. The relationship of job scope with in-role performance was positively significant (b=0.45, p<0.001) and explained 17 percent (DR2=0.17, p<0.001) variance in in-role performance. In order to test H2, we used the regression method suggested by Preacher and Hayes (2004), which deploys bootstrap models to test moderation. Bias-corrected confidence intervals (CI) were computed at two selected levels of supervisor support, with 5,000 random samples and replacement from the full sample. Results of this analysis are presented in Table III, which show that the interaction term of job scope and supervisor support was significant in predicting in-role performance (B=-0.31, t=-4.0, p<0.01). The formal two-tailed significance test which assumed normal distribution demonstrated that the conditional direct effect was significant (Effect=0.49, Boot SE=0.02, t=5.4, p<0.001). Bootstrap results confirmed these effects with a bootstrapped 95 percent CI around the conditional direct effect not containing zero (0.11, 0.31). Figure 1 shows that the relationship of job scope with in-role performance was positive and high when supervisor support was high and low when supervisor support was low. Therefore, H2 received support. Overall, there were two hypotheses and these hypotheses were confirmed. More specifically, our results pertaining to the hypothesized relations can be summarized as follows; all predictions related to direct relations of job scope with in-role performance, interaction effects of job scope and supervisor's support were confirmed. Since the emergence of studies on job design, researchers have failed to properly recognize the significance of interpersonal interactions as a crucial aspect that contributes to the meaningfulness of the job (Humphrey et al., 2007). This paper attempts to address this gap by considering the role of supervisor support as a moderator. In support of our hypotheses, job scope was found to be positively related with in-role performance. Furthermore, this positive relationship is more pronounced when supervisor support is strong. Theoretical implications Our study makes key theoretical contributions to our current knowledge about enriched job and the supervisor role to enhance performance of employees. First, we contribute to the research by highlighting two major determinants of performance simultaneously including "system factors" and "person factors" The former factor is related with organizational environment such as work design, later aspects is related with motivation of individuals through different sources such as supervisor support or co-worker support that may affect individual performance (Williams, 2002). Second, the results from our study reveal a positive relationship between job scope and in-role job performance for those employees who are experiencing high levels of supervisor support. Hence, supervisor support seems to be a boundary condition for the positive influence that high job scope has on in-role performance (Grant and Parker, 2009). By incorporating social/relational characteristics of the work context into theories of job design, we have responded to earlier calls of scholars to extend JCM by considering characteristics that are pivotal in many current jobs (Grant and Parker, 2009). Hence through social support theory it is proved that employees will outperform in challenging jobs when they are receiving managerial/supervisor support (Eisenberger et al., 2002). Third, previous research studies were mainly based on data collected in the developed Western countries, and there is still a need to test and analyze the job characteristics in different cultural settings to enhance the generalizability of JCM. Parker et al. (2001) advocated for untraditional cultural settings for generalizability of study's findings. Keeping in view such factors in mind we conducted this study in Pakistan, a country very different from Western countries in various terms. Practical implications The findings from this inquiry offer useful insights for managers and consultants. First, it highlights the importance of social characteristics in a job which are likely to influence a range of work outcomes. Increased interaction between supervisor and employees strengthens interpersonal relationships at work and increases opportunities for communication which is a fundamental key to enhancing job performance. Social activities have a positive quality in that they generate feelings of energy, enthusiasm and positive affect while at work (Watson, 2000). Second, strong social support provides opportunities through which employees can acquire assistance from their supervisors. This form of interaction is likely to help employees clarify and make sense of their roles and quickly address their concerns and issues. Third, such kind of interactions provides additional insight to employees about their performance and important feedback for continuous learning and personal development (Berman et al., 2002). Managerial implications Multiple sources for data collection were being adopted. Thus, the results of the study may applicable to the various sectors from where data were collected. The results of the study have provided inputs on how to enhance perceived organizational support of supervisors for employees in highly challenging jobs in various organizations. In public or private sector performance level of employees can be boosted when the employees feel content. That state can only be appeared when informal and formal organization policies like mentoring, socialization and open-door conflict management intend to encourage workplace supervisor supportive policies (Frazier and Tupper, 2016). According to the Job Demand-Resource model framework, in highly competitive work environment quality of the relationship with supervisor matters a lot in the company at various levels, not only in the board room but also with employees at the team or department level, or in focus groups. Such kind of supporting relations of supervisors with employees are crucially important to build commitment and credibility of employees for the various types of organizations (Bauer and Hammig, 2014). Similarly, organizations are likely to reap many benefits by providing to new entrants fresh work trends at the workplace where they feel comfortable sharing their ideas and feel valued (Bhatnagar, 2014; Kim, 2014). Limitations and suggestions for future research There are several limitations that should be considered when assessing this study's findings. First, the data collected at one point in time for all the variables, limiting the causal inferences that can be drawn based on our findings. For stronger and more convincing results, future research should consider the use of experimental, quasi-experimental or longitudinal design. Also, in order to reduce concerns about common method bias, data were collected by using two different sources: self and supervisory reports (Podsakoff et al., 2003). Second, the study was conducted in Pakistan. For the generalizability of the study's findings, these results should be validated in other cultural settings because Pakistan is very different in terms of socio-political, cultural and legal aspects in comparison to Western countries (Parker et al., 2001). Finally, this study only focused on one side of the interpersonal support: the critical role that supervisors play in determining employees' workplace perceptions, motivations, attitudes and behaviors. The support employees get from their peers was not considered in this study but may also be a crucial source that moderates the relationship between job scope and performance. Co-worker/peer support involves actions by co-workers that are either helpful or intended to be helpful (Hongvichit, 2015). This study proposed that the relational context of jobs, particularly in the form of social support by supervisors, can motivate employees to perform their duties more productively. It also highlights how the structure of an employee's work with the manifestation of social support plays a critical role in shaping the employee's relationships with their supervisors. This study highlights that both job design and in-role performance have a strong link with relational contexts that can motivate employees to achieve their stated targets. Thus, this research improves our understanding of how social context at the workplace can make a difference for employees and their organizations.
|
The purpose of this paper is to explore how the relationship between job scope and in-role performance is contingent upon the level of social support (i.e. supervisor support) received in the workplace.
|
[SECTION: Method] Organizations of various sectors are not giving proper attention toward the importance of social context in highly challenging working environment for employees. The social context in highly challenging work settings plays a vital role in shaping employees' proficiencies and behaviors. Social context at work captures the interpersonal interactions that develop as employees perform their jobs, roles and tasks (Chou, 2015; Grant and Parker, 2009). The relational perspective of work design focuses on how the social aspects of a job combined with job roles and/or tasks advance an employee's understanding of the job and how this understanding can impact organizational outcomes (Grant and Parker, 2009). In this regard, social exchange theory is a very useful approach for integrating relational or social work context variables into JCM. The JCM posits that five core job characteristics make jobs intrinsically motivating and satisfying and encourage the achievement of high performance (Hackman and Oldham, 1980). Despite the popularity and continuous use of JCM in job enrichment literature, the model has received multiple criticisms from other scholars (Grant et al., 2010). Two main critiques of the model stand out. First, although the JCM advocates that "Growth Need Strength" (GNS), a trait associated with the need for achievement, moderates the relationship between job characteristics and outcome variables, empirical research has not been conclusive about the moderating effects of GNS (Johns, 2006). Second, the commonly used job design measure in the model, known as the Job Diagnostic Survey (JDS) (Hackman and Oldham, 1980) focused on narrow set of motivational job characteristics which may be less applicable in current work contexts (Parker et al., 2001). If scholars continue to ignore broader range of work design literature, the JCM model runs the risk of being deficient (Morgeson and Humphrey, 2006). More recently, studies in the theory and research of work design have emphasized the importance of the relational aspects of job design (Morgeson and Humphrey, 2006). More specifically, they have suggested that there is a need to explore the role of social support in modifying the effects of job characteristics on outcomes. The research on social networks points out that interpersonal relationships often enhance employees' motivations, opportunities and resources (Frazier and Tupper, 2016; Hongvichit, 2015). Economically constrained research communities require a careful assessment of responsibilities and options for researchers and research subjects alike (Rafique, 2015). Very few research studies are available for managers and practitioners in Pakistan at job design and its affirmative relatedness to enhance the performance of employees in various sectors of Pakistan. Thus, managers in Pakistan are facing hindrances in finding the ways to improve the quality of job design through focusing on various factors which ultimately boost the performance of employees (Ali and Zia-ur-Rehman, 2014). Drawing upon social exchange theory and related literature, in this paper we explore how supervisor support, an important aspect of social context, can act as a moderator in JCM. We propose that social support is a critical factor that will enhance the five core job characteristics for individuals in organizations. Strong social support should encourage employees to take up the challenges associated with high scope jobs and lead to improved productivity. The JCM framework of Hackman and Oldham (1975) includes the five job characteristics that influence work-attitudes and behaviors: skills variety, task identity, task significance, autonomy and feedback. They also suggested that GNS, a trait associated with internal energy achievement orientation, acts as moderator in the job characteristics/outcomes relationship. Although JCM continues to enjoy strong support among scholars for the proposed direct relationship between core job characteristics and outcomes, there is mixed evidence confirming the moderator effects of GNS or other related variables such as personality (e.g. Xie and Johns, 1995). We will first briefly review the research on the relationship between job characteristics and job performance and then discuss the proposed moderating effects of supervisor support in shaping this relationship. Relationship of job characteristics with job performance Standard job performance states the degree to which a management employee accomplish organizationally prescribed work role prospects (Katz and Kahn, 1966). In order to get efficiency at workplace, inventive job performance considered as critical element for organizations (Kanter, 1988). Different research studies conducted in the field of Psychology and Organizational Behavior propose that individual job performance gets affected by characteristics of the work environment, which may in turn interact, with employees' personal characteristics (Hyatt and Prawitt, 2001). The Job Characteristics Model (Hackman and Oldham, 1976) reinforced the value of job enrichment that generates positive effects on work outcomes such as job performance, job satisfaction, motivation and work effectiveness. Job performance is one of the most important individual behaviors that concern organizations and the relationship between job characteristics and job performance has caught the attention of many researchers seeking to enhance scholarly and practical understanding (Schleicher et al., 2004). Furthermore, previous research was mainly based on data collected in the developed Western countries, and there is still a need to test and analyze the job characteristics in different cultural settings to test the efficacy and thus generalizability of JCM. According to Parker et al. (2001), untraditional cultural settings could affect a study's findings. With this need in mind, our study was conducted in Pakistan, a country very different from Western countries in terms of socio-political, cultural and legal aspects. Designing jobs and roles appropriately are imperative in elevating the performance of employees, which is addressed through job characteristic model. So this motivates us to test that how job design can enhance the performance of the employees (Ali and Zia-ur-Rehman, 2014). Designing positions with enriched job characteristics may help with the development of effective performance in a challenging organizational environment. Enriched jobs often promote the creativity that leads to high job performance (Elsbach and Hargadon, 2006). The paper is based on a questionnaire survey. The data were subjected to t-test and correlation analysis using SPSS software. Thus, that was an empirical study. In quintessence, enriched and complex jobs are associated with positive attitudinal/behavioral outcomes. Therefore, we suggest the following hypothesis: H1. Job scope will be positively related to job performance. The moderating role of supervisor support Meta-analytic studies suggest that a substantial amount of variance in job performance is not explained by job characteristics. For example, Fried and Ferris (1987) indicated that the five dimensions of job characteristics are all weakly related to job performance. In a more recent meta-analysis, however, Humphrey and colleagues (2007) found a strong relationship between job characteristics and job performance. This variation between scholars in the observed strength of the relationship between core job characteristics and job performance may be indicative that moderators are at work (Parker and Wall, 1998). In this inquiry, we explore the specific moderating role of supervisor support. Social exchange theory asserts that employees making contributions to the organization in the form of effort and time expect that these investments will be reciprocated by the organization. One way to reciprocate is the provision of support through manager to the employees which they feel that supervisors value their contributions and care about their well-being (Chou, 2015; O'Donnell et al., 2012). Social support is positively linked with different work outcomes (Morgeson and Humphrey, 2006), and also acts as an important buffer against job-related demands (Thomas and Wright, 2013). According to recent findings, employees who receive career-related suggestions and psychological support from their supervisors reciprocate by stretching their performance and going that extra mile (Sun et al., 2014). For employees to be able to cope with a challenging work environment characterized by a need for continuous change and innovation, appropriate supervisor support and guidance is crucial so that employees can accomplish challenging goals with high levels of performance (Carter et al., 2013). Thus, supervisor support in a challenging work environment operates as a pivotal catalyst to improve the in-role performance of employees (Rooney et al., 2008). Supervisor support depends on the interpersonal skills of supervisors and in a challenging job environment it displays in terms of credibility and a deep concern for their subordinates' needs (Hongvichit, 2015). In challenging job situations, strong social bonds between management and employees reflects that supervisors, through their support, are able to adjust individual job conditions in order to facilitate better in-role performance. Therefore, we suggest the following hypothesis: H2. Supervisor support moderates the relationship of job scope with job performance such that the relationship will be stronger when supervisor support is high. Data collection and sample This study was conducted in Pakistan, where people attitude toward research is not serious. Thus, for the collection of quality data it was decided that those sectors should be focused from where the chance of collection of quality data would be higher. Different sectors were chosen to look at the variability of responses regarding job design in them. Various professions were represented in the sample, with respondents belonging to all designations levels ranging from entry level office work to top management for checking the impact of supervisor support in challenging work environment at broader level. The questionnaires were distributed by hand and through mail as well. Data were collected through survey distributed to eight different organizations of five miscellaneous sectors to capture maximum variations. The business activities and operations of the organizations include the banks/financial institutions, the educational sector, the health sector, water and power development authority and Pakistan International Airlines. Reminders were given to the respondents of that concerned organization to fill out questionnaires. Total of 640 distributed questionnaires yielded 426 returns for a response rate of 66 percent. After removing incomplete questionnaires and ones with missing peer-reports, 328 complete useable pairs of responses were available for analyses. To avoid problems associated with common method bias and social desirability, we collected the data from two different sources: self and supervisor-reports. While job scope characteristics, supervisor support and all demographic data were measured with self-reports and job performance data were collected through supervisor-reports. The self and supervisor report versions of the survey received a similar code to match survey responses. A cover letter accompanied all questionnaires that explained the study's scope, the voluntary nature of the survey, and assured participants of strict confidentiality for the responses. Completed responses were personally collected by the first author. Measures In Pakistan, English is the language of instruction in all higher education institutes and it is the official language of correspondence in companies. In general, white collar workers in Pakistan are proficient in English according to previous workplace research conducted in Pakistan (e.g. Raja and Johns, 2010). Therefore, it was decided to distribute the questionnaires in English. Job scope Job scope was measured with the self-reported measures from the JDS (Hackman and Oldham, 1976). Sample items with anchors of 1= very little and 7= very much included for all five characteristics. The reliability of overall job scope was 0.86. In-role performance Job performance was measured using supervisor-reports based on the seven-item measure developed by Williams and Anderson (1991). Five point Likert scale used to measure the responses of respondents. The job performance had an excellent Cronbach's a coefficient of 0.89. Supervisor support We measured supervisor support using the nine-item scale developed by Greenhaus et al. (1990). Sample items with anchors of 1= strongly disagree and 5= strongly agree. The reliability of this measure was 0.83. Control variables Information about demographic factors were collected through self-reports. One-way analysis of variance (ANOVA) revealed significant differences between organizations and age in performance. Therefore, we controlled for the effects of organizations and age in analyses. Confirmatory factor analyses The factor structure of the variables was checked through confirmatory factor analysis (CFA) using AMOS 22 with maximum likelihood estimation. CFA was conducted for each measure in order to examine the model fit and the uniqueness of the measures. CFA analysis ran on a five-factor model of job scope, which resulted in a very good model fit (kh2=194, df= 99, RMSEA=0.04, CFI=0.96, GFI=0.94, AGFI=0.90). The results of the CFA for the one-factor model of performance revealed that the one-factor model produced satisfactory fit (kh2=159, df=27, CFI=0.79, GFI=0.94, AGFI=0.92, RMSEA=0.07). The results of standardized factor loadings indicated that all the items loaded above 0.30. The results of the CFA for the one-factor model of self-reported supervisor support revealed that the one-factor model yielded better fit (kh2=87, df=12, CFI=0.86, GFI=0.94, AGFI=0.90, RMSEA=0.03). The results of standardized factor loadings indicated that all the items loaded above 0.30. Results Table I displays the mean values, standard deviations and correlation between the variables. The highest correlation obtained in the current study was between job scope and in-role performance (r=0.46, p<0.001). With respect to the associations between job scope, supervisor support, job scope correlated significantly with supervisor support (r=0.34, p<0.01). We used simple multiple regression analysis to test H1. Results of this analysis are displayed in Table II. The relationship of job scope with in-role performance was positively significant (b=0.45, p<0.001) and explained 17 percent (DR2=0.17, p<0.001) variance in in-role performance. In order to test H2, we used the regression method suggested by Preacher and Hayes (2004), which deploys bootstrap models to test moderation. Bias-corrected confidence intervals (CI) were computed at two selected levels of supervisor support, with 5,000 random samples and replacement from the full sample. Results of this analysis are presented in Table III, which show that the interaction term of job scope and supervisor support was significant in predicting in-role performance (B=-0.31, t=-4.0, p<0.01). The formal two-tailed significance test which assumed normal distribution demonstrated that the conditional direct effect was significant (Effect=0.49, Boot SE=0.02, t=5.4, p<0.001). Bootstrap results confirmed these effects with a bootstrapped 95 percent CI around the conditional direct effect not containing zero (0.11, 0.31). Figure 1 shows that the relationship of job scope with in-role performance was positive and high when supervisor support was high and low when supervisor support was low. Therefore, H2 received support. Overall, there were two hypotheses and these hypotheses were confirmed. More specifically, our results pertaining to the hypothesized relations can be summarized as follows; all predictions related to direct relations of job scope with in-role performance, interaction effects of job scope and supervisor's support were confirmed. Since the emergence of studies on job design, researchers have failed to properly recognize the significance of interpersonal interactions as a crucial aspect that contributes to the meaningfulness of the job (Humphrey et al., 2007). This paper attempts to address this gap by considering the role of supervisor support as a moderator. In support of our hypotheses, job scope was found to be positively related with in-role performance. Furthermore, this positive relationship is more pronounced when supervisor support is strong. Theoretical implications Our study makes key theoretical contributions to our current knowledge about enriched job and the supervisor role to enhance performance of employees. First, we contribute to the research by highlighting two major determinants of performance simultaneously including "system factors" and "person factors" The former factor is related with organizational environment such as work design, later aspects is related with motivation of individuals through different sources such as supervisor support or co-worker support that may affect individual performance (Williams, 2002). Second, the results from our study reveal a positive relationship between job scope and in-role job performance for those employees who are experiencing high levels of supervisor support. Hence, supervisor support seems to be a boundary condition for the positive influence that high job scope has on in-role performance (Grant and Parker, 2009). By incorporating social/relational characteristics of the work context into theories of job design, we have responded to earlier calls of scholars to extend JCM by considering characteristics that are pivotal in many current jobs (Grant and Parker, 2009). Hence through social support theory it is proved that employees will outperform in challenging jobs when they are receiving managerial/supervisor support (Eisenberger et al., 2002). Third, previous research studies were mainly based on data collected in the developed Western countries, and there is still a need to test and analyze the job characteristics in different cultural settings to enhance the generalizability of JCM. Parker et al. (2001) advocated for untraditional cultural settings for generalizability of study's findings. Keeping in view such factors in mind we conducted this study in Pakistan, a country very different from Western countries in various terms. Practical implications The findings from this inquiry offer useful insights for managers and consultants. First, it highlights the importance of social characteristics in a job which are likely to influence a range of work outcomes. Increased interaction between supervisor and employees strengthens interpersonal relationships at work and increases opportunities for communication which is a fundamental key to enhancing job performance. Social activities have a positive quality in that they generate feelings of energy, enthusiasm and positive affect while at work (Watson, 2000). Second, strong social support provides opportunities through which employees can acquire assistance from their supervisors. This form of interaction is likely to help employees clarify and make sense of their roles and quickly address their concerns and issues. Third, such kind of interactions provides additional insight to employees about their performance and important feedback for continuous learning and personal development (Berman et al., 2002). Managerial implications Multiple sources for data collection were being adopted. Thus, the results of the study may applicable to the various sectors from where data were collected. The results of the study have provided inputs on how to enhance perceived organizational support of supervisors for employees in highly challenging jobs in various organizations. In public or private sector performance level of employees can be boosted when the employees feel content. That state can only be appeared when informal and formal organization policies like mentoring, socialization and open-door conflict management intend to encourage workplace supervisor supportive policies (Frazier and Tupper, 2016). According to the Job Demand-Resource model framework, in highly competitive work environment quality of the relationship with supervisor matters a lot in the company at various levels, not only in the board room but also with employees at the team or department level, or in focus groups. Such kind of supporting relations of supervisors with employees are crucially important to build commitment and credibility of employees for the various types of organizations (Bauer and Hammig, 2014). Similarly, organizations are likely to reap many benefits by providing to new entrants fresh work trends at the workplace where they feel comfortable sharing their ideas and feel valued (Bhatnagar, 2014; Kim, 2014). Limitations and suggestions for future research There are several limitations that should be considered when assessing this study's findings. First, the data collected at one point in time for all the variables, limiting the causal inferences that can be drawn based on our findings. For stronger and more convincing results, future research should consider the use of experimental, quasi-experimental or longitudinal design. Also, in order to reduce concerns about common method bias, data were collected by using two different sources: self and supervisory reports (Podsakoff et al., 2003). Second, the study was conducted in Pakistan. For the generalizability of the study's findings, these results should be validated in other cultural settings because Pakistan is very different in terms of socio-political, cultural and legal aspects in comparison to Western countries (Parker et al., 2001). Finally, this study only focused on one side of the interpersonal support: the critical role that supervisors play in determining employees' workplace perceptions, motivations, attitudes and behaviors. The support employees get from their peers was not considered in this study but may also be a crucial source that moderates the relationship between job scope and performance. Co-worker/peer support involves actions by co-workers that are either helpful or intended to be helpful (Hongvichit, 2015). This study proposed that the relational context of jobs, particularly in the form of social support by supervisors, can motivate employees to perform their duties more productively. It also highlights how the structure of an employee's work with the manifestation of social support plays a critical role in shaping the employee's relationships with their supervisors. This study highlights that both job design and in-role performance have a strong link with relational contexts that can motivate employees to achieve their stated targets. Thus, this research improves our understanding of how social context at the workplace can make a difference for employees and their organizations.
|
A total of 640 questionnaires were distributed to employees of Pakistani companies, yielding 328 useable responses for analysis. Regression analysis was used to test for both hypotheses.
|
[SECTION: Findings] Organizations of various sectors are not giving proper attention toward the importance of social context in highly challenging working environment for employees. The social context in highly challenging work settings plays a vital role in shaping employees' proficiencies and behaviors. Social context at work captures the interpersonal interactions that develop as employees perform their jobs, roles and tasks (Chou, 2015; Grant and Parker, 2009). The relational perspective of work design focuses on how the social aspects of a job combined with job roles and/or tasks advance an employee's understanding of the job and how this understanding can impact organizational outcomes (Grant and Parker, 2009). In this regard, social exchange theory is a very useful approach for integrating relational or social work context variables into JCM. The JCM posits that five core job characteristics make jobs intrinsically motivating and satisfying and encourage the achievement of high performance (Hackman and Oldham, 1980). Despite the popularity and continuous use of JCM in job enrichment literature, the model has received multiple criticisms from other scholars (Grant et al., 2010). Two main critiques of the model stand out. First, although the JCM advocates that "Growth Need Strength" (GNS), a trait associated with the need for achievement, moderates the relationship between job characteristics and outcome variables, empirical research has not been conclusive about the moderating effects of GNS (Johns, 2006). Second, the commonly used job design measure in the model, known as the Job Diagnostic Survey (JDS) (Hackman and Oldham, 1980) focused on narrow set of motivational job characteristics which may be less applicable in current work contexts (Parker et al., 2001). If scholars continue to ignore broader range of work design literature, the JCM model runs the risk of being deficient (Morgeson and Humphrey, 2006). More recently, studies in the theory and research of work design have emphasized the importance of the relational aspects of job design (Morgeson and Humphrey, 2006). More specifically, they have suggested that there is a need to explore the role of social support in modifying the effects of job characteristics on outcomes. The research on social networks points out that interpersonal relationships often enhance employees' motivations, opportunities and resources (Frazier and Tupper, 2016; Hongvichit, 2015). Economically constrained research communities require a careful assessment of responsibilities and options for researchers and research subjects alike (Rafique, 2015). Very few research studies are available for managers and practitioners in Pakistan at job design and its affirmative relatedness to enhance the performance of employees in various sectors of Pakistan. Thus, managers in Pakistan are facing hindrances in finding the ways to improve the quality of job design through focusing on various factors which ultimately boost the performance of employees (Ali and Zia-ur-Rehman, 2014). Drawing upon social exchange theory and related literature, in this paper we explore how supervisor support, an important aspect of social context, can act as a moderator in JCM. We propose that social support is a critical factor that will enhance the five core job characteristics for individuals in organizations. Strong social support should encourage employees to take up the challenges associated with high scope jobs and lead to improved productivity. The JCM framework of Hackman and Oldham (1975) includes the five job characteristics that influence work-attitudes and behaviors: skills variety, task identity, task significance, autonomy and feedback. They also suggested that GNS, a trait associated with internal energy achievement orientation, acts as moderator in the job characteristics/outcomes relationship. Although JCM continues to enjoy strong support among scholars for the proposed direct relationship between core job characteristics and outcomes, there is mixed evidence confirming the moderator effects of GNS or other related variables such as personality (e.g. Xie and Johns, 1995). We will first briefly review the research on the relationship between job characteristics and job performance and then discuss the proposed moderating effects of supervisor support in shaping this relationship. Relationship of job characteristics with job performance Standard job performance states the degree to which a management employee accomplish organizationally prescribed work role prospects (Katz and Kahn, 1966). In order to get efficiency at workplace, inventive job performance considered as critical element for organizations (Kanter, 1988). Different research studies conducted in the field of Psychology and Organizational Behavior propose that individual job performance gets affected by characteristics of the work environment, which may in turn interact, with employees' personal characteristics (Hyatt and Prawitt, 2001). The Job Characteristics Model (Hackman and Oldham, 1976) reinforced the value of job enrichment that generates positive effects on work outcomes such as job performance, job satisfaction, motivation and work effectiveness. Job performance is one of the most important individual behaviors that concern organizations and the relationship between job characteristics and job performance has caught the attention of many researchers seeking to enhance scholarly and practical understanding (Schleicher et al., 2004). Furthermore, previous research was mainly based on data collected in the developed Western countries, and there is still a need to test and analyze the job characteristics in different cultural settings to test the efficacy and thus generalizability of JCM. According to Parker et al. (2001), untraditional cultural settings could affect a study's findings. With this need in mind, our study was conducted in Pakistan, a country very different from Western countries in terms of socio-political, cultural and legal aspects. Designing jobs and roles appropriately are imperative in elevating the performance of employees, which is addressed through job characteristic model. So this motivates us to test that how job design can enhance the performance of the employees (Ali and Zia-ur-Rehman, 2014). Designing positions with enriched job characteristics may help with the development of effective performance in a challenging organizational environment. Enriched jobs often promote the creativity that leads to high job performance (Elsbach and Hargadon, 2006). The paper is based on a questionnaire survey. The data were subjected to t-test and correlation analysis using SPSS software. Thus, that was an empirical study. In quintessence, enriched and complex jobs are associated with positive attitudinal/behavioral outcomes. Therefore, we suggest the following hypothesis: H1. Job scope will be positively related to job performance. The moderating role of supervisor support Meta-analytic studies suggest that a substantial amount of variance in job performance is not explained by job characteristics. For example, Fried and Ferris (1987) indicated that the five dimensions of job characteristics are all weakly related to job performance. In a more recent meta-analysis, however, Humphrey and colleagues (2007) found a strong relationship between job characteristics and job performance. This variation between scholars in the observed strength of the relationship between core job characteristics and job performance may be indicative that moderators are at work (Parker and Wall, 1998). In this inquiry, we explore the specific moderating role of supervisor support. Social exchange theory asserts that employees making contributions to the organization in the form of effort and time expect that these investments will be reciprocated by the organization. One way to reciprocate is the provision of support through manager to the employees which they feel that supervisors value their contributions and care about their well-being (Chou, 2015; O'Donnell et al., 2012). Social support is positively linked with different work outcomes (Morgeson and Humphrey, 2006), and also acts as an important buffer against job-related demands (Thomas and Wright, 2013). According to recent findings, employees who receive career-related suggestions and psychological support from their supervisors reciprocate by stretching their performance and going that extra mile (Sun et al., 2014). For employees to be able to cope with a challenging work environment characterized by a need for continuous change and innovation, appropriate supervisor support and guidance is crucial so that employees can accomplish challenging goals with high levels of performance (Carter et al., 2013). Thus, supervisor support in a challenging work environment operates as a pivotal catalyst to improve the in-role performance of employees (Rooney et al., 2008). Supervisor support depends on the interpersonal skills of supervisors and in a challenging job environment it displays in terms of credibility and a deep concern for their subordinates' needs (Hongvichit, 2015). In challenging job situations, strong social bonds between management and employees reflects that supervisors, through their support, are able to adjust individual job conditions in order to facilitate better in-role performance. Therefore, we suggest the following hypothesis: H2. Supervisor support moderates the relationship of job scope with job performance such that the relationship will be stronger when supervisor support is high. Data collection and sample This study was conducted in Pakistan, where people attitude toward research is not serious. Thus, for the collection of quality data it was decided that those sectors should be focused from where the chance of collection of quality data would be higher. Different sectors were chosen to look at the variability of responses regarding job design in them. Various professions were represented in the sample, with respondents belonging to all designations levels ranging from entry level office work to top management for checking the impact of supervisor support in challenging work environment at broader level. The questionnaires were distributed by hand and through mail as well. Data were collected through survey distributed to eight different organizations of five miscellaneous sectors to capture maximum variations. The business activities and operations of the organizations include the banks/financial institutions, the educational sector, the health sector, water and power development authority and Pakistan International Airlines. Reminders were given to the respondents of that concerned organization to fill out questionnaires. Total of 640 distributed questionnaires yielded 426 returns for a response rate of 66 percent. After removing incomplete questionnaires and ones with missing peer-reports, 328 complete useable pairs of responses were available for analyses. To avoid problems associated with common method bias and social desirability, we collected the data from two different sources: self and supervisor-reports. While job scope characteristics, supervisor support and all demographic data were measured with self-reports and job performance data were collected through supervisor-reports. The self and supervisor report versions of the survey received a similar code to match survey responses. A cover letter accompanied all questionnaires that explained the study's scope, the voluntary nature of the survey, and assured participants of strict confidentiality for the responses. Completed responses were personally collected by the first author. Measures In Pakistan, English is the language of instruction in all higher education institutes and it is the official language of correspondence in companies. In general, white collar workers in Pakistan are proficient in English according to previous workplace research conducted in Pakistan (e.g. Raja and Johns, 2010). Therefore, it was decided to distribute the questionnaires in English. Job scope Job scope was measured with the self-reported measures from the JDS (Hackman and Oldham, 1976). Sample items with anchors of 1= very little and 7= very much included for all five characteristics. The reliability of overall job scope was 0.86. In-role performance Job performance was measured using supervisor-reports based on the seven-item measure developed by Williams and Anderson (1991). Five point Likert scale used to measure the responses of respondents. The job performance had an excellent Cronbach's a coefficient of 0.89. Supervisor support We measured supervisor support using the nine-item scale developed by Greenhaus et al. (1990). Sample items with anchors of 1= strongly disagree and 5= strongly agree. The reliability of this measure was 0.83. Control variables Information about demographic factors were collected through self-reports. One-way analysis of variance (ANOVA) revealed significant differences between organizations and age in performance. Therefore, we controlled for the effects of organizations and age in analyses. Confirmatory factor analyses The factor structure of the variables was checked through confirmatory factor analysis (CFA) using AMOS 22 with maximum likelihood estimation. CFA was conducted for each measure in order to examine the model fit and the uniqueness of the measures. CFA analysis ran on a five-factor model of job scope, which resulted in a very good model fit (kh2=194, df= 99, RMSEA=0.04, CFI=0.96, GFI=0.94, AGFI=0.90). The results of the CFA for the one-factor model of performance revealed that the one-factor model produced satisfactory fit (kh2=159, df=27, CFI=0.79, GFI=0.94, AGFI=0.92, RMSEA=0.07). The results of standardized factor loadings indicated that all the items loaded above 0.30. The results of the CFA for the one-factor model of self-reported supervisor support revealed that the one-factor model yielded better fit (kh2=87, df=12, CFI=0.86, GFI=0.94, AGFI=0.90, RMSEA=0.03). The results of standardized factor loadings indicated that all the items loaded above 0.30. Results Table I displays the mean values, standard deviations and correlation between the variables. The highest correlation obtained in the current study was between job scope and in-role performance (r=0.46, p<0.001). With respect to the associations between job scope, supervisor support, job scope correlated significantly with supervisor support (r=0.34, p<0.01). We used simple multiple regression analysis to test H1. Results of this analysis are displayed in Table II. The relationship of job scope with in-role performance was positively significant (b=0.45, p<0.001) and explained 17 percent (DR2=0.17, p<0.001) variance in in-role performance. In order to test H2, we used the regression method suggested by Preacher and Hayes (2004), which deploys bootstrap models to test moderation. Bias-corrected confidence intervals (CI) were computed at two selected levels of supervisor support, with 5,000 random samples and replacement from the full sample. Results of this analysis are presented in Table III, which show that the interaction term of job scope and supervisor support was significant in predicting in-role performance (B=-0.31, t=-4.0, p<0.01). The formal two-tailed significance test which assumed normal distribution demonstrated that the conditional direct effect was significant (Effect=0.49, Boot SE=0.02, t=5.4, p<0.001). Bootstrap results confirmed these effects with a bootstrapped 95 percent CI around the conditional direct effect not containing zero (0.11, 0.31). Figure 1 shows that the relationship of job scope with in-role performance was positive and high when supervisor support was high and low when supervisor support was low. Therefore, H2 received support. Overall, there were two hypotheses and these hypotheses were confirmed. More specifically, our results pertaining to the hypothesized relations can be summarized as follows; all predictions related to direct relations of job scope with in-role performance, interaction effects of job scope and supervisor's support were confirmed. Since the emergence of studies on job design, researchers have failed to properly recognize the significance of interpersonal interactions as a crucial aspect that contributes to the meaningfulness of the job (Humphrey et al., 2007). This paper attempts to address this gap by considering the role of supervisor support as a moderator. In support of our hypotheses, job scope was found to be positively related with in-role performance. Furthermore, this positive relationship is more pronounced when supervisor support is strong. Theoretical implications Our study makes key theoretical contributions to our current knowledge about enriched job and the supervisor role to enhance performance of employees. First, we contribute to the research by highlighting two major determinants of performance simultaneously including "system factors" and "person factors" The former factor is related with organizational environment such as work design, later aspects is related with motivation of individuals through different sources such as supervisor support or co-worker support that may affect individual performance (Williams, 2002). Second, the results from our study reveal a positive relationship between job scope and in-role job performance for those employees who are experiencing high levels of supervisor support. Hence, supervisor support seems to be a boundary condition for the positive influence that high job scope has on in-role performance (Grant and Parker, 2009). By incorporating social/relational characteristics of the work context into theories of job design, we have responded to earlier calls of scholars to extend JCM by considering characteristics that are pivotal in many current jobs (Grant and Parker, 2009). Hence through social support theory it is proved that employees will outperform in challenging jobs when they are receiving managerial/supervisor support (Eisenberger et al., 2002). Third, previous research studies were mainly based on data collected in the developed Western countries, and there is still a need to test and analyze the job characteristics in different cultural settings to enhance the generalizability of JCM. Parker et al. (2001) advocated for untraditional cultural settings for generalizability of study's findings. Keeping in view such factors in mind we conducted this study in Pakistan, a country very different from Western countries in various terms. Practical implications The findings from this inquiry offer useful insights for managers and consultants. First, it highlights the importance of social characteristics in a job which are likely to influence a range of work outcomes. Increased interaction between supervisor and employees strengthens interpersonal relationships at work and increases opportunities for communication which is a fundamental key to enhancing job performance. Social activities have a positive quality in that they generate feelings of energy, enthusiasm and positive affect while at work (Watson, 2000). Second, strong social support provides opportunities through which employees can acquire assistance from their supervisors. This form of interaction is likely to help employees clarify and make sense of their roles and quickly address their concerns and issues. Third, such kind of interactions provides additional insight to employees about their performance and important feedback for continuous learning and personal development (Berman et al., 2002). Managerial implications Multiple sources for data collection were being adopted. Thus, the results of the study may applicable to the various sectors from where data were collected. The results of the study have provided inputs on how to enhance perceived organizational support of supervisors for employees in highly challenging jobs in various organizations. In public or private sector performance level of employees can be boosted when the employees feel content. That state can only be appeared when informal and formal organization policies like mentoring, socialization and open-door conflict management intend to encourage workplace supervisor supportive policies (Frazier and Tupper, 2016). According to the Job Demand-Resource model framework, in highly competitive work environment quality of the relationship with supervisor matters a lot in the company at various levels, not only in the board room but also with employees at the team or department level, or in focus groups. Such kind of supporting relations of supervisors with employees are crucially important to build commitment and credibility of employees for the various types of organizations (Bauer and Hammig, 2014). Similarly, organizations are likely to reap many benefits by providing to new entrants fresh work trends at the workplace where they feel comfortable sharing their ideas and feel valued (Bhatnagar, 2014; Kim, 2014). Limitations and suggestions for future research There are several limitations that should be considered when assessing this study's findings. First, the data collected at one point in time for all the variables, limiting the causal inferences that can be drawn based on our findings. For stronger and more convincing results, future research should consider the use of experimental, quasi-experimental or longitudinal design. Also, in order to reduce concerns about common method bias, data were collected by using two different sources: self and supervisory reports (Podsakoff et al., 2003). Second, the study was conducted in Pakistan. For the generalizability of the study's findings, these results should be validated in other cultural settings because Pakistan is very different in terms of socio-political, cultural and legal aspects in comparison to Western countries (Parker et al., 2001). Finally, this study only focused on one side of the interpersonal support: the critical role that supervisors play in determining employees' workplace perceptions, motivations, attitudes and behaviors. The support employees get from their peers was not considered in this study but may also be a crucial source that moderates the relationship between job scope and performance. Co-worker/peer support involves actions by co-workers that are either helpful or intended to be helpful (Hongvichit, 2015). This study proposed that the relational context of jobs, particularly in the form of social support by supervisors, can motivate employees to perform their duties more productively. It also highlights how the structure of an employee's work with the manifestation of social support plays a critical role in shaping the employee's relationships with their supervisors. This study highlights that both job design and in-role performance have a strong link with relational contexts that can motivate employees to achieve their stated targets. Thus, this research improves our understanding of how social context at the workplace can make a difference for employees and their organizations.
|
The results support the role of supervisor support as a moderator in the relationship between in-role performance, a dimension of job performance and job scope. The findings show that a higher job scope would facilitate higher job performance from employees who receive high levels of supervisor support.
|
[SECTION: Value] Organizations of various sectors are not giving proper attention toward the importance of social context in highly challenging working environment for employees. The social context in highly challenging work settings plays a vital role in shaping employees' proficiencies and behaviors. Social context at work captures the interpersonal interactions that develop as employees perform their jobs, roles and tasks (Chou, 2015; Grant and Parker, 2009). The relational perspective of work design focuses on how the social aspects of a job combined with job roles and/or tasks advance an employee's understanding of the job and how this understanding can impact organizational outcomes (Grant and Parker, 2009). In this regard, social exchange theory is a very useful approach for integrating relational or social work context variables into JCM. The JCM posits that five core job characteristics make jobs intrinsically motivating and satisfying and encourage the achievement of high performance (Hackman and Oldham, 1980). Despite the popularity and continuous use of JCM in job enrichment literature, the model has received multiple criticisms from other scholars (Grant et al., 2010). Two main critiques of the model stand out. First, although the JCM advocates that "Growth Need Strength" (GNS), a trait associated with the need for achievement, moderates the relationship between job characteristics and outcome variables, empirical research has not been conclusive about the moderating effects of GNS (Johns, 2006). Second, the commonly used job design measure in the model, known as the Job Diagnostic Survey (JDS) (Hackman and Oldham, 1980) focused on narrow set of motivational job characteristics which may be less applicable in current work contexts (Parker et al., 2001). If scholars continue to ignore broader range of work design literature, the JCM model runs the risk of being deficient (Morgeson and Humphrey, 2006). More recently, studies in the theory and research of work design have emphasized the importance of the relational aspects of job design (Morgeson and Humphrey, 2006). More specifically, they have suggested that there is a need to explore the role of social support in modifying the effects of job characteristics on outcomes. The research on social networks points out that interpersonal relationships often enhance employees' motivations, opportunities and resources (Frazier and Tupper, 2016; Hongvichit, 2015). Economically constrained research communities require a careful assessment of responsibilities and options for researchers and research subjects alike (Rafique, 2015). Very few research studies are available for managers and practitioners in Pakistan at job design and its affirmative relatedness to enhance the performance of employees in various sectors of Pakistan. Thus, managers in Pakistan are facing hindrances in finding the ways to improve the quality of job design through focusing on various factors which ultimately boost the performance of employees (Ali and Zia-ur-Rehman, 2014). Drawing upon social exchange theory and related literature, in this paper we explore how supervisor support, an important aspect of social context, can act as a moderator in JCM. We propose that social support is a critical factor that will enhance the five core job characteristics for individuals in organizations. Strong social support should encourage employees to take up the challenges associated with high scope jobs and lead to improved productivity. The JCM framework of Hackman and Oldham (1975) includes the five job characteristics that influence work-attitudes and behaviors: skills variety, task identity, task significance, autonomy and feedback. They also suggested that GNS, a trait associated with internal energy achievement orientation, acts as moderator in the job characteristics/outcomes relationship. Although JCM continues to enjoy strong support among scholars for the proposed direct relationship between core job characteristics and outcomes, there is mixed evidence confirming the moderator effects of GNS or other related variables such as personality (e.g. Xie and Johns, 1995). We will first briefly review the research on the relationship between job characteristics and job performance and then discuss the proposed moderating effects of supervisor support in shaping this relationship. Relationship of job characteristics with job performance Standard job performance states the degree to which a management employee accomplish organizationally prescribed work role prospects (Katz and Kahn, 1966). In order to get efficiency at workplace, inventive job performance considered as critical element for organizations (Kanter, 1988). Different research studies conducted in the field of Psychology and Organizational Behavior propose that individual job performance gets affected by characteristics of the work environment, which may in turn interact, with employees' personal characteristics (Hyatt and Prawitt, 2001). The Job Characteristics Model (Hackman and Oldham, 1976) reinforced the value of job enrichment that generates positive effects on work outcomes such as job performance, job satisfaction, motivation and work effectiveness. Job performance is one of the most important individual behaviors that concern organizations and the relationship between job characteristics and job performance has caught the attention of many researchers seeking to enhance scholarly and practical understanding (Schleicher et al., 2004). Furthermore, previous research was mainly based on data collected in the developed Western countries, and there is still a need to test and analyze the job characteristics in different cultural settings to test the efficacy and thus generalizability of JCM. According to Parker et al. (2001), untraditional cultural settings could affect a study's findings. With this need in mind, our study was conducted in Pakistan, a country very different from Western countries in terms of socio-political, cultural and legal aspects. Designing jobs and roles appropriately are imperative in elevating the performance of employees, which is addressed through job characteristic model. So this motivates us to test that how job design can enhance the performance of the employees (Ali and Zia-ur-Rehman, 2014). Designing positions with enriched job characteristics may help with the development of effective performance in a challenging organizational environment. Enriched jobs often promote the creativity that leads to high job performance (Elsbach and Hargadon, 2006). The paper is based on a questionnaire survey. The data were subjected to t-test and correlation analysis using SPSS software. Thus, that was an empirical study. In quintessence, enriched and complex jobs are associated with positive attitudinal/behavioral outcomes. Therefore, we suggest the following hypothesis: H1. Job scope will be positively related to job performance. The moderating role of supervisor support Meta-analytic studies suggest that a substantial amount of variance in job performance is not explained by job characteristics. For example, Fried and Ferris (1987) indicated that the five dimensions of job characteristics are all weakly related to job performance. In a more recent meta-analysis, however, Humphrey and colleagues (2007) found a strong relationship between job characteristics and job performance. This variation between scholars in the observed strength of the relationship between core job characteristics and job performance may be indicative that moderators are at work (Parker and Wall, 1998). In this inquiry, we explore the specific moderating role of supervisor support. Social exchange theory asserts that employees making contributions to the organization in the form of effort and time expect that these investments will be reciprocated by the organization. One way to reciprocate is the provision of support through manager to the employees which they feel that supervisors value their contributions and care about their well-being (Chou, 2015; O'Donnell et al., 2012). Social support is positively linked with different work outcomes (Morgeson and Humphrey, 2006), and also acts as an important buffer against job-related demands (Thomas and Wright, 2013). According to recent findings, employees who receive career-related suggestions and psychological support from their supervisors reciprocate by stretching their performance and going that extra mile (Sun et al., 2014). For employees to be able to cope with a challenging work environment characterized by a need for continuous change and innovation, appropriate supervisor support and guidance is crucial so that employees can accomplish challenging goals with high levels of performance (Carter et al., 2013). Thus, supervisor support in a challenging work environment operates as a pivotal catalyst to improve the in-role performance of employees (Rooney et al., 2008). Supervisor support depends on the interpersonal skills of supervisors and in a challenging job environment it displays in terms of credibility and a deep concern for their subordinates' needs (Hongvichit, 2015). In challenging job situations, strong social bonds between management and employees reflects that supervisors, through their support, are able to adjust individual job conditions in order to facilitate better in-role performance. Therefore, we suggest the following hypothesis: H2. Supervisor support moderates the relationship of job scope with job performance such that the relationship will be stronger when supervisor support is high. Data collection and sample This study was conducted in Pakistan, where people attitude toward research is not serious. Thus, for the collection of quality data it was decided that those sectors should be focused from where the chance of collection of quality data would be higher. Different sectors were chosen to look at the variability of responses regarding job design in them. Various professions were represented in the sample, with respondents belonging to all designations levels ranging from entry level office work to top management for checking the impact of supervisor support in challenging work environment at broader level. The questionnaires were distributed by hand and through mail as well. Data were collected through survey distributed to eight different organizations of five miscellaneous sectors to capture maximum variations. The business activities and operations of the organizations include the banks/financial institutions, the educational sector, the health sector, water and power development authority and Pakistan International Airlines. Reminders were given to the respondents of that concerned organization to fill out questionnaires. Total of 640 distributed questionnaires yielded 426 returns for a response rate of 66 percent. After removing incomplete questionnaires and ones with missing peer-reports, 328 complete useable pairs of responses were available for analyses. To avoid problems associated with common method bias and social desirability, we collected the data from two different sources: self and supervisor-reports. While job scope characteristics, supervisor support and all demographic data were measured with self-reports and job performance data were collected through supervisor-reports. The self and supervisor report versions of the survey received a similar code to match survey responses. A cover letter accompanied all questionnaires that explained the study's scope, the voluntary nature of the survey, and assured participants of strict confidentiality for the responses. Completed responses were personally collected by the first author. Measures In Pakistan, English is the language of instruction in all higher education institutes and it is the official language of correspondence in companies. In general, white collar workers in Pakistan are proficient in English according to previous workplace research conducted in Pakistan (e.g. Raja and Johns, 2010). Therefore, it was decided to distribute the questionnaires in English. Job scope Job scope was measured with the self-reported measures from the JDS (Hackman and Oldham, 1976). Sample items with anchors of 1= very little and 7= very much included for all five characteristics. The reliability of overall job scope was 0.86. In-role performance Job performance was measured using supervisor-reports based on the seven-item measure developed by Williams and Anderson (1991). Five point Likert scale used to measure the responses of respondents. The job performance had an excellent Cronbach's a coefficient of 0.89. Supervisor support We measured supervisor support using the nine-item scale developed by Greenhaus et al. (1990). Sample items with anchors of 1= strongly disagree and 5= strongly agree. The reliability of this measure was 0.83. Control variables Information about demographic factors were collected through self-reports. One-way analysis of variance (ANOVA) revealed significant differences between organizations and age in performance. Therefore, we controlled for the effects of organizations and age in analyses. Confirmatory factor analyses The factor structure of the variables was checked through confirmatory factor analysis (CFA) using AMOS 22 with maximum likelihood estimation. CFA was conducted for each measure in order to examine the model fit and the uniqueness of the measures. CFA analysis ran on a five-factor model of job scope, which resulted in a very good model fit (kh2=194, df= 99, RMSEA=0.04, CFI=0.96, GFI=0.94, AGFI=0.90). The results of the CFA for the one-factor model of performance revealed that the one-factor model produced satisfactory fit (kh2=159, df=27, CFI=0.79, GFI=0.94, AGFI=0.92, RMSEA=0.07). The results of standardized factor loadings indicated that all the items loaded above 0.30. The results of the CFA for the one-factor model of self-reported supervisor support revealed that the one-factor model yielded better fit (kh2=87, df=12, CFI=0.86, GFI=0.94, AGFI=0.90, RMSEA=0.03). The results of standardized factor loadings indicated that all the items loaded above 0.30. Results Table I displays the mean values, standard deviations and correlation between the variables. The highest correlation obtained in the current study was between job scope and in-role performance (r=0.46, p<0.001). With respect to the associations between job scope, supervisor support, job scope correlated significantly with supervisor support (r=0.34, p<0.01). We used simple multiple regression analysis to test H1. Results of this analysis are displayed in Table II. The relationship of job scope with in-role performance was positively significant (b=0.45, p<0.001) and explained 17 percent (DR2=0.17, p<0.001) variance in in-role performance. In order to test H2, we used the regression method suggested by Preacher and Hayes (2004), which deploys bootstrap models to test moderation. Bias-corrected confidence intervals (CI) were computed at two selected levels of supervisor support, with 5,000 random samples and replacement from the full sample. Results of this analysis are presented in Table III, which show that the interaction term of job scope and supervisor support was significant in predicting in-role performance (B=-0.31, t=-4.0, p<0.01). The formal two-tailed significance test which assumed normal distribution demonstrated that the conditional direct effect was significant (Effect=0.49, Boot SE=0.02, t=5.4, p<0.001). Bootstrap results confirmed these effects with a bootstrapped 95 percent CI around the conditional direct effect not containing zero (0.11, 0.31). Figure 1 shows that the relationship of job scope with in-role performance was positive and high when supervisor support was high and low when supervisor support was low. Therefore, H2 received support. Overall, there were two hypotheses and these hypotheses were confirmed. More specifically, our results pertaining to the hypothesized relations can be summarized as follows; all predictions related to direct relations of job scope with in-role performance, interaction effects of job scope and supervisor's support were confirmed. Since the emergence of studies on job design, researchers have failed to properly recognize the significance of interpersonal interactions as a crucial aspect that contributes to the meaningfulness of the job (Humphrey et al., 2007). This paper attempts to address this gap by considering the role of supervisor support as a moderator. In support of our hypotheses, job scope was found to be positively related with in-role performance. Furthermore, this positive relationship is more pronounced when supervisor support is strong. Theoretical implications Our study makes key theoretical contributions to our current knowledge about enriched job and the supervisor role to enhance performance of employees. First, we contribute to the research by highlighting two major determinants of performance simultaneously including "system factors" and "person factors" The former factor is related with organizational environment such as work design, later aspects is related with motivation of individuals through different sources such as supervisor support or co-worker support that may affect individual performance (Williams, 2002). Second, the results from our study reveal a positive relationship between job scope and in-role job performance for those employees who are experiencing high levels of supervisor support. Hence, supervisor support seems to be a boundary condition for the positive influence that high job scope has on in-role performance (Grant and Parker, 2009). By incorporating social/relational characteristics of the work context into theories of job design, we have responded to earlier calls of scholars to extend JCM by considering characteristics that are pivotal in many current jobs (Grant and Parker, 2009). Hence through social support theory it is proved that employees will outperform in challenging jobs when they are receiving managerial/supervisor support (Eisenberger et al., 2002). Third, previous research studies were mainly based on data collected in the developed Western countries, and there is still a need to test and analyze the job characteristics in different cultural settings to enhance the generalizability of JCM. Parker et al. (2001) advocated for untraditional cultural settings for generalizability of study's findings. Keeping in view such factors in mind we conducted this study in Pakistan, a country very different from Western countries in various terms. Practical implications The findings from this inquiry offer useful insights for managers and consultants. First, it highlights the importance of social characteristics in a job which are likely to influence a range of work outcomes. Increased interaction between supervisor and employees strengthens interpersonal relationships at work and increases opportunities for communication which is a fundamental key to enhancing job performance. Social activities have a positive quality in that they generate feelings of energy, enthusiasm and positive affect while at work (Watson, 2000). Second, strong social support provides opportunities through which employees can acquire assistance from their supervisors. This form of interaction is likely to help employees clarify and make sense of their roles and quickly address their concerns and issues. Third, such kind of interactions provides additional insight to employees about their performance and important feedback for continuous learning and personal development (Berman et al., 2002). Managerial implications Multiple sources for data collection were being adopted. Thus, the results of the study may applicable to the various sectors from where data were collected. The results of the study have provided inputs on how to enhance perceived organizational support of supervisors for employees in highly challenging jobs in various organizations. In public or private sector performance level of employees can be boosted when the employees feel content. That state can only be appeared when informal and formal organization policies like mentoring, socialization and open-door conflict management intend to encourage workplace supervisor supportive policies (Frazier and Tupper, 2016). According to the Job Demand-Resource model framework, in highly competitive work environment quality of the relationship with supervisor matters a lot in the company at various levels, not only in the board room but also with employees at the team or department level, or in focus groups. Such kind of supporting relations of supervisors with employees are crucially important to build commitment and credibility of employees for the various types of organizations (Bauer and Hammig, 2014). Similarly, organizations are likely to reap many benefits by providing to new entrants fresh work trends at the workplace where they feel comfortable sharing their ideas and feel valued (Bhatnagar, 2014; Kim, 2014). Limitations and suggestions for future research There are several limitations that should be considered when assessing this study's findings. First, the data collected at one point in time for all the variables, limiting the causal inferences that can be drawn based on our findings. For stronger and more convincing results, future research should consider the use of experimental, quasi-experimental or longitudinal design. Also, in order to reduce concerns about common method bias, data were collected by using two different sources: self and supervisory reports (Podsakoff et al., 2003). Second, the study was conducted in Pakistan. For the generalizability of the study's findings, these results should be validated in other cultural settings because Pakistan is very different in terms of socio-political, cultural and legal aspects in comparison to Western countries (Parker et al., 2001). Finally, this study only focused on one side of the interpersonal support: the critical role that supervisors play in determining employees' workplace perceptions, motivations, attitudes and behaviors. The support employees get from their peers was not considered in this study but may also be a crucial source that moderates the relationship between job scope and performance. Co-worker/peer support involves actions by co-workers that are either helpful or intended to be helpful (Hongvichit, 2015). This study proposed that the relational context of jobs, particularly in the form of social support by supervisors, can motivate employees to perform their duties more productively. It also highlights how the structure of an employee's work with the manifestation of social support plays a critical role in shaping the employee's relationships with their supervisors. This study highlights that both job design and in-role performance have a strong link with relational contexts that can motivate employees to achieve their stated targets. Thus, this research improves our understanding of how social context at the workplace can make a difference for employees and their organizations.
|
This paper makes three key contributions to the literature on job design. First, this inquiry shows that a strong link does exist between job scope and job performance; previous studies have failed to find a strong relationship. Second, it highlights how social context, especially in highly challenging work settings, can shape employees' proficiencies and behaviors. Third, this paper offers a novel perspective in job design research by incorporating a contextual moderator (i.e. supervisor support).
|
[SECTION: Purpose] Family businesses are the most dominant among publically traded firms across the world (Shleifer and Vishny, 1986; Burkart et al., 2003; Anderson and Reeb, 2003; La Porta et al., 1999). In Continental Europe, about 44 per cent of publicly held firms are family-controlled (Faccio and Lang, 2002). In the USA, the equity ownership concentration is modest; among the Fortune 500 firms, around one-third is family firms (Anderson and Reeb, 2003). The concentration of ownership was found to be higher in other developed nations (Faccio and Lang, 2002; Franks and Mayer, 2001; Gorton and Schmid, 1996). Family businesses dominate many developing economies with about two-third of the firms in Asian countries owned by families or individuals (Claessens et al., 2000). In India, around 60 per cent of the listed top 500 firms are family firms (Chakrabarti et al., 2008). These family firms hold large equity stakes and more often than not have family representation on the board of directors. Family equity stake in Indian firms may be divided across individual holdings by promoters and their family members, privately held firms and through cross-holdings from other listed group businesses. The control and influence exerted by family firms may lead to performance difference with the non-family firms (Anderson and Reeb, 2003). The impact of organizational form (separate ownership and control or combined ownership and control) on firm performance has been mixed empirically. Berle and Means (1932) and Jensen and Meckling (1976) asserted that with the separation of ownership and control, there arises a potential for conflict of interest, which leads to the reduction in firm value. Managers may pursue activities in self-interest at the expense of profit maximization creating agency costs. The costs involved in monitoring the management and the difficulty of developing contracts to accurately specify the actions of the managers in the interest of the shareholders leads to lower market valuation of the firm. Contrary to this, Fama and Jensen (1983) argue that separation of ownership and control is advantageous as the efficiency costs outweigh the agency costs. Institutional mechanisms evolve over time to align with the objectives of managers and the shareholders (James, 1999). Firms with combined ownership and control may choose to pursue different investment rule and may have incentive to forgo profitable investment opportunity thereby reducing the firm value (Fama and Jensen, 1985; James, 1999). Active shareholders act as a mechanism for monitoring the firm management. However, shareholders who own very few shares of a firm have little incentive to do so. On the other hand, blockholders or large shareholders who have a substantial investment in a firm can play a key role to mitigate the agency problem. Blockholders can be the family or promoters of the company or institutional investors. Owing to the percentage of votes they hold, blockholders can pressurize the firm management toward improving investor protection (Shleifer and Vishny, 1997). Blockholders can force change the management that fails to deliver the expected results either in an AGM or through proxy fights in the case of a staggered board. A positive correlation between institutional investor's involvement and firm performance has been found in the literature (Agrawal and Knoeber, 1996; Strickland et al., 1996; Shome and Singh, 1995). However, a major problem is that the voting by the institutional investors has been low historically (Hampel, 1998) and those who vote also may be just "Box Tickers" who vote without considering the issues appropriately (Solomon, 2007). Others, like the mutual fund houses and the insurance companies, opt to sell their shares rather than have their say on the management of the firm. Another issue is that these institutional investors may hold a highly diversified portfolio, making it difficult for them to focus on any one company. In some cases, these institutional investors may exert pressure on the management to further their own gains rather than of all the shareholders (Shleifer and Vishny, 1997). This potential tradeoff between high agency cost of monitoring and clash of interest between managers and shareholders may not exist in family firms (James, 1999). Family blockholders can mitigate the agency problem associated with the separation of ownership and management (Villalonga and Amit, 2006; Anderson and Reeb, 2003) and, hence, should have a positive effect on firm value (Berle and Means, 1932). Combining ownership and management can help mitigate managerial opportunism (Demsetz and Lehn, 1985). Family blockholders have a strong incentive, by way of protecting their family wealth, to monitor managers and minimize free rider problem associated with minority shareholders (Anderson and Reeb, 2003; Barontini and Caprio, 2006). At the same time concentrated shareholders can, in exchange for profits, extract private benefits from the firm (Fama and Jensen, 1983; Shleifer and Vishny, 1997; Demsetz, 1983). Thus, concentrated shareholders can generate conflict of interest with minority shareholders (Burkart et al., 2003). The impact of blockholdings by family members on firm performance has yielded mixed empirical results. Family firms were found to perform better than nonfamily firms more so when a family member serves as CEO (Anderson and Reeb, 2003; Maury, 2006). Contrary to these results, family firms were not found to have better firm performance (measured by Tobin's q) as compared to nonfamily firms (Holderness and Sheehan, 1988; Miller et al., 2007). The corporate governance practices of family firms are also found to differ from those of nonfamily firms (Bartholomeusz and Tanewski, 2006). So, does family firms improves firm performance or destroys the firm value? Also, how effective are corporate governance mechanisms in controlling the agency costs in family firms. Is there any significant difference in the relationship between board structure and firm performance among family and nonfamily firms? We explore these questions, by studying the board structure and ownership pattern using a sample of top Indian firms from 2002 to 2012 and comparing family firms and nonfamily firms. Most governance studies on family firms have used data from the developed countries (Jameson et al., 2014). Given the predominance of controlling shareholders in emerging economies, a country level study in the Indian context will help to improve our understanding of the impact of controlling shareholders on corporate governance and firm performance. To provide a context for this study, we briefly describe the institutional background in India. India inherited a financial market developed by the British India and a legal system based on British laws that had excellent provision for shareholder protection (La Porta et al., 1998). Despite these advantages, law enforcement remained a problem. India had its share of issues arising out of poor monitoring of Indian firms, law enforcement, limited information dissemination and licensing requirements. Those issues resulted in the stock market scandals, insider trading issues and allotment of preferential shares to family members at a discount (Khanna and Palepu, 2004; Rajagopalan and Zhang, 2008). Indian legal system today is characterized by long delays due to limitations in resources. A case in point is the bankruptcy resolution, which is deemed to take the longest time to resolve and has the lowest rate of recovery (Kang and Nayar, 2004). Post-independence, India adopted the development strategy of import substituting industrialization. The set of regulations laid out to implement this strategy in the Industries Act 1951 resulted in a system that required businesses to obtain multiple licenses and bureaucratic approvals (Chakrabarti et al., 2008). This system labeled as "license-permit raj" created institutional barriers for smaller businesses, thereby weakening the competition (Reed, 2002). Further, the equity markets were illiquid, and the banks were reluctant to fund private sector due to weak creditor protection. Thus, the business environment benefitted those with government connection while discouraging market-based interactions. This institutional business environment led to the growth of the family business group and concentrated shareholdings (Chakrabarti et al., 2008). These behavioral and institutional differences in emerging markets can provide different insights into corporate governance (Fan et al., 2011). More so for the Asian countries; the diverse institutional development background makes the corporate governance issues unique (Peng and Jiang, 2010). As mentioned in the literature above, Indian firms are characterized by large shareholders and more so by the founding families. This presents a unique setting for studying their impact or lack of it on the firm's corporate governance. This study provides, to our knowledge, the first multi-year comparative analysis of the relationship between board structure and firm performance of family firms in India. The paper is subsequently structured as follows. Section 2 outlines the potential benefits and negative effects of family firms. Section 3 details the sample data and the methodology used in the analysis. Section 4 reports the results of the analysis and Section 5 discusses the results and concludes the paper. Controlling blockholders like family can have potential benefits and competitive advantage. The extent literature on family firms focuses mostly on the agency problem. The problem in widely held firms includes limited ability to select reliable agents, monitor the selected agents and ensure performance (Fama and Jensen, 1983). An effective monitoring can improve firm performance and reduce agency cost (Fama and Jensen, 1983; Jensen and Meckling, 1976). The principle agent conflict associated with widely held firms may be reduced by concentrated blockholders, specifically the family blockholders, as they have a significant economic incentive to monitor the management (Demsetz and Lehn, 1985). Specifically, the substantial intergenerational ownership stake and having a majority of their wealth invested in a single business, the family firms have strong incentive to monitor the management. The lengthy involvement of family members in the business, in some cases spanning generations, permits them to move further along the learning curve. This superior knowledge allows the family members to monitor the managers better (Anderson and Reeb, 2003). Also, this long-term presence of family members in their firm leads to longer investment horizons than other shareholders and may provide an incentive to invest according to market rules (James, 1999). This willingness of family firms to invest in the long-term project leads to greater investment efficiency (Anderson and Reeb, 2003; James, 1999; Stein, 1988). Another advantage of having a long-term presence of family firms is that the external suppliers, dealers, lenders, etc., are more likely to have favorable dealings with the same governing bodies, owing to the long-term dealings and reputation, than with nonfamily firms. This sustained presence of family necessitates having their reputation maintained (Anderson and Reeb, 2003). Thus, family firms have several advantages while monitoring managers, superior knowledge, longer investment horizons and the need to pass on the business to future generation. Consistent with this argument, several studies have shown that family firms have superior performance compared to nonfamily firms. In large public US firms, stronger firm performance was found in family firms compared to nonfamily firms (Anderson and Reeb, 2003). In Japan, the family firms owned or managed by the founder showed superior performance (Saito, 2008). In a study of 13 Western European countries, it was found that family control was associated with higher valuation (Maury, 2006). In listed Swiss firms, Isakov and Weisskopf (2014) found that family firms were more profitable than non-family firms. The controlling families are often involved in the management of the firms, reducing the agency problem of owner-manager conflict. Combining ownership and managerial control can help mitigate the managerial expropriation (Demsetz and Lehn, 1985). This nonseparation of management and ownership also reduces the agency costs involved in establishing the mechanism to monitor the management (Garcia-Ramos and Garcia-Olalla, 2011). Morck et al. (1988) suggest that founder CEOs bring in innovation and value enhancing expertise. Family firms intend to maintain continuity by passing on the business to the next generation of the family (Casson, 1999; Chami, 1999). Thus, to improve the viability of the company for a longer time horizon, the present generation family managers may take decisions to maximize the long-term firm value. The empirical results on the impact of family management on firm performance have been mixed. In the US firms, CEOs who are family members had a positive relationship with accounting measures of performance (Anderson and Reeb, 2003; McConaughy et al., 1998). Family firms, where the founding family is on the supervisory or the executive board, showed better firm performance in German firms (Andres, 2008). Villalonga and Amit (2006) found that family ownership creates value when a family member serves as CEO or the chairman of the firm. The literature suggests that having a family CEO may be beneficial to the value of the firm. Family firms are also said to be fraught with nepotism, family disputes, capital restrictions, exploitation of minority shareholders and executive entrenchment, all of which adversely affect firm performance (Allen and Panian, 1982; Chandler, 1990; Faccio et al., 2001; Gomez-Mejia et al., 2001; Perez-Gonzalez, 2006; Schulze et al., 2001, 2003). Expropriation of minority shareholders by large shareholders may generate additional agency problem (Faccio et al., 2001). Concentrated shareholders, by virtue of their controlling position in the firm, may extract private benefits at the expense of minority shareholders (Burkart et al., 2003). Capital expenditure can be affected by the families' preference for special dividends (DeAngelo and DeAngelo, 2000). Employee productivity can also be adversely affected by family shareholders acting on their own behalf (Burkart et al., 1997). Family firms are found to show biases toward business ventures of other family members resulting in suboptimal investment (Singell, 1997). Family shareholders tend to forgo profit maximization activities owing to their financial preferences which often conflicts with those of minority shareholders (Anderson and Reeb, 2003). The empirical evidence has also backed these negative aspects of controlling shareholding as family firms have been found to have an adverse effect on firm performance. Family ownership structure in US firms is perceived as less efficient and profitable than dispersed ownership (Anderson and Reeb, 2003; Morck, 2007; Singell, 1997). Holderness and Sheehan (1988) using Tobin's q as performance measure found that family firms have lower firm performance than nonfamily firms. The proportion of family shareholdings has also been found to have an impact on the firm performance. Anderson and Reeb (2003) show that when the family ownership exceeds 30 per cent, the gains from family control starts to taper off. Studies from other countries have also provided evidence for the adverse effect of family ownership on firm value. In Canadian and Swedish firms, family shareholders were found to have a negative effect on firm performance (Morck, 2007; Cronqvist and Nilsson, 2003). In East Asian countries where transparency is low, the family firms are shown to harm minority shareholders (Faccio et al., 2001). Preference for family members for the executive positions in family firms suggests a restricted talent pool for selection of qualified candidates (Anderson and Reeb, 2003). Family CEO may also lead to resentment among other senior outside executives of the firm as talent, merit and tenure are not seen as criteria for top management positions (Schulze et al., 2001). This limitation in talent pool may lead to competitive disadvantage for the family-run businesses when compared to nonfamily firms. Also, the accountability of a family CEO toward minority shareholders may be less than that exhibited by professional managers (Gomez-Mejia et al., 2001). The role of family members in selection of managers and the desire to remain in control of the firms may lead to greater managerial entrenchment (Gomez-Mejia et al., 2001). Shleifer and Vishny (1997) opine that the greatest cost imposed by large shareholders is by continuing to run the business in spite of being incompetent and unqualified. The mixed results on the impact of the family firms on firm performance raise several questions. Does family firms perform better compared to non-family firms? How do family firms impact the relationship between board structure and firm performance? Does having a family CEO improve firm performance compared to a professionally managed firm? Does having a family CEO impact the relationship between board structure and firm performance? Having analyzed the difference between family and nonfamily firms, the following question can be asked for within family firms analysis. Within the family firms, does having a higher proportion of family ownership improve firm performance? How does higher proportion of family representative directors impact the firm performance of family firms? Again, does having a professional CEO improve firm performance compared to family CEO in family firms? How did family firms cope with the recession period? Also, does having a larger board, higher proportion of independent directors and independent chairman improve the performance of family firm? These are some of the questions that we address in this study. 4.1 Sample For our investigation, we use top Indian firms listed on the Bombay Stock Exchange (BSE), specifically the firms forming a part of the BSE 200 index. We exclude financial institutions and public sector units as strong government regulations may potentially affect performance of these firms. Firm-specific control variables were collected from the ACE Equity database. Governance data comprising of board characteristics such as independent directors, chairman independence, role duality, non-executive directors, board busyness, board size, board meetings and board attendance, was manually collected from the annual reports of the firms from ReportJunction database. As per Clause 49 of listing agreement, firms are required to file the corporate governance report quarterly; however, for our empirical analysis, we considered annual performance of the firm and select the financial year-end corporate governance reports. We excluded the firms where the governance data was not available for more than one year. Our final sample consisted of 100 firms yielding 1,100 firm-years or observations as our sample was restricted by the publically available data from 2002 to 2012. The Variance inflation factor (VIF) was calculated for each independent variable, and all VIF values were less than 10 suggesting that there is no multicollinearity issue (Myers, 1990). 4.2 Variables and data source We identify the family firms based on the fractional equity ownership of the family members and the presence of family members on the board of directors. We were also guided by the classification of the family firms for listed Indian firms given by the ACE Equity database. We use a dummy variable that equals one for family firms and zero otherwise. Our study incorporates variables such as family management, family shareholdings and family representative directors on the board. We measure family management using a dummy that indicates the presence of a family CEO. Family shareholdings represent the percentage of equity held by the family members in the firm. Family representative directors are the proportion of family members on the board. We also study the recession period and the impact it had on the family firms. According to National Bureau of Economic Research (NBER), the recession that began in December 2007 officially ended in June 2009. We use a dummy variable to specify the period as pre-recession from the year 2002 to 2008 with a value 0, and recession period from the year 2008 to 2012 with a value 1 (Table I). 4.3 Dependent variable For our analysis, we use accounting measures of performance namely return on equity (ROE) and return on capital used (ROCE). As several studies on family firms use Tobin's q as a measure of firm performance, we also use Tobin's q measured using the book value of debt plus market value of common stock divided by the book value of assets as a proxy for Tobin's q (Black et al., 2014; Khanna and Palepu, 2000). 4.4 Explanatory variables We study the role played by the directors of a firm by categorizing the various board-related factors into two broad categories, namely, board leadership and board activity. On the basis of the agency theory, we investigate the board leadership roles played by the directors of the firm and its impact on the firm performance. Board activity is examined using the resource dependency theory. Board Leadership comprising of factors relating to the board leadership that includes role duality, independent directors, chairman independence and non-executive directors and Board Activity is measured by variables: board meeting, board size, board attendance and board busyness. We operationalize the various board parameters as follows: For our board index, based on the percentage of Independent Directors on the board, we divide the firms into three sets, the top, middle and bottom, and we respectively assign a score of 0, 1, 2 for <50 per cent, 50-75 per cent and >75 per cent independent directors on the board. Similarly, we assign the scores of 0, 1 and 2 for Non-Executive Directors on the board. For Role Duality, we assign a score of one for nonduality (splitting of the role of CEO and chairman) and zero for role duality. Similarly, for Chairman Independence, we use dummy variable by assigning a score of 1 if the chairman is independent, and zero otherwise. The dummy for Board Size is assigned a score of zero if board size is less than eight and greater than 15 and one otherwise. For the number of Board Meetings, we use a dummy variable that is equal to zero if the number of board meetings is less than four, equal to one if board meetings are between four to eight and two otherwise. For Board Attendance, we assign a score of 1 if the average board attendance is greater than 75 per cent and zero otherwise. To measure Board Busyness we assign a value zero if average outside board membership is greater than 10; one if it is between five and 10 and two if it is less than two. The governance index is constructed by adding the points for each board parameter discussed above. First, we divide the board parameters into two broad sets, namely, board leadership and board activity. We then assign weights to the two sets to form governance index. We assign more weigh to board leadership by assigning 60 per cent weightage in the overall score. This index is then converted to dummy variables by assigning a value of 1 to the top 60 per cent of the firms having the strongest board parameters and zero otherwise. We do this to compare the top firms with bottom firms with respect to the board parameters. Though this simple index may not reflect the impact of individual board parameters, it does, however, help in distinguishing between the firms with a stronger board and those with weaker board parameters. We use the board index (dummy variable that takes the value of 1 for the top 60 per cent of the firms having the strongest board parameters and zero otherwise) as the explanatory variable. Additionally, we also use the proportion of independent directors, board size and chairman independence (dummy variable that takes the value 1, if the chairman is independent else 0) as explanatory variables in our subset analysis. 4.5 Control variables To control for industry and firm characteristics, we use several control variables in our analysis, namely, leverage (ratio of debt to total assets), firm age (number of years since the date of incorporation) and firm size (log of total assets), sales growth (moving geometric average of net sales growth over the past three years), asset tangibility (fixed asset to total asset), stock volatility (4 year average of monthly standard deviation of stock price returns) and Industry classification based on two-digit National Industrial Classification (NIC) code. The data for these financial parameters were collected from ACE Equity database (Table I). 4.6 Method In our study, we observe the same firm at different point of time, and hence, we use the random effects model for our regression analysis (Arellano and Bond, 1991; Blundell and Bond, 1998). Black et al. (2014) opines that in developing countries a good approach to study this kind of relationship would be to build a panel data and use fixed or random effects model. Our primary interest is in analyzing the relationship between board structure and firm performance in a family firm. The analysis also includes additional variables such as family management, family shareholdings and family representative directors on the board. Serial correlation and heteroskedasticity are controlled by using the Huber White Sandwich Estimator (clustered) for the variance. Panel data analysis was conducted using the R statistical package (Version 3.1.1) with add-on package "PLM" (Croissant and Millo, 2008). The regression we use for the multivariate analysis takes the form: (1) Performance=b0+b1(Family Firm)+b2(CG Score)+b3(Family Firm) * (CG score)+b4(Control Variables)+b5-34(Two Digit SIC Code)+e where: Firm Performance is the accounting measures of performance ROA and ROE and Tobin's q. Family Firms is the dummy variable with value 1 for family firms, and zero otherwise. Control Variables are leverage, age, firm size, sales growth, tangibility, stock volatility, two-digit NIC code. 5.1 Summary statistics Descriptive statistics for the sample data shown in Table II is broken down into the two groups: family and nonfamily firms. The last two columns show the difference of means and the t-values between the family and nonfamily firms. Family firms constitute 73 per cent of our sample. Among family firms, about half of the sample had promoter CEO or CEO from the promoter family and about 30 per cent firms had both the CEO and the chairman from the founder family. This suggests an active participation of founders and founder families in the management of Indian firms. The governance characteristic of our sample shows that the mean independent directors for family firms (5.423) is higher than that for nonfamily firms (4.878), and the difference is significant at 1 per cent level. The mean independent chairman and mean role duality of nonfamily firms is higher than family firms and the difference is significant at the 1 per cent level in both cases. There is not much of a difference in the mean board size and non-executive directors between the two groups. The mean for the frequency of board meeting is slightly higher for family firms, but the mean board attendance is higher for nonfamily firms. The mean board busyness for family firms (5.206) was significantly higher compared to nonfamily firms (4.424). The firm level characteristics are reported in Table II. The mean performance (ROE, ROCE and Tobin's q) of nonfamily firms is significantly higher than family firms indicating that nonfamily firms on an average perform better than family firms. Consistent with this observation we find that the net sales growth is also higher for nonfamily firms. Family firms exhibit higher stock volatility compared to nonfamily firms. Family firms are leveraged more compared to nonfamily firms. The mean total assets for family firms are also significantly higher than those for nonfamily firms. These results suggest that the nonfamily firms are better performers than family firms. 5.2 Multivariate analysis The result of the panel data analysis to examine the impact of board index on firm performance in family firms is presented in Table III. The dummy variable board index takes the value of 1 for the top 60 per cent of the firms having the strongest board parameters and zero otherwise. We divide the firms into two groups representing the family and nonfamily firms using a dummy variable, which equals one for family firms and zero for nonfamily firms. The interpretations of the coefficients of the regression results are carried out as follows: The coefficient of the interaction term (family firm x board index) represents the incremental effect of board index on the performance of family firms with respect to the reference group non-family firm. Similarly, the coefficient of the board index represents the incremental effect of board index on the reference group non-family firms. The incremental effect of board index on the group family firm is given by the sum of the coefficients of board index and coefficient of the interaction between the family firm and board index [Board Index + (FF x Board Index)]. From Table III Column 1, we find that the interaction variable (Family Firms x Board Index) is negatively and significantly associated with firm performance measured by both ROE and ROCE. This result suggests that the incremental effect of the board score, for family firms with respect to nonfamily firms, is having a negative impact on the firm performance. The coefficient of board index shows that there is a positive effect of board index on firm performance of the nonfamily firms. Thus, the incremental effect of board index on family firm (b for ROE = 3.789 + (-4.439) = -0.65) is negative. Similar results are shown with the other two performance measures ROCE and Tobin's q (Columns 3 and 5). Thus, it is shown that for family firms, any improvement in board score decreases the firm performance. This result could be because the family firm management may endeavor for the improvement of the board structure as a reaction to the poor firm performance in the past. In Columns 2, 4 and 6 of Table III, the Family Firm indicator variable of Columns 1, 3 and 5 is replaced by the indicator variable Family CEO. When we segregate the firms based on the family CEO and nonfamily professional CEO managed firms, we find no significant difference in the relationship between board index and firm performance. This shows that having a family CEO or professional CEO does not make any significant impact on the relationship between corporate governance and firm performance. This result is contrary to the studies which find that family firms managed by family CEO tend to be less profitable than professionally managed firms (Gomez-Mejia et al., 2001; Singell, 1997). Our results are also contrary to studies that have showed positive relationship between family management and firm performance (Anderson and Reeb, 2003). Our results provide evidence on the irrelevance of family management on the relationship between board structure and firm performance in large firms in Indian context. The control variables, leverage, was found to be negatively and significantly associated with firm performance for both the models and all the three performance measures which are consistent with other studies (Titman and Wessels, 1988; Friend and Lang, 1988). Firm age was found to have a positive and significant effect on firm performance indicating experienced firms have an edge over newer firms. Firm size was negatively related to firm performance indicating that larger firms are less profitable. Sales growth was positive and significant. Asset tangibility was found to be negative, significant only in case of ROCE and Tobin's q as the performance measure. Stock volatility as expected was negatively related to firm performance. In Table IV, we also analyze the result of subsample containing only the family firms to study the effects of having professional CEO for the family firm, the proportion of family ownership and family representative directors on the firm performance. Columns 1, 5 and 9 show the results for family CEO for performance measures ROE, ROCE and Tobin's q respectively. Similar to our earlier findings, the results for family CEO with respect to professional CEO did not show any significant difference for all the three measures of performance. Thus, in a family firm having a professional CEO does not seem to impact the relationship between board score and firm performance. This result confirms to the study done by others that find no difference between family-managed and professionally managed firms (Daily and Dalton, 1992; Willard et al., 1992). This result may be possibly due to professional CEO being socially connected to the controlling family of the firm who controls the actions of the CEO. Another reason could be that the role of family members in the selection of managers and the desire to remain in control of the firms may lead to greater managerial entrenchment (Gomez-Mejia et al., 2001). Next, we analyze the results for the proportion of family shareholding. It is expected that the substantial intergenerational ownership stake and having a majority of their wealth invested in a single business present a strong incentive for the family firms to monitor the management (Demsetz and Lehn, 1985). Thus, having higher family ownership and the need to pass on the business to the future generation, it is expected that the firm performance would be positively associated with family ownership. Columns 2, 6 and 10 in Table IV show the results for family ownership. Within the family firms, having a higher proportion of shareholdings by family members did not show any significant impact on the firm performance. This result could be because despite having less than majority shareholding, the family still retains control in the firm by virtue of highly dispersed outside shareholdings or by being in the management of the firm. We also examine the impact of family members on the board of directors. Advocates of stewardship theory propose that the concentration of power and authority to a single person in a firm will lead to better performance (Donaldson and Davis, 1991). Thus, having more members in the board would strengthen family control over the management of the firm. Alternatively, the tendency of families to forgo profit maximization activities owing to their financial preferences (Anderson and Reeb, 2003) and biases toward business ventures of other family members (Singell, 1997) may lead to lower firm performance. However, as reported in Table IV, Columns 3, 7 and 11, we find that for family representative directors, there is no significant association with all the three measures of performance. To analyze the impact the recession period had on the performance of the family firms, we add the coefficient for board score and board score interacted with recession dummy variable for the sample of family firm. Columns 4, 8 and 12 of Table IV presents the effect of the recession on the performance of family firms. The results show a negative effect of recession compared to non-recession period; however, it was not significant. This could be due the caution exhibited by the controlling family, as family wealth would be at stake in the case of any substantial losses to the firm. Also, to improve the viability of the company for a longer time horizon, the present generation family managers may take decisions to maximize the long-term firm value. Again, the long-term presence and reputation of family firms helps them to have favorable dealings with external suppliers, dealers, lenders, etc., which are likely to continue even during any economic downturn. Hence, no discernable difference in performance is noticed before and during the recession period. Additionally, we sought to study the impact of specific governance mechanism in improving the firm performance in family firms in India. We create subsample based on above and below the median proportion of independent directors, board size and chairman independence. Having more independent directors on the board has generally been associated with better monitoring and improved firm performance. Table V reports the results for the subsample analysis. From Table V, Column 1, we find that median board independence does not show any significant relationship with firm performance. The reason for this result could be that the independent directors may not be truly independent due to allegiance toward the family who got them on the board. The results were similar for subsamples of board size and firms having independent chairman shown in Columns 2-3. Thus, this study does not find any evidence that having a higher proportion of independent directors, larger board size or an independent chairman, improves the relationship between family firms and firm performance. In this study, we analyze the interaction effect of the family firms and board governance factors on firm performance in a sample of top publically traded Indian firms. We contribute to the growing literature on family firms by providing a multi-year analysis on the influence of board structure on firm performance in a family firm vis-a-vis nonfamily firm in the Indian context. In this study, we endeavor to expand our understanding of corporate governance and to shed light on the impact of the proportion of shareholding, family representative directors and having a professional CEO in large family firms in India. The result of the panel data analysis shows that the interaction variable (family firms x board score) has a statistically significant negative associated with firm performance measured by both Tobin's q and ROE. This is consistent with the results by Garcia-Ramos and Garcia-Olalla (2011) in the European context. Our result suggests that the incremental effect of the board index score, for family firms with respect to nonfamily firms, is having a negative impact on the firm performance. When we segregate the firms based on the family CEO and professionally managed firms, we do not find any significant difference in the relationship between board structure and firm performance. We obtained similar results with subsample analysis containing only the family firms. This result is consistent with the results of Daily and Dalton (1992) and Willard et al. (1992) confirming the irrelevance of the management structure but contrary to the studies by Gomez-Mejia et al. (2001) and Singell (1997) which find that family firms managed by family CEO tend to be less profitable than professionally managed firms. Our results also contradict the studies that show CEO's who are family members had a positive relationship with accounting measures of performance (Anderson and Reeb, 2003; McConaughy et al., 1998). In the sub-sample analysis containing only the family firms, having a higher proportion of family ownership, family representative directors and having professional CEO for family firms did not show any significant impact on board structure and the firm performance. This result implies that family management and family ownership are irrelevant to the value of the firm. The impact of the recession on family firms was also not found to be significant compared to non-recession period. Contrary to the popular belief, we find that having an independent board and independent chairman does not improve firm performance. Results for the subsample based on above and below the median proportion of independent directors do not show any significant relationship with firm performance. The results were similar for subsamples of board size and firms having an independent chairman. Thus, this study does not find any evidence that having a higher proportion of independent directors, board size or an independent chairman, improves the firm performance in a family firm. The research has several limitations that need to be kept in mind while interpreting the results. This study is limited to the analysis of publically listed top Indian firm and hence further research needs to be conducted on a sample having a mix of large and small firms to improve the generalizability of the results. The difficulty in obtaining reliable data on exhaustive governance variables limits this study to board parameters. A natural extension of this study would be to include a larger number of governance parameters. The widespread pyramiding in Indian family firms makes it difficult to ascertain the quantum of actual family ownership and hence limits the accuracy of any study on family ownership. A cross-country research can also be conducted to analyze how the legal and regulatory institutional differences impact the findings presented in this study. The future extension of this study may analyze the family firms based on the founder led family firms and firms led by the subsequent generation of the founder family.
|
Given the prevalence of family-run businesses in India, this paper aims to empirically investigate the impact of family firms on the relationship between firm performance and board characteristics. The effectiveness of board characteristics such as independent directors, chairman independence, role duality, non-executive directors, board busyness, board size, board meetings and board attendance are studied in the Indian context.
|
[SECTION: Method] Family businesses are the most dominant among publically traded firms across the world (Shleifer and Vishny, 1986; Burkart et al., 2003; Anderson and Reeb, 2003; La Porta et al., 1999). In Continental Europe, about 44 per cent of publicly held firms are family-controlled (Faccio and Lang, 2002). In the USA, the equity ownership concentration is modest; among the Fortune 500 firms, around one-third is family firms (Anderson and Reeb, 2003). The concentration of ownership was found to be higher in other developed nations (Faccio and Lang, 2002; Franks and Mayer, 2001; Gorton and Schmid, 1996). Family businesses dominate many developing economies with about two-third of the firms in Asian countries owned by families or individuals (Claessens et al., 2000). In India, around 60 per cent of the listed top 500 firms are family firms (Chakrabarti et al., 2008). These family firms hold large equity stakes and more often than not have family representation on the board of directors. Family equity stake in Indian firms may be divided across individual holdings by promoters and their family members, privately held firms and through cross-holdings from other listed group businesses. The control and influence exerted by family firms may lead to performance difference with the non-family firms (Anderson and Reeb, 2003). The impact of organizational form (separate ownership and control or combined ownership and control) on firm performance has been mixed empirically. Berle and Means (1932) and Jensen and Meckling (1976) asserted that with the separation of ownership and control, there arises a potential for conflict of interest, which leads to the reduction in firm value. Managers may pursue activities in self-interest at the expense of profit maximization creating agency costs. The costs involved in monitoring the management and the difficulty of developing contracts to accurately specify the actions of the managers in the interest of the shareholders leads to lower market valuation of the firm. Contrary to this, Fama and Jensen (1983) argue that separation of ownership and control is advantageous as the efficiency costs outweigh the agency costs. Institutional mechanisms evolve over time to align with the objectives of managers and the shareholders (James, 1999). Firms with combined ownership and control may choose to pursue different investment rule and may have incentive to forgo profitable investment opportunity thereby reducing the firm value (Fama and Jensen, 1985; James, 1999). Active shareholders act as a mechanism for monitoring the firm management. However, shareholders who own very few shares of a firm have little incentive to do so. On the other hand, blockholders or large shareholders who have a substantial investment in a firm can play a key role to mitigate the agency problem. Blockholders can be the family or promoters of the company or institutional investors. Owing to the percentage of votes they hold, blockholders can pressurize the firm management toward improving investor protection (Shleifer and Vishny, 1997). Blockholders can force change the management that fails to deliver the expected results either in an AGM or through proxy fights in the case of a staggered board. A positive correlation between institutional investor's involvement and firm performance has been found in the literature (Agrawal and Knoeber, 1996; Strickland et al., 1996; Shome and Singh, 1995). However, a major problem is that the voting by the institutional investors has been low historically (Hampel, 1998) and those who vote also may be just "Box Tickers" who vote without considering the issues appropriately (Solomon, 2007). Others, like the mutual fund houses and the insurance companies, opt to sell their shares rather than have their say on the management of the firm. Another issue is that these institutional investors may hold a highly diversified portfolio, making it difficult for them to focus on any one company. In some cases, these institutional investors may exert pressure on the management to further their own gains rather than of all the shareholders (Shleifer and Vishny, 1997). This potential tradeoff between high agency cost of monitoring and clash of interest between managers and shareholders may not exist in family firms (James, 1999). Family blockholders can mitigate the agency problem associated with the separation of ownership and management (Villalonga and Amit, 2006; Anderson and Reeb, 2003) and, hence, should have a positive effect on firm value (Berle and Means, 1932). Combining ownership and management can help mitigate managerial opportunism (Demsetz and Lehn, 1985). Family blockholders have a strong incentive, by way of protecting their family wealth, to monitor managers and minimize free rider problem associated with minority shareholders (Anderson and Reeb, 2003; Barontini and Caprio, 2006). At the same time concentrated shareholders can, in exchange for profits, extract private benefits from the firm (Fama and Jensen, 1983; Shleifer and Vishny, 1997; Demsetz, 1983). Thus, concentrated shareholders can generate conflict of interest with minority shareholders (Burkart et al., 2003). The impact of blockholdings by family members on firm performance has yielded mixed empirical results. Family firms were found to perform better than nonfamily firms more so when a family member serves as CEO (Anderson and Reeb, 2003; Maury, 2006). Contrary to these results, family firms were not found to have better firm performance (measured by Tobin's q) as compared to nonfamily firms (Holderness and Sheehan, 1988; Miller et al., 2007). The corporate governance practices of family firms are also found to differ from those of nonfamily firms (Bartholomeusz and Tanewski, 2006). So, does family firms improves firm performance or destroys the firm value? Also, how effective are corporate governance mechanisms in controlling the agency costs in family firms. Is there any significant difference in the relationship between board structure and firm performance among family and nonfamily firms? We explore these questions, by studying the board structure and ownership pattern using a sample of top Indian firms from 2002 to 2012 and comparing family firms and nonfamily firms. Most governance studies on family firms have used data from the developed countries (Jameson et al., 2014). Given the predominance of controlling shareholders in emerging economies, a country level study in the Indian context will help to improve our understanding of the impact of controlling shareholders on corporate governance and firm performance. To provide a context for this study, we briefly describe the institutional background in India. India inherited a financial market developed by the British India and a legal system based on British laws that had excellent provision for shareholder protection (La Porta et al., 1998). Despite these advantages, law enforcement remained a problem. India had its share of issues arising out of poor monitoring of Indian firms, law enforcement, limited information dissemination and licensing requirements. Those issues resulted in the stock market scandals, insider trading issues and allotment of preferential shares to family members at a discount (Khanna and Palepu, 2004; Rajagopalan and Zhang, 2008). Indian legal system today is characterized by long delays due to limitations in resources. A case in point is the bankruptcy resolution, which is deemed to take the longest time to resolve and has the lowest rate of recovery (Kang and Nayar, 2004). Post-independence, India adopted the development strategy of import substituting industrialization. The set of regulations laid out to implement this strategy in the Industries Act 1951 resulted in a system that required businesses to obtain multiple licenses and bureaucratic approvals (Chakrabarti et al., 2008). This system labeled as "license-permit raj" created institutional barriers for smaller businesses, thereby weakening the competition (Reed, 2002). Further, the equity markets were illiquid, and the banks were reluctant to fund private sector due to weak creditor protection. Thus, the business environment benefitted those with government connection while discouraging market-based interactions. This institutional business environment led to the growth of the family business group and concentrated shareholdings (Chakrabarti et al., 2008). These behavioral and institutional differences in emerging markets can provide different insights into corporate governance (Fan et al., 2011). More so for the Asian countries; the diverse institutional development background makes the corporate governance issues unique (Peng and Jiang, 2010). As mentioned in the literature above, Indian firms are characterized by large shareholders and more so by the founding families. This presents a unique setting for studying their impact or lack of it on the firm's corporate governance. This study provides, to our knowledge, the first multi-year comparative analysis of the relationship between board structure and firm performance of family firms in India. The paper is subsequently structured as follows. Section 2 outlines the potential benefits and negative effects of family firms. Section 3 details the sample data and the methodology used in the analysis. Section 4 reports the results of the analysis and Section 5 discusses the results and concludes the paper. Controlling blockholders like family can have potential benefits and competitive advantage. The extent literature on family firms focuses mostly on the agency problem. The problem in widely held firms includes limited ability to select reliable agents, monitor the selected agents and ensure performance (Fama and Jensen, 1983). An effective monitoring can improve firm performance and reduce agency cost (Fama and Jensen, 1983; Jensen and Meckling, 1976). The principle agent conflict associated with widely held firms may be reduced by concentrated blockholders, specifically the family blockholders, as they have a significant economic incentive to monitor the management (Demsetz and Lehn, 1985). Specifically, the substantial intergenerational ownership stake and having a majority of their wealth invested in a single business, the family firms have strong incentive to monitor the management. The lengthy involvement of family members in the business, in some cases spanning generations, permits them to move further along the learning curve. This superior knowledge allows the family members to monitor the managers better (Anderson and Reeb, 2003). Also, this long-term presence of family members in their firm leads to longer investment horizons than other shareholders and may provide an incentive to invest according to market rules (James, 1999). This willingness of family firms to invest in the long-term project leads to greater investment efficiency (Anderson and Reeb, 2003; James, 1999; Stein, 1988). Another advantage of having a long-term presence of family firms is that the external suppliers, dealers, lenders, etc., are more likely to have favorable dealings with the same governing bodies, owing to the long-term dealings and reputation, than with nonfamily firms. This sustained presence of family necessitates having their reputation maintained (Anderson and Reeb, 2003). Thus, family firms have several advantages while monitoring managers, superior knowledge, longer investment horizons and the need to pass on the business to future generation. Consistent with this argument, several studies have shown that family firms have superior performance compared to nonfamily firms. In large public US firms, stronger firm performance was found in family firms compared to nonfamily firms (Anderson and Reeb, 2003). In Japan, the family firms owned or managed by the founder showed superior performance (Saito, 2008). In a study of 13 Western European countries, it was found that family control was associated with higher valuation (Maury, 2006). In listed Swiss firms, Isakov and Weisskopf (2014) found that family firms were more profitable than non-family firms. The controlling families are often involved in the management of the firms, reducing the agency problem of owner-manager conflict. Combining ownership and managerial control can help mitigate the managerial expropriation (Demsetz and Lehn, 1985). This nonseparation of management and ownership also reduces the agency costs involved in establishing the mechanism to monitor the management (Garcia-Ramos and Garcia-Olalla, 2011). Morck et al. (1988) suggest that founder CEOs bring in innovation and value enhancing expertise. Family firms intend to maintain continuity by passing on the business to the next generation of the family (Casson, 1999; Chami, 1999). Thus, to improve the viability of the company for a longer time horizon, the present generation family managers may take decisions to maximize the long-term firm value. The empirical results on the impact of family management on firm performance have been mixed. In the US firms, CEOs who are family members had a positive relationship with accounting measures of performance (Anderson and Reeb, 2003; McConaughy et al., 1998). Family firms, where the founding family is on the supervisory or the executive board, showed better firm performance in German firms (Andres, 2008). Villalonga and Amit (2006) found that family ownership creates value when a family member serves as CEO or the chairman of the firm. The literature suggests that having a family CEO may be beneficial to the value of the firm. Family firms are also said to be fraught with nepotism, family disputes, capital restrictions, exploitation of minority shareholders and executive entrenchment, all of which adversely affect firm performance (Allen and Panian, 1982; Chandler, 1990; Faccio et al., 2001; Gomez-Mejia et al., 2001; Perez-Gonzalez, 2006; Schulze et al., 2001, 2003). Expropriation of minority shareholders by large shareholders may generate additional agency problem (Faccio et al., 2001). Concentrated shareholders, by virtue of their controlling position in the firm, may extract private benefits at the expense of minority shareholders (Burkart et al., 2003). Capital expenditure can be affected by the families' preference for special dividends (DeAngelo and DeAngelo, 2000). Employee productivity can also be adversely affected by family shareholders acting on their own behalf (Burkart et al., 1997). Family firms are found to show biases toward business ventures of other family members resulting in suboptimal investment (Singell, 1997). Family shareholders tend to forgo profit maximization activities owing to their financial preferences which often conflicts with those of minority shareholders (Anderson and Reeb, 2003). The empirical evidence has also backed these negative aspects of controlling shareholding as family firms have been found to have an adverse effect on firm performance. Family ownership structure in US firms is perceived as less efficient and profitable than dispersed ownership (Anderson and Reeb, 2003; Morck, 2007; Singell, 1997). Holderness and Sheehan (1988) using Tobin's q as performance measure found that family firms have lower firm performance than nonfamily firms. The proportion of family shareholdings has also been found to have an impact on the firm performance. Anderson and Reeb (2003) show that when the family ownership exceeds 30 per cent, the gains from family control starts to taper off. Studies from other countries have also provided evidence for the adverse effect of family ownership on firm value. In Canadian and Swedish firms, family shareholders were found to have a negative effect on firm performance (Morck, 2007; Cronqvist and Nilsson, 2003). In East Asian countries where transparency is low, the family firms are shown to harm minority shareholders (Faccio et al., 2001). Preference for family members for the executive positions in family firms suggests a restricted talent pool for selection of qualified candidates (Anderson and Reeb, 2003). Family CEO may also lead to resentment among other senior outside executives of the firm as talent, merit and tenure are not seen as criteria for top management positions (Schulze et al., 2001). This limitation in talent pool may lead to competitive disadvantage for the family-run businesses when compared to nonfamily firms. Also, the accountability of a family CEO toward minority shareholders may be less than that exhibited by professional managers (Gomez-Mejia et al., 2001). The role of family members in selection of managers and the desire to remain in control of the firms may lead to greater managerial entrenchment (Gomez-Mejia et al., 2001). Shleifer and Vishny (1997) opine that the greatest cost imposed by large shareholders is by continuing to run the business in spite of being incompetent and unqualified. The mixed results on the impact of the family firms on firm performance raise several questions. Does family firms perform better compared to non-family firms? How do family firms impact the relationship between board structure and firm performance? Does having a family CEO improve firm performance compared to a professionally managed firm? Does having a family CEO impact the relationship between board structure and firm performance? Having analyzed the difference between family and nonfamily firms, the following question can be asked for within family firms analysis. Within the family firms, does having a higher proportion of family ownership improve firm performance? How does higher proportion of family representative directors impact the firm performance of family firms? Again, does having a professional CEO improve firm performance compared to family CEO in family firms? How did family firms cope with the recession period? Also, does having a larger board, higher proportion of independent directors and independent chairman improve the performance of family firm? These are some of the questions that we address in this study. 4.1 Sample For our investigation, we use top Indian firms listed on the Bombay Stock Exchange (BSE), specifically the firms forming a part of the BSE 200 index. We exclude financial institutions and public sector units as strong government regulations may potentially affect performance of these firms. Firm-specific control variables were collected from the ACE Equity database. Governance data comprising of board characteristics such as independent directors, chairman independence, role duality, non-executive directors, board busyness, board size, board meetings and board attendance, was manually collected from the annual reports of the firms from ReportJunction database. As per Clause 49 of listing agreement, firms are required to file the corporate governance report quarterly; however, for our empirical analysis, we considered annual performance of the firm and select the financial year-end corporate governance reports. We excluded the firms where the governance data was not available for more than one year. Our final sample consisted of 100 firms yielding 1,100 firm-years or observations as our sample was restricted by the publically available data from 2002 to 2012. The Variance inflation factor (VIF) was calculated for each independent variable, and all VIF values were less than 10 suggesting that there is no multicollinearity issue (Myers, 1990). 4.2 Variables and data source We identify the family firms based on the fractional equity ownership of the family members and the presence of family members on the board of directors. We were also guided by the classification of the family firms for listed Indian firms given by the ACE Equity database. We use a dummy variable that equals one for family firms and zero otherwise. Our study incorporates variables such as family management, family shareholdings and family representative directors on the board. We measure family management using a dummy that indicates the presence of a family CEO. Family shareholdings represent the percentage of equity held by the family members in the firm. Family representative directors are the proportion of family members on the board. We also study the recession period and the impact it had on the family firms. According to National Bureau of Economic Research (NBER), the recession that began in December 2007 officially ended in June 2009. We use a dummy variable to specify the period as pre-recession from the year 2002 to 2008 with a value 0, and recession period from the year 2008 to 2012 with a value 1 (Table I). 4.3 Dependent variable For our analysis, we use accounting measures of performance namely return on equity (ROE) and return on capital used (ROCE). As several studies on family firms use Tobin's q as a measure of firm performance, we also use Tobin's q measured using the book value of debt plus market value of common stock divided by the book value of assets as a proxy for Tobin's q (Black et al., 2014; Khanna and Palepu, 2000). 4.4 Explanatory variables We study the role played by the directors of a firm by categorizing the various board-related factors into two broad categories, namely, board leadership and board activity. On the basis of the agency theory, we investigate the board leadership roles played by the directors of the firm and its impact on the firm performance. Board activity is examined using the resource dependency theory. Board Leadership comprising of factors relating to the board leadership that includes role duality, independent directors, chairman independence and non-executive directors and Board Activity is measured by variables: board meeting, board size, board attendance and board busyness. We operationalize the various board parameters as follows: For our board index, based on the percentage of Independent Directors on the board, we divide the firms into three sets, the top, middle and bottom, and we respectively assign a score of 0, 1, 2 for <50 per cent, 50-75 per cent and >75 per cent independent directors on the board. Similarly, we assign the scores of 0, 1 and 2 for Non-Executive Directors on the board. For Role Duality, we assign a score of one for nonduality (splitting of the role of CEO and chairman) and zero for role duality. Similarly, for Chairman Independence, we use dummy variable by assigning a score of 1 if the chairman is independent, and zero otherwise. The dummy for Board Size is assigned a score of zero if board size is less than eight and greater than 15 and one otherwise. For the number of Board Meetings, we use a dummy variable that is equal to zero if the number of board meetings is less than four, equal to one if board meetings are between four to eight and two otherwise. For Board Attendance, we assign a score of 1 if the average board attendance is greater than 75 per cent and zero otherwise. To measure Board Busyness we assign a value zero if average outside board membership is greater than 10; one if it is between five and 10 and two if it is less than two. The governance index is constructed by adding the points for each board parameter discussed above. First, we divide the board parameters into two broad sets, namely, board leadership and board activity. We then assign weights to the two sets to form governance index. We assign more weigh to board leadership by assigning 60 per cent weightage in the overall score. This index is then converted to dummy variables by assigning a value of 1 to the top 60 per cent of the firms having the strongest board parameters and zero otherwise. We do this to compare the top firms with bottom firms with respect to the board parameters. Though this simple index may not reflect the impact of individual board parameters, it does, however, help in distinguishing between the firms with a stronger board and those with weaker board parameters. We use the board index (dummy variable that takes the value of 1 for the top 60 per cent of the firms having the strongest board parameters and zero otherwise) as the explanatory variable. Additionally, we also use the proportion of independent directors, board size and chairman independence (dummy variable that takes the value 1, if the chairman is independent else 0) as explanatory variables in our subset analysis. 4.5 Control variables To control for industry and firm characteristics, we use several control variables in our analysis, namely, leverage (ratio of debt to total assets), firm age (number of years since the date of incorporation) and firm size (log of total assets), sales growth (moving geometric average of net sales growth over the past three years), asset tangibility (fixed asset to total asset), stock volatility (4 year average of monthly standard deviation of stock price returns) and Industry classification based on two-digit National Industrial Classification (NIC) code. The data for these financial parameters were collected from ACE Equity database (Table I). 4.6 Method In our study, we observe the same firm at different point of time, and hence, we use the random effects model for our regression analysis (Arellano and Bond, 1991; Blundell and Bond, 1998). Black et al. (2014) opines that in developing countries a good approach to study this kind of relationship would be to build a panel data and use fixed or random effects model. Our primary interest is in analyzing the relationship between board structure and firm performance in a family firm. The analysis also includes additional variables such as family management, family shareholdings and family representative directors on the board. Serial correlation and heteroskedasticity are controlled by using the Huber White Sandwich Estimator (clustered) for the variance. Panel data analysis was conducted using the R statistical package (Version 3.1.1) with add-on package "PLM" (Croissant and Millo, 2008). The regression we use for the multivariate analysis takes the form: (1) Performance=b0+b1(Family Firm)+b2(CG Score)+b3(Family Firm) * (CG score)+b4(Control Variables)+b5-34(Two Digit SIC Code)+e where: Firm Performance is the accounting measures of performance ROA and ROE and Tobin's q. Family Firms is the dummy variable with value 1 for family firms, and zero otherwise. Control Variables are leverage, age, firm size, sales growth, tangibility, stock volatility, two-digit NIC code. 5.1 Summary statistics Descriptive statistics for the sample data shown in Table II is broken down into the two groups: family and nonfamily firms. The last two columns show the difference of means and the t-values between the family and nonfamily firms. Family firms constitute 73 per cent of our sample. Among family firms, about half of the sample had promoter CEO or CEO from the promoter family and about 30 per cent firms had both the CEO and the chairman from the founder family. This suggests an active participation of founders and founder families in the management of Indian firms. The governance characteristic of our sample shows that the mean independent directors for family firms (5.423) is higher than that for nonfamily firms (4.878), and the difference is significant at 1 per cent level. The mean independent chairman and mean role duality of nonfamily firms is higher than family firms and the difference is significant at the 1 per cent level in both cases. There is not much of a difference in the mean board size and non-executive directors between the two groups. The mean for the frequency of board meeting is slightly higher for family firms, but the mean board attendance is higher for nonfamily firms. The mean board busyness for family firms (5.206) was significantly higher compared to nonfamily firms (4.424). The firm level characteristics are reported in Table II. The mean performance (ROE, ROCE and Tobin's q) of nonfamily firms is significantly higher than family firms indicating that nonfamily firms on an average perform better than family firms. Consistent with this observation we find that the net sales growth is also higher for nonfamily firms. Family firms exhibit higher stock volatility compared to nonfamily firms. Family firms are leveraged more compared to nonfamily firms. The mean total assets for family firms are also significantly higher than those for nonfamily firms. These results suggest that the nonfamily firms are better performers than family firms. 5.2 Multivariate analysis The result of the panel data analysis to examine the impact of board index on firm performance in family firms is presented in Table III. The dummy variable board index takes the value of 1 for the top 60 per cent of the firms having the strongest board parameters and zero otherwise. We divide the firms into two groups representing the family and nonfamily firms using a dummy variable, which equals one for family firms and zero for nonfamily firms. The interpretations of the coefficients of the regression results are carried out as follows: The coefficient of the interaction term (family firm x board index) represents the incremental effect of board index on the performance of family firms with respect to the reference group non-family firm. Similarly, the coefficient of the board index represents the incremental effect of board index on the reference group non-family firms. The incremental effect of board index on the group family firm is given by the sum of the coefficients of board index and coefficient of the interaction between the family firm and board index [Board Index + (FF x Board Index)]. From Table III Column 1, we find that the interaction variable (Family Firms x Board Index) is negatively and significantly associated with firm performance measured by both ROE and ROCE. This result suggests that the incremental effect of the board score, for family firms with respect to nonfamily firms, is having a negative impact on the firm performance. The coefficient of board index shows that there is a positive effect of board index on firm performance of the nonfamily firms. Thus, the incremental effect of board index on family firm (b for ROE = 3.789 + (-4.439) = -0.65) is negative. Similar results are shown with the other two performance measures ROCE and Tobin's q (Columns 3 and 5). Thus, it is shown that for family firms, any improvement in board score decreases the firm performance. This result could be because the family firm management may endeavor for the improvement of the board structure as a reaction to the poor firm performance in the past. In Columns 2, 4 and 6 of Table III, the Family Firm indicator variable of Columns 1, 3 and 5 is replaced by the indicator variable Family CEO. When we segregate the firms based on the family CEO and nonfamily professional CEO managed firms, we find no significant difference in the relationship between board index and firm performance. This shows that having a family CEO or professional CEO does not make any significant impact on the relationship between corporate governance and firm performance. This result is contrary to the studies which find that family firms managed by family CEO tend to be less profitable than professionally managed firms (Gomez-Mejia et al., 2001; Singell, 1997). Our results are also contrary to studies that have showed positive relationship between family management and firm performance (Anderson and Reeb, 2003). Our results provide evidence on the irrelevance of family management on the relationship between board structure and firm performance in large firms in Indian context. The control variables, leverage, was found to be negatively and significantly associated with firm performance for both the models and all the three performance measures which are consistent with other studies (Titman and Wessels, 1988; Friend and Lang, 1988). Firm age was found to have a positive and significant effect on firm performance indicating experienced firms have an edge over newer firms. Firm size was negatively related to firm performance indicating that larger firms are less profitable. Sales growth was positive and significant. Asset tangibility was found to be negative, significant only in case of ROCE and Tobin's q as the performance measure. Stock volatility as expected was negatively related to firm performance. In Table IV, we also analyze the result of subsample containing only the family firms to study the effects of having professional CEO for the family firm, the proportion of family ownership and family representative directors on the firm performance. Columns 1, 5 and 9 show the results for family CEO for performance measures ROE, ROCE and Tobin's q respectively. Similar to our earlier findings, the results for family CEO with respect to professional CEO did not show any significant difference for all the three measures of performance. Thus, in a family firm having a professional CEO does not seem to impact the relationship between board score and firm performance. This result confirms to the study done by others that find no difference between family-managed and professionally managed firms (Daily and Dalton, 1992; Willard et al., 1992). This result may be possibly due to professional CEO being socially connected to the controlling family of the firm who controls the actions of the CEO. Another reason could be that the role of family members in the selection of managers and the desire to remain in control of the firms may lead to greater managerial entrenchment (Gomez-Mejia et al., 2001). Next, we analyze the results for the proportion of family shareholding. It is expected that the substantial intergenerational ownership stake and having a majority of their wealth invested in a single business present a strong incentive for the family firms to monitor the management (Demsetz and Lehn, 1985). Thus, having higher family ownership and the need to pass on the business to the future generation, it is expected that the firm performance would be positively associated with family ownership. Columns 2, 6 and 10 in Table IV show the results for family ownership. Within the family firms, having a higher proportion of shareholdings by family members did not show any significant impact on the firm performance. This result could be because despite having less than majority shareholding, the family still retains control in the firm by virtue of highly dispersed outside shareholdings or by being in the management of the firm. We also examine the impact of family members on the board of directors. Advocates of stewardship theory propose that the concentration of power and authority to a single person in a firm will lead to better performance (Donaldson and Davis, 1991). Thus, having more members in the board would strengthen family control over the management of the firm. Alternatively, the tendency of families to forgo profit maximization activities owing to their financial preferences (Anderson and Reeb, 2003) and biases toward business ventures of other family members (Singell, 1997) may lead to lower firm performance. However, as reported in Table IV, Columns 3, 7 and 11, we find that for family representative directors, there is no significant association with all the three measures of performance. To analyze the impact the recession period had on the performance of the family firms, we add the coefficient for board score and board score interacted with recession dummy variable for the sample of family firm. Columns 4, 8 and 12 of Table IV presents the effect of the recession on the performance of family firms. The results show a negative effect of recession compared to non-recession period; however, it was not significant. This could be due the caution exhibited by the controlling family, as family wealth would be at stake in the case of any substantial losses to the firm. Also, to improve the viability of the company for a longer time horizon, the present generation family managers may take decisions to maximize the long-term firm value. Again, the long-term presence and reputation of family firms helps them to have favorable dealings with external suppliers, dealers, lenders, etc., which are likely to continue even during any economic downturn. Hence, no discernable difference in performance is noticed before and during the recession period. Additionally, we sought to study the impact of specific governance mechanism in improving the firm performance in family firms in India. We create subsample based on above and below the median proportion of independent directors, board size and chairman independence. Having more independent directors on the board has generally been associated with better monitoring and improved firm performance. Table V reports the results for the subsample analysis. From Table V, Column 1, we find that median board independence does not show any significant relationship with firm performance. The reason for this result could be that the independent directors may not be truly independent due to allegiance toward the family who got them on the board. The results were similar for subsamples of board size and firms having independent chairman shown in Columns 2-3. Thus, this study does not find any evidence that having a higher proportion of independent directors, larger board size or an independent chairman, improves the relationship between family firms and firm performance. In this study, we analyze the interaction effect of the family firms and board governance factors on firm performance in a sample of top publically traded Indian firms. We contribute to the growing literature on family firms by providing a multi-year analysis on the influence of board structure on firm performance in a family firm vis-a-vis nonfamily firm in the Indian context. In this study, we endeavor to expand our understanding of corporate governance and to shed light on the impact of the proportion of shareholding, family representative directors and having a professional CEO in large family firms in India. The result of the panel data analysis shows that the interaction variable (family firms x board score) has a statistically significant negative associated with firm performance measured by both Tobin's q and ROE. This is consistent with the results by Garcia-Ramos and Garcia-Olalla (2011) in the European context. Our result suggests that the incremental effect of the board index score, for family firms with respect to nonfamily firms, is having a negative impact on the firm performance. When we segregate the firms based on the family CEO and professionally managed firms, we do not find any significant difference in the relationship between board structure and firm performance. We obtained similar results with subsample analysis containing only the family firms. This result is consistent with the results of Daily and Dalton (1992) and Willard et al. (1992) confirming the irrelevance of the management structure but contrary to the studies by Gomez-Mejia et al. (2001) and Singell (1997) which find that family firms managed by family CEO tend to be less profitable than professionally managed firms. Our results also contradict the studies that show CEO's who are family members had a positive relationship with accounting measures of performance (Anderson and Reeb, 2003; McConaughy et al., 1998). In the sub-sample analysis containing only the family firms, having a higher proportion of family ownership, family representative directors and having professional CEO for family firms did not show any significant impact on board structure and the firm performance. This result implies that family management and family ownership are irrelevant to the value of the firm. The impact of the recession on family firms was also not found to be significant compared to non-recession period. Contrary to the popular belief, we find that having an independent board and independent chairman does not improve firm performance. Results for the subsample based on above and below the median proportion of independent directors do not show any significant relationship with firm performance. The results were similar for subsamples of board size and firms having an independent chairman. Thus, this study does not find any evidence that having a higher proportion of independent directors, board size or an independent chairman, improves the firm performance in a family firm. The research has several limitations that need to be kept in mind while interpreting the results. This study is limited to the analysis of publically listed top Indian firm and hence further research needs to be conducted on a sample having a mix of large and small firms to improve the generalizability of the results. The difficulty in obtaining reliable data on exhaustive governance variables limits this study to board parameters. A natural extension of this study would be to include a larger number of governance parameters. The widespread pyramiding in Indian family firms makes it difficult to ascertain the quantum of actual family ownership and hence limits the accuracy of any study on family ownership. A cross-country research can also be conducted to analyze how the legal and regulatory institutional differences impact the findings presented in this study. The future extension of this study may analyze the family firms based on the founder led family firms and firms led by the subsequent generation of the founder family.
|
The sample consists of top-listed firms in India for the period 2002 to 2012. Board index was constructed to capture the governance quality of the firm. The authors also study the relationship between board structure and firm performance by segregating the sample based on family management, family ownership and family representative directors. Random effects model was used for the regression analysis in the study.
|
[SECTION: Findings] Family businesses are the most dominant among publically traded firms across the world (Shleifer and Vishny, 1986; Burkart et al., 2003; Anderson and Reeb, 2003; La Porta et al., 1999). In Continental Europe, about 44 per cent of publicly held firms are family-controlled (Faccio and Lang, 2002). In the USA, the equity ownership concentration is modest; among the Fortune 500 firms, around one-third is family firms (Anderson and Reeb, 2003). The concentration of ownership was found to be higher in other developed nations (Faccio and Lang, 2002; Franks and Mayer, 2001; Gorton and Schmid, 1996). Family businesses dominate many developing economies with about two-third of the firms in Asian countries owned by families or individuals (Claessens et al., 2000). In India, around 60 per cent of the listed top 500 firms are family firms (Chakrabarti et al., 2008). These family firms hold large equity stakes and more often than not have family representation on the board of directors. Family equity stake in Indian firms may be divided across individual holdings by promoters and their family members, privately held firms and through cross-holdings from other listed group businesses. The control and influence exerted by family firms may lead to performance difference with the non-family firms (Anderson and Reeb, 2003). The impact of organizational form (separate ownership and control or combined ownership and control) on firm performance has been mixed empirically. Berle and Means (1932) and Jensen and Meckling (1976) asserted that with the separation of ownership and control, there arises a potential for conflict of interest, which leads to the reduction in firm value. Managers may pursue activities in self-interest at the expense of profit maximization creating agency costs. The costs involved in monitoring the management and the difficulty of developing contracts to accurately specify the actions of the managers in the interest of the shareholders leads to lower market valuation of the firm. Contrary to this, Fama and Jensen (1983) argue that separation of ownership and control is advantageous as the efficiency costs outweigh the agency costs. Institutional mechanisms evolve over time to align with the objectives of managers and the shareholders (James, 1999). Firms with combined ownership and control may choose to pursue different investment rule and may have incentive to forgo profitable investment opportunity thereby reducing the firm value (Fama and Jensen, 1985; James, 1999). Active shareholders act as a mechanism for monitoring the firm management. However, shareholders who own very few shares of a firm have little incentive to do so. On the other hand, blockholders or large shareholders who have a substantial investment in a firm can play a key role to mitigate the agency problem. Blockholders can be the family or promoters of the company or institutional investors. Owing to the percentage of votes they hold, blockholders can pressurize the firm management toward improving investor protection (Shleifer and Vishny, 1997). Blockholders can force change the management that fails to deliver the expected results either in an AGM or through proxy fights in the case of a staggered board. A positive correlation between institutional investor's involvement and firm performance has been found in the literature (Agrawal and Knoeber, 1996; Strickland et al., 1996; Shome and Singh, 1995). However, a major problem is that the voting by the institutional investors has been low historically (Hampel, 1998) and those who vote also may be just "Box Tickers" who vote without considering the issues appropriately (Solomon, 2007). Others, like the mutual fund houses and the insurance companies, opt to sell their shares rather than have their say on the management of the firm. Another issue is that these institutional investors may hold a highly diversified portfolio, making it difficult for them to focus on any one company. In some cases, these institutional investors may exert pressure on the management to further their own gains rather than of all the shareholders (Shleifer and Vishny, 1997). This potential tradeoff between high agency cost of monitoring and clash of interest between managers and shareholders may not exist in family firms (James, 1999). Family blockholders can mitigate the agency problem associated with the separation of ownership and management (Villalonga and Amit, 2006; Anderson and Reeb, 2003) and, hence, should have a positive effect on firm value (Berle and Means, 1932). Combining ownership and management can help mitigate managerial opportunism (Demsetz and Lehn, 1985). Family blockholders have a strong incentive, by way of protecting their family wealth, to monitor managers and minimize free rider problem associated with minority shareholders (Anderson and Reeb, 2003; Barontini and Caprio, 2006). At the same time concentrated shareholders can, in exchange for profits, extract private benefits from the firm (Fama and Jensen, 1983; Shleifer and Vishny, 1997; Demsetz, 1983). Thus, concentrated shareholders can generate conflict of interest with minority shareholders (Burkart et al., 2003). The impact of blockholdings by family members on firm performance has yielded mixed empirical results. Family firms were found to perform better than nonfamily firms more so when a family member serves as CEO (Anderson and Reeb, 2003; Maury, 2006). Contrary to these results, family firms were not found to have better firm performance (measured by Tobin's q) as compared to nonfamily firms (Holderness and Sheehan, 1988; Miller et al., 2007). The corporate governance practices of family firms are also found to differ from those of nonfamily firms (Bartholomeusz and Tanewski, 2006). So, does family firms improves firm performance or destroys the firm value? Also, how effective are corporate governance mechanisms in controlling the agency costs in family firms. Is there any significant difference in the relationship between board structure and firm performance among family and nonfamily firms? We explore these questions, by studying the board structure and ownership pattern using a sample of top Indian firms from 2002 to 2012 and comparing family firms and nonfamily firms. Most governance studies on family firms have used data from the developed countries (Jameson et al., 2014). Given the predominance of controlling shareholders in emerging economies, a country level study in the Indian context will help to improve our understanding of the impact of controlling shareholders on corporate governance and firm performance. To provide a context for this study, we briefly describe the institutional background in India. India inherited a financial market developed by the British India and a legal system based on British laws that had excellent provision for shareholder protection (La Porta et al., 1998). Despite these advantages, law enforcement remained a problem. India had its share of issues arising out of poor monitoring of Indian firms, law enforcement, limited information dissemination and licensing requirements. Those issues resulted in the stock market scandals, insider trading issues and allotment of preferential shares to family members at a discount (Khanna and Palepu, 2004; Rajagopalan and Zhang, 2008). Indian legal system today is characterized by long delays due to limitations in resources. A case in point is the bankruptcy resolution, which is deemed to take the longest time to resolve and has the lowest rate of recovery (Kang and Nayar, 2004). Post-independence, India adopted the development strategy of import substituting industrialization. The set of regulations laid out to implement this strategy in the Industries Act 1951 resulted in a system that required businesses to obtain multiple licenses and bureaucratic approvals (Chakrabarti et al., 2008). This system labeled as "license-permit raj" created institutional barriers for smaller businesses, thereby weakening the competition (Reed, 2002). Further, the equity markets were illiquid, and the banks were reluctant to fund private sector due to weak creditor protection. Thus, the business environment benefitted those with government connection while discouraging market-based interactions. This institutional business environment led to the growth of the family business group and concentrated shareholdings (Chakrabarti et al., 2008). These behavioral and institutional differences in emerging markets can provide different insights into corporate governance (Fan et al., 2011). More so for the Asian countries; the diverse institutional development background makes the corporate governance issues unique (Peng and Jiang, 2010). As mentioned in the literature above, Indian firms are characterized by large shareholders and more so by the founding families. This presents a unique setting for studying their impact or lack of it on the firm's corporate governance. This study provides, to our knowledge, the first multi-year comparative analysis of the relationship between board structure and firm performance of family firms in India. The paper is subsequently structured as follows. Section 2 outlines the potential benefits and negative effects of family firms. Section 3 details the sample data and the methodology used in the analysis. Section 4 reports the results of the analysis and Section 5 discusses the results and concludes the paper. Controlling blockholders like family can have potential benefits and competitive advantage. The extent literature on family firms focuses mostly on the agency problem. The problem in widely held firms includes limited ability to select reliable agents, monitor the selected agents and ensure performance (Fama and Jensen, 1983). An effective monitoring can improve firm performance and reduce agency cost (Fama and Jensen, 1983; Jensen and Meckling, 1976). The principle agent conflict associated with widely held firms may be reduced by concentrated blockholders, specifically the family blockholders, as they have a significant economic incentive to monitor the management (Demsetz and Lehn, 1985). Specifically, the substantial intergenerational ownership stake and having a majority of their wealth invested in a single business, the family firms have strong incentive to monitor the management. The lengthy involvement of family members in the business, in some cases spanning generations, permits them to move further along the learning curve. This superior knowledge allows the family members to monitor the managers better (Anderson and Reeb, 2003). Also, this long-term presence of family members in their firm leads to longer investment horizons than other shareholders and may provide an incentive to invest according to market rules (James, 1999). This willingness of family firms to invest in the long-term project leads to greater investment efficiency (Anderson and Reeb, 2003; James, 1999; Stein, 1988). Another advantage of having a long-term presence of family firms is that the external suppliers, dealers, lenders, etc., are more likely to have favorable dealings with the same governing bodies, owing to the long-term dealings and reputation, than with nonfamily firms. This sustained presence of family necessitates having their reputation maintained (Anderson and Reeb, 2003). Thus, family firms have several advantages while monitoring managers, superior knowledge, longer investment horizons and the need to pass on the business to future generation. Consistent with this argument, several studies have shown that family firms have superior performance compared to nonfamily firms. In large public US firms, stronger firm performance was found in family firms compared to nonfamily firms (Anderson and Reeb, 2003). In Japan, the family firms owned or managed by the founder showed superior performance (Saito, 2008). In a study of 13 Western European countries, it was found that family control was associated with higher valuation (Maury, 2006). In listed Swiss firms, Isakov and Weisskopf (2014) found that family firms were more profitable than non-family firms. The controlling families are often involved in the management of the firms, reducing the agency problem of owner-manager conflict. Combining ownership and managerial control can help mitigate the managerial expropriation (Demsetz and Lehn, 1985). This nonseparation of management and ownership also reduces the agency costs involved in establishing the mechanism to monitor the management (Garcia-Ramos and Garcia-Olalla, 2011). Morck et al. (1988) suggest that founder CEOs bring in innovation and value enhancing expertise. Family firms intend to maintain continuity by passing on the business to the next generation of the family (Casson, 1999; Chami, 1999). Thus, to improve the viability of the company for a longer time horizon, the present generation family managers may take decisions to maximize the long-term firm value. The empirical results on the impact of family management on firm performance have been mixed. In the US firms, CEOs who are family members had a positive relationship with accounting measures of performance (Anderson and Reeb, 2003; McConaughy et al., 1998). Family firms, where the founding family is on the supervisory or the executive board, showed better firm performance in German firms (Andres, 2008). Villalonga and Amit (2006) found that family ownership creates value when a family member serves as CEO or the chairman of the firm. The literature suggests that having a family CEO may be beneficial to the value of the firm. Family firms are also said to be fraught with nepotism, family disputes, capital restrictions, exploitation of minority shareholders and executive entrenchment, all of which adversely affect firm performance (Allen and Panian, 1982; Chandler, 1990; Faccio et al., 2001; Gomez-Mejia et al., 2001; Perez-Gonzalez, 2006; Schulze et al., 2001, 2003). Expropriation of minority shareholders by large shareholders may generate additional agency problem (Faccio et al., 2001). Concentrated shareholders, by virtue of their controlling position in the firm, may extract private benefits at the expense of minority shareholders (Burkart et al., 2003). Capital expenditure can be affected by the families' preference for special dividends (DeAngelo and DeAngelo, 2000). Employee productivity can also be adversely affected by family shareholders acting on their own behalf (Burkart et al., 1997). Family firms are found to show biases toward business ventures of other family members resulting in suboptimal investment (Singell, 1997). Family shareholders tend to forgo profit maximization activities owing to their financial preferences which often conflicts with those of minority shareholders (Anderson and Reeb, 2003). The empirical evidence has also backed these negative aspects of controlling shareholding as family firms have been found to have an adverse effect on firm performance. Family ownership structure in US firms is perceived as less efficient and profitable than dispersed ownership (Anderson and Reeb, 2003; Morck, 2007; Singell, 1997). Holderness and Sheehan (1988) using Tobin's q as performance measure found that family firms have lower firm performance than nonfamily firms. The proportion of family shareholdings has also been found to have an impact on the firm performance. Anderson and Reeb (2003) show that when the family ownership exceeds 30 per cent, the gains from family control starts to taper off. Studies from other countries have also provided evidence for the adverse effect of family ownership on firm value. In Canadian and Swedish firms, family shareholders were found to have a negative effect on firm performance (Morck, 2007; Cronqvist and Nilsson, 2003). In East Asian countries where transparency is low, the family firms are shown to harm minority shareholders (Faccio et al., 2001). Preference for family members for the executive positions in family firms suggests a restricted talent pool for selection of qualified candidates (Anderson and Reeb, 2003). Family CEO may also lead to resentment among other senior outside executives of the firm as talent, merit and tenure are not seen as criteria for top management positions (Schulze et al., 2001). This limitation in talent pool may lead to competitive disadvantage for the family-run businesses when compared to nonfamily firms. Also, the accountability of a family CEO toward minority shareholders may be less than that exhibited by professional managers (Gomez-Mejia et al., 2001). The role of family members in selection of managers and the desire to remain in control of the firms may lead to greater managerial entrenchment (Gomez-Mejia et al., 2001). Shleifer and Vishny (1997) opine that the greatest cost imposed by large shareholders is by continuing to run the business in spite of being incompetent and unqualified. The mixed results on the impact of the family firms on firm performance raise several questions. Does family firms perform better compared to non-family firms? How do family firms impact the relationship between board structure and firm performance? Does having a family CEO improve firm performance compared to a professionally managed firm? Does having a family CEO impact the relationship between board structure and firm performance? Having analyzed the difference between family and nonfamily firms, the following question can be asked for within family firms analysis. Within the family firms, does having a higher proportion of family ownership improve firm performance? How does higher proportion of family representative directors impact the firm performance of family firms? Again, does having a professional CEO improve firm performance compared to family CEO in family firms? How did family firms cope with the recession period? Also, does having a larger board, higher proportion of independent directors and independent chairman improve the performance of family firm? These are some of the questions that we address in this study. 4.1 Sample For our investigation, we use top Indian firms listed on the Bombay Stock Exchange (BSE), specifically the firms forming a part of the BSE 200 index. We exclude financial institutions and public sector units as strong government regulations may potentially affect performance of these firms. Firm-specific control variables were collected from the ACE Equity database. Governance data comprising of board characteristics such as independent directors, chairman independence, role duality, non-executive directors, board busyness, board size, board meetings and board attendance, was manually collected from the annual reports of the firms from ReportJunction database. As per Clause 49 of listing agreement, firms are required to file the corporate governance report quarterly; however, for our empirical analysis, we considered annual performance of the firm and select the financial year-end corporate governance reports. We excluded the firms where the governance data was not available for more than one year. Our final sample consisted of 100 firms yielding 1,100 firm-years or observations as our sample was restricted by the publically available data from 2002 to 2012. The Variance inflation factor (VIF) was calculated for each independent variable, and all VIF values were less than 10 suggesting that there is no multicollinearity issue (Myers, 1990). 4.2 Variables and data source We identify the family firms based on the fractional equity ownership of the family members and the presence of family members on the board of directors. We were also guided by the classification of the family firms for listed Indian firms given by the ACE Equity database. We use a dummy variable that equals one for family firms and zero otherwise. Our study incorporates variables such as family management, family shareholdings and family representative directors on the board. We measure family management using a dummy that indicates the presence of a family CEO. Family shareholdings represent the percentage of equity held by the family members in the firm. Family representative directors are the proportion of family members on the board. We also study the recession period and the impact it had on the family firms. According to National Bureau of Economic Research (NBER), the recession that began in December 2007 officially ended in June 2009. We use a dummy variable to specify the period as pre-recession from the year 2002 to 2008 with a value 0, and recession period from the year 2008 to 2012 with a value 1 (Table I). 4.3 Dependent variable For our analysis, we use accounting measures of performance namely return on equity (ROE) and return on capital used (ROCE). As several studies on family firms use Tobin's q as a measure of firm performance, we also use Tobin's q measured using the book value of debt plus market value of common stock divided by the book value of assets as a proxy for Tobin's q (Black et al., 2014; Khanna and Palepu, 2000). 4.4 Explanatory variables We study the role played by the directors of a firm by categorizing the various board-related factors into two broad categories, namely, board leadership and board activity. On the basis of the agency theory, we investigate the board leadership roles played by the directors of the firm and its impact on the firm performance. Board activity is examined using the resource dependency theory. Board Leadership comprising of factors relating to the board leadership that includes role duality, independent directors, chairman independence and non-executive directors and Board Activity is measured by variables: board meeting, board size, board attendance and board busyness. We operationalize the various board parameters as follows: For our board index, based on the percentage of Independent Directors on the board, we divide the firms into three sets, the top, middle and bottom, and we respectively assign a score of 0, 1, 2 for <50 per cent, 50-75 per cent and >75 per cent independent directors on the board. Similarly, we assign the scores of 0, 1 and 2 for Non-Executive Directors on the board. For Role Duality, we assign a score of one for nonduality (splitting of the role of CEO and chairman) and zero for role duality. Similarly, for Chairman Independence, we use dummy variable by assigning a score of 1 if the chairman is independent, and zero otherwise. The dummy for Board Size is assigned a score of zero if board size is less than eight and greater than 15 and one otherwise. For the number of Board Meetings, we use a dummy variable that is equal to zero if the number of board meetings is less than four, equal to one if board meetings are between four to eight and two otherwise. For Board Attendance, we assign a score of 1 if the average board attendance is greater than 75 per cent and zero otherwise. To measure Board Busyness we assign a value zero if average outside board membership is greater than 10; one if it is between five and 10 and two if it is less than two. The governance index is constructed by adding the points for each board parameter discussed above. First, we divide the board parameters into two broad sets, namely, board leadership and board activity. We then assign weights to the two sets to form governance index. We assign more weigh to board leadership by assigning 60 per cent weightage in the overall score. This index is then converted to dummy variables by assigning a value of 1 to the top 60 per cent of the firms having the strongest board parameters and zero otherwise. We do this to compare the top firms with bottom firms with respect to the board parameters. Though this simple index may not reflect the impact of individual board parameters, it does, however, help in distinguishing between the firms with a stronger board and those with weaker board parameters. We use the board index (dummy variable that takes the value of 1 for the top 60 per cent of the firms having the strongest board parameters and zero otherwise) as the explanatory variable. Additionally, we also use the proportion of independent directors, board size and chairman independence (dummy variable that takes the value 1, if the chairman is independent else 0) as explanatory variables in our subset analysis. 4.5 Control variables To control for industry and firm characteristics, we use several control variables in our analysis, namely, leverage (ratio of debt to total assets), firm age (number of years since the date of incorporation) and firm size (log of total assets), sales growth (moving geometric average of net sales growth over the past three years), asset tangibility (fixed asset to total asset), stock volatility (4 year average of monthly standard deviation of stock price returns) and Industry classification based on two-digit National Industrial Classification (NIC) code. The data for these financial parameters were collected from ACE Equity database (Table I). 4.6 Method In our study, we observe the same firm at different point of time, and hence, we use the random effects model for our regression analysis (Arellano and Bond, 1991; Blundell and Bond, 1998). Black et al. (2014) opines that in developing countries a good approach to study this kind of relationship would be to build a panel data and use fixed or random effects model. Our primary interest is in analyzing the relationship between board structure and firm performance in a family firm. The analysis also includes additional variables such as family management, family shareholdings and family representative directors on the board. Serial correlation and heteroskedasticity are controlled by using the Huber White Sandwich Estimator (clustered) for the variance. Panel data analysis was conducted using the R statistical package (Version 3.1.1) with add-on package "PLM" (Croissant and Millo, 2008). The regression we use for the multivariate analysis takes the form: (1) Performance=b0+b1(Family Firm)+b2(CG Score)+b3(Family Firm) * (CG score)+b4(Control Variables)+b5-34(Two Digit SIC Code)+e where: Firm Performance is the accounting measures of performance ROA and ROE and Tobin's q. Family Firms is the dummy variable with value 1 for family firms, and zero otherwise. Control Variables are leverage, age, firm size, sales growth, tangibility, stock volatility, two-digit NIC code. 5.1 Summary statistics Descriptive statistics for the sample data shown in Table II is broken down into the two groups: family and nonfamily firms. The last two columns show the difference of means and the t-values between the family and nonfamily firms. Family firms constitute 73 per cent of our sample. Among family firms, about half of the sample had promoter CEO or CEO from the promoter family and about 30 per cent firms had both the CEO and the chairman from the founder family. This suggests an active participation of founders and founder families in the management of Indian firms. The governance characteristic of our sample shows that the mean independent directors for family firms (5.423) is higher than that for nonfamily firms (4.878), and the difference is significant at 1 per cent level. The mean independent chairman and mean role duality of nonfamily firms is higher than family firms and the difference is significant at the 1 per cent level in both cases. There is not much of a difference in the mean board size and non-executive directors between the two groups. The mean for the frequency of board meeting is slightly higher for family firms, but the mean board attendance is higher for nonfamily firms. The mean board busyness for family firms (5.206) was significantly higher compared to nonfamily firms (4.424). The firm level characteristics are reported in Table II. The mean performance (ROE, ROCE and Tobin's q) of nonfamily firms is significantly higher than family firms indicating that nonfamily firms on an average perform better than family firms. Consistent with this observation we find that the net sales growth is also higher for nonfamily firms. Family firms exhibit higher stock volatility compared to nonfamily firms. Family firms are leveraged more compared to nonfamily firms. The mean total assets for family firms are also significantly higher than those for nonfamily firms. These results suggest that the nonfamily firms are better performers than family firms. 5.2 Multivariate analysis The result of the panel data analysis to examine the impact of board index on firm performance in family firms is presented in Table III. The dummy variable board index takes the value of 1 for the top 60 per cent of the firms having the strongest board parameters and zero otherwise. We divide the firms into two groups representing the family and nonfamily firms using a dummy variable, which equals one for family firms and zero for nonfamily firms. The interpretations of the coefficients of the regression results are carried out as follows: The coefficient of the interaction term (family firm x board index) represents the incremental effect of board index on the performance of family firms with respect to the reference group non-family firm. Similarly, the coefficient of the board index represents the incremental effect of board index on the reference group non-family firms. The incremental effect of board index on the group family firm is given by the sum of the coefficients of board index and coefficient of the interaction between the family firm and board index [Board Index + (FF x Board Index)]. From Table III Column 1, we find that the interaction variable (Family Firms x Board Index) is negatively and significantly associated with firm performance measured by both ROE and ROCE. This result suggests that the incremental effect of the board score, for family firms with respect to nonfamily firms, is having a negative impact on the firm performance. The coefficient of board index shows that there is a positive effect of board index on firm performance of the nonfamily firms. Thus, the incremental effect of board index on family firm (b for ROE = 3.789 + (-4.439) = -0.65) is negative. Similar results are shown with the other two performance measures ROCE and Tobin's q (Columns 3 and 5). Thus, it is shown that for family firms, any improvement in board score decreases the firm performance. This result could be because the family firm management may endeavor for the improvement of the board structure as a reaction to the poor firm performance in the past. In Columns 2, 4 and 6 of Table III, the Family Firm indicator variable of Columns 1, 3 and 5 is replaced by the indicator variable Family CEO. When we segregate the firms based on the family CEO and nonfamily professional CEO managed firms, we find no significant difference in the relationship between board index and firm performance. This shows that having a family CEO or professional CEO does not make any significant impact on the relationship between corporate governance and firm performance. This result is contrary to the studies which find that family firms managed by family CEO tend to be less profitable than professionally managed firms (Gomez-Mejia et al., 2001; Singell, 1997). Our results are also contrary to studies that have showed positive relationship between family management and firm performance (Anderson and Reeb, 2003). Our results provide evidence on the irrelevance of family management on the relationship between board structure and firm performance in large firms in Indian context. The control variables, leverage, was found to be negatively and significantly associated with firm performance for both the models and all the three performance measures which are consistent with other studies (Titman and Wessels, 1988; Friend and Lang, 1988). Firm age was found to have a positive and significant effect on firm performance indicating experienced firms have an edge over newer firms. Firm size was negatively related to firm performance indicating that larger firms are less profitable. Sales growth was positive and significant. Asset tangibility was found to be negative, significant only in case of ROCE and Tobin's q as the performance measure. Stock volatility as expected was negatively related to firm performance. In Table IV, we also analyze the result of subsample containing only the family firms to study the effects of having professional CEO for the family firm, the proportion of family ownership and family representative directors on the firm performance. Columns 1, 5 and 9 show the results for family CEO for performance measures ROE, ROCE and Tobin's q respectively. Similar to our earlier findings, the results for family CEO with respect to professional CEO did not show any significant difference for all the three measures of performance. Thus, in a family firm having a professional CEO does not seem to impact the relationship between board score and firm performance. This result confirms to the study done by others that find no difference between family-managed and professionally managed firms (Daily and Dalton, 1992; Willard et al., 1992). This result may be possibly due to professional CEO being socially connected to the controlling family of the firm who controls the actions of the CEO. Another reason could be that the role of family members in the selection of managers and the desire to remain in control of the firms may lead to greater managerial entrenchment (Gomez-Mejia et al., 2001). Next, we analyze the results for the proportion of family shareholding. It is expected that the substantial intergenerational ownership stake and having a majority of their wealth invested in a single business present a strong incentive for the family firms to monitor the management (Demsetz and Lehn, 1985). Thus, having higher family ownership and the need to pass on the business to the future generation, it is expected that the firm performance would be positively associated with family ownership. Columns 2, 6 and 10 in Table IV show the results for family ownership. Within the family firms, having a higher proportion of shareholdings by family members did not show any significant impact on the firm performance. This result could be because despite having less than majority shareholding, the family still retains control in the firm by virtue of highly dispersed outside shareholdings or by being in the management of the firm. We also examine the impact of family members on the board of directors. Advocates of stewardship theory propose that the concentration of power and authority to a single person in a firm will lead to better performance (Donaldson and Davis, 1991). Thus, having more members in the board would strengthen family control over the management of the firm. Alternatively, the tendency of families to forgo profit maximization activities owing to their financial preferences (Anderson and Reeb, 2003) and biases toward business ventures of other family members (Singell, 1997) may lead to lower firm performance. However, as reported in Table IV, Columns 3, 7 and 11, we find that for family representative directors, there is no significant association with all the three measures of performance. To analyze the impact the recession period had on the performance of the family firms, we add the coefficient for board score and board score interacted with recession dummy variable for the sample of family firm. Columns 4, 8 and 12 of Table IV presents the effect of the recession on the performance of family firms. The results show a negative effect of recession compared to non-recession period; however, it was not significant. This could be due the caution exhibited by the controlling family, as family wealth would be at stake in the case of any substantial losses to the firm. Also, to improve the viability of the company for a longer time horizon, the present generation family managers may take decisions to maximize the long-term firm value. Again, the long-term presence and reputation of family firms helps them to have favorable dealings with external suppliers, dealers, lenders, etc., which are likely to continue even during any economic downturn. Hence, no discernable difference in performance is noticed before and during the recession period. Additionally, we sought to study the impact of specific governance mechanism in improving the firm performance in family firms in India. We create subsample based on above and below the median proportion of independent directors, board size and chairman independence. Having more independent directors on the board has generally been associated with better monitoring and improved firm performance. Table V reports the results for the subsample analysis. From Table V, Column 1, we find that median board independence does not show any significant relationship with firm performance. The reason for this result could be that the independent directors may not be truly independent due to allegiance toward the family who got them on the board. The results were similar for subsamples of board size and firms having independent chairman shown in Columns 2-3. Thus, this study does not find any evidence that having a higher proportion of independent directors, larger board size or an independent chairman, improves the relationship between family firms and firm performance. In this study, we analyze the interaction effect of the family firms and board governance factors on firm performance in a sample of top publically traded Indian firms. We contribute to the growing literature on family firms by providing a multi-year analysis on the influence of board structure on firm performance in a family firm vis-a-vis nonfamily firm in the Indian context. In this study, we endeavor to expand our understanding of corporate governance and to shed light on the impact of the proportion of shareholding, family representative directors and having a professional CEO in large family firms in India. The result of the panel data analysis shows that the interaction variable (family firms x board score) has a statistically significant negative associated with firm performance measured by both Tobin's q and ROE. This is consistent with the results by Garcia-Ramos and Garcia-Olalla (2011) in the European context. Our result suggests that the incremental effect of the board index score, for family firms with respect to nonfamily firms, is having a negative impact on the firm performance. When we segregate the firms based on the family CEO and professionally managed firms, we do not find any significant difference in the relationship between board structure and firm performance. We obtained similar results with subsample analysis containing only the family firms. This result is consistent with the results of Daily and Dalton (1992) and Willard et al. (1992) confirming the irrelevance of the management structure but contrary to the studies by Gomez-Mejia et al. (2001) and Singell (1997) which find that family firms managed by family CEO tend to be less profitable than professionally managed firms. Our results also contradict the studies that show CEO's who are family members had a positive relationship with accounting measures of performance (Anderson and Reeb, 2003; McConaughy et al., 1998). In the sub-sample analysis containing only the family firms, having a higher proportion of family ownership, family representative directors and having professional CEO for family firms did not show any significant impact on board structure and the firm performance. This result implies that family management and family ownership are irrelevant to the value of the firm. The impact of the recession on family firms was also not found to be significant compared to non-recession period. Contrary to the popular belief, we find that having an independent board and independent chairman does not improve firm performance. Results for the subsample based on above and below the median proportion of independent directors do not show any significant relationship with firm performance. The results were similar for subsamples of board size and firms having an independent chairman. Thus, this study does not find any evidence that having a higher proportion of independent directors, board size or an independent chairman, improves the firm performance in a family firm. The research has several limitations that need to be kept in mind while interpreting the results. This study is limited to the analysis of publically listed top Indian firm and hence further research needs to be conducted on a sample having a mix of large and small firms to improve the generalizability of the results. The difficulty in obtaining reliable data on exhaustive governance variables limits this study to board parameters. A natural extension of this study would be to include a larger number of governance parameters. The widespread pyramiding in Indian family firms makes it difficult to ascertain the quantum of actual family ownership and hence limits the accuracy of any study on family ownership. A cross-country research can also be conducted to analyze how the legal and regulatory institutional differences impact the findings presented in this study. The future extension of this study may analyze the family firms based on the founder led family firms and firms led by the subsequent generation of the founder family.
|
The authors find a negative effect of board structure on firm performance in family firms compared to non-family firms. Contrary to the most Western literature, family management was not found to significantly affect firm performance as compared to that of professionally managed firms. In the subset analysis of family firms, higher proportion of family ownership and family representative directors did not show any significant impact on the firm performance. Having a higher proportion of independent directors, larger board size or an independent chairman does not appear to improve this insignificant relationship between family firms and firm performance. Also, in family firms, no significant difference in performance is noticed before and during recession period.
|
[SECTION: Value] Family businesses are the most dominant among publically traded firms across the world (Shleifer and Vishny, 1986; Burkart et al., 2003; Anderson and Reeb, 2003; La Porta et al., 1999). In Continental Europe, about 44 per cent of publicly held firms are family-controlled (Faccio and Lang, 2002). In the USA, the equity ownership concentration is modest; among the Fortune 500 firms, around one-third is family firms (Anderson and Reeb, 2003). The concentration of ownership was found to be higher in other developed nations (Faccio and Lang, 2002; Franks and Mayer, 2001; Gorton and Schmid, 1996). Family businesses dominate many developing economies with about two-third of the firms in Asian countries owned by families or individuals (Claessens et al., 2000). In India, around 60 per cent of the listed top 500 firms are family firms (Chakrabarti et al., 2008). These family firms hold large equity stakes and more often than not have family representation on the board of directors. Family equity stake in Indian firms may be divided across individual holdings by promoters and their family members, privately held firms and through cross-holdings from other listed group businesses. The control and influence exerted by family firms may lead to performance difference with the non-family firms (Anderson and Reeb, 2003). The impact of organizational form (separate ownership and control or combined ownership and control) on firm performance has been mixed empirically. Berle and Means (1932) and Jensen and Meckling (1976) asserted that with the separation of ownership and control, there arises a potential for conflict of interest, which leads to the reduction in firm value. Managers may pursue activities in self-interest at the expense of profit maximization creating agency costs. The costs involved in monitoring the management and the difficulty of developing contracts to accurately specify the actions of the managers in the interest of the shareholders leads to lower market valuation of the firm. Contrary to this, Fama and Jensen (1983) argue that separation of ownership and control is advantageous as the efficiency costs outweigh the agency costs. Institutional mechanisms evolve over time to align with the objectives of managers and the shareholders (James, 1999). Firms with combined ownership and control may choose to pursue different investment rule and may have incentive to forgo profitable investment opportunity thereby reducing the firm value (Fama and Jensen, 1985; James, 1999). Active shareholders act as a mechanism for monitoring the firm management. However, shareholders who own very few shares of a firm have little incentive to do so. On the other hand, blockholders or large shareholders who have a substantial investment in a firm can play a key role to mitigate the agency problem. Blockholders can be the family or promoters of the company or institutional investors. Owing to the percentage of votes they hold, blockholders can pressurize the firm management toward improving investor protection (Shleifer and Vishny, 1997). Blockholders can force change the management that fails to deliver the expected results either in an AGM or through proxy fights in the case of a staggered board. A positive correlation between institutional investor's involvement and firm performance has been found in the literature (Agrawal and Knoeber, 1996; Strickland et al., 1996; Shome and Singh, 1995). However, a major problem is that the voting by the institutional investors has been low historically (Hampel, 1998) and those who vote also may be just "Box Tickers" who vote without considering the issues appropriately (Solomon, 2007). Others, like the mutual fund houses and the insurance companies, opt to sell their shares rather than have their say on the management of the firm. Another issue is that these institutional investors may hold a highly diversified portfolio, making it difficult for them to focus on any one company. In some cases, these institutional investors may exert pressure on the management to further their own gains rather than of all the shareholders (Shleifer and Vishny, 1997). This potential tradeoff between high agency cost of monitoring and clash of interest between managers and shareholders may not exist in family firms (James, 1999). Family blockholders can mitigate the agency problem associated with the separation of ownership and management (Villalonga and Amit, 2006; Anderson and Reeb, 2003) and, hence, should have a positive effect on firm value (Berle and Means, 1932). Combining ownership and management can help mitigate managerial opportunism (Demsetz and Lehn, 1985). Family blockholders have a strong incentive, by way of protecting their family wealth, to monitor managers and minimize free rider problem associated with minority shareholders (Anderson and Reeb, 2003; Barontini and Caprio, 2006). At the same time concentrated shareholders can, in exchange for profits, extract private benefits from the firm (Fama and Jensen, 1983; Shleifer and Vishny, 1997; Demsetz, 1983). Thus, concentrated shareholders can generate conflict of interest with minority shareholders (Burkart et al., 2003). The impact of blockholdings by family members on firm performance has yielded mixed empirical results. Family firms were found to perform better than nonfamily firms more so when a family member serves as CEO (Anderson and Reeb, 2003; Maury, 2006). Contrary to these results, family firms were not found to have better firm performance (measured by Tobin's q) as compared to nonfamily firms (Holderness and Sheehan, 1988; Miller et al., 2007). The corporate governance practices of family firms are also found to differ from those of nonfamily firms (Bartholomeusz and Tanewski, 2006). So, does family firms improves firm performance or destroys the firm value? Also, how effective are corporate governance mechanisms in controlling the agency costs in family firms. Is there any significant difference in the relationship between board structure and firm performance among family and nonfamily firms? We explore these questions, by studying the board structure and ownership pattern using a sample of top Indian firms from 2002 to 2012 and comparing family firms and nonfamily firms. Most governance studies on family firms have used data from the developed countries (Jameson et al., 2014). Given the predominance of controlling shareholders in emerging economies, a country level study in the Indian context will help to improve our understanding of the impact of controlling shareholders on corporate governance and firm performance. To provide a context for this study, we briefly describe the institutional background in India. India inherited a financial market developed by the British India and a legal system based on British laws that had excellent provision for shareholder protection (La Porta et al., 1998). Despite these advantages, law enforcement remained a problem. India had its share of issues arising out of poor monitoring of Indian firms, law enforcement, limited information dissemination and licensing requirements. Those issues resulted in the stock market scandals, insider trading issues and allotment of preferential shares to family members at a discount (Khanna and Palepu, 2004; Rajagopalan and Zhang, 2008). Indian legal system today is characterized by long delays due to limitations in resources. A case in point is the bankruptcy resolution, which is deemed to take the longest time to resolve and has the lowest rate of recovery (Kang and Nayar, 2004). Post-independence, India adopted the development strategy of import substituting industrialization. The set of regulations laid out to implement this strategy in the Industries Act 1951 resulted in a system that required businesses to obtain multiple licenses and bureaucratic approvals (Chakrabarti et al., 2008). This system labeled as "license-permit raj" created institutional barriers for smaller businesses, thereby weakening the competition (Reed, 2002). Further, the equity markets were illiquid, and the banks were reluctant to fund private sector due to weak creditor protection. Thus, the business environment benefitted those with government connection while discouraging market-based interactions. This institutional business environment led to the growth of the family business group and concentrated shareholdings (Chakrabarti et al., 2008). These behavioral and institutional differences in emerging markets can provide different insights into corporate governance (Fan et al., 2011). More so for the Asian countries; the diverse institutional development background makes the corporate governance issues unique (Peng and Jiang, 2010). As mentioned in the literature above, Indian firms are characterized by large shareholders and more so by the founding families. This presents a unique setting for studying their impact or lack of it on the firm's corporate governance. This study provides, to our knowledge, the first multi-year comparative analysis of the relationship between board structure and firm performance of family firms in India. The paper is subsequently structured as follows. Section 2 outlines the potential benefits and negative effects of family firms. Section 3 details the sample data and the methodology used in the analysis. Section 4 reports the results of the analysis and Section 5 discusses the results and concludes the paper. Controlling blockholders like family can have potential benefits and competitive advantage. The extent literature on family firms focuses mostly on the agency problem. The problem in widely held firms includes limited ability to select reliable agents, monitor the selected agents and ensure performance (Fama and Jensen, 1983). An effective monitoring can improve firm performance and reduce agency cost (Fama and Jensen, 1983; Jensen and Meckling, 1976). The principle agent conflict associated with widely held firms may be reduced by concentrated blockholders, specifically the family blockholders, as they have a significant economic incentive to monitor the management (Demsetz and Lehn, 1985). Specifically, the substantial intergenerational ownership stake and having a majority of their wealth invested in a single business, the family firms have strong incentive to monitor the management. The lengthy involvement of family members in the business, in some cases spanning generations, permits them to move further along the learning curve. This superior knowledge allows the family members to monitor the managers better (Anderson and Reeb, 2003). Also, this long-term presence of family members in their firm leads to longer investment horizons than other shareholders and may provide an incentive to invest according to market rules (James, 1999). This willingness of family firms to invest in the long-term project leads to greater investment efficiency (Anderson and Reeb, 2003; James, 1999; Stein, 1988). Another advantage of having a long-term presence of family firms is that the external suppliers, dealers, lenders, etc., are more likely to have favorable dealings with the same governing bodies, owing to the long-term dealings and reputation, than with nonfamily firms. This sustained presence of family necessitates having their reputation maintained (Anderson and Reeb, 2003). Thus, family firms have several advantages while monitoring managers, superior knowledge, longer investment horizons and the need to pass on the business to future generation. Consistent with this argument, several studies have shown that family firms have superior performance compared to nonfamily firms. In large public US firms, stronger firm performance was found in family firms compared to nonfamily firms (Anderson and Reeb, 2003). In Japan, the family firms owned or managed by the founder showed superior performance (Saito, 2008). In a study of 13 Western European countries, it was found that family control was associated with higher valuation (Maury, 2006). In listed Swiss firms, Isakov and Weisskopf (2014) found that family firms were more profitable than non-family firms. The controlling families are often involved in the management of the firms, reducing the agency problem of owner-manager conflict. Combining ownership and managerial control can help mitigate the managerial expropriation (Demsetz and Lehn, 1985). This nonseparation of management and ownership also reduces the agency costs involved in establishing the mechanism to monitor the management (Garcia-Ramos and Garcia-Olalla, 2011). Morck et al. (1988) suggest that founder CEOs bring in innovation and value enhancing expertise. Family firms intend to maintain continuity by passing on the business to the next generation of the family (Casson, 1999; Chami, 1999). Thus, to improve the viability of the company for a longer time horizon, the present generation family managers may take decisions to maximize the long-term firm value. The empirical results on the impact of family management on firm performance have been mixed. In the US firms, CEOs who are family members had a positive relationship with accounting measures of performance (Anderson and Reeb, 2003; McConaughy et al., 1998). Family firms, where the founding family is on the supervisory or the executive board, showed better firm performance in German firms (Andres, 2008). Villalonga and Amit (2006) found that family ownership creates value when a family member serves as CEO or the chairman of the firm. The literature suggests that having a family CEO may be beneficial to the value of the firm. Family firms are also said to be fraught with nepotism, family disputes, capital restrictions, exploitation of minority shareholders and executive entrenchment, all of which adversely affect firm performance (Allen and Panian, 1982; Chandler, 1990; Faccio et al., 2001; Gomez-Mejia et al., 2001; Perez-Gonzalez, 2006; Schulze et al., 2001, 2003). Expropriation of minority shareholders by large shareholders may generate additional agency problem (Faccio et al., 2001). Concentrated shareholders, by virtue of their controlling position in the firm, may extract private benefits at the expense of minority shareholders (Burkart et al., 2003). Capital expenditure can be affected by the families' preference for special dividends (DeAngelo and DeAngelo, 2000). Employee productivity can also be adversely affected by family shareholders acting on their own behalf (Burkart et al., 1997). Family firms are found to show biases toward business ventures of other family members resulting in suboptimal investment (Singell, 1997). Family shareholders tend to forgo profit maximization activities owing to their financial preferences which often conflicts with those of minority shareholders (Anderson and Reeb, 2003). The empirical evidence has also backed these negative aspects of controlling shareholding as family firms have been found to have an adverse effect on firm performance. Family ownership structure in US firms is perceived as less efficient and profitable than dispersed ownership (Anderson and Reeb, 2003; Morck, 2007; Singell, 1997). Holderness and Sheehan (1988) using Tobin's q as performance measure found that family firms have lower firm performance than nonfamily firms. The proportion of family shareholdings has also been found to have an impact on the firm performance. Anderson and Reeb (2003) show that when the family ownership exceeds 30 per cent, the gains from family control starts to taper off. Studies from other countries have also provided evidence for the adverse effect of family ownership on firm value. In Canadian and Swedish firms, family shareholders were found to have a negative effect on firm performance (Morck, 2007; Cronqvist and Nilsson, 2003). In East Asian countries where transparency is low, the family firms are shown to harm minority shareholders (Faccio et al., 2001). Preference for family members for the executive positions in family firms suggests a restricted talent pool for selection of qualified candidates (Anderson and Reeb, 2003). Family CEO may also lead to resentment among other senior outside executives of the firm as talent, merit and tenure are not seen as criteria for top management positions (Schulze et al., 2001). This limitation in talent pool may lead to competitive disadvantage for the family-run businesses when compared to nonfamily firms. Also, the accountability of a family CEO toward minority shareholders may be less than that exhibited by professional managers (Gomez-Mejia et al., 2001). The role of family members in selection of managers and the desire to remain in control of the firms may lead to greater managerial entrenchment (Gomez-Mejia et al., 2001). Shleifer and Vishny (1997) opine that the greatest cost imposed by large shareholders is by continuing to run the business in spite of being incompetent and unqualified. The mixed results on the impact of the family firms on firm performance raise several questions. Does family firms perform better compared to non-family firms? How do family firms impact the relationship between board structure and firm performance? Does having a family CEO improve firm performance compared to a professionally managed firm? Does having a family CEO impact the relationship between board structure and firm performance? Having analyzed the difference between family and nonfamily firms, the following question can be asked for within family firms analysis. Within the family firms, does having a higher proportion of family ownership improve firm performance? How does higher proportion of family representative directors impact the firm performance of family firms? Again, does having a professional CEO improve firm performance compared to family CEO in family firms? How did family firms cope with the recession period? Also, does having a larger board, higher proportion of independent directors and independent chairman improve the performance of family firm? These are some of the questions that we address in this study. 4.1 Sample For our investigation, we use top Indian firms listed on the Bombay Stock Exchange (BSE), specifically the firms forming a part of the BSE 200 index. We exclude financial institutions and public sector units as strong government regulations may potentially affect performance of these firms. Firm-specific control variables were collected from the ACE Equity database. Governance data comprising of board characteristics such as independent directors, chairman independence, role duality, non-executive directors, board busyness, board size, board meetings and board attendance, was manually collected from the annual reports of the firms from ReportJunction database. As per Clause 49 of listing agreement, firms are required to file the corporate governance report quarterly; however, for our empirical analysis, we considered annual performance of the firm and select the financial year-end corporate governance reports. We excluded the firms where the governance data was not available for more than one year. Our final sample consisted of 100 firms yielding 1,100 firm-years or observations as our sample was restricted by the publically available data from 2002 to 2012. The Variance inflation factor (VIF) was calculated for each independent variable, and all VIF values were less than 10 suggesting that there is no multicollinearity issue (Myers, 1990). 4.2 Variables and data source We identify the family firms based on the fractional equity ownership of the family members and the presence of family members on the board of directors. We were also guided by the classification of the family firms for listed Indian firms given by the ACE Equity database. We use a dummy variable that equals one for family firms and zero otherwise. Our study incorporates variables such as family management, family shareholdings and family representative directors on the board. We measure family management using a dummy that indicates the presence of a family CEO. Family shareholdings represent the percentage of equity held by the family members in the firm. Family representative directors are the proportion of family members on the board. We also study the recession period and the impact it had on the family firms. According to National Bureau of Economic Research (NBER), the recession that began in December 2007 officially ended in June 2009. We use a dummy variable to specify the period as pre-recession from the year 2002 to 2008 with a value 0, and recession period from the year 2008 to 2012 with a value 1 (Table I). 4.3 Dependent variable For our analysis, we use accounting measures of performance namely return on equity (ROE) and return on capital used (ROCE). As several studies on family firms use Tobin's q as a measure of firm performance, we also use Tobin's q measured using the book value of debt plus market value of common stock divided by the book value of assets as a proxy for Tobin's q (Black et al., 2014; Khanna and Palepu, 2000). 4.4 Explanatory variables We study the role played by the directors of a firm by categorizing the various board-related factors into two broad categories, namely, board leadership and board activity. On the basis of the agency theory, we investigate the board leadership roles played by the directors of the firm and its impact on the firm performance. Board activity is examined using the resource dependency theory. Board Leadership comprising of factors relating to the board leadership that includes role duality, independent directors, chairman independence and non-executive directors and Board Activity is measured by variables: board meeting, board size, board attendance and board busyness. We operationalize the various board parameters as follows: For our board index, based on the percentage of Independent Directors on the board, we divide the firms into three sets, the top, middle and bottom, and we respectively assign a score of 0, 1, 2 for <50 per cent, 50-75 per cent and >75 per cent independent directors on the board. Similarly, we assign the scores of 0, 1 and 2 for Non-Executive Directors on the board. For Role Duality, we assign a score of one for nonduality (splitting of the role of CEO and chairman) and zero for role duality. Similarly, for Chairman Independence, we use dummy variable by assigning a score of 1 if the chairman is independent, and zero otherwise. The dummy for Board Size is assigned a score of zero if board size is less than eight and greater than 15 and one otherwise. For the number of Board Meetings, we use a dummy variable that is equal to zero if the number of board meetings is less than four, equal to one if board meetings are between four to eight and two otherwise. For Board Attendance, we assign a score of 1 if the average board attendance is greater than 75 per cent and zero otherwise. To measure Board Busyness we assign a value zero if average outside board membership is greater than 10; one if it is between five and 10 and two if it is less than two. The governance index is constructed by adding the points for each board parameter discussed above. First, we divide the board parameters into two broad sets, namely, board leadership and board activity. We then assign weights to the two sets to form governance index. We assign more weigh to board leadership by assigning 60 per cent weightage in the overall score. This index is then converted to dummy variables by assigning a value of 1 to the top 60 per cent of the firms having the strongest board parameters and zero otherwise. We do this to compare the top firms with bottom firms with respect to the board parameters. Though this simple index may not reflect the impact of individual board parameters, it does, however, help in distinguishing between the firms with a stronger board and those with weaker board parameters. We use the board index (dummy variable that takes the value of 1 for the top 60 per cent of the firms having the strongest board parameters and zero otherwise) as the explanatory variable. Additionally, we also use the proportion of independent directors, board size and chairman independence (dummy variable that takes the value 1, if the chairman is independent else 0) as explanatory variables in our subset analysis. 4.5 Control variables To control for industry and firm characteristics, we use several control variables in our analysis, namely, leverage (ratio of debt to total assets), firm age (number of years since the date of incorporation) and firm size (log of total assets), sales growth (moving geometric average of net sales growth over the past three years), asset tangibility (fixed asset to total asset), stock volatility (4 year average of monthly standard deviation of stock price returns) and Industry classification based on two-digit National Industrial Classification (NIC) code. The data for these financial parameters were collected from ACE Equity database (Table I). 4.6 Method In our study, we observe the same firm at different point of time, and hence, we use the random effects model for our regression analysis (Arellano and Bond, 1991; Blundell and Bond, 1998). Black et al. (2014) opines that in developing countries a good approach to study this kind of relationship would be to build a panel data and use fixed or random effects model. Our primary interest is in analyzing the relationship between board structure and firm performance in a family firm. The analysis also includes additional variables such as family management, family shareholdings and family representative directors on the board. Serial correlation and heteroskedasticity are controlled by using the Huber White Sandwich Estimator (clustered) for the variance. Panel data analysis was conducted using the R statistical package (Version 3.1.1) with add-on package "PLM" (Croissant and Millo, 2008). The regression we use for the multivariate analysis takes the form: (1) Performance=b0+b1(Family Firm)+b2(CG Score)+b3(Family Firm) * (CG score)+b4(Control Variables)+b5-34(Two Digit SIC Code)+e where: Firm Performance is the accounting measures of performance ROA and ROE and Tobin's q. Family Firms is the dummy variable with value 1 for family firms, and zero otherwise. Control Variables are leverage, age, firm size, sales growth, tangibility, stock volatility, two-digit NIC code. 5.1 Summary statistics Descriptive statistics for the sample data shown in Table II is broken down into the two groups: family and nonfamily firms. The last two columns show the difference of means and the t-values between the family and nonfamily firms. Family firms constitute 73 per cent of our sample. Among family firms, about half of the sample had promoter CEO or CEO from the promoter family and about 30 per cent firms had both the CEO and the chairman from the founder family. This suggests an active participation of founders and founder families in the management of Indian firms. The governance characteristic of our sample shows that the mean independent directors for family firms (5.423) is higher than that for nonfamily firms (4.878), and the difference is significant at 1 per cent level. The mean independent chairman and mean role duality of nonfamily firms is higher than family firms and the difference is significant at the 1 per cent level in both cases. There is not much of a difference in the mean board size and non-executive directors between the two groups. The mean for the frequency of board meeting is slightly higher for family firms, but the mean board attendance is higher for nonfamily firms. The mean board busyness for family firms (5.206) was significantly higher compared to nonfamily firms (4.424). The firm level characteristics are reported in Table II. The mean performance (ROE, ROCE and Tobin's q) of nonfamily firms is significantly higher than family firms indicating that nonfamily firms on an average perform better than family firms. Consistent with this observation we find that the net sales growth is also higher for nonfamily firms. Family firms exhibit higher stock volatility compared to nonfamily firms. Family firms are leveraged more compared to nonfamily firms. The mean total assets for family firms are also significantly higher than those for nonfamily firms. These results suggest that the nonfamily firms are better performers than family firms. 5.2 Multivariate analysis The result of the panel data analysis to examine the impact of board index on firm performance in family firms is presented in Table III. The dummy variable board index takes the value of 1 for the top 60 per cent of the firms having the strongest board parameters and zero otherwise. We divide the firms into two groups representing the family and nonfamily firms using a dummy variable, which equals one for family firms and zero for nonfamily firms. The interpretations of the coefficients of the regression results are carried out as follows: The coefficient of the interaction term (family firm x board index) represents the incremental effect of board index on the performance of family firms with respect to the reference group non-family firm. Similarly, the coefficient of the board index represents the incremental effect of board index on the reference group non-family firms. The incremental effect of board index on the group family firm is given by the sum of the coefficients of board index and coefficient of the interaction between the family firm and board index [Board Index + (FF x Board Index)]. From Table III Column 1, we find that the interaction variable (Family Firms x Board Index) is negatively and significantly associated with firm performance measured by both ROE and ROCE. This result suggests that the incremental effect of the board score, for family firms with respect to nonfamily firms, is having a negative impact on the firm performance. The coefficient of board index shows that there is a positive effect of board index on firm performance of the nonfamily firms. Thus, the incremental effect of board index on family firm (b for ROE = 3.789 + (-4.439) = -0.65) is negative. Similar results are shown with the other two performance measures ROCE and Tobin's q (Columns 3 and 5). Thus, it is shown that for family firms, any improvement in board score decreases the firm performance. This result could be because the family firm management may endeavor for the improvement of the board structure as a reaction to the poor firm performance in the past. In Columns 2, 4 and 6 of Table III, the Family Firm indicator variable of Columns 1, 3 and 5 is replaced by the indicator variable Family CEO. When we segregate the firms based on the family CEO and nonfamily professional CEO managed firms, we find no significant difference in the relationship between board index and firm performance. This shows that having a family CEO or professional CEO does not make any significant impact on the relationship between corporate governance and firm performance. This result is contrary to the studies which find that family firms managed by family CEO tend to be less profitable than professionally managed firms (Gomez-Mejia et al., 2001; Singell, 1997). Our results are also contrary to studies that have showed positive relationship between family management and firm performance (Anderson and Reeb, 2003). Our results provide evidence on the irrelevance of family management on the relationship between board structure and firm performance in large firms in Indian context. The control variables, leverage, was found to be negatively and significantly associated with firm performance for both the models and all the three performance measures which are consistent with other studies (Titman and Wessels, 1988; Friend and Lang, 1988). Firm age was found to have a positive and significant effect on firm performance indicating experienced firms have an edge over newer firms. Firm size was negatively related to firm performance indicating that larger firms are less profitable. Sales growth was positive and significant. Asset tangibility was found to be negative, significant only in case of ROCE and Tobin's q as the performance measure. Stock volatility as expected was negatively related to firm performance. In Table IV, we also analyze the result of subsample containing only the family firms to study the effects of having professional CEO for the family firm, the proportion of family ownership and family representative directors on the firm performance. Columns 1, 5 and 9 show the results for family CEO for performance measures ROE, ROCE and Tobin's q respectively. Similar to our earlier findings, the results for family CEO with respect to professional CEO did not show any significant difference for all the three measures of performance. Thus, in a family firm having a professional CEO does not seem to impact the relationship between board score and firm performance. This result confirms to the study done by others that find no difference between family-managed and professionally managed firms (Daily and Dalton, 1992; Willard et al., 1992). This result may be possibly due to professional CEO being socially connected to the controlling family of the firm who controls the actions of the CEO. Another reason could be that the role of family members in the selection of managers and the desire to remain in control of the firms may lead to greater managerial entrenchment (Gomez-Mejia et al., 2001). Next, we analyze the results for the proportion of family shareholding. It is expected that the substantial intergenerational ownership stake and having a majority of their wealth invested in a single business present a strong incentive for the family firms to monitor the management (Demsetz and Lehn, 1985). Thus, having higher family ownership and the need to pass on the business to the future generation, it is expected that the firm performance would be positively associated with family ownership. Columns 2, 6 and 10 in Table IV show the results for family ownership. Within the family firms, having a higher proportion of shareholdings by family members did not show any significant impact on the firm performance. This result could be because despite having less than majority shareholding, the family still retains control in the firm by virtue of highly dispersed outside shareholdings or by being in the management of the firm. We also examine the impact of family members on the board of directors. Advocates of stewardship theory propose that the concentration of power and authority to a single person in a firm will lead to better performance (Donaldson and Davis, 1991). Thus, having more members in the board would strengthen family control over the management of the firm. Alternatively, the tendency of families to forgo profit maximization activities owing to their financial preferences (Anderson and Reeb, 2003) and biases toward business ventures of other family members (Singell, 1997) may lead to lower firm performance. However, as reported in Table IV, Columns 3, 7 and 11, we find that for family representative directors, there is no significant association with all the three measures of performance. To analyze the impact the recession period had on the performance of the family firms, we add the coefficient for board score and board score interacted with recession dummy variable for the sample of family firm. Columns 4, 8 and 12 of Table IV presents the effect of the recession on the performance of family firms. The results show a negative effect of recession compared to non-recession period; however, it was not significant. This could be due the caution exhibited by the controlling family, as family wealth would be at stake in the case of any substantial losses to the firm. Also, to improve the viability of the company for a longer time horizon, the present generation family managers may take decisions to maximize the long-term firm value. Again, the long-term presence and reputation of family firms helps them to have favorable dealings with external suppliers, dealers, lenders, etc., which are likely to continue even during any economic downturn. Hence, no discernable difference in performance is noticed before and during the recession period. Additionally, we sought to study the impact of specific governance mechanism in improving the firm performance in family firms in India. We create subsample based on above and below the median proportion of independent directors, board size and chairman independence. Having more independent directors on the board has generally been associated with better monitoring and improved firm performance. Table V reports the results for the subsample analysis. From Table V, Column 1, we find that median board independence does not show any significant relationship with firm performance. The reason for this result could be that the independent directors may not be truly independent due to allegiance toward the family who got them on the board. The results were similar for subsamples of board size and firms having independent chairman shown in Columns 2-3. Thus, this study does not find any evidence that having a higher proportion of independent directors, larger board size or an independent chairman, improves the relationship between family firms and firm performance. In this study, we analyze the interaction effect of the family firms and board governance factors on firm performance in a sample of top publically traded Indian firms. We contribute to the growing literature on family firms by providing a multi-year analysis on the influence of board structure on firm performance in a family firm vis-a-vis nonfamily firm in the Indian context. In this study, we endeavor to expand our understanding of corporate governance and to shed light on the impact of the proportion of shareholding, family representative directors and having a professional CEO in large family firms in India. The result of the panel data analysis shows that the interaction variable (family firms x board score) has a statistically significant negative associated with firm performance measured by both Tobin's q and ROE. This is consistent with the results by Garcia-Ramos and Garcia-Olalla (2011) in the European context. Our result suggests that the incremental effect of the board index score, for family firms with respect to nonfamily firms, is having a negative impact on the firm performance. When we segregate the firms based on the family CEO and professionally managed firms, we do not find any significant difference in the relationship between board structure and firm performance. We obtained similar results with subsample analysis containing only the family firms. This result is consistent with the results of Daily and Dalton (1992) and Willard et al. (1992) confirming the irrelevance of the management structure but contrary to the studies by Gomez-Mejia et al. (2001) and Singell (1997) which find that family firms managed by family CEO tend to be less profitable than professionally managed firms. Our results also contradict the studies that show CEO's who are family members had a positive relationship with accounting measures of performance (Anderson and Reeb, 2003; McConaughy et al., 1998). In the sub-sample analysis containing only the family firms, having a higher proportion of family ownership, family representative directors and having professional CEO for family firms did not show any significant impact on board structure and the firm performance. This result implies that family management and family ownership are irrelevant to the value of the firm. The impact of the recession on family firms was also not found to be significant compared to non-recession period. Contrary to the popular belief, we find that having an independent board and independent chairman does not improve firm performance. Results for the subsample based on above and below the median proportion of independent directors do not show any significant relationship with firm performance. The results were similar for subsamples of board size and firms having an independent chairman. Thus, this study does not find any evidence that having a higher proportion of independent directors, board size or an independent chairman, improves the firm performance in a family firm. The research has several limitations that need to be kept in mind while interpreting the results. This study is limited to the analysis of publically listed top Indian firm and hence further research needs to be conducted on a sample having a mix of large and small firms to improve the generalizability of the results. The difficulty in obtaining reliable data on exhaustive governance variables limits this study to board parameters. A natural extension of this study would be to include a larger number of governance parameters. The widespread pyramiding in Indian family firms makes it difficult to ascertain the quantum of actual family ownership and hence limits the accuracy of any study on family ownership. A cross-country research can also be conducted to analyze how the legal and regulatory institutional differences impact the findings presented in this study. The future extension of this study may analyze the family firms based on the founder led family firms and firms led by the subsequent generation of the founder family.
|
The study uses a self-defined corporate governance index to measure the governance parameters, specifically the board characteristics. The results documented in this study adds to the debate on the generalizability of the findings in Western governance studies in emerging markets like India with unique institutional development background.
|
[SECTION: Purpose] Unstable and dynamic societal and business environments are making it more imperative than ever that organisations have the ability to operate its learning processes in order to optimise competence development, performance and innovation (Brinkerhoff, 2006). Research-based knowledge on the subject has taught us that for enterprises to optimise the implementation of learning in practice, they need to take care of elements related to the learner, the learning situation and the organisational setting where new knowledge and/or competences are to be integrated and used. Researchers and practitioners, however, from the organisational learning field have, since the first theoretical and empirical studies, struggled to understand how organisations are to optimise its design and models for the sharing and use of new external and internal knowledge and competences in an organisational context. A vast bulk of research underscores that enterprises still see very poor results when it comes to the transferring and sharing of new competences and knowledge in practice (Burke and Hutchins, 2007; Kazbour et al., 2013). The business challenge is that enterprises often have insufficient data and tools to support the organisational learning process aimed at integrating and employing new knowledge and/or competences to improve competence development, performance and innovation. New knowledge and competences often end unused in practice due to relapse of different kinds, e.g., routine take over, no room/time to use the new knowledge, no support from peers or managers and task irrelevant. This challenge leads to the research question for this paper: how are enterprises to arrange its learning processes in order to optimise the integration and create sustainable organisational learning processes and results? In this paper, we present the form and content of an innovative learning technology tool that can assist enterprises in the continuous evaluation of their capacity to integrate and use new knowledge, skills and/or competences in practice using real time and feedback data. Often, launching learning evaluation tools in enterprises can be a cumbersome and - at times - expensive procedure that focuses on verifying the individual outcome and effects after participation in a specific in-service or external learning activity, e.g., a course or seminar. We argue that in order to qualify the transfer of knowledge and organisational learning, enterprises, its leaders and HR should be responsible and need to be better equipped at keeping track and understand the learning processes at an organisational level. An aspiration that does not necessarily comprise comprehensively imbued measurements techniques aimed solely at the end result but instead focuses on how organisations can support and optimise the learning process as it happens at work and in this way build sustainable organisational learning. Today's learning research focuses on how to understand, design and make learning activities, formal as well as informal, an integral part of the existing organisational HRM and management system and practice. A classical theoretical understanding of learning in organisations and workplaces departs from an idea of learning as a matter of designing a learning and practice situation characterised by mutually shared or identical elements. This behavioural - and later developed within a cognitive frame - theory of transfer and sharing of knowledge typically settled in the common elements theory on shared traits between the learning and practice environments or symbolic representations. Later, Baldwin and Ford (1988) advanced a model that has had an immense influence on theoretical and empirical research contributions within the field of learning studies in adult learning studies and human resource. The distinction between three broad dimensions, the learner's characteristics, the design of the learning situation and the work environment settings, has guided research- and practice-oriented contributions within the organisational and workplace learning field for decades. The influence of each dimension and sub-dimensions for the learning process in organisations has been explored, theoretically and empirically, in a vast number of articles and books. These all sources emphasise that we still need to develop knowledge on the topic and to build improved tools and methods in order to optimise and build better guides for the management of learning processes and outcomes in organisations (Burke and Hutchins, 2007). For example, Blume et al. (2010) demonstrate that existing research is characterised by inconsistent measurement data and that the learning transfer "problem" still needs attention from both the research community and responsible agents for the integration of new knowledge and competences in organisations. For example, Blume et al. (2010) underscore the importance of the workplace and organisational settings for improving the concrete integration of new knowledge and competences. One suggestion is to continue scrutinise how the work environment, design of work and absorptive capacities located in the receiving organisation as regard support and management of knowledge, experiences and competences can be leveraged mediated by learning evaluation tools to optimise and better integrate learning (Blume et al., 2010) in practice, which we turn our attention to in the following section. The studied and described tool of this paper (named SuperInsight) is an evaluation and development-oriented organisational learning technology devised with the aim of strengthening a sustainable integration and anchoring of new knowledge and/or competences in practice. The evaluation tool is deliberately designed as a lite version of an evaluation tool in order to avoid, as we often experience, bulky and highly bureaucratized (red tape) procedures when it comes to working with measurement and integration of learning activities in a practice setting. The purpose of the lite tool is two-folded. First, the lite tool aims to reinforce enterprises ability to keep track of, on a running basis, the impact of learning activities in practice based on real-time data. Second, the lite tool intends to support an organisational learning process among the participating enterprises based on continuous input and feedback on the quality of the learning process combined with unique and specific insight on where to intervene - if needed - in order to optimise the learning transfer process and result. The evaluation tool is implemented through a ten weeks pulse survey that keeps track and continuously evaluates a completed learning activity, for example, an external competence development programme or an internal training course. In practice, after the completion of the learning activities, participants receive four questions each week related to the ongoing implementation efforts of new knowledge and/or competences. It takes approximately 45 s for each participant to complete the pulse survey. Collected data and analysis operate with three general indicators to categorise the implementation level of new knowledge and competences. More concretely, the responsible person for the implementation (HR, trainer, manager, leader, etc.) log into a dashboard on her/his computer where the person gets an overview of individuals and/or groups/larger teams learning status as regard the level of implementation. "Green" indicator (strong implementation) colour represents a participant who experiences that she/he is succeeding with the use of new knowledge and competences in practice. "Yellow" colour indicator (hesitant implementation) denotes a participant who experiences that she/he will be successful, nevertheless is hesitating when it comes to the integration and use of new knowledge and competences in their work life. "Red" colour indicator (weak implementation) characterizes a participant who experiences that she/he has come to a complete stop, i.e., no use of the new knowledge and/or competences. The theoretical model behind the lite tool builds on Baldwin and Fords's (1988) "Learning Transfer model". In our model and devised lite tool, we apply the three dimensions and ten factors from the model that is associated with learning transfer. Thus, the lite tool is organised around ten learning factors that have been identified in research as having a strong positive correlation with integration and use of new knowledge and competences in practice (Baldwin and Ford, 1988; Blume et al., 2010; Burke and Hutchins, 2007). Learner characteristics Learner characteristics include elements like self-efficacy, motivation and perceived utility of training. Learner characteristic scores on this indicator should be very high. If not, we typically see that course owner is trying to get a participant to do training where they lack context or a more clear purpose. Training design Training design includes elements like behavioural modelling, error management and realistic case-based learning environments where there is access to real-time feedback and supervision. Training design scores and comments are primarily concentrated on how to better support the implementation process in the organisation (not the actual training). The work environment The work environmentelements like transfer climate, support from leaders and peers, opportunity to perform and follow-up procedures. The work environment scores are (not surprisingly) where we see the largest potential for improvement (especially after week 3). Analysis of data from a case study in a large Danish enterprise from the telecommunications industry conducting leadership and sales training shows that participants who are in the red category, usually are drowning in operations and do not have the possibility to focus on learning implementation on new knowledge or competences in a work context. A characteristic comment from a participant in the red category is: "I will start to work with the new knowledge, when my department get more resources" or "We are firefighting all the time". Here, the role of the closest leader or accountable HR staff is to support the participant in prioritising their work tasks to facilitate the actual use of new knowledge or competences - if the organisation truly wants to be serious about their approach to organisational learning. The analysis shows that constraints in employing new knowledge and competences are not related to the actual course/training activity, nor personal characteristics, but to the concrete work processes and tasks. Participants with an uncertain implementation behaviour (yellow) will typically have obtained new knowledge and have the appropriate motivation and thus be ready to incorporate new knowledge and competences to develop their work practices. However, participants are hesitant when it comes to the concrete implementation. It is important that the organisation and its leaders identify participants in the yellow category at the right time, and subsequently offer proper guidance and support. Comments in this group are typically: "It was a very good course. During the week, I have thought a lot about what we learned [...] But used any of the new skills - I have not". For this category, the task of the closest leader or HR manager is to offer guidance and support to instill the needed implementation. It is specifically in relation to the yellow category that the strength of real-time and pulse survey data shows its aptitude as a learning evaluation technology compared to more evidence-based measurement tools. The lite tool makes processes of learning transparent and provides fast continuous feedback to HRM and the organisation building grounds for actualising sustainable organisational learning throughout the implementation process. Participants in the green category are the organisational champions of integrating and implementing new knowledge and competences in practice. It is important to acknowledge this group of participant's high focus and capacity to implement new knowledge and competences to improve and innovate existing tasks or operations. Participants identified as showing high focus and capacity to create organisational learning can be utilised in the follow-up process as ambassadors, coaches or even learning consultants in the organisation - thereby adding to their development as well Figure 1. It varies from project to project, and from enterprise to enterprise, how many participants are to be categorised as red, yellow or green at a given time in the follow-up evaluation process. However, the patterns we do identify are that an organisation typically contains 5-10 per cent in the red category at any given time while the large majority of participants in green rapidly turn yellow in the weeks following a formal learning event. After two to three weeks after a completed learning activity, a large portion of the participants in the green category will have turned yellow. Seen from a value creation's perspective, the idea is quite simple: by launching weekly pulse surveys, the organisation, HRM and the nearest leaders have the tool to identify those in yellow (hesitating) and quickly react and adapt its learning implementation practices to offer help/support/refocus. Enterprises that use a real-time learning evaluation tool see early in the process a strong implementation behaviour at levels of 60+ per cent and still see a huge potential for further improving in its capacity to learn as an organisation. The lite learning tool includes participants and the participant's leaders in the evaluation process. In this way, human resource management, or another programme shareholder in an enterprise, has the option to engage in a more active role in the facilitation of the participants and leaders learning processes and outcomes. HRM/programme owners can do follow-ups with participants that have engaged in external or internal formal and informal learning activities if they have had one or more weeks in the red/yellow category. In this way, learning responsible employees can keep their focus on being actively engaged in supporting the integration and use of new knowledge and/or competences when learning processes seem impeded or hindered. The lite tool works, thus, not only as an evaluation method for where implementation does not work optimally but also as a method to initiate dialogue and discussions on why and where new knowledge or competences are not used/implemented and in this way help the participants/teams and organisations strengthen the capacity to work with absorbing and using external and internal knowledge and competences. In this paper, we argue for a double effect from using and deploying the described lite evaluation tools that connect more clearly to the real-time learning processes than a classical "baseline-midline-endline" effect evaluation offer. We see in our case study a positive effect on the individual level using new knowledge and/or competence in a work context. In addition, we see a positive effect on an organisational level. Enterprises that employ process evaluation tools with pulse measures learn to develop, create, integrate and share new knowledge and competences, thus creating a foundation for organisational learning through new routines and practices. In the following section, we will elaborate on lessons learned from our case study. Further, we will present recommendations for practice. Exploit your "ten weeks window of opportunity" Analysis of aggregated case data shows that after a learning activity has been completed, there is a ten weeks "window of opportunity". In this "ten weeks of opportunity" the conditions for influencing the learning process and concrete implementation, e.g., changes in actions or practices of the learner, are most inclined to be pushed in a positive direction. Thus, we recommend leaders and organisations not to wait or hope for learning to automatically happen but actively use the evaluation process data intelligently and in due time. Focus on the learning process ensures better value creation Use of process evaluation tools is productive especially for creating behavioural changes from strategic learning programmes, for example, sales training, negotiation courses, leadership and project management programmes. Weekly pulse surveys make the learning process transparent and adjustable, building optimised circumstances for reaching aims and targets. Mobilise the leaders Organisations (responsible from HR, learning consultants, etc.) are often challenged by lack of commitment from the learners nearest leader when it comes to keeping track and supervision during the learning process. Learning research shows that the nearest leaders behaviour influence in a high degree how the learner experiences the learning situation and whether the learner deems it worth to use energy on integrating and using new knowledge and/or competences in practice. This means that a lack of support and commitment to actively work with integrating knowledge and/or competences from the nearest leader is associated with a corresponding lack of focus by the learner and vice versa. To work actively with learning and development is not the only goal of a team or group in an organisation. Learning and development are closely tied to organisational goal and strategies - incorporating organisational intentionality - thus all involved, including the nearest leader, must ensure that learners (and teams) should be supported in their learning processes. Do not hesitate - activate A real-time learning evaluation tool provides dynamic decision-making routines that cultivate a learning system where organisations and its leaders on a continuous basis can choose where, when and how to do follow-ups and facilitation as regard organisational learning. The three types of implementation behaviour demand support in different ways, which the responsible learning leader and the nearest leader need to pay attention to. For example, learners in the red category may need help to prioritise their portfolio of tasks in order to focus on the learning process while greens can be used to share their knowledge and competences to bolster the learning system. The most important type, however, is the yellows - the group of employees that over a span of time have shown a hesitant and undecided behaviour when it comes to using new knowledge and/or competences in practice. For this group, the pulse measures are used to initiate dialogue with the learners to help them with the learning implementation. Know your learners! This lesson learned reverberates around questions as "who are the learners really?" and "how does employees work with learning implementation in a work context?". Too often the answer to these questions is a black box, given that enterprises mostly rely on previous experiences and hope following the "learning happens automatically" principle. The described and discussed learning evaluation tool equips organisations to work more structured and more informed with competence development. Data-driven leadership From our case studies, when we look at learning implementation programme owners in enterprises today, they have no routines for working with larger sets of concrete knowledge (data) about how participants learning trajectories unfold. Jeopardising the entire learning investment by not ensuring a strong data-driven follow-up process creating the grounds for sustainable organisational learning processes and results creates uncertainty. Why not set the same standards and norms when it comes to learning and development of new knowledge and/or competence development as we do for operations, marketing and sales, service, technology development, etc. Be sensitive and ethical At last, a very important lesson learned is that we need to be sensitive and ethical in our use of data. Working more focussed and strategic with learning evaluation tools and data in organisations, both on an individual, group and organisational level, creates an ethical call for all involved, especially leaders and learning responsible persons with access to data. We need to constantly remind ourselves to use data constructively to support, facilitate and help creating strong learning paths for all involved and not see this type of learning evaluation as a new tool to manage and control careers, bonus, or demote people. We have asked how enterprises are to arrange its learning processes in order to optimise the integration and creation of sustainable organisational learning. Based on a learning evaluation tool that makes a processual real-time evaluation of learning implementation of new knowledge and/or competences, we explored from the case study data how this type of tool influences learning processes and results in organisations. We found that the developed learning evaluation tool had consequences for how organisations work with learning and development on two levels. On the individual level, processual real-time lite learning evaluation tools create transparency and adaptability working with the integration and use of new knowledge and/or competences in a work context. Learning research has reiterated that transfer and sharing of knowledge from learning activities to practice is a challenging process. Often, learning implementation sees very meagre results and missing value creation for the investing enterprise. With a real-time processual tool, organisations have a device learner, leaders and responsible learning agents, which can be used to improve and influence the concrete integration, and use of new knowledge and/or competences as learning happens. On the organisational level, using the processual tool shapes the organisational capacity to improve routines and practices for how to work with organisational learning and learning data in general. Instead of treating learning and development as something that happens "automatically", organisations now have a tool that can help them create the grounds for informed decisions aimed at creating sustainable organisational learning processes and results. Our method and empirical results prompt insights that call for enterprise practice to enhance focus, dialogue and real-time work on learning processes at a multi-stakeholder level in organisations. The design and deployment of a real-time lite evaluation tool in organisations are key to bolstering learning and development, so that organisations and societies can become more responsive in responding to the challenges posed by today's knowledge economy.
|
The purpose of this paper is to explore how enterprises are to arrange its learning processes in order to optimise the integration and creation of sustainable organisational learning. The paper describes a lite learning evaluation technology that makes processual real-time evaluation of implementation of new knowledge and competences in practice context.
|
[SECTION: Method] Unstable and dynamic societal and business environments are making it more imperative than ever that organisations have the ability to operate its learning processes in order to optimise competence development, performance and innovation (Brinkerhoff, 2006). Research-based knowledge on the subject has taught us that for enterprises to optimise the implementation of learning in practice, they need to take care of elements related to the learner, the learning situation and the organisational setting where new knowledge and/or competences are to be integrated and used. Researchers and practitioners, however, from the organisational learning field have, since the first theoretical and empirical studies, struggled to understand how organisations are to optimise its design and models for the sharing and use of new external and internal knowledge and competences in an organisational context. A vast bulk of research underscores that enterprises still see very poor results when it comes to the transferring and sharing of new competences and knowledge in practice (Burke and Hutchins, 2007; Kazbour et al., 2013). The business challenge is that enterprises often have insufficient data and tools to support the organisational learning process aimed at integrating and employing new knowledge and/or competences to improve competence development, performance and innovation. New knowledge and competences often end unused in practice due to relapse of different kinds, e.g., routine take over, no room/time to use the new knowledge, no support from peers or managers and task irrelevant. This challenge leads to the research question for this paper: how are enterprises to arrange its learning processes in order to optimise the integration and create sustainable organisational learning processes and results? In this paper, we present the form and content of an innovative learning technology tool that can assist enterprises in the continuous evaluation of their capacity to integrate and use new knowledge, skills and/or competences in practice using real time and feedback data. Often, launching learning evaluation tools in enterprises can be a cumbersome and - at times - expensive procedure that focuses on verifying the individual outcome and effects after participation in a specific in-service or external learning activity, e.g., a course or seminar. We argue that in order to qualify the transfer of knowledge and organisational learning, enterprises, its leaders and HR should be responsible and need to be better equipped at keeping track and understand the learning processes at an organisational level. An aspiration that does not necessarily comprise comprehensively imbued measurements techniques aimed solely at the end result but instead focuses on how organisations can support and optimise the learning process as it happens at work and in this way build sustainable organisational learning. Today's learning research focuses on how to understand, design and make learning activities, formal as well as informal, an integral part of the existing organisational HRM and management system and practice. A classical theoretical understanding of learning in organisations and workplaces departs from an idea of learning as a matter of designing a learning and practice situation characterised by mutually shared or identical elements. This behavioural - and later developed within a cognitive frame - theory of transfer and sharing of knowledge typically settled in the common elements theory on shared traits between the learning and practice environments or symbolic representations. Later, Baldwin and Ford (1988) advanced a model that has had an immense influence on theoretical and empirical research contributions within the field of learning studies in adult learning studies and human resource. The distinction between three broad dimensions, the learner's characteristics, the design of the learning situation and the work environment settings, has guided research- and practice-oriented contributions within the organisational and workplace learning field for decades. The influence of each dimension and sub-dimensions for the learning process in organisations has been explored, theoretically and empirically, in a vast number of articles and books. These all sources emphasise that we still need to develop knowledge on the topic and to build improved tools and methods in order to optimise and build better guides for the management of learning processes and outcomes in organisations (Burke and Hutchins, 2007). For example, Blume et al. (2010) demonstrate that existing research is characterised by inconsistent measurement data and that the learning transfer "problem" still needs attention from both the research community and responsible agents for the integration of new knowledge and competences in organisations. For example, Blume et al. (2010) underscore the importance of the workplace and organisational settings for improving the concrete integration of new knowledge and competences. One suggestion is to continue scrutinise how the work environment, design of work and absorptive capacities located in the receiving organisation as regard support and management of knowledge, experiences and competences can be leveraged mediated by learning evaluation tools to optimise and better integrate learning (Blume et al., 2010) in practice, which we turn our attention to in the following section. The studied and described tool of this paper (named SuperInsight) is an evaluation and development-oriented organisational learning technology devised with the aim of strengthening a sustainable integration and anchoring of new knowledge and/or competences in practice. The evaluation tool is deliberately designed as a lite version of an evaluation tool in order to avoid, as we often experience, bulky and highly bureaucratized (red tape) procedures when it comes to working with measurement and integration of learning activities in a practice setting. The purpose of the lite tool is two-folded. First, the lite tool aims to reinforce enterprises ability to keep track of, on a running basis, the impact of learning activities in practice based on real-time data. Second, the lite tool intends to support an organisational learning process among the participating enterprises based on continuous input and feedback on the quality of the learning process combined with unique and specific insight on where to intervene - if needed - in order to optimise the learning transfer process and result. The evaluation tool is implemented through a ten weeks pulse survey that keeps track and continuously evaluates a completed learning activity, for example, an external competence development programme or an internal training course. In practice, after the completion of the learning activities, participants receive four questions each week related to the ongoing implementation efforts of new knowledge and/or competences. It takes approximately 45 s for each participant to complete the pulse survey. Collected data and analysis operate with three general indicators to categorise the implementation level of new knowledge and competences. More concretely, the responsible person for the implementation (HR, trainer, manager, leader, etc.) log into a dashboard on her/his computer where the person gets an overview of individuals and/or groups/larger teams learning status as regard the level of implementation. "Green" indicator (strong implementation) colour represents a participant who experiences that she/he is succeeding with the use of new knowledge and competences in practice. "Yellow" colour indicator (hesitant implementation) denotes a participant who experiences that she/he will be successful, nevertheless is hesitating when it comes to the integration and use of new knowledge and competences in their work life. "Red" colour indicator (weak implementation) characterizes a participant who experiences that she/he has come to a complete stop, i.e., no use of the new knowledge and/or competences. The theoretical model behind the lite tool builds on Baldwin and Fords's (1988) "Learning Transfer model". In our model and devised lite tool, we apply the three dimensions and ten factors from the model that is associated with learning transfer. Thus, the lite tool is organised around ten learning factors that have been identified in research as having a strong positive correlation with integration and use of new knowledge and competences in practice (Baldwin and Ford, 1988; Blume et al., 2010; Burke and Hutchins, 2007). Learner characteristics Learner characteristics include elements like self-efficacy, motivation and perceived utility of training. Learner characteristic scores on this indicator should be very high. If not, we typically see that course owner is trying to get a participant to do training where they lack context or a more clear purpose. Training design Training design includes elements like behavioural modelling, error management and realistic case-based learning environments where there is access to real-time feedback and supervision. Training design scores and comments are primarily concentrated on how to better support the implementation process in the organisation (not the actual training). The work environment The work environmentelements like transfer climate, support from leaders and peers, opportunity to perform and follow-up procedures. The work environment scores are (not surprisingly) where we see the largest potential for improvement (especially after week 3). Analysis of data from a case study in a large Danish enterprise from the telecommunications industry conducting leadership and sales training shows that participants who are in the red category, usually are drowning in operations and do not have the possibility to focus on learning implementation on new knowledge or competences in a work context. A characteristic comment from a participant in the red category is: "I will start to work with the new knowledge, when my department get more resources" or "We are firefighting all the time". Here, the role of the closest leader or accountable HR staff is to support the participant in prioritising their work tasks to facilitate the actual use of new knowledge or competences - if the organisation truly wants to be serious about their approach to organisational learning. The analysis shows that constraints in employing new knowledge and competences are not related to the actual course/training activity, nor personal characteristics, but to the concrete work processes and tasks. Participants with an uncertain implementation behaviour (yellow) will typically have obtained new knowledge and have the appropriate motivation and thus be ready to incorporate new knowledge and competences to develop their work practices. However, participants are hesitant when it comes to the concrete implementation. It is important that the organisation and its leaders identify participants in the yellow category at the right time, and subsequently offer proper guidance and support. Comments in this group are typically: "It was a very good course. During the week, I have thought a lot about what we learned [...] But used any of the new skills - I have not". For this category, the task of the closest leader or HR manager is to offer guidance and support to instill the needed implementation. It is specifically in relation to the yellow category that the strength of real-time and pulse survey data shows its aptitude as a learning evaluation technology compared to more evidence-based measurement tools. The lite tool makes processes of learning transparent and provides fast continuous feedback to HRM and the organisation building grounds for actualising sustainable organisational learning throughout the implementation process. Participants in the green category are the organisational champions of integrating and implementing new knowledge and competences in practice. It is important to acknowledge this group of participant's high focus and capacity to implement new knowledge and competences to improve and innovate existing tasks or operations. Participants identified as showing high focus and capacity to create organisational learning can be utilised in the follow-up process as ambassadors, coaches or even learning consultants in the organisation - thereby adding to their development as well Figure 1. It varies from project to project, and from enterprise to enterprise, how many participants are to be categorised as red, yellow or green at a given time in the follow-up evaluation process. However, the patterns we do identify are that an organisation typically contains 5-10 per cent in the red category at any given time while the large majority of participants in green rapidly turn yellow in the weeks following a formal learning event. After two to three weeks after a completed learning activity, a large portion of the participants in the green category will have turned yellow. Seen from a value creation's perspective, the idea is quite simple: by launching weekly pulse surveys, the organisation, HRM and the nearest leaders have the tool to identify those in yellow (hesitating) and quickly react and adapt its learning implementation practices to offer help/support/refocus. Enterprises that use a real-time learning evaluation tool see early in the process a strong implementation behaviour at levels of 60+ per cent and still see a huge potential for further improving in its capacity to learn as an organisation. The lite learning tool includes participants and the participant's leaders in the evaluation process. In this way, human resource management, or another programme shareholder in an enterprise, has the option to engage in a more active role in the facilitation of the participants and leaders learning processes and outcomes. HRM/programme owners can do follow-ups with participants that have engaged in external or internal formal and informal learning activities if they have had one or more weeks in the red/yellow category. In this way, learning responsible employees can keep their focus on being actively engaged in supporting the integration and use of new knowledge and/or competences when learning processes seem impeded or hindered. The lite tool works, thus, not only as an evaluation method for where implementation does not work optimally but also as a method to initiate dialogue and discussions on why and where new knowledge or competences are not used/implemented and in this way help the participants/teams and organisations strengthen the capacity to work with absorbing and using external and internal knowledge and competences. In this paper, we argue for a double effect from using and deploying the described lite evaluation tools that connect more clearly to the real-time learning processes than a classical "baseline-midline-endline" effect evaluation offer. We see in our case study a positive effect on the individual level using new knowledge and/or competence in a work context. In addition, we see a positive effect on an organisational level. Enterprises that employ process evaluation tools with pulse measures learn to develop, create, integrate and share new knowledge and competences, thus creating a foundation for organisational learning through new routines and practices. In the following section, we will elaborate on lessons learned from our case study. Further, we will present recommendations for practice. Exploit your "ten weeks window of opportunity" Analysis of aggregated case data shows that after a learning activity has been completed, there is a ten weeks "window of opportunity". In this "ten weeks of opportunity" the conditions for influencing the learning process and concrete implementation, e.g., changes in actions or practices of the learner, are most inclined to be pushed in a positive direction. Thus, we recommend leaders and organisations not to wait or hope for learning to automatically happen but actively use the evaluation process data intelligently and in due time. Focus on the learning process ensures better value creation Use of process evaluation tools is productive especially for creating behavioural changes from strategic learning programmes, for example, sales training, negotiation courses, leadership and project management programmes. Weekly pulse surveys make the learning process transparent and adjustable, building optimised circumstances for reaching aims and targets. Mobilise the leaders Organisations (responsible from HR, learning consultants, etc.) are often challenged by lack of commitment from the learners nearest leader when it comes to keeping track and supervision during the learning process. Learning research shows that the nearest leaders behaviour influence in a high degree how the learner experiences the learning situation and whether the learner deems it worth to use energy on integrating and using new knowledge and/or competences in practice. This means that a lack of support and commitment to actively work with integrating knowledge and/or competences from the nearest leader is associated with a corresponding lack of focus by the learner and vice versa. To work actively with learning and development is not the only goal of a team or group in an organisation. Learning and development are closely tied to organisational goal and strategies - incorporating organisational intentionality - thus all involved, including the nearest leader, must ensure that learners (and teams) should be supported in their learning processes. Do not hesitate - activate A real-time learning evaluation tool provides dynamic decision-making routines that cultivate a learning system where organisations and its leaders on a continuous basis can choose where, when and how to do follow-ups and facilitation as regard organisational learning. The three types of implementation behaviour demand support in different ways, which the responsible learning leader and the nearest leader need to pay attention to. For example, learners in the red category may need help to prioritise their portfolio of tasks in order to focus on the learning process while greens can be used to share their knowledge and competences to bolster the learning system. The most important type, however, is the yellows - the group of employees that over a span of time have shown a hesitant and undecided behaviour when it comes to using new knowledge and/or competences in practice. For this group, the pulse measures are used to initiate dialogue with the learners to help them with the learning implementation. Know your learners! This lesson learned reverberates around questions as "who are the learners really?" and "how does employees work with learning implementation in a work context?". Too often the answer to these questions is a black box, given that enterprises mostly rely on previous experiences and hope following the "learning happens automatically" principle. The described and discussed learning evaluation tool equips organisations to work more structured and more informed with competence development. Data-driven leadership From our case studies, when we look at learning implementation programme owners in enterprises today, they have no routines for working with larger sets of concrete knowledge (data) about how participants learning trajectories unfold. Jeopardising the entire learning investment by not ensuring a strong data-driven follow-up process creating the grounds for sustainable organisational learning processes and results creates uncertainty. Why not set the same standards and norms when it comes to learning and development of new knowledge and/or competence development as we do for operations, marketing and sales, service, technology development, etc. Be sensitive and ethical At last, a very important lesson learned is that we need to be sensitive and ethical in our use of data. Working more focussed and strategic with learning evaluation tools and data in organisations, both on an individual, group and organisational level, creates an ethical call for all involved, especially leaders and learning responsible persons with access to data. We need to constantly remind ourselves to use data constructively to support, facilitate and help creating strong learning paths for all involved and not see this type of learning evaluation as a new tool to manage and control careers, bonus, or demote people. We have asked how enterprises are to arrange its learning processes in order to optimise the integration and creation of sustainable organisational learning. Based on a learning evaluation tool that makes a processual real-time evaluation of learning implementation of new knowledge and/or competences, we explored from the case study data how this type of tool influences learning processes and results in organisations. We found that the developed learning evaluation tool had consequences for how organisations work with learning and development on two levels. On the individual level, processual real-time lite learning evaluation tools create transparency and adaptability working with the integration and use of new knowledge and/or competences in a work context. Learning research has reiterated that transfer and sharing of knowledge from learning activities to practice is a challenging process. Often, learning implementation sees very meagre results and missing value creation for the investing enterprise. With a real-time processual tool, organisations have a device learner, leaders and responsible learning agents, which can be used to improve and influence the concrete integration, and use of new knowledge and/or competences as learning happens. On the organisational level, using the processual tool shapes the organisational capacity to improve routines and practices for how to work with organisational learning and learning data in general. Instead of treating learning and development as something that happens "automatically", organisations now have a tool that can help them create the grounds for informed decisions aimed at creating sustainable organisational learning processes and results. Our method and empirical results prompt insights that call for enterprise practice to enhance focus, dialogue and real-time work on learning processes at a multi-stakeholder level in organisations. The design and deployment of a real-time lite evaluation tool in organisations are key to bolstering learning and development, so that organisations and societies can become more responsive in responding to the challenges posed by today's knowledge economy.
|
The research is based on a case study that is designed and planned as a mixed method inquiry. The empirical case study is based on data from a large Danish enterprise from the telecommunication industry conducting a leadership and sales training programme. Case study analysis uses data drawn from the implemented pulse survey followed up with qualitative interviews with the course participants.
|
[SECTION: Findings] Unstable and dynamic societal and business environments are making it more imperative than ever that organisations have the ability to operate its learning processes in order to optimise competence development, performance and innovation (Brinkerhoff, 2006). Research-based knowledge on the subject has taught us that for enterprises to optimise the implementation of learning in practice, they need to take care of elements related to the learner, the learning situation and the organisational setting where new knowledge and/or competences are to be integrated and used. Researchers and practitioners, however, from the organisational learning field have, since the first theoretical and empirical studies, struggled to understand how organisations are to optimise its design and models for the sharing and use of new external and internal knowledge and competences in an organisational context. A vast bulk of research underscores that enterprises still see very poor results when it comes to the transferring and sharing of new competences and knowledge in practice (Burke and Hutchins, 2007; Kazbour et al., 2013). The business challenge is that enterprises often have insufficient data and tools to support the organisational learning process aimed at integrating and employing new knowledge and/or competences to improve competence development, performance and innovation. New knowledge and competences often end unused in practice due to relapse of different kinds, e.g., routine take over, no room/time to use the new knowledge, no support from peers or managers and task irrelevant. This challenge leads to the research question for this paper: how are enterprises to arrange its learning processes in order to optimise the integration and create sustainable organisational learning processes and results? In this paper, we present the form and content of an innovative learning technology tool that can assist enterprises in the continuous evaluation of their capacity to integrate and use new knowledge, skills and/or competences in practice using real time and feedback data. Often, launching learning evaluation tools in enterprises can be a cumbersome and - at times - expensive procedure that focuses on verifying the individual outcome and effects after participation in a specific in-service or external learning activity, e.g., a course or seminar. We argue that in order to qualify the transfer of knowledge and organisational learning, enterprises, its leaders and HR should be responsible and need to be better equipped at keeping track and understand the learning processes at an organisational level. An aspiration that does not necessarily comprise comprehensively imbued measurements techniques aimed solely at the end result but instead focuses on how organisations can support and optimise the learning process as it happens at work and in this way build sustainable organisational learning. Today's learning research focuses on how to understand, design and make learning activities, formal as well as informal, an integral part of the existing organisational HRM and management system and practice. A classical theoretical understanding of learning in organisations and workplaces departs from an idea of learning as a matter of designing a learning and practice situation characterised by mutually shared or identical elements. This behavioural - and later developed within a cognitive frame - theory of transfer and sharing of knowledge typically settled in the common elements theory on shared traits between the learning and practice environments or symbolic representations. Later, Baldwin and Ford (1988) advanced a model that has had an immense influence on theoretical and empirical research contributions within the field of learning studies in adult learning studies and human resource. The distinction between three broad dimensions, the learner's characteristics, the design of the learning situation and the work environment settings, has guided research- and practice-oriented contributions within the organisational and workplace learning field for decades. The influence of each dimension and sub-dimensions for the learning process in organisations has been explored, theoretically and empirically, in a vast number of articles and books. These all sources emphasise that we still need to develop knowledge on the topic and to build improved tools and methods in order to optimise and build better guides for the management of learning processes and outcomes in organisations (Burke and Hutchins, 2007). For example, Blume et al. (2010) demonstrate that existing research is characterised by inconsistent measurement data and that the learning transfer "problem" still needs attention from both the research community and responsible agents for the integration of new knowledge and competences in organisations. For example, Blume et al. (2010) underscore the importance of the workplace and organisational settings for improving the concrete integration of new knowledge and competences. One suggestion is to continue scrutinise how the work environment, design of work and absorptive capacities located in the receiving organisation as regard support and management of knowledge, experiences and competences can be leveraged mediated by learning evaluation tools to optimise and better integrate learning (Blume et al., 2010) in practice, which we turn our attention to in the following section. The studied and described tool of this paper (named SuperInsight) is an evaluation and development-oriented organisational learning technology devised with the aim of strengthening a sustainable integration and anchoring of new knowledge and/or competences in practice. The evaluation tool is deliberately designed as a lite version of an evaluation tool in order to avoid, as we often experience, bulky and highly bureaucratized (red tape) procedures when it comes to working with measurement and integration of learning activities in a practice setting. The purpose of the lite tool is two-folded. First, the lite tool aims to reinforce enterprises ability to keep track of, on a running basis, the impact of learning activities in practice based on real-time data. Second, the lite tool intends to support an organisational learning process among the participating enterprises based on continuous input and feedback on the quality of the learning process combined with unique and specific insight on where to intervene - if needed - in order to optimise the learning transfer process and result. The evaluation tool is implemented through a ten weeks pulse survey that keeps track and continuously evaluates a completed learning activity, for example, an external competence development programme or an internal training course. In practice, after the completion of the learning activities, participants receive four questions each week related to the ongoing implementation efforts of new knowledge and/or competences. It takes approximately 45 s for each participant to complete the pulse survey. Collected data and analysis operate with three general indicators to categorise the implementation level of new knowledge and competences. More concretely, the responsible person for the implementation (HR, trainer, manager, leader, etc.) log into a dashboard on her/his computer where the person gets an overview of individuals and/or groups/larger teams learning status as regard the level of implementation. "Green" indicator (strong implementation) colour represents a participant who experiences that she/he is succeeding with the use of new knowledge and competences in practice. "Yellow" colour indicator (hesitant implementation) denotes a participant who experiences that she/he will be successful, nevertheless is hesitating when it comes to the integration and use of new knowledge and competences in their work life. "Red" colour indicator (weak implementation) characterizes a participant who experiences that she/he has come to a complete stop, i.e., no use of the new knowledge and/or competences. The theoretical model behind the lite tool builds on Baldwin and Fords's (1988) "Learning Transfer model". In our model and devised lite tool, we apply the three dimensions and ten factors from the model that is associated with learning transfer. Thus, the lite tool is organised around ten learning factors that have been identified in research as having a strong positive correlation with integration and use of new knowledge and competences in practice (Baldwin and Ford, 1988; Blume et al., 2010; Burke and Hutchins, 2007). Learner characteristics Learner characteristics include elements like self-efficacy, motivation and perceived utility of training. Learner characteristic scores on this indicator should be very high. If not, we typically see that course owner is trying to get a participant to do training where they lack context or a more clear purpose. Training design Training design includes elements like behavioural modelling, error management and realistic case-based learning environments where there is access to real-time feedback and supervision. Training design scores and comments are primarily concentrated on how to better support the implementation process in the organisation (not the actual training). The work environment The work environmentelements like transfer climate, support from leaders and peers, opportunity to perform and follow-up procedures. The work environment scores are (not surprisingly) where we see the largest potential for improvement (especially after week 3). Analysis of data from a case study in a large Danish enterprise from the telecommunications industry conducting leadership and sales training shows that participants who are in the red category, usually are drowning in operations and do not have the possibility to focus on learning implementation on new knowledge or competences in a work context. A characteristic comment from a participant in the red category is: "I will start to work with the new knowledge, when my department get more resources" or "We are firefighting all the time". Here, the role of the closest leader or accountable HR staff is to support the participant in prioritising their work tasks to facilitate the actual use of new knowledge or competences - if the organisation truly wants to be serious about their approach to organisational learning. The analysis shows that constraints in employing new knowledge and competences are not related to the actual course/training activity, nor personal characteristics, but to the concrete work processes and tasks. Participants with an uncertain implementation behaviour (yellow) will typically have obtained new knowledge and have the appropriate motivation and thus be ready to incorporate new knowledge and competences to develop their work practices. However, participants are hesitant when it comes to the concrete implementation. It is important that the organisation and its leaders identify participants in the yellow category at the right time, and subsequently offer proper guidance and support. Comments in this group are typically: "It was a very good course. During the week, I have thought a lot about what we learned [...] But used any of the new skills - I have not". For this category, the task of the closest leader or HR manager is to offer guidance and support to instill the needed implementation. It is specifically in relation to the yellow category that the strength of real-time and pulse survey data shows its aptitude as a learning evaluation technology compared to more evidence-based measurement tools. The lite tool makes processes of learning transparent and provides fast continuous feedback to HRM and the organisation building grounds for actualising sustainable organisational learning throughout the implementation process. Participants in the green category are the organisational champions of integrating and implementing new knowledge and competences in practice. It is important to acknowledge this group of participant's high focus and capacity to implement new knowledge and competences to improve and innovate existing tasks or operations. Participants identified as showing high focus and capacity to create organisational learning can be utilised in the follow-up process as ambassadors, coaches or even learning consultants in the organisation - thereby adding to their development as well Figure 1. It varies from project to project, and from enterprise to enterprise, how many participants are to be categorised as red, yellow or green at a given time in the follow-up evaluation process. However, the patterns we do identify are that an organisation typically contains 5-10 per cent in the red category at any given time while the large majority of participants in green rapidly turn yellow in the weeks following a formal learning event. After two to three weeks after a completed learning activity, a large portion of the participants in the green category will have turned yellow. Seen from a value creation's perspective, the idea is quite simple: by launching weekly pulse surveys, the organisation, HRM and the nearest leaders have the tool to identify those in yellow (hesitating) and quickly react and adapt its learning implementation practices to offer help/support/refocus. Enterprises that use a real-time learning evaluation tool see early in the process a strong implementation behaviour at levels of 60+ per cent and still see a huge potential for further improving in its capacity to learn as an organisation. The lite learning tool includes participants and the participant's leaders in the evaluation process. In this way, human resource management, or another programme shareholder in an enterprise, has the option to engage in a more active role in the facilitation of the participants and leaders learning processes and outcomes. HRM/programme owners can do follow-ups with participants that have engaged in external or internal formal and informal learning activities if they have had one or more weeks in the red/yellow category. In this way, learning responsible employees can keep their focus on being actively engaged in supporting the integration and use of new knowledge and/or competences when learning processes seem impeded or hindered. The lite tool works, thus, not only as an evaluation method for where implementation does not work optimally but also as a method to initiate dialogue and discussions on why and where new knowledge or competences are not used/implemented and in this way help the participants/teams and organisations strengthen the capacity to work with absorbing and using external and internal knowledge and competences. In this paper, we argue for a double effect from using and deploying the described lite evaluation tools that connect more clearly to the real-time learning processes than a classical "baseline-midline-endline" effect evaluation offer. We see in our case study a positive effect on the individual level using new knowledge and/or competence in a work context. In addition, we see a positive effect on an organisational level. Enterprises that employ process evaluation tools with pulse measures learn to develop, create, integrate and share new knowledge and competences, thus creating a foundation for organisational learning through new routines and practices. In the following section, we will elaborate on lessons learned from our case study. Further, we will present recommendations for practice. Exploit your "ten weeks window of opportunity" Analysis of aggregated case data shows that after a learning activity has been completed, there is a ten weeks "window of opportunity". In this "ten weeks of opportunity" the conditions for influencing the learning process and concrete implementation, e.g., changes in actions or practices of the learner, are most inclined to be pushed in a positive direction. Thus, we recommend leaders and organisations not to wait or hope for learning to automatically happen but actively use the evaluation process data intelligently and in due time. Focus on the learning process ensures better value creation Use of process evaluation tools is productive especially for creating behavioural changes from strategic learning programmes, for example, sales training, negotiation courses, leadership and project management programmes. Weekly pulse surveys make the learning process transparent and adjustable, building optimised circumstances for reaching aims and targets. Mobilise the leaders Organisations (responsible from HR, learning consultants, etc.) are often challenged by lack of commitment from the learners nearest leader when it comes to keeping track and supervision during the learning process. Learning research shows that the nearest leaders behaviour influence in a high degree how the learner experiences the learning situation and whether the learner deems it worth to use energy on integrating and using new knowledge and/or competences in practice. This means that a lack of support and commitment to actively work with integrating knowledge and/or competences from the nearest leader is associated with a corresponding lack of focus by the learner and vice versa. To work actively with learning and development is not the only goal of a team or group in an organisation. Learning and development are closely tied to organisational goal and strategies - incorporating organisational intentionality - thus all involved, including the nearest leader, must ensure that learners (and teams) should be supported in their learning processes. Do not hesitate - activate A real-time learning evaluation tool provides dynamic decision-making routines that cultivate a learning system where organisations and its leaders on a continuous basis can choose where, when and how to do follow-ups and facilitation as regard organisational learning. The three types of implementation behaviour demand support in different ways, which the responsible learning leader and the nearest leader need to pay attention to. For example, learners in the red category may need help to prioritise their portfolio of tasks in order to focus on the learning process while greens can be used to share their knowledge and competences to bolster the learning system. The most important type, however, is the yellows - the group of employees that over a span of time have shown a hesitant and undecided behaviour when it comes to using new knowledge and/or competences in practice. For this group, the pulse measures are used to initiate dialogue with the learners to help them with the learning implementation. Know your learners! This lesson learned reverberates around questions as "who are the learners really?" and "how does employees work with learning implementation in a work context?". Too often the answer to these questions is a black box, given that enterprises mostly rely on previous experiences and hope following the "learning happens automatically" principle. The described and discussed learning evaluation tool equips organisations to work more structured and more informed with competence development. Data-driven leadership From our case studies, when we look at learning implementation programme owners in enterprises today, they have no routines for working with larger sets of concrete knowledge (data) about how participants learning trajectories unfold. Jeopardising the entire learning investment by not ensuring a strong data-driven follow-up process creating the grounds for sustainable organisational learning processes and results creates uncertainty. Why not set the same standards and norms when it comes to learning and development of new knowledge and/or competence development as we do for operations, marketing and sales, service, technology development, etc. Be sensitive and ethical At last, a very important lesson learned is that we need to be sensitive and ethical in our use of data. Working more focussed and strategic with learning evaluation tools and data in organisations, both on an individual, group and organisational level, creates an ethical call for all involved, especially leaders and learning responsible persons with access to data. We need to constantly remind ourselves to use data constructively to support, facilitate and help creating strong learning paths for all involved and not see this type of learning evaluation as a new tool to manage and control careers, bonus, or demote people. We have asked how enterprises are to arrange its learning processes in order to optimise the integration and creation of sustainable organisational learning. Based on a learning evaluation tool that makes a processual real-time evaluation of learning implementation of new knowledge and/or competences, we explored from the case study data how this type of tool influences learning processes and results in organisations. We found that the developed learning evaluation tool had consequences for how organisations work with learning and development on two levels. On the individual level, processual real-time lite learning evaluation tools create transparency and adaptability working with the integration and use of new knowledge and/or competences in a work context. Learning research has reiterated that transfer and sharing of knowledge from learning activities to practice is a challenging process. Often, learning implementation sees very meagre results and missing value creation for the investing enterprise. With a real-time processual tool, organisations have a device learner, leaders and responsible learning agents, which can be used to improve and influence the concrete integration, and use of new knowledge and/or competences as learning happens. On the organisational level, using the processual tool shapes the organisational capacity to improve routines and practices for how to work with organisational learning and learning data in general. Instead of treating learning and development as something that happens "automatically", organisations now have a tool that can help them create the grounds for informed decisions aimed at creating sustainable organisational learning processes and results. Our method and empirical results prompt insights that call for enterprise practice to enhance focus, dialogue and real-time work on learning processes at a multi-stakeholder level in organisations. The design and deployment of a real-time lite evaluation tool in organisations are key to bolstering learning and development, so that organisations and societies can become more responsive in responding to the challenges posed by today's knowledge economy.
|
The authors show results on two levels. On the individual level, processual real-time lite learning evaluation tools create transparency and adaptability. On the organisational level, tool shapes the organisational capacity to improve routines and practices for how to work with organisational learning and learning data in general. Instead of treating learning and development as something that happens "automatically", organisations now have a tool for informed decisions aimed at creating sustainable organisational learning processes and results.
|
[SECTION: Value] Unstable and dynamic societal and business environments are making it more imperative than ever that organisations have the ability to operate its learning processes in order to optimise competence development, performance and innovation (Brinkerhoff, 2006). Research-based knowledge on the subject has taught us that for enterprises to optimise the implementation of learning in practice, they need to take care of elements related to the learner, the learning situation and the organisational setting where new knowledge and/or competences are to be integrated and used. Researchers and practitioners, however, from the organisational learning field have, since the first theoretical and empirical studies, struggled to understand how organisations are to optimise its design and models for the sharing and use of new external and internal knowledge and competences in an organisational context. A vast bulk of research underscores that enterprises still see very poor results when it comes to the transferring and sharing of new competences and knowledge in practice (Burke and Hutchins, 2007; Kazbour et al., 2013). The business challenge is that enterprises often have insufficient data and tools to support the organisational learning process aimed at integrating and employing new knowledge and/or competences to improve competence development, performance and innovation. New knowledge and competences often end unused in practice due to relapse of different kinds, e.g., routine take over, no room/time to use the new knowledge, no support from peers or managers and task irrelevant. This challenge leads to the research question for this paper: how are enterprises to arrange its learning processes in order to optimise the integration and create sustainable organisational learning processes and results? In this paper, we present the form and content of an innovative learning technology tool that can assist enterprises in the continuous evaluation of their capacity to integrate and use new knowledge, skills and/or competences in practice using real time and feedback data. Often, launching learning evaluation tools in enterprises can be a cumbersome and - at times - expensive procedure that focuses on verifying the individual outcome and effects after participation in a specific in-service or external learning activity, e.g., a course or seminar. We argue that in order to qualify the transfer of knowledge and organisational learning, enterprises, its leaders and HR should be responsible and need to be better equipped at keeping track and understand the learning processes at an organisational level. An aspiration that does not necessarily comprise comprehensively imbued measurements techniques aimed solely at the end result but instead focuses on how organisations can support and optimise the learning process as it happens at work and in this way build sustainable organisational learning. Today's learning research focuses on how to understand, design and make learning activities, formal as well as informal, an integral part of the existing organisational HRM and management system and practice. A classical theoretical understanding of learning in organisations and workplaces departs from an idea of learning as a matter of designing a learning and practice situation characterised by mutually shared or identical elements. This behavioural - and later developed within a cognitive frame - theory of transfer and sharing of knowledge typically settled in the common elements theory on shared traits between the learning and practice environments or symbolic representations. Later, Baldwin and Ford (1988) advanced a model that has had an immense influence on theoretical and empirical research contributions within the field of learning studies in adult learning studies and human resource. The distinction between three broad dimensions, the learner's characteristics, the design of the learning situation and the work environment settings, has guided research- and practice-oriented contributions within the organisational and workplace learning field for decades. The influence of each dimension and sub-dimensions for the learning process in organisations has been explored, theoretically and empirically, in a vast number of articles and books. These all sources emphasise that we still need to develop knowledge on the topic and to build improved tools and methods in order to optimise and build better guides for the management of learning processes and outcomes in organisations (Burke and Hutchins, 2007). For example, Blume et al. (2010) demonstrate that existing research is characterised by inconsistent measurement data and that the learning transfer "problem" still needs attention from both the research community and responsible agents for the integration of new knowledge and competences in organisations. For example, Blume et al. (2010) underscore the importance of the workplace and organisational settings for improving the concrete integration of new knowledge and competences. One suggestion is to continue scrutinise how the work environment, design of work and absorptive capacities located in the receiving organisation as regard support and management of knowledge, experiences and competences can be leveraged mediated by learning evaluation tools to optimise and better integrate learning (Blume et al., 2010) in practice, which we turn our attention to in the following section. The studied and described tool of this paper (named SuperInsight) is an evaluation and development-oriented organisational learning technology devised with the aim of strengthening a sustainable integration and anchoring of new knowledge and/or competences in practice. The evaluation tool is deliberately designed as a lite version of an evaluation tool in order to avoid, as we often experience, bulky and highly bureaucratized (red tape) procedures when it comes to working with measurement and integration of learning activities in a practice setting. The purpose of the lite tool is two-folded. First, the lite tool aims to reinforce enterprises ability to keep track of, on a running basis, the impact of learning activities in practice based on real-time data. Second, the lite tool intends to support an organisational learning process among the participating enterprises based on continuous input and feedback on the quality of the learning process combined with unique and specific insight on where to intervene - if needed - in order to optimise the learning transfer process and result. The evaluation tool is implemented through a ten weeks pulse survey that keeps track and continuously evaluates a completed learning activity, for example, an external competence development programme or an internal training course. In practice, after the completion of the learning activities, participants receive four questions each week related to the ongoing implementation efforts of new knowledge and/or competences. It takes approximately 45 s for each participant to complete the pulse survey. Collected data and analysis operate with three general indicators to categorise the implementation level of new knowledge and competences. More concretely, the responsible person for the implementation (HR, trainer, manager, leader, etc.) log into a dashboard on her/his computer where the person gets an overview of individuals and/or groups/larger teams learning status as regard the level of implementation. "Green" indicator (strong implementation) colour represents a participant who experiences that she/he is succeeding with the use of new knowledge and competences in practice. "Yellow" colour indicator (hesitant implementation) denotes a participant who experiences that she/he will be successful, nevertheless is hesitating when it comes to the integration and use of new knowledge and competences in their work life. "Red" colour indicator (weak implementation) characterizes a participant who experiences that she/he has come to a complete stop, i.e., no use of the new knowledge and/or competences. The theoretical model behind the lite tool builds on Baldwin and Fords's (1988) "Learning Transfer model". In our model and devised lite tool, we apply the three dimensions and ten factors from the model that is associated with learning transfer. Thus, the lite tool is organised around ten learning factors that have been identified in research as having a strong positive correlation with integration and use of new knowledge and competences in practice (Baldwin and Ford, 1988; Blume et al., 2010; Burke and Hutchins, 2007). Learner characteristics Learner characteristics include elements like self-efficacy, motivation and perceived utility of training. Learner characteristic scores on this indicator should be very high. If not, we typically see that course owner is trying to get a participant to do training where they lack context or a more clear purpose. Training design Training design includes elements like behavioural modelling, error management and realistic case-based learning environments where there is access to real-time feedback and supervision. Training design scores and comments are primarily concentrated on how to better support the implementation process in the organisation (not the actual training). The work environment The work environmentelements like transfer climate, support from leaders and peers, opportunity to perform and follow-up procedures. The work environment scores are (not surprisingly) where we see the largest potential for improvement (especially after week 3). Analysis of data from a case study in a large Danish enterprise from the telecommunications industry conducting leadership and sales training shows that participants who are in the red category, usually are drowning in operations and do not have the possibility to focus on learning implementation on new knowledge or competences in a work context. A characteristic comment from a participant in the red category is: "I will start to work with the new knowledge, when my department get more resources" or "We are firefighting all the time". Here, the role of the closest leader or accountable HR staff is to support the participant in prioritising their work tasks to facilitate the actual use of new knowledge or competences - if the organisation truly wants to be serious about their approach to organisational learning. The analysis shows that constraints in employing new knowledge and competences are not related to the actual course/training activity, nor personal characteristics, but to the concrete work processes and tasks. Participants with an uncertain implementation behaviour (yellow) will typically have obtained new knowledge and have the appropriate motivation and thus be ready to incorporate new knowledge and competences to develop their work practices. However, participants are hesitant when it comes to the concrete implementation. It is important that the organisation and its leaders identify participants in the yellow category at the right time, and subsequently offer proper guidance and support. Comments in this group are typically: "It was a very good course. During the week, I have thought a lot about what we learned [...] But used any of the new skills - I have not". For this category, the task of the closest leader or HR manager is to offer guidance and support to instill the needed implementation. It is specifically in relation to the yellow category that the strength of real-time and pulse survey data shows its aptitude as a learning evaluation technology compared to more evidence-based measurement tools. The lite tool makes processes of learning transparent and provides fast continuous feedback to HRM and the organisation building grounds for actualising sustainable organisational learning throughout the implementation process. Participants in the green category are the organisational champions of integrating and implementing new knowledge and competences in practice. It is important to acknowledge this group of participant's high focus and capacity to implement new knowledge and competences to improve and innovate existing tasks or operations. Participants identified as showing high focus and capacity to create organisational learning can be utilised in the follow-up process as ambassadors, coaches or even learning consultants in the organisation - thereby adding to their development as well Figure 1. It varies from project to project, and from enterprise to enterprise, how many participants are to be categorised as red, yellow or green at a given time in the follow-up evaluation process. However, the patterns we do identify are that an organisation typically contains 5-10 per cent in the red category at any given time while the large majority of participants in green rapidly turn yellow in the weeks following a formal learning event. After two to three weeks after a completed learning activity, a large portion of the participants in the green category will have turned yellow. Seen from a value creation's perspective, the idea is quite simple: by launching weekly pulse surveys, the organisation, HRM and the nearest leaders have the tool to identify those in yellow (hesitating) and quickly react and adapt its learning implementation practices to offer help/support/refocus. Enterprises that use a real-time learning evaluation tool see early in the process a strong implementation behaviour at levels of 60+ per cent and still see a huge potential for further improving in its capacity to learn as an organisation. The lite learning tool includes participants and the participant's leaders in the evaluation process. In this way, human resource management, or another programme shareholder in an enterprise, has the option to engage in a more active role in the facilitation of the participants and leaders learning processes and outcomes. HRM/programme owners can do follow-ups with participants that have engaged in external or internal formal and informal learning activities if they have had one or more weeks in the red/yellow category. In this way, learning responsible employees can keep their focus on being actively engaged in supporting the integration and use of new knowledge and/or competences when learning processes seem impeded or hindered. The lite tool works, thus, not only as an evaluation method for where implementation does not work optimally but also as a method to initiate dialogue and discussions on why and where new knowledge or competences are not used/implemented and in this way help the participants/teams and organisations strengthen the capacity to work with absorbing and using external and internal knowledge and competences. In this paper, we argue for a double effect from using and deploying the described lite evaluation tools that connect more clearly to the real-time learning processes than a classical "baseline-midline-endline" effect evaluation offer. We see in our case study a positive effect on the individual level using new knowledge and/or competence in a work context. In addition, we see a positive effect on an organisational level. Enterprises that employ process evaluation tools with pulse measures learn to develop, create, integrate and share new knowledge and competences, thus creating a foundation for organisational learning through new routines and practices. In the following section, we will elaborate on lessons learned from our case study. Further, we will present recommendations for practice. Exploit your "ten weeks window of opportunity" Analysis of aggregated case data shows that after a learning activity has been completed, there is a ten weeks "window of opportunity". In this "ten weeks of opportunity" the conditions for influencing the learning process and concrete implementation, e.g., changes in actions or practices of the learner, are most inclined to be pushed in a positive direction. Thus, we recommend leaders and organisations not to wait or hope for learning to automatically happen but actively use the evaluation process data intelligently and in due time. Focus on the learning process ensures better value creation Use of process evaluation tools is productive especially for creating behavioural changes from strategic learning programmes, for example, sales training, negotiation courses, leadership and project management programmes. Weekly pulse surveys make the learning process transparent and adjustable, building optimised circumstances for reaching aims and targets. Mobilise the leaders Organisations (responsible from HR, learning consultants, etc.) are often challenged by lack of commitment from the learners nearest leader when it comes to keeping track and supervision during the learning process. Learning research shows that the nearest leaders behaviour influence in a high degree how the learner experiences the learning situation and whether the learner deems it worth to use energy on integrating and using new knowledge and/or competences in practice. This means that a lack of support and commitment to actively work with integrating knowledge and/or competences from the nearest leader is associated with a corresponding lack of focus by the learner and vice versa. To work actively with learning and development is not the only goal of a team or group in an organisation. Learning and development are closely tied to organisational goal and strategies - incorporating organisational intentionality - thus all involved, including the nearest leader, must ensure that learners (and teams) should be supported in their learning processes. Do not hesitate - activate A real-time learning evaluation tool provides dynamic decision-making routines that cultivate a learning system where organisations and its leaders on a continuous basis can choose where, when and how to do follow-ups and facilitation as regard organisational learning. The three types of implementation behaviour demand support in different ways, which the responsible learning leader and the nearest leader need to pay attention to. For example, learners in the red category may need help to prioritise their portfolio of tasks in order to focus on the learning process while greens can be used to share their knowledge and competences to bolster the learning system. The most important type, however, is the yellows - the group of employees that over a span of time have shown a hesitant and undecided behaviour when it comes to using new knowledge and/or competences in practice. For this group, the pulse measures are used to initiate dialogue with the learners to help them with the learning implementation. Know your learners! This lesson learned reverberates around questions as "who are the learners really?" and "how does employees work with learning implementation in a work context?". Too often the answer to these questions is a black box, given that enterprises mostly rely on previous experiences and hope following the "learning happens automatically" principle. The described and discussed learning evaluation tool equips organisations to work more structured and more informed with competence development. Data-driven leadership From our case studies, when we look at learning implementation programme owners in enterprises today, they have no routines for working with larger sets of concrete knowledge (data) about how participants learning trajectories unfold. Jeopardising the entire learning investment by not ensuring a strong data-driven follow-up process creating the grounds for sustainable organisational learning processes and results creates uncertainty. Why not set the same standards and norms when it comes to learning and development of new knowledge and/or competence development as we do for operations, marketing and sales, service, technology development, etc. Be sensitive and ethical At last, a very important lesson learned is that we need to be sensitive and ethical in our use of data. Working more focussed and strategic with learning evaluation tools and data in organisations, both on an individual, group and organisational level, creates an ethical call for all involved, especially leaders and learning responsible persons with access to data. We need to constantly remind ourselves to use data constructively to support, facilitate and help creating strong learning paths for all involved and not see this type of learning evaluation as a new tool to manage and control careers, bonus, or demote people. We have asked how enterprises are to arrange its learning processes in order to optimise the integration and creation of sustainable organisational learning. Based on a learning evaluation tool that makes a processual real-time evaluation of learning implementation of new knowledge and/or competences, we explored from the case study data how this type of tool influences learning processes and results in organisations. We found that the developed learning evaluation tool had consequences for how organisations work with learning and development on two levels. On the individual level, processual real-time lite learning evaluation tools create transparency and adaptability working with the integration and use of new knowledge and/or competences in a work context. Learning research has reiterated that transfer and sharing of knowledge from learning activities to practice is a challenging process. Often, learning implementation sees very meagre results and missing value creation for the investing enterprise. With a real-time processual tool, organisations have a device learner, leaders and responsible learning agents, which can be used to improve and influence the concrete integration, and use of new knowledge and/or competences as learning happens. On the organisational level, using the processual tool shapes the organisational capacity to improve routines and practices for how to work with organisational learning and learning data in general. Instead of treating learning and development as something that happens "automatically", organisations now have a tool that can help them create the grounds for informed decisions aimed at creating sustainable organisational learning processes and results. Our method and empirical results prompt insights that call for enterprise practice to enhance focus, dialogue and real-time work on learning processes at a multi-stakeholder level in organisations. The design and deployment of a real-time lite evaluation tool in organisations are key to bolstering learning and development, so that organisations and societies can become more responsive in responding to the challenges posed by today's knowledge economy.
|
The paper prompts insights that call for enterprises to enhance focus and dialogue on how to work in new and smart ways with learning at a multi-stakeholder level in organisations. The design and deployment of a real-time lite evaluation tool in organisations are key to bolster learning and competence development, so that organisations and societies can become more responsive in responding to the challenges posed by today's knowledge economy.
|
[SECTION: Purpose] The Irish wine market has experienced unprecedented growth in the last from 15 to 20 years. From 1990 to 2007, total wine sales in Ireland have more than quadrupled, increasing from 1.7 to 7.6 million cases. In the 13 years between 1994 and 2007, wine's proportion of the Irish alcohol market more than doubled from 8 per cent to 17.9 per cent (Wine Development Board, 2007). Growth in wine consumption is forecasted to continue with a growth of 15 per cent expected by 2012 (Euromonitor, 2008). As the wine drinking culture in Ireland is relatively new, the segmentation of the market and brand positioning is in its infancy. Further study into segmentation is required to improve the profitability of the industry, and to develop choice and the accessibility of wine for Irish consumers. The specific purpose of the paper is to examine how the Irish wine market may be effectively segmented for improved brand positioning in Ireland. Thus, the paper aims to determine the key trends in the Irish wine market, examine the state of marketing in the wine industry, evaluate different approaches to segmenting the Irish wine market, and develop profiles of the resulting segments. The Irish wine market has experienced remarkable growth with the number of wine drinkers in Ireland doubling since 1990 and with over five times as much table wine being consumed in 2007, as was consumed in 1990 (WDB, 2007). The increased consumption of wine in Ireland over the last 15 years is attributed to the improved accessibility, affordability and branding of wine (Moran, 2002). To emphasise the significance of the growth in wine consumption in Ireland, the level of growth in wine buying is compared with growth in the overall food and beverage sector. Practically all wine bought in Ireland is imported. Between 2000 and 2004, wine sales (and therefore imports) increased by 56 per cent (WDB, 2004), while the overall imports of the food, beverages and other animal products category increased by only 18 per cent in the same time frame (CSO, 2006).For a marketer assessing the Irish wine market, equally important to the growth in consumption, is the huge shift in the type of wine preferred by the Irish wine drinker. Specifically, there is a notable shift towards New World wines, with diminishing preference for Old World wines (WDB, 2007). New World wines refer to wines from regions outside of Europe. Prominent New World wine producing regions include South Africa, Australia, New Zealand, Chile, Argentina and California. Old World countries refer to European countries with a long history of wine production, such as France, Italy, Germany and Spain (Fielden, 1994). Up to 1990, the majority of wine consumed in Ireland was Old World wine, and accounted for 94 per cent of the market (WDB, 2007). Since 1990, there has been a steady shift in demand towards New World wine, which in 1990 accounted for 6 per cent of the market and in 2007 held a 71 per cent market share (WDB, 2007). Historically, French wines were the market leader in the Irish wine market, but since 2001, Australia now has the largest market share, accounting for 26 per cent of wine consumed in Ireland (WDB, 2007). In 2006, the top ten wine brands in the Irish still light wine market accounted for nearly 25 per cent of total sales. These top ten wine brands and their respective brand shares are Jacob's Creek (3.2 per cent), Blossom Hill (2.7 per cent), Rosemount (2.6 per cent), E&J Gallo (2.5 per cent), Wolf Blass (2.5 per cent), Hardys (2.3 per cent), Concha Y Toro (2.3 per cent), Long Mountain (2.2 per cent), Santa Rita (2.1 per cent) and Carmen (1.9 per cent), all of which are from New World countries (Euromonitor, 2008). Understanding the shift in the country of origin preference is important as it represents an important shift in preference for style, taste, brand, price and other wine variables. Testing country of origin preference is, therefore, an essential element in this Irish wine market study. The production of wine is a specialised area, and the wine industry has traditionally adopted a production-focused mindset with the complexities of viticulture and vinification having occupied the attention of specialists in the area (Thomas and Pickering, 2003). Bruwer et al. (2002) note the agricultural basis of the wine value chain, and the industry is often criticised as employing mass marketing campaigns (Gluckman, 1990; Spawton, 1991a; Hall and Winchester, 1999; Bruwer et al., 2002).According to Thomas and Pickering (2003), the marketing of wine is in its infancy, relative to the long history of wine making and wine drinking. Aggressive marketing was uncommon in the industry with vineyard operators relying on the strength of their reputation to compete in the marketplace (Hall, 2004). In the early 1990s, an interest in branding emerged in the industry as a method of coping with changes in distribution and the growth of wine retailing. In 1991, the European Journal of Marketing dedicated an issue to the marketing of wine, where Spawton (1991b, c, d, e) through a series of articles, provides an insight into the state of wine marketing. This special issue served as an introduction to marketing for the industry, with the purpose of illustrating the advantage and necessity in developing a customer mindset.There is a realisation in the industry that its future is geared to meeting the expectations of the wine consumer. That has contributed to the growing importance of wine marketing within the industry (Spawton, 1991b, p. 6).Gluckman (1990), in a frequently cited article, presents a number of challenges facing the wine marketer in branding wines. Most notably, a move towards own brand labelling by retailers reinforces the necessity for strong wine brands. This need for improved branding of wine has prompted the undertaking of research in wine markets. In an industry which is relatively new to marketing, and in the Irish wine market, which has seen tremendous growth and transformation, there is a need for greater understanding of the market dynamics. Market research in general, and market segmentation in particular, has a potentially pivotal role to play in assisting wine marketers to position their wine brands effectively. Weir (1960, p. 95, as cited in Yankelovich, 1964) provides the following description of what a market is, and more importantly, what it is not:"The market" is not a single, cohesive unit; it is a seething, disparate, pullulating, antagonistic, infinitely varied sea of differing human beings - every one of them as distinct from every other one as fingerprints; everyone of them living in circumstances different in countless ways from those in which every other one of them is living.This description of a market is a colourful representation of popular marketing thought on the composition of markets in the late 1950s and 1960s. Smith (1956), in what is considered a landmark article (Reynolds, 1965; Haley, 1968; Wind, 1978; Green and Krieger, 1991; Lin, 2002) introduces a marketing strategy labelled market segmentation, as an approach to competing successfully in the reality of an environment of imperfect competition. The original article by Smith (1956) introduces market segmentation as a strategy. Market segmentation strategy was considered an alternative to product differentiation strategy to deal with diversity in the market. While the initial representation of the market segmentation strategy is based in economic theory, market segmentation developed as one of the most foremost concepts in marketing thought (Wind, 1978; Johnson et al., 1991; Lin, 2002).At a broad level, market segmentation provides a marketer with a clearer focus on customer needs, and thereby aids decision making for improved competitive advantage (McDonald and Dunbar, 1992; Croft, 1994; Kotler and Keller, 2003). While Croft (1994) highlights market segmentation as aiding decision making in general, Yankelovich (1964) specifies what exactly segmentation analysis can achieve. In identifying groups of customers with similar needs, a marketer has the information required to target the most profitable group with the most potential. With this knowledge a marketer can develop product lines and promotion activities, choose advertising media, advance positioning of offerings and improve timing of advertising to appeal to the segment of the market whose needs possess the greatest profit potential.A critical decision to be made in conducting segmentation research is choosing an appropriate segmentation base. A segmentation base is the criteria used to divide the defined market into groups of consumers with similarities. At the most basic level a market can be split up according to the profiles of the consumers. Variables such as demographics, geographic location of consumers and the socio economic class to which they belong, are considered profile segmentation bases. The behavioural segmentation category includes bases such as usage occasion, benefits sought, perceptions and beliefs, while the psychographic bases category includes lifestyle and personality variables as a means for identifying groups of consumers with similarities. The more abstract and less concrete the information required for the segmentation base, the more difficult it is to measure responses and their link with behaviour. In choosing a segmentation base for a wine market study there are a number aspects to the market which need to be considered. Literature on wine consumer behaviour focuses on two areas; the factors influencing wine consumer behaviour; and the wine consumer's purchasing decision-making process. An effective wine segmentation study would be one which aids understanding of these two areas and aids marketers in evaluating how stimuli, such as brand positioning strategies, influence wine choice.According to Bruwer et al. (2002), wine markets have been segmented using all the bases identified above. For the purpose of an Irish wine market segmentation study, behavioural segmentation with an involvement basis proves a suitable choice as it is an approach, which yields insight into consumer behaviour, but is not overly difficult to measure (Lockshin et al., 2001). Employing a behavioural segmentation base allows for the decision making process and the influencing factors of the Irish wine consumer to be tested and makes the process and the factors the basis for splitting up the market into meaningful and actionable segments.A key consideration in exploring the wine consumer decision-making process and consumers' evaluation of alternatives, is that wine attributes represent intrinsic and extrinsic cues for the consumer. Sanchez and Gill (1998) illustrate how consumers have preferences according to the bundle of benefits they are seeking. The challenge in understanding these preferences is the large number of wine attributes which exist, and therefore, the greater number of possible bundles of benefits that are present. Wine attributes include: brand name, producer, grape variety, blend of grape varieties, vintage, region of origin, price, label, bottle type, cork type, bottle size, colour of wine, style of wine and level of alcohol. Due to the large number of wine attributes, wine consumers have a wider range of considerations in making purchasing decisions. Examining the hierarchy of importance of wine attributes to Irish wine buyers is a central consideration in segmenting the market. The research design is primarily descriptive in nature as similar investigations into other wine markets have been descriptive in design (Orth et al., 2005; Johnson et al., 1991; Hall, 2004; Bruwer et al. Li and Reid, 2002; Johnson, 2003; Thomas and Pickering, 2003). Due to the descriptive nature of this research, a quantitative approach to primary data collection is most suitable. Quantitative data is appropriate for determining and understanding the behaviours and characteristics of a large sample of wine drinkers. Specifically, survey data collection was undertaken, with a questionnaire collection instrument administered through a personal interview. The questionnaire was two pages in length with 15 tick box questions. The questionnaire posed questions to gather data on four topics: volume of usage, buying preference, product involvement, and demographic information.As an accurate sampling frame was unavailable for the population of the 1,451,000 wine drinkers in Ireland (WDB, 2004), non-probability sampling was undertaken. The sampling type was convenience sampling, as wine buyers were approached at the point of purchase. Convenience sampling has been employed in previous wine segmentation studies, namely, Australian wine market research by Hall (2004) and Bruwer et al. (2002). To ensure the sample was as representative of the population as possible, a large sample size of 300 was chosen and the questionnaire was administered in a variety of outlets to gather information from wine drinkers with wide ranging involvement levels.The fieldwork took place over three weeks in June 2006, in eight wine selling outlets in Galway City and County. This approach is similar to other wine segmentation studies (Bruwer et al., 2000; Hall and Winchester, 1999) where the fieldwork was limited to one region of the market being researched. Research by Bruwer et al. (2000) in segmenting the Australian wine market using a wine-related lifestyle approach is based on fieldwork conducted in Adelaide, while Hall and Winchester's (1999) findings, confirming empirically segments in the Australian wine market, are derived from questionnaires administered in Melbourne. The fieldwork locations for this research consisted of four supermarkets, two off licences and two wine shops. The supermarkets and off licences were in both Galway City and Galway County locations, and the wine shops are situated in Galway City. In total, 316 questionnaires were collected, and after nine were removed for failing a screening question or being incomplete, 307 questionnaires were coded and inputted into SPSS for analysis. There are two sets of findings resulting from the primary research: an overall wine sample analysis and a segment analysis. The overall wine sample is composed of 64 per cent female respondents and 36 per cent male respondents. In terms of age group, over 75 per cent of the sample is aged between 25 and 54 years. The consumer behaviour data reveals the average Irish wine drinker buys seven bottles of wine per month, spending EUR10.57 a bottle, with an average monthly spend of EUR80. Wine is usually bought in a standard size bottle of 75 cl (95 per cent) in either a supermarket (42 per cent) or off licence (35 per cent) and is mostly red wine (43 per cent). Wine is most frequently consumed when dining at home (50 per cent ), followed by when dining out (20 per cent). The five most important product attributes when buying wine are: price per bottle, style of wine (e.g. fruity), region of origin (e.g. Burgundy) and brand name (e.g. Jacob's Creek). The most popular wine is Australian wine, followed by Chilean, French and South African wine.A comparison of the overall wine sample findings with the Wine Development Board's (2004) national statistic shows the characteristics of the research sample are similar to the characteristics of the national market. The sample findings for the country of origin preference are notably similar to the WDB (2004) statistics (see Figure 1). One exception is US wine, which is preferred by just 3 per cent of the sample, but 13 per cent of the national statistics. The similarities between the two sets of statistics suggest the sample data findings on consumer behaviour are representative of the population.Similarly, Figure 2 compares the age category of the sample finding with the age categories of the WDB (2004) population. With the exception of the "65 or older category", the age profile of the two sets of data are similar. Discrepancies between the current research and the WDB (2004) findings may be explained by the time lapse between the two studies, or by the WDB research population being wine consumers, while the current research population was wine buyers.Segment analysis Similar to other wine segmentation studies (Johnson et al., 1993; Bruwer et al., 2002), a k-clustering approach to segmentation was adopted. The k-clustering approach is an exploratory approach, which examines the relationships between more than one variable and produces clusters based on these relationships. The benefit of employing a k-clustering approach is that clusters are formed based on actual relationships within the data set, and not prior assumptions by the researcher.Using the k-means algorithm is a hierarchical method of clustering which reassigns cases to segments "whose centroid is closest to the case" (Punj and Stewart, 1983, p. 139). Table I illustrates the final cluster centres of 17 variables for three clusters. Three of the variables relate to buyer behaviour characteristics (i.e. volumes and average spends). An involvement variable is included in the k-clustering analysis. The involvement variable was developed using Lockshin et al. (2001) product measurement instrument. Fourteen variables in the k-cluster analysis were variables quantifying the importance to the sample of fourteen wine characteristics when choosing wine. The final two variables are the type of wine most frequently purchased, and the age category of the sample. Figure 2 compares the age category results for the WDB (2004) and the total sample of this research. As there is a difference in the two sets of results, age category was included in the k-clustering.The researcher chooses how many clusters are desirable before conducting the k-clustering analysis. As highlighted by Bruwer et al. (2002), there can be difficulty in determining how many clusters are appropriate. A trial and error approach was adopted and the k-clustering was conducted four times to firstly produce two clusters, then three, four and finally five clusters. The four sets of results were assessed to determine which set of results yielded the most sizeable and actionable clusters. Of the four sets of results, the k-clustering which yielded three clusters was found to produce three distinctive and substantial clusters, and these clusters were accepted as the three final segments. To develop three segment profiles, the data for the three clusters is cross tabulated. Three profiles of the average wine buyer in each segment are developed below, followed by a commentary on how the findings are of use to wine marketers.Segment profiles Segment 1: Casual wine buyer (52 per cent of the sample) The average wine buyer in this segment is female, aged from 45 to 54 years, buys on average six bottles of wine per month, spends on average EUR10 a bottle, resulting in an average monthly spend of about EUR60. She is moderately involved in the purchase, with the region of origin, price per bottle, and grape variety being the most important criteria when buying wine. This wine buyer chooses wine they like over more concrete attributes such as alcohol level and price. The style of wine and the brand name are also important. She shows a preference for red wine, and buys in the supermarket usually choosing Australian or French wine. As the Segment 1 wine buyer chooses wine according to more ethereal qualities like style and brand, over more concrete wine attributes, this segment possibly has a casual approach to choosing wine, and is therefore labelled a "Casual Wine Buyer".Marketing to the casual wine buyer The casual wine buyer Segment has the greatest the degree of variability of country of origin preference, with the top two wine countries, Australia and France, being from both Old and New Worlds of wine. The segment is therefore suitable as a target market for wine brands from all countries in both the Old and New World, particularly, Australian, France, Chile, South Africa and Spain. For the casual wine buyer, region of origin is the most important product attribute in choosing wine. Emphasising in the brand communication message, the country from which the wine originates, may be an effective method for wine brands to gain membership in the target market's evoked set of alternative brands.The style of wine preferred by the casual wine buyer is red wine (58 per cent of the segment) with a low preference for white wine (13 per cent of the segment). In promotion activities, such as special price offers when buying more than one bottle, marketers targeting the casual wine buyer Segment, should avoid combining both red and white varieties of the same brand, as the casual wine buyer shows a high preference for red wine. In reaching the casual wine buyer, wine marketers must be conscious, that while wine is usually bought in supermarkets, off licences and wine shops are also popular places of purchase. The variety in country of origin and place of purchase in this segment, gives this segment an image of being casual and informal in their approach to wine buying. A possible point of difference for wine brands targeting this segment could be a casual and informal brand identity, to mirror the possible sentiments of its target market.Segment 2: Value seeking wine buyer (35 per cent of the sample) The average wine buyer in this segment is also female, but slightly younger than the casual wine buyer, being in the age category of 25 to 34 years. Similar to the casual wine buyer, she is employed, with secondary or third level education. She spends on average EUR55 a month on wine, on five bottles. She is the least involved in the wine purchase process of the three segments. For the average buyer in Segment 2, wine is chosen in a purchase situation based on price, style of wine, and brand name, alcohol level and grape variety. Price is the most important aspect of the purchase decision, with alcohol level also of importance, which might suggest she looks for measurable indicators of getting value for her money. This wine drinker buys wine in the supermarket or off licence, drinks the majority of wine at home and buys both red and white wine. This segment shows a preference for New World wine, with 80 per cent of the countries of origin of wine bought by this segment being New World countries. The importance of easily comparable indicators of value (price and alcohol level) in choosing wine gives this segment the title "Value Seeking Wine Buyer".Marketing to the value seeking wine buyer The value seeking segment shows greater preference for wine from New World countries, and proves a suitable target market for wine brands from Chile, South Africa, Australia and the USA. Attracting the value seeking segment requires marketers to emphasise product attributes that are easily measurable. As price is the most important product attribute, attracting this segment with a low price is an obvious positioning strategy. The level of alcohol is also one of the top important attributes in choosing wine. Brands that communicate the value for money qualities of the wine, and resonate with the segment's desire to find affordable wines, may succeed in attracting this segment.The value seeking segment is an attractive segment, despite buying less and spending less than either of the other segments. The profile highlights a segment with consistent preferences and behaviours, which might suggest a greater propensity to being brand loyal, making the segment attractive to wine marketers with a brand image of value for money.Segment 3: Wine traditionalist (13 per cent of the sample) Unlike the other two segments, the average wine buyer in Segment 3 is male, aged between 35 to 44 years, and self-employed with third level education. He is greatly involved in his purchases and buys on average15 bottles of wine per month, spending on average EUR16 a bottle, resulting in an expensive average monthly spend of EUR235. He chooses wine according to traditionally perceived indicators of wine quality such as grape variety, style, region, classification and vintage. Price does not appear in his top five criteria for choosing wine. The Segment 3 wine buyer prefers Old World wines, which account for 63 per cent of the segment's country of origin preferences. The top three wines of preference are from France, Spain and Italy. This wine buyer predominantly buys his wine in wine shops. Despite adopting a connoisseur approach to choosing wine, he is similar to buyers in the other segments, in that wine is consumed most often when dining at home, and is usually bought in standard size bottles. Due to the importance of traditional indicators of wine quality (i.e. classification and vintage), the average wine buyer in this segment may be referred to as a "Wine Traditionalist".Marketing to the wine traditionalist While the wine traditionalist segment accounts for just 13 per cent of the market, in the context of average spend per month, the segment is equally large as the casual wine buyer segment. The wine traditionalist segment is a suitable target market, for Old World wine brands, especially French, Italian and Spanish brands. The wine traditionalist segment may also be suitable as a target market for New World wine brands, especially Australian and Chilean wine, provided the brand message communicates specific indicators of quality, namely, the grape variety, classification, vintage, region of origin and the style of wine.Marketers targeting the wine traditionalists should note the segment may also buy wine by the case, and is most likely to buy both red and white wine. Similar to the casual wine buyer segment, the wine traditionalist shows variety in the type of wine bought, which may suggest, gaining brand loyalty may be difficult. However, the large average spend per month (EUR234.03) deems the wine traditionalist segment an attractive target market for wine brands which command a higher price. Figure 3 illustrates the proportion of the sample in each segment. The research examines how the Irish wine market can be effectively segmented to improve brand positioning. The increases in consumption of wine and increases in preference for wine from New World countries are key trends in the Irish wine market. Consumer behaviour, particularly involvement in wine purchases, and the importance of wine attributes, are necessary considerations in a wine market segmentation study. In terms of relevance, substance, and accessibility, a k-clustering segmentation design, with a behavioural basis including an involvement variable, proves to be an appropriate approach to segmenting the Irish wine market. The profiles of the three resulting segments: casual wine buyer, value seeking wine buyer and wine traditionalist, are sizeable, accessible, relevant and actionable.The profiles developed as a result of the primary research, provides wine marketers with an insight in Irish wine consumer behaviour. Specifically, marketers are provided with accessible and sizeable segments, with meaningfully distinctions and similarities drawn between them. Brand positioning can be improved by ensuring the brand communicates and emphasises the product attributes, which the targeted segments values the most when choosing wine. The demographic information and the buyer behaviour data provide marketers with points of access to their target market. The involvement base, when used in conjunction with other behaviour variables, proves effective in producing sizeable, accessible and actionable segments. A limitation of adopting a behavioural basis in conducting the segmentation is the highly descriptive nature of the resulting data. Examining behaviours give an insight into how consumers act, but fails to take into account the underlying motivations and rationale for consumer actions. The use of more complex segmentation bases, such as value systems and lifestyles would wield a richer understanding of the Irish wine consumer. A second suggestion for future research is an empirically tested wine market behavioural segmentation study, to confirm the findings in this research at a national level.In answering the research question, the Irish wine market can be effectively segmented, with a k-cluster design, with a behavioural basis. Effectively segmenting the Irish wine market requires more than the involvement variable, and calls for other behavioural variables, including the importance of product attributes and country of origin preferences to be included in the segmentation process.Opens in a new window.Figure 1 Wine country of origin preferenceOpens in a new window.Figure 2 Age category bar chartOpens in a new window.Figure 3 Final segmentsOpens in a new window.Table I k-cluster analysis: final cluster centres
|
- The Irish wine market has experienced unprecedented growth in the last 15 years; drivers of growth include increased affordability and accessibility of wine and improved branding. The purpose of this paper is to examine how the Irish wine market may be meaningfully segmented for successful brand positioning.
|
[SECTION: Method] The Irish wine market has experienced unprecedented growth in the last from 15 to 20 years. From 1990 to 2007, total wine sales in Ireland have more than quadrupled, increasing from 1.7 to 7.6 million cases. In the 13 years between 1994 and 2007, wine's proportion of the Irish alcohol market more than doubled from 8 per cent to 17.9 per cent (Wine Development Board, 2007). Growth in wine consumption is forecasted to continue with a growth of 15 per cent expected by 2012 (Euromonitor, 2008). As the wine drinking culture in Ireland is relatively new, the segmentation of the market and brand positioning is in its infancy. Further study into segmentation is required to improve the profitability of the industry, and to develop choice and the accessibility of wine for Irish consumers. The specific purpose of the paper is to examine how the Irish wine market may be effectively segmented for improved brand positioning in Ireland. Thus, the paper aims to determine the key trends in the Irish wine market, examine the state of marketing in the wine industry, evaluate different approaches to segmenting the Irish wine market, and develop profiles of the resulting segments. The Irish wine market has experienced remarkable growth with the number of wine drinkers in Ireland doubling since 1990 and with over five times as much table wine being consumed in 2007, as was consumed in 1990 (WDB, 2007). The increased consumption of wine in Ireland over the last 15 years is attributed to the improved accessibility, affordability and branding of wine (Moran, 2002). To emphasise the significance of the growth in wine consumption in Ireland, the level of growth in wine buying is compared with growth in the overall food and beverage sector. Practically all wine bought in Ireland is imported. Between 2000 and 2004, wine sales (and therefore imports) increased by 56 per cent (WDB, 2004), while the overall imports of the food, beverages and other animal products category increased by only 18 per cent in the same time frame (CSO, 2006).For a marketer assessing the Irish wine market, equally important to the growth in consumption, is the huge shift in the type of wine preferred by the Irish wine drinker. Specifically, there is a notable shift towards New World wines, with diminishing preference for Old World wines (WDB, 2007). New World wines refer to wines from regions outside of Europe. Prominent New World wine producing regions include South Africa, Australia, New Zealand, Chile, Argentina and California. Old World countries refer to European countries with a long history of wine production, such as France, Italy, Germany and Spain (Fielden, 1994). Up to 1990, the majority of wine consumed in Ireland was Old World wine, and accounted for 94 per cent of the market (WDB, 2007). Since 1990, there has been a steady shift in demand towards New World wine, which in 1990 accounted for 6 per cent of the market and in 2007 held a 71 per cent market share (WDB, 2007). Historically, French wines were the market leader in the Irish wine market, but since 2001, Australia now has the largest market share, accounting for 26 per cent of wine consumed in Ireland (WDB, 2007). In 2006, the top ten wine brands in the Irish still light wine market accounted for nearly 25 per cent of total sales. These top ten wine brands and their respective brand shares are Jacob's Creek (3.2 per cent), Blossom Hill (2.7 per cent), Rosemount (2.6 per cent), E&J Gallo (2.5 per cent), Wolf Blass (2.5 per cent), Hardys (2.3 per cent), Concha Y Toro (2.3 per cent), Long Mountain (2.2 per cent), Santa Rita (2.1 per cent) and Carmen (1.9 per cent), all of which are from New World countries (Euromonitor, 2008). Understanding the shift in the country of origin preference is important as it represents an important shift in preference for style, taste, brand, price and other wine variables. Testing country of origin preference is, therefore, an essential element in this Irish wine market study. The production of wine is a specialised area, and the wine industry has traditionally adopted a production-focused mindset with the complexities of viticulture and vinification having occupied the attention of specialists in the area (Thomas and Pickering, 2003). Bruwer et al. (2002) note the agricultural basis of the wine value chain, and the industry is often criticised as employing mass marketing campaigns (Gluckman, 1990; Spawton, 1991a; Hall and Winchester, 1999; Bruwer et al., 2002).According to Thomas and Pickering (2003), the marketing of wine is in its infancy, relative to the long history of wine making and wine drinking. Aggressive marketing was uncommon in the industry with vineyard operators relying on the strength of their reputation to compete in the marketplace (Hall, 2004). In the early 1990s, an interest in branding emerged in the industry as a method of coping with changes in distribution and the growth of wine retailing. In 1991, the European Journal of Marketing dedicated an issue to the marketing of wine, where Spawton (1991b, c, d, e) through a series of articles, provides an insight into the state of wine marketing. This special issue served as an introduction to marketing for the industry, with the purpose of illustrating the advantage and necessity in developing a customer mindset.There is a realisation in the industry that its future is geared to meeting the expectations of the wine consumer. That has contributed to the growing importance of wine marketing within the industry (Spawton, 1991b, p. 6).Gluckman (1990), in a frequently cited article, presents a number of challenges facing the wine marketer in branding wines. Most notably, a move towards own brand labelling by retailers reinforces the necessity for strong wine brands. This need for improved branding of wine has prompted the undertaking of research in wine markets. In an industry which is relatively new to marketing, and in the Irish wine market, which has seen tremendous growth and transformation, there is a need for greater understanding of the market dynamics. Market research in general, and market segmentation in particular, has a potentially pivotal role to play in assisting wine marketers to position their wine brands effectively. Weir (1960, p. 95, as cited in Yankelovich, 1964) provides the following description of what a market is, and more importantly, what it is not:"The market" is not a single, cohesive unit; it is a seething, disparate, pullulating, antagonistic, infinitely varied sea of differing human beings - every one of them as distinct from every other one as fingerprints; everyone of them living in circumstances different in countless ways from those in which every other one of them is living.This description of a market is a colourful representation of popular marketing thought on the composition of markets in the late 1950s and 1960s. Smith (1956), in what is considered a landmark article (Reynolds, 1965; Haley, 1968; Wind, 1978; Green and Krieger, 1991; Lin, 2002) introduces a marketing strategy labelled market segmentation, as an approach to competing successfully in the reality of an environment of imperfect competition. The original article by Smith (1956) introduces market segmentation as a strategy. Market segmentation strategy was considered an alternative to product differentiation strategy to deal with diversity in the market. While the initial representation of the market segmentation strategy is based in economic theory, market segmentation developed as one of the most foremost concepts in marketing thought (Wind, 1978; Johnson et al., 1991; Lin, 2002).At a broad level, market segmentation provides a marketer with a clearer focus on customer needs, and thereby aids decision making for improved competitive advantage (McDonald and Dunbar, 1992; Croft, 1994; Kotler and Keller, 2003). While Croft (1994) highlights market segmentation as aiding decision making in general, Yankelovich (1964) specifies what exactly segmentation analysis can achieve. In identifying groups of customers with similar needs, a marketer has the information required to target the most profitable group with the most potential. With this knowledge a marketer can develop product lines and promotion activities, choose advertising media, advance positioning of offerings and improve timing of advertising to appeal to the segment of the market whose needs possess the greatest profit potential.A critical decision to be made in conducting segmentation research is choosing an appropriate segmentation base. A segmentation base is the criteria used to divide the defined market into groups of consumers with similarities. At the most basic level a market can be split up according to the profiles of the consumers. Variables such as demographics, geographic location of consumers and the socio economic class to which they belong, are considered profile segmentation bases. The behavioural segmentation category includes bases such as usage occasion, benefits sought, perceptions and beliefs, while the psychographic bases category includes lifestyle and personality variables as a means for identifying groups of consumers with similarities. The more abstract and less concrete the information required for the segmentation base, the more difficult it is to measure responses and their link with behaviour. In choosing a segmentation base for a wine market study there are a number aspects to the market which need to be considered. Literature on wine consumer behaviour focuses on two areas; the factors influencing wine consumer behaviour; and the wine consumer's purchasing decision-making process. An effective wine segmentation study would be one which aids understanding of these two areas and aids marketers in evaluating how stimuli, such as brand positioning strategies, influence wine choice.According to Bruwer et al. (2002), wine markets have been segmented using all the bases identified above. For the purpose of an Irish wine market segmentation study, behavioural segmentation with an involvement basis proves a suitable choice as it is an approach, which yields insight into consumer behaviour, but is not overly difficult to measure (Lockshin et al., 2001). Employing a behavioural segmentation base allows for the decision making process and the influencing factors of the Irish wine consumer to be tested and makes the process and the factors the basis for splitting up the market into meaningful and actionable segments.A key consideration in exploring the wine consumer decision-making process and consumers' evaluation of alternatives, is that wine attributes represent intrinsic and extrinsic cues for the consumer. Sanchez and Gill (1998) illustrate how consumers have preferences according to the bundle of benefits they are seeking. The challenge in understanding these preferences is the large number of wine attributes which exist, and therefore, the greater number of possible bundles of benefits that are present. Wine attributes include: brand name, producer, grape variety, blend of grape varieties, vintage, region of origin, price, label, bottle type, cork type, bottle size, colour of wine, style of wine and level of alcohol. Due to the large number of wine attributes, wine consumers have a wider range of considerations in making purchasing decisions. Examining the hierarchy of importance of wine attributes to Irish wine buyers is a central consideration in segmenting the market. The research design is primarily descriptive in nature as similar investigations into other wine markets have been descriptive in design (Orth et al., 2005; Johnson et al., 1991; Hall, 2004; Bruwer et al. Li and Reid, 2002; Johnson, 2003; Thomas and Pickering, 2003). Due to the descriptive nature of this research, a quantitative approach to primary data collection is most suitable. Quantitative data is appropriate for determining and understanding the behaviours and characteristics of a large sample of wine drinkers. Specifically, survey data collection was undertaken, with a questionnaire collection instrument administered through a personal interview. The questionnaire was two pages in length with 15 tick box questions. The questionnaire posed questions to gather data on four topics: volume of usage, buying preference, product involvement, and demographic information.As an accurate sampling frame was unavailable for the population of the 1,451,000 wine drinkers in Ireland (WDB, 2004), non-probability sampling was undertaken. The sampling type was convenience sampling, as wine buyers were approached at the point of purchase. Convenience sampling has been employed in previous wine segmentation studies, namely, Australian wine market research by Hall (2004) and Bruwer et al. (2002). To ensure the sample was as representative of the population as possible, a large sample size of 300 was chosen and the questionnaire was administered in a variety of outlets to gather information from wine drinkers with wide ranging involvement levels.The fieldwork took place over three weeks in June 2006, in eight wine selling outlets in Galway City and County. This approach is similar to other wine segmentation studies (Bruwer et al., 2000; Hall and Winchester, 1999) where the fieldwork was limited to one region of the market being researched. Research by Bruwer et al. (2000) in segmenting the Australian wine market using a wine-related lifestyle approach is based on fieldwork conducted in Adelaide, while Hall and Winchester's (1999) findings, confirming empirically segments in the Australian wine market, are derived from questionnaires administered in Melbourne. The fieldwork locations for this research consisted of four supermarkets, two off licences and two wine shops. The supermarkets and off licences were in both Galway City and Galway County locations, and the wine shops are situated in Galway City. In total, 316 questionnaires were collected, and after nine were removed for failing a screening question or being incomplete, 307 questionnaires were coded and inputted into SPSS for analysis. There are two sets of findings resulting from the primary research: an overall wine sample analysis and a segment analysis. The overall wine sample is composed of 64 per cent female respondents and 36 per cent male respondents. In terms of age group, over 75 per cent of the sample is aged between 25 and 54 years. The consumer behaviour data reveals the average Irish wine drinker buys seven bottles of wine per month, spending EUR10.57 a bottle, with an average monthly spend of EUR80. Wine is usually bought in a standard size bottle of 75 cl (95 per cent) in either a supermarket (42 per cent) or off licence (35 per cent) and is mostly red wine (43 per cent). Wine is most frequently consumed when dining at home (50 per cent ), followed by when dining out (20 per cent). The five most important product attributes when buying wine are: price per bottle, style of wine (e.g. fruity), region of origin (e.g. Burgundy) and brand name (e.g. Jacob's Creek). The most popular wine is Australian wine, followed by Chilean, French and South African wine.A comparison of the overall wine sample findings with the Wine Development Board's (2004) national statistic shows the characteristics of the research sample are similar to the characteristics of the national market. The sample findings for the country of origin preference are notably similar to the WDB (2004) statistics (see Figure 1). One exception is US wine, which is preferred by just 3 per cent of the sample, but 13 per cent of the national statistics. The similarities between the two sets of statistics suggest the sample data findings on consumer behaviour are representative of the population.Similarly, Figure 2 compares the age category of the sample finding with the age categories of the WDB (2004) population. With the exception of the "65 or older category", the age profile of the two sets of data are similar. Discrepancies between the current research and the WDB (2004) findings may be explained by the time lapse between the two studies, or by the WDB research population being wine consumers, while the current research population was wine buyers.Segment analysis Similar to other wine segmentation studies (Johnson et al., 1993; Bruwer et al., 2002), a k-clustering approach to segmentation was adopted. The k-clustering approach is an exploratory approach, which examines the relationships between more than one variable and produces clusters based on these relationships. The benefit of employing a k-clustering approach is that clusters are formed based on actual relationships within the data set, and not prior assumptions by the researcher.Using the k-means algorithm is a hierarchical method of clustering which reassigns cases to segments "whose centroid is closest to the case" (Punj and Stewart, 1983, p. 139). Table I illustrates the final cluster centres of 17 variables for three clusters. Three of the variables relate to buyer behaviour characteristics (i.e. volumes and average spends). An involvement variable is included in the k-clustering analysis. The involvement variable was developed using Lockshin et al. (2001) product measurement instrument. Fourteen variables in the k-cluster analysis were variables quantifying the importance to the sample of fourteen wine characteristics when choosing wine. The final two variables are the type of wine most frequently purchased, and the age category of the sample. Figure 2 compares the age category results for the WDB (2004) and the total sample of this research. As there is a difference in the two sets of results, age category was included in the k-clustering.The researcher chooses how many clusters are desirable before conducting the k-clustering analysis. As highlighted by Bruwer et al. (2002), there can be difficulty in determining how many clusters are appropriate. A trial and error approach was adopted and the k-clustering was conducted four times to firstly produce two clusters, then three, four and finally five clusters. The four sets of results were assessed to determine which set of results yielded the most sizeable and actionable clusters. Of the four sets of results, the k-clustering which yielded three clusters was found to produce three distinctive and substantial clusters, and these clusters were accepted as the three final segments. To develop three segment profiles, the data for the three clusters is cross tabulated. Three profiles of the average wine buyer in each segment are developed below, followed by a commentary on how the findings are of use to wine marketers.Segment profiles Segment 1: Casual wine buyer (52 per cent of the sample) The average wine buyer in this segment is female, aged from 45 to 54 years, buys on average six bottles of wine per month, spends on average EUR10 a bottle, resulting in an average monthly spend of about EUR60. She is moderately involved in the purchase, with the region of origin, price per bottle, and grape variety being the most important criteria when buying wine. This wine buyer chooses wine they like over more concrete attributes such as alcohol level and price. The style of wine and the brand name are also important. She shows a preference for red wine, and buys in the supermarket usually choosing Australian or French wine. As the Segment 1 wine buyer chooses wine according to more ethereal qualities like style and brand, over more concrete wine attributes, this segment possibly has a casual approach to choosing wine, and is therefore labelled a "Casual Wine Buyer".Marketing to the casual wine buyer The casual wine buyer Segment has the greatest the degree of variability of country of origin preference, with the top two wine countries, Australia and France, being from both Old and New Worlds of wine. The segment is therefore suitable as a target market for wine brands from all countries in both the Old and New World, particularly, Australian, France, Chile, South Africa and Spain. For the casual wine buyer, region of origin is the most important product attribute in choosing wine. Emphasising in the brand communication message, the country from which the wine originates, may be an effective method for wine brands to gain membership in the target market's evoked set of alternative brands.The style of wine preferred by the casual wine buyer is red wine (58 per cent of the segment) with a low preference for white wine (13 per cent of the segment). In promotion activities, such as special price offers when buying more than one bottle, marketers targeting the casual wine buyer Segment, should avoid combining both red and white varieties of the same brand, as the casual wine buyer shows a high preference for red wine. In reaching the casual wine buyer, wine marketers must be conscious, that while wine is usually bought in supermarkets, off licences and wine shops are also popular places of purchase. The variety in country of origin and place of purchase in this segment, gives this segment an image of being casual and informal in their approach to wine buying. A possible point of difference for wine brands targeting this segment could be a casual and informal brand identity, to mirror the possible sentiments of its target market.Segment 2: Value seeking wine buyer (35 per cent of the sample) The average wine buyer in this segment is also female, but slightly younger than the casual wine buyer, being in the age category of 25 to 34 years. Similar to the casual wine buyer, she is employed, with secondary or third level education. She spends on average EUR55 a month on wine, on five bottles. She is the least involved in the wine purchase process of the three segments. For the average buyer in Segment 2, wine is chosen in a purchase situation based on price, style of wine, and brand name, alcohol level and grape variety. Price is the most important aspect of the purchase decision, with alcohol level also of importance, which might suggest she looks for measurable indicators of getting value for her money. This wine drinker buys wine in the supermarket or off licence, drinks the majority of wine at home and buys both red and white wine. This segment shows a preference for New World wine, with 80 per cent of the countries of origin of wine bought by this segment being New World countries. The importance of easily comparable indicators of value (price and alcohol level) in choosing wine gives this segment the title "Value Seeking Wine Buyer".Marketing to the value seeking wine buyer The value seeking segment shows greater preference for wine from New World countries, and proves a suitable target market for wine brands from Chile, South Africa, Australia and the USA. Attracting the value seeking segment requires marketers to emphasise product attributes that are easily measurable. As price is the most important product attribute, attracting this segment with a low price is an obvious positioning strategy. The level of alcohol is also one of the top important attributes in choosing wine. Brands that communicate the value for money qualities of the wine, and resonate with the segment's desire to find affordable wines, may succeed in attracting this segment.The value seeking segment is an attractive segment, despite buying less and spending less than either of the other segments. The profile highlights a segment with consistent preferences and behaviours, which might suggest a greater propensity to being brand loyal, making the segment attractive to wine marketers with a brand image of value for money.Segment 3: Wine traditionalist (13 per cent of the sample) Unlike the other two segments, the average wine buyer in Segment 3 is male, aged between 35 to 44 years, and self-employed with third level education. He is greatly involved in his purchases and buys on average15 bottles of wine per month, spending on average EUR16 a bottle, resulting in an expensive average monthly spend of EUR235. He chooses wine according to traditionally perceived indicators of wine quality such as grape variety, style, region, classification and vintage. Price does not appear in his top five criteria for choosing wine. The Segment 3 wine buyer prefers Old World wines, which account for 63 per cent of the segment's country of origin preferences. The top three wines of preference are from France, Spain and Italy. This wine buyer predominantly buys his wine in wine shops. Despite adopting a connoisseur approach to choosing wine, he is similar to buyers in the other segments, in that wine is consumed most often when dining at home, and is usually bought in standard size bottles. Due to the importance of traditional indicators of wine quality (i.e. classification and vintage), the average wine buyer in this segment may be referred to as a "Wine Traditionalist".Marketing to the wine traditionalist While the wine traditionalist segment accounts for just 13 per cent of the market, in the context of average spend per month, the segment is equally large as the casual wine buyer segment. The wine traditionalist segment is a suitable target market, for Old World wine brands, especially French, Italian and Spanish brands. The wine traditionalist segment may also be suitable as a target market for New World wine brands, especially Australian and Chilean wine, provided the brand message communicates specific indicators of quality, namely, the grape variety, classification, vintage, region of origin and the style of wine.Marketers targeting the wine traditionalists should note the segment may also buy wine by the case, and is most likely to buy both red and white wine. Similar to the casual wine buyer segment, the wine traditionalist shows variety in the type of wine bought, which may suggest, gaining brand loyalty may be difficult. However, the large average spend per month (EUR234.03) deems the wine traditionalist segment an attractive target market for wine brands which command a higher price. Figure 3 illustrates the proportion of the sample in each segment. The research examines how the Irish wine market can be effectively segmented to improve brand positioning. The increases in consumption of wine and increases in preference for wine from New World countries are key trends in the Irish wine market. Consumer behaviour, particularly involvement in wine purchases, and the importance of wine attributes, are necessary considerations in a wine market segmentation study. In terms of relevance, substance, and accessibility, a k-clustering segmentation design, with a behavioural basis including an involvement variable, proves to be an appropriate approach to segmenting the Irish wine market. The profiles of the three resulting segments: casual wine buyer, value seeking wine buyer and wine traditionalist, are sizeable, accessible, relevant and actionable.The profiles developed as a result of the primary research, provides wine marketers with an insight in Irish wine consumer behaviour. Specifically, marketers are provided with accessible and sizeable segments, with meaningfully distinctions and similarities drawn between them. Brand positioning can be improved by ensuring the brand communicates and emphasises the product attributes, which the targeted segments values the most when choosing wine. The demographic information and the buyer behaviour data provide marketers with points of access to their target market. The involvement base, when used in conjunction with other behaviour variables, proves effective in producing sizeable, accessible and actionable segments. A limitation of adopting a behavioural basis in conducting the segmentation is the highly descriptive nature of the resulting data. Examining behaviours give an insight into how consumers act, but fails to take into account the underlying motivations and rationale for consumer actions. The use of more complex segmentation bases, such as value systems and lifestyles would wield a richer understanding of the Irish wine consumer. A second suggestion for future research is an empirically tested wine market behavioural segmentation study, to confirm the findings in this research at a national level.In answering the research question, the Irish wine market can be effectively segmented, with a k-cluster design, with a behavioural basis. Effectively segmenting the Irish wine market requires more than the involvement variable, and calls for other behavioural variables, including the importance of product attributes and country of origin preferences to be included in the segmentation process.Opens in a new window.Figure 1 Wine country of origin preferenceOpens in a new window.Figure 2 Age category bar chartOpens in a new window.Figure 3 Final segmentsOpens in a new window.Table I k-cluster analysis: final cluster centres
|
- This research adopts a lifestyle segmentation approach by linking lifestyle values, product attributes and buying and consumption patterns. The primary research is descriptive in design, employing a self-administered questionnaire to collect quantitative data on wine consumer behaviour. Efforts made to ensure a highly representative sample included choosing a large sample size, administering the questionnaire in a range of outlets, and gathering information from wine drinkers with wide ranging involvement levels.
|
[SECTION: Findings] The Irish wine market has experienced unprecedented growth in the last from 15 to 20 years. From 1990 to 2007, total wine sales in Ireland have more than quadrupled, increasing from 1.7 to 7.6 million cases. In the 13 years between 1994 and 2007, wine's proportion of the Irish alcohol market more than doubled from 8 per cent to 17.9 per cent (Wine Development Board, 2007). Growth in wine consumption is forecasted to continue with a growth of 15 per cent expected by 2012 (Euromonitor, 2008). As the wine drinking culture in Ireland is relatively new, the segmentation of the market and brand positioning is in its infancy. Further study into segmentation is required to improve the profitability of the industry, and to develop choice and the accessibility of wine for Irish consumers. The specific purpose of the paper is to examine how the Irish wine market may be effectively segmented for improved brand positioning in Ireland. Thus, the paper aims to determine the key trends in the Irish wine market, examine the state of marketing in the wine industry, evaluate different approaches to segmenting the Irish wine market, and develop profiles of the resulting segments. The Irish wine market has experienced remarkable growth with the number of wine drinkers in Ireland doubling since 1990 and with over five times as much table wine being consumed in 2007, as was consumed in 1990 (WDB, 2007). The increased consumption of wine in Ireland over the last 15 years is attributed to the improved accessibility, affordability and branding of wine (Moran, 2002). To emphasise the significance of the growth in wine consumption in Ireland, the level of growth in wine buying is compared with growth in the overall food and beverage sector. Practically all wine bought in Ireland is imported. Between 2000 and 2004, wine sales (and therefore imports) increased by 56 per cent (WDB, 2004), while the overall imports of the food, beverages and other animal products category increased by only 18 per cent in the same time frame (CSO, 2006).For a marketer assessing the Irish wine market, equally important to the growth in consumption, is the huge shift in the type of wine preferred by the Irish wine drinker. Specifically, there is a notable shift towards New World wines, with diminishing preference for Old World wines (WDB, 2007). New World wines refer to wines from regions outside of Europe. Prominent New World wine producing regions include South Africa, Australia, New Zealand, Chile, Argentina and California. Old World countries refer to European countries with a long history of wine production, such as France, Italy, Germany and Spain (Fielden, 1994). Up to 1990, the majority of wine consumed in Ireland was Old World wine, and accounted for 94 per cent of the market (WDB, 2007). Since 1990, there has been a steady shift in demand towards New World wine, which in 1990 accounted for 6 per cent of the market and in 2007 held a 71 per cent market share (WDB, 2007). Historically, French wines were the market leader in the Irish wine market, but since 2001, Australia now has the largest market share, accounting for 26 per cent of wine consumed in Ireland (WDB, 2007). In 2006, the top ten wine brands in the Irish still light wine market accounted for nearly 25 per cent of total sales. These top ten wine brands and their respective brand shares are Jacob's Creek (3.2 per cent), Blossom Hill (2.7 per cent), Rosemount (2.6 per cent), E&J Gallo (2.5 per cent), Wolf Blass (2.5 per cent), Hardys (2.3 per cent), Concha Y Toro (2.3 per cent), Long Mountain (2.2 per cent), Santa Rita (2.1 per cent) and Carmen (1.9 per cent), all of which are from New World countries (Euromonitor, 2008). Understanding the shift in the country of origin preference is important as it represents an important shift in preference for style, taste, brand, price and other wine variables. Testing country of origin preference is, therefore, an essential element in this Irish wine market study. The production of wine is a specialised area, and the wine industry has traditionally adopted a production-focused mindset with the complexities of viticulture and vinification having occupied the attention of specialists in the area (Thomas and Pickering, 2003). Bruwer et al. (2002) note the agricultural basis of the wine value chain, and the industry is often criticised as employing mass marketing campaigns (Gluckman, 1990; Spawton, 1991a; Hall and Winchester, 1999; Bruwer et al., 2002).According to Thomas and Pickering (2003), the marketing of wine is in its infancy, relative to the long history of wine making and wine drinking. Aggressive marketing was uncommon in the industry with vineyard operators relying on the strength of their reputation to compete in the marketplace (Hall, 2004). In the early 1990s, an interest in branding emerged in the industry as a method of coping with changes in distribution and the growth of wine retailing. In 1991, the European Journal of Marketing dedicated an issue to the marketing of wine, where Spawton (1991b, c, d, e) through a series of articles, provides an insight into the state of wine marketing. This special issue served as an introduction to marketing for the industry, with the purpose of illustrating the advantage and necessity in developing a customer mindset.There is a realisation in the industry that its future is geared to meeting the expectations of the wine consumer. That has contributed to the growing importance of wine marketing within the industry (Spawton, 1991b, p. 6).Gluckman (1990), in a frequently cited article, presents a number of challenges facing the wine marketer in branding wines. Most notably, a move towards own brand labelling by retailers reinforces the necessity for strong wine brands. This need for improved branding of wine has prompted the undertaking of research in wine markets. In an industry which is relatively new to marketing, and in the Irish wine market, which has seen tremendous growth and transformation, there is a need for greater understanding of the market dynamics. Market research in general, and market segmentation in particular, has a potentially pivotal role to play in assisting wine marketers to position their wine brands effectively. Weir (1960, p. 95, as cited in Yankelovich, 1964) provides the following description of what a market is, and more importantly, what it is not:"The market" is not a single, cohesive unit; it is a seething, disparate, pullulating, antagonistic, infinitely varied sea of differing human beings - every one of them as distinct from every other one as fingerprints; everyone of them living in circumstances different in countless ways from those in which every other one of them is living.This description of a market is a colourful representation of popular marketing thought on the composition of markets in the late 1950s and 1960s. Smith (1956), in what is considered a landmark article (Reynolds, 1965; Haley, 1968; Wind, 1978; Green and Krieger, 1991; Lin, 2002) introduces a marketing strategy labelled market segmentation, as an approach to competing successfully in the reality of an environment of imperfect competition. The original article by Smith (1956) introduces market segmentation as a strategy. Market segmentation strategy was considered an alternative to product differentiation strategy to deal with diversity in the market. While the initial representation of the market segmentation strategy is based in economic theory, market segmentation developed as one of the most foremost concepts in marketing thought (Wind, 1978; Johnson et al., 1991; Lin, 2002).At a broad level, market segmentation provides a marketer with a clearer focus on customer needs, and thereby aids decision making for improved competitive advantage (McDonald and Dunbar, 1992; Croft, 1994; Kotler and Keller, 2003). While Croft (1994) highlights market segmentation as aiding decision making in general, Yankelovich (1964) specifies what exactly segmentation analysis can achieve. In identifying groups of customers with similar needs, a marketer has the information required to target the most profitable group with the most potential. With this knowledge a marketer can develop product lines and promotion activities, choose advertising media, advance positioning of offerings and improve timing of advertising to appeal to the segment of the market whose needs possess the greatest profit potential.A critical decision to be made in conducting segmentation research is choosing an appropriate segmentation base. A segmentation base is the criteria used to divide the defined market into groups of consumers with similarities. At the most basic level a market can be split up according to the profiles of the consumers. Variables such as demographics, geographic location of consumers and the socio economic class to which they belong, are considered profile segmentation bases. The behavioural segmentation category includes bases such as usage occasion, benefits sought, perceptions and beliefs, while the psychographic bases category includes lifestyle and personality variables as a means for identifying groups of consumers with similarities. The more abstract and less concrete the information required for the segmentation base, the more difficult it is to measure responses and their link with behaviour. In choosing a segmentation base for a wine market study there are a number aspects to the market which need to be considered. Literature on wine consumer behaviour focuses on two areas; the factors influencing wine consumer behaviour; and the wine consumer's purchasing decision-making process. An effective wine segmentation study would be one which aids understanding of these two areas and aids marketers in evaluating how stimuli, such as brand positioning strategies, influence wine choice.According to Bruwer et al. (2002), wine markets have been segmented using all the bases identified above. For the purpose of an Irish wine market segmentation study, behavioural segmentation with an involvement basis proves a suitable choice as it is an approach, which yields insight into consumer behaviour, but is not overly difficult to measure (Lockshin et al., 2001). Employing a behavioural segmentation base allows for the decision making process and the influencing factors of the Irish wine consumer to be tested and makes the process and the factors the basis for splitting up the market into meaningful and actionable segments.A key consideration in exploring the wine consumer decision-making process and consumers' evaluation of alternatives, is that wine attributes represent intrinsic and extrinsic cues for the consumer. Sanchez and Gill (1998) illustrate how consumers have preferences according to the bundle of benefits they are seeking. The challenge in understanding these preferences is the large number of wine attributes which exist, and therefore, the greater number of possible bundles of benefits that are present. Wine attributes include: brand name, producer, grape variety, blend of grape varieties, vintage, region of origin, price, label, bottle type, cork type, bottle size, colour of wine, style of wine and level of alcohol. Due to the large number of wine attributes, wine consumers have a wider range of considerations in making purchasing decisions. Examining the hierarchy of importance of wine attributes to Irish wine buyers is a central consideration in segmenting the market. The research design is primarily descriptive in nature as similar investigations into other wine markets have been descriptive in design (Orth et al., 2005; Johnson et al., 1991; Hall, 2004; Bruwer et al. Li and Reid, 2002; Johnson, 2003; Thomas and Pickering, 2003). Due to the descriptive nature of this research, a quantitative approach to primary data collection is most suitable. Quantitative data is appropriate for determining and understanding the behaviours and characteristics of a large sample of wine drinkers. Specifically, survey data collection was undertaken, with a questionnaire collection instrument administered through a personal interview. The questionnaire was two pages in length with 15 tick box questions. The questionnaire posed questions to gather data on four topics: volume of usage, buying preference, product involvement, and demographic information.As an accurate sampling frame was unavailable for the population of the 1,451,000 wine drinkers in Ireland (WDB, 2004), non-probability sampling was undertaken. The sampling type was convenience sampling, as wine buyers were approached at the point of purchase. Convenience sampling has been employed in previous wine segmentation studies, namely, Australian wine market research by Hall (2004) and Bruwer et al. (2002). To ensure the sample was as representative of the population as possible, a large sample size of 300 was chosen and the questionnaire was administered in a variety of outlets to gather information from wine drinkers with wide ranging involvement levels.The fieldwork took place over three weeks in June 2006, in eight wine selling outlets in Galway City and County. This approach is similar to other wine segmentation studies (Bruwer et al., 2000; Hall and Winchester, 1999) where the fieldwork was limited to one region of the market being researched. Research by Bruwer et al. (2000) in segmenting the Australian wine market using a wine-related lifestyle approach is based on fieldwork conducted in Adelaide, while Hall and Winchester's (1999) findings, confirming empirically segments in the Australian wine market, are derived from questionnaires administered in Melbourne. The fieldwork locations for this research consisted of four supermarkets, two off licences and two wine shops. The supermarkets and off licences were in both Galway City and Galway County locations, and the wine shops are situated in Galway City. In total, 316 questionnaires were collected, and after nine were removed for failing a screening question or being incomplete, 307 questionnaires were coded and inputted into SPSS for analysis. There are two sets of findings resulting from the primary research: an overall wine sample analysis and a segment analysis. The overall wine sample is composed of 64 per cent female respondents and 36 per cent male respondents. In terms of age group, over 75 per cent of the sample is aged between 25 and 54 years. The consumer behaviour data reveals the average Irish wine drinker buys seven bottles of wine per month, spending EUR10.57 a bottle, with an average monthly spend of EUR80. Wine is usually bought in a standard size bottle of 75 cl (95 per cent) in either a supermarket (42 per cent) or off licence (35 per cent) and is mostly red wine (43 per cent). Wine is most frequently consumed when dining at home (50 per cent ), followed by when dining out (20 per cent). The five most important product attributes when buying wine are: price per bottle, style of wine (e.g. fruity), region of origin (e.g. Burgundy) and brand name (e.g. Jacob's Creek). The most popular wine is Australian wine, followed by Chilean, French and South African wine.A comparison of the overall wine sample findings with the Wine Development Board's (2004) national statistic shows the characteristics of the research sample are similar to the characteristics of the national market. The sample findings for the country of origin preference are notably similar to the WDB (2004) statistics (see Figure 1). One exception is US wine, which is preferred by just 3 per cent of the sample, but 13 per cent of the national statistics. The similarities between the two sets of statistics suggest the sample data findings on consumer behaviour are representative of the population.Similarly, Figure 2 compares the age category of the sample finding with the age categories of the WDB (2004) population. With the exception of the "65 or older category", the age profile of the two sets of data are similar. Discrepancies between the current research and the WDB (2004) findings may be explained by the time lapse between the two studies, or by the WDB research population being wine consumers, while the current research population was wine buyers.Segment analysis Similar to other wine segmentation studies (Johnson et al., 1993; Bruwer et al., 2002), a k-clustering approach to segmentation was adopted. The k-clustering approach is an exploratory approach, which examines the relationships between more than one variable and produces clusters based on these relationships. The benefit of employing a k-clustering approach is that clusters are formed based on actual relationships within the data set, and not prior assumptions by the researcher.Using the k-means algorithm is a hierarchical method of clustering which reassigns cases to segments "whose centroid is closest to the case" (Punj and Stewart, 1983, p. 139). Table I illustrates the final cluster centres of 17 variables for three clusters. Three of the variables relate to buyer behaviour characteristics (i.e. volumes and average spends). An involvement variable is included in the k-clustering analysis. The involvement variable was developed using Lockshin et al. (2001) product measurement instrument. Fourteen variables in the k-cluster analysis were variables quantifying the importance to the sample of fourteen wine characteristics when choosing wine. The final two variables are the type of wine most frequently purchased, and the age category of the sample. Figure 2 compares the age category results for the WDB (2004) and the total sample of this research. As there is a difference in the two sets of results, age category was included in the k-clustering.The researcher chooses how many clusters are desirable before conducting the k-clustering analysis. As highlighted by Bruwer et al. (2002), there can be difficulty in determining how many clusters are appropriate. A trial and error approach was adopted and the k-clustering was conducted four times to firstly produce two clusters, then three, four and finally five clusters. The four sets of results were assessed to determine which set of results yielded the most sizeable and actionable clusters. Of the four sets of results, the k-clustering which yielded three clusters was found to produce three distinctive and substantial clusters, and these clusters were accepted as the three final segments. To develop three segment profiles, the data for the three clusters is cross tabulated. Three profiles of the average wine buyer in each segment are developed below, followed by a commentary on how the findings are of use to wine marketers.Segment profiles Segment 1: Casual wine buyer (52 per cent of the sample) The average wine buyer in this segment is female, aged from 45 to 54 years, buys on average six bottles of wine per month, spends on average EUR10 a bottle, resulting in an average monthly spend of about EUR60. She is moderately involved in the purchase, with the region of origin, price per bottle, and grape variety being the most important criteria when buying wine. This wine buyer chooses wine they like over more concrete attributes such as alcohol level and price. The style of wine and the brand name are also important. She shows a preference for red wine, and buys in the supermarket usually choosing Australian or French wine. As the Segment 1 wine buyer chooses wine according to more ethereal qualities like style and brand, over more concrete wine attributes, this segment possibly has a casual approach to choosing wine, and is therefore labelled a "Casual Wine Buyer".Marketing to the casual wine buyer The casual wine buyer Segment has the greatest the degree of variability of country of origin preference, with the top two wine countries, Australia and France, being from both Old and New Worlds of wine. The segment is therefore suitable as a target market for wine brands from all countries in both the Old and New World, particularly, Australian, France, Chile, South Africa and Spain. For the casual wine buyer, region of origin is the most important product attribute in choosing wine. Emphasising in the brand communication message, the country from which the wine originates, may be an effective method for wine brands to gain membership in the target market's evoked set of alternative brands.The style of wine preferred by the casual wine buyer is red wine (58 per cent of the segment) with a low preference for white wine (13 per cent of the segment). In promotion activities, such as special price offers when buying more than one bottle, marketers targeting the casual wine buyer Segment, should avoid combining both red and white varieties of the same brand, as the casual wine buyer shows a high preference for red wine. In reaching the casual wine buyer, wine marketers must be conscious, that while wine is usually bought in supermarkets, off licences and wine shops are also popular places of purchase. The variety in country of origin and place of purchase in this segment, gives this segment an image of being casual and informal in their approach to wine buying. A possible point of difference for wine brands targeting this segment could be a casual and informal brand identity, to mirror the possible sentiments of its target market.Segment 2: Value seeking wine buyer (35 per cent of the sample) The average wine buyer in this segment is also female, but slightly younger than the casual wine buyer, being in the age category of 25 to 34 years. Similar to the casual wine buyer, she is employed, with secondary or third level education. She spends on average EUR55 a month on wine, on five bottles. She is the least involved in the wine purchase process of the three segments. For the average buyer in Segment 2, wine is chosen in a purchase situation based on price, style of wine, and brand name, alcohol level and grape variety. Price is the most important aspect of the purchase decision, with alcohol level also of importance, which might suggest she looks for measurable indicators of getting value for her money. This wine drinker buys wine in the supermarket or off licence, drinks the majority of wine at home and buys both red and white wine. This segment shows a preference for New World wine, with 80 per cent of the countries of origin of wine bought by this segment being New World countries. The importance of easily comparable indicators of value (price and alcohol level) in choosing wine gives this segment the title "Value Seeking Wine Buyer".Marketing to the value seeking wine buyer The value seeking segment shows greater preference for wine from New World countries, and proves a suitable target market for wine brands from Chile, South Africa, Australia and the USA. Attracting the value seeking segment requires marketers to emphasise product attributes that are easily measurable. As price is the most important product attribute, attracting this segment with a low price is an obvious positioning strategy. The level of alcohol is also one of the top important attributes in choosing wine. Brands that communicate the value for money qualities of the wine, and resonate with the segment's desire to find affordable wines, may succeed in attracting this segment.The value seeking segment is an attractive segment, despite buying less and spending less than either of the other segments. The profile highlights a segment with consistent preferences and behaviours, which might suggest a greater propensity to being brand loyal, making the segment attractive to wine marketers with a brand image of value for money.Segment 3: Wine traditionalist (13 per cent of the sample) Unlike the other two segments, the average wine buyer in Segment 3 is male, aged between 35 to 44 years, and self-employed with third level education. He is greatly involved in his purchases and buys on average15 bottles of wine per month, spending on average EUR16 a bottle, resulting in an expensive average monthly spend of EUR235. He chooses wine according to traditionally perceived indicators of wine quality such as grape variety, style, region, classification and vintage. Price does not appear in his top five criteria for choosing wine. The Segment 3 wine buyer prefers Old World wines, which account for 63 per cent of the segment's country of origin preferences. The top three wines of preference are from France, Spain and Italy. This wine buyer predominantly buys his wine in wine shops. Despite adopting a connoisseur approach to choosing wine, he is similar to buyers in the other segments, in that wine is consumed most often when dining at home, and is usually bought in standard size bottles. Due to the importance of traditional indicators of wine quality (i.e. classification and vintage), the average wine buyer in this segment may be referred to as a "Wine Traditionalist".Marketing to the wine traditionalist While the wine traditionalist segment accounts for just 13 per cent of the market, in the context of average spend per month, the segment is equally large as the casual wine buyer segment. The wine traditionalist segment is a suitable target market, for Old World wine brands, especially French, Italian and Spanish brands. The wine traditionalist segment may also be suitable as a target market for New World wine brands, especially Australian and Chilean wine, provided the brand message communicates specific indicators of quality, namely, the grape variety, classification, vintage, region of origin and the style of wine.Marketers targeting the wine traditionalists should note the segment may also buy wine by the case, and is most likely to buy both red and white wine. Similar to the casual wine buyer segment, the wine traditionalist shows variety in the type of wine bought, which may suggest, gaining brand loyalty may be difficult. However, the large average spend per month (EUR234.03) deems the wine traditionalist segment an attractive target market for wine brands which command a higher price. Figure 3 illustrates the proportion of the sample in each segment. The research examines how the Irish wine market can be effectively segmented to improve brand positioning. The increases in consumption of wine and increases in preference for wine from New World countries are key trends in the Irish wine market. Consumer behaviour, particularly involvement in wine purchases, and the importance of wine attributes, are necessary considerations in a wine market segmentation study. In terms of relevance, substance, and accessibility, a k-clustering segmentation design, with a behavioural basis including an involvement variable, proves to be an appropriate approach to segmenting the Irish wine market. The profiles of the three resulting segments: casual wine buyer, value seeking wine buyer and wine traditionalist, are sizeable, accessible, relevant and actionable.The profiles developed as a result of the primary research, provides wine marketers with an insight in Irish wine consumer behaviour. Specifically, marketers are provided with accessible and sizeable segments, with meaningfully distinctions and similarities drawn between them. Brand positioning can be improved by ensuring the brand communicates and emphasises the product attributes, which the targeted segments values the most when choosing wine. The demographic information and the buyer behaviour data provide marketers with points of access to their target market. The involvement base, when used in conjunction with other behaviour variables, proves effective in producing sizeable, accessible and actionable segments. A limitation of adopting a behavioural basis in conducting the segmentation is the highly descriptive nature of the resulting data. Examining behaviours give an insight into how consumers act, but fails to take into account the underlying motivations and rationale for consumer actions. The use of more complex segmentation bases, such as value systems and lifestyles would wield a richer understanding of the Irish wine consumer. A second suggestion for future research is an empirically tested wine market behavioural segmentation study, to confirm the findings in this research at a national level.In answering the research question, the Irish wine market can be effectively segmented, with a k-cluster design, with a behavioural basis. Effectively segmenting the Irish wine market requires more than the involvement variable, and calls for other behavioural variables, including the importance of product attributes and country of origin preferences to be included in the segmentation process.Opens in a new window.Figure 1 Wine country of origin preferenceOpens in a new window.Figure 2 Age category bar chartOpens in a new window.Figure 3 Final segmentsOpens in a new window.Table I k-cluster analysis: final cluster centres
|
- The research identifies three clusters of wine consumers: casual wine buyer, value seeking wine buyer, and wine traditionalist. Together, the clusters provide an insight into consumers' behaviour. The high correlation of statistics from this research with those of The Wine Development Board suggests the data findings are representative of the population.
|
[SECTION: Value] The Irish wine market has experienced unprecedented growth in the last from 15 to 20 years. From 1990 to 2007, total wine sales in Ireland have more than quadrupled, increasing from 1.7 to 7.6 million cases. In the 13 years between 1994 and 2007, wine's proportion of the Irish alcohol market more than doubled from 8 per cent to 17.9 per cent (Wine Development Board, 2007). Growth in wine consumption is forecasted to continue with a growth of 15 per cent expected by 2012 (Euromonitor, 2008). As the wine drinking culture in Ireland is relatively new, the segmentation of the market and brand positioning is in its infancy. Further study into segmentation is required to improve the profitability of the industry, and to develop choice and the accessibility of wine for Irish consumers. The specific purpose of the paper is to examine how the Irish wine market may be effectively segmented for improved brand positioning in Ireland. Thus, the paper aims to determine the key trends in the Irish wine market, examine the state of marketing in the wine industry, evaluate different approaches to segmenting the Irish wine market, and develop profiles of the resulting segments. The Irish wine market has experienced remarkable growth with the number of wine drinkers in Ireland doubling since 1990 and with over five times as much table wine being consumed in 2007, as was consumed in 1990 (WDB, 2007). The increased consumption of wine in Ireland over the last 15 years is attributed to the improved accessibility, affordability and branding of wine (Moran, 2002). To emphasise the significance of the growth in wine consumption in Ireland, the level of growth in wine buying is compared with growth in the overall food and beverage sector. Practically all wine bought in Ireland is imported. Between 2000 and 2004, wine sales (and therefore imports) increased by 56 per cent (WDB, 2004), while the overall imports of the food, beverages and other animal products category increased by only 18 per cent in the same time frame (CSO, 2006).For a marketer assessing the Irish wine market, equally important to the growth in consumption, is the huge shift in the type of wine preferred by the Irish wine drinker. Specifically, there is a notable shift towards New World wines, with diminishing preference for Old World wines (WDB, 2007). New World wines refer to wines from regions outside of Europe. Prominent New World wine producing regions include South Africa, Australia, New Zealand, Chile, Argentina and California. Old World countries refer to European countries with a long history of wine production, such as France, Italy, Germany and Spain (Fielden, 1994). Up to 1990, the majority of wine consumed in Ireland was Old World wine, and accounted for 94 per cent of the market (WDB, 2007). Since 1990, there has been a steady shift in demand towards New World wine, which in 1990 accounted for 6 per cent of the market and in 2007 held a 71 per cent market share (WDB, 2007). Historically, French wines were the market leader in the Irish wine market, but since 2001, Australia now has the largest market share, accounting for 26 per cent of wine consumed in Ireland (WDB, 2007). In 2006, the top ten wine brands in the Irish still light wine market accounted for nearly 25 per cent of total sales. These top ten wine brands and their respective brand shares are Jacob's Creek (3.2 per cent), Blossom Hill (2.7 per cent), Rosemount (2.6 per cent), E&J Gallo (2.5 per cent), Wolf Blass (2.5 per cent), Hardys (2.3 per cent), Concha Y Toro (2.3 per cent), Long Mountain (2.2 per cent), Santa Rita (2.1 per cent) and Carmen (1.9 per cent), all of which are from New World countries (Euromonitor, 2008). Understanding the shift in the country of origin preference is important as it represents an important shift in preference for style, taste, brand, price and other wine variables. Testing country of origin preference is, therefore, an essential element in this Irish wine market study. The production of wine is a specialised area, and the wine industry has traditionally adopted a production-focused mindset with the complexities of viticulture and vinification having occupied the attention of specialists in the area (Thomas and Pickering, 2003). Bruwer et al. (2002) note the agricultural basis of the wine value chain, and the industry is often criticised as employing mass marketing campaigns (Gluckman, 1990; Spawton, 1991a; Hall and Winchester, 1999; Bruwer et al., 2002).According to Thomas and Pickering (2003), the marketing of wine is in its infancy, relative to the long history of wine making and wine drinking. Aggressive marketing was uncommon in the industry with vineyard operators relying on the strength of their reputation to compete in the marketplace (Hall, 2004). In the early 1990s, an interest in branding emerged in the industry as a method of coping with changes in distribution and the growth of wine retailing. In 1991, the European Journal of Marketing dedicated an issue to the marketing of wine, where Spawton (1991b, c, d, e) through a series of articles, provides an insight into the state of wine marketing. This special issue served as an introduction to marketing for the industry, with the purpose of illustrating the advantage and necessity in developing a customer mindset.There is a realisation in the industry that its future is geared to meeting the expectations of the wine consumer. That has contributed to the growing importance of wine marketing within the industry (Spawton, 1991b, p. 6).Gluckman (1990), in a frequently cited article, presents a number of challenges facing the wine marketer in branding wines. Most notably, a move towards own brand labelling by retailers reinforces the necessity for strong wine brands. This need for improved branding of wine has prompted the undertaking of research in wine markets. In an industry which is relatively new to marketing, and in the Irish wine market, which has seen tremendous growth and transformation, there is a need for greater understanding of the market dynamics. Market research in general, and market segmentation in particular, has a potentially pivotal role to play in assisting wine marketers to position their wine brands effectively. Weir (1960, p. 95, as cited in Yankelovich, 1964) provides the following description of what a market is, and more importantly, what it is not:"The market" is not a single, cohesive unit; it is a seething, disparate, pullulating, antagonistic, infinitely varied sea of differing human beings - every one of them as distinct from every other one as fingerprints; everyone of them living in circumstances different in countless ways from those in which every other one of them is living.This description of a market is a colourful representation of popular marketing thought on the composition of markets in the late 1950s and 1960s. Smith (1956), in what is considered a landmark article (Reynolds, 1965; Haley, 1968; Wind, 1978; Green and Krieger, 1991; Lin, 2002) introduces a marketing strategy labelled market segmentation, as an approach to competing successfully in the reality of an environment of imperfect competition. The original article by Smith (1956) introduces market segmentation as a strategy. Market segmentation strategy was considered an alternative to product differentiation strategy to deal with diversity in the market. While the initial representation of the market segmentation strategy is based in economic theory, market segmentation developed as one of the most foremost concepts in marketing thought (Wind, 1978; Johnson et al., 1991; Lin, 2002).At a broad level, market segmentation provides a marketer with a clearer focus on customer needs, and thereby aids decision making for improved competitive advantage (McDonald and Dunbar, 1992; Croft, 1994; Kotler and Keller, 2003). While Croft (1994) highlights market segmentation as aiding decision making in general, Yankelovich (1964) specifies what exactly segmentation analysis can achieve. In identifying groups of customers with similar needs, a marketer has the information required to target the most profitable group with the most potential. With this knowledge a marketer can develop product lines and promotion activities, choose advertising media, advance positioning of offerings and improve timing of advertising to appeal to the segment of the market whose needs possess the greatest profit potential.A critical decision to be made in conducting segmentation research is choosing an appropriate segmentation base. A segmentation base is the criteria used to divide the defined market into groups of consumers with similarities. At the most basic level a market can be split up according to the profiles of the consumers. Variables such as demographics, geographic location of consumers and the socio economic class to which they belong, are considered profile segmentation bases. The behavioural segmentation category includes bases such as usage occasion, benefits sought, perceptions and beliefs, while the psychographic bases category includes lifestyle and personality variables as a means for identifying groups of consumers with similarities. The more abstract and less concrete the information required for the segmentation base, the more difficult it is to measure responses and their link with behaviour. In choosing a segmentation base for a wine market study there are a number aspects to the market which need to be considered. Literature on wine consumer behaviour focuses on two areas; the factors influencing wine consumer behaviour; and the wine consumer's purchasing decision-making process. An effective wine segmentation study would be one which aids understanding of these two areas and aids marketers in evaluating how stimuli, such as brand positioning strategies, influence wine choice.According to Bruwer et al. (2002), wine markets have been segmented using all the bases identified above. For the purpose of an Irish wine market segmentation study, behavioural segmentation with an involvement basis proves a suitable choice as it is an approach, which yields insight into consumer behaviour, but is not overly difficult to measure (Lockshin et al., 2001). Employing a behavioural segmentation base allows for the decision making process and the influencing factors of the Irish wine consumer to be tested and makes the process and the factors the basis for splitting up the market into meaningful and actionable segments.A key consideration in exploring the wine consumer decision-making process and consumers' evaluation of alternatives, is that wine attributes represent intrinsic and extrinsic cues for the consumer. Sanchez and Gill (1998) illustrate how consumers have preferences according to the bundle of benefits they are seeking. The challenge in understanding these preferences is the large number of wine attributes which exist, and therefore, the greater number of possible bundles of benefits that are present. Wine attributes include: brand name, producer, grape variety, blend of grape varieties, vintage, region of origin, price, label, bottle type, cork type, bottle size, colour of wine, style of wine and level of alcohol. Due to the large number of wine attributes, wine consumers have a wider range of considerations in making purchasing decisions. Examining the hierarchy of importance of wine attributes to Irish wine buyers is a central consideration in segmenting the market. The research design is primarily descriptive in nature as similar investigations into other wine markets have been descriptive in design (Orth et al., 2005; Johnson et al., 1991; Hall, 2004; Bruwer et al. Li and Reid, 2002; Johnson, 2003; Thomas and Pickering, 2003). Due to the descriptive nature of this research, a quantitative approach to primary data collection is most suitable. Quantitative data is appropriate for determining and understanding the behaviours and characteristics of a large sample of wine drinkers. Specifically, survey data collection was undertaken, with a questionnaire collection instrument administered through a personal interview. The questionnaire was two pages in length with 15 tick box questions. The questionnaire posed questions to gather data on four topics: volume of usage, buying preference, product involvement, and demographic information.As an accurate sampling frame was unavailable for the population of the 1,451,000 wine drinkers in Ireland (WDB, 2004), non-probability sampling was undertaken. The sampling type was convenience sampling, as wine buyers were approached at the point of purchase. Convenience sampling has been employed in previous wine segmentation studies, namely, Australian wine market research by Hall (2004) and Bruwer et al. (2002). To ensure the sample was as representative of the population as possible, a large sample size of 300 was chosen and the questionnaire was administered in a variety of outlets to gather information from wine drinkers with wide ranging involvement levels.The fieldwork took place over three weeks in June 2006, in eight wine selling outlets in Galway City and County. This approach is similar to other wine segmentation studies (Bruwer et al., 2000; Hall and Winchester, 1999) where the fieldwork was limited to one region of the market being researched. Research by Bruwer et al. (2000) in segmenting the Australian wine market using a wine-related lifestyle approach is based on fieldwork conducted in Adelaide, while Hall and Winchester's (1999) findings, confirming empirically segments in the Australian wine market, are derived from questionnaires administered in Melbourne. The fieldwork locations for this research consisted of four supermarkets, two off licences and two wine shops. The supermarkets and off licences were in both Galway City and Galway County locations, and the wine shops are situated in Galway City. In total, 316 questionnaires were collected, and after nine were removed for failing a screening question or being incomplete, 307 questionnaires were coded and inputted into SPSS for analysis. There are two sets of findings resulting from the primary research: an overall wine sample analysis and a segment analysis. The overall wine sample is composed of 64 per cent female respondents and 36 per cent male respondents. In terms of age group, over 75 per cent of the sample is aged between 25 and 54 years. The consumer behaviour data reveals the average Irish wine drinker buys seven bottles of wine per month, spending EUR10.57 a bottle, with an average monthly spend of EUR80. Wine is usually bought in a standard size bottle of 75 cl (95 per cent) in either a supermarket (42 per cent) or off licence (35 per cent) and is mostly red wine (43 per cent). Wine is most frequently consumed when dining at home (50 per cent ), followed by when dining out (20 per cent). The five most important product attributes when buying wine are: price per bottle, style of wine (e.g. fruity), region of origin (e.g. Burgundy) and brand name (e.g. Jacob's Creek). The most popular wine is Australian wine, followed by Chilean, French and South African wine.A comparison of the overall wine sample findings with the Wine Development Board's (2004) national statistic shows the characteristics of the research sample are similar to the characteristics of the national market. The sample findings for the country of origin preference are notably similar to the WDB (2004) statistics (see Figure 1). One exception is US wine, which is preferred by just 3 per cent of the sample, but 13 per cent of the national statistics. The similarities between the two sets of statistics suggest the sample data findings on consumer behaviour are representative of the population.Similarly, Figure 2 compares the age category of the sample finding with the age categories of the WDB (2004) population. With the exception of the "65 or older category", the age profile of the two sets of data are similar. Discrepancies between the current research and the WDB (2004) findings may be explained by the time lapse between the two studies, or by the WDB research population being wine consumers, while the current research population was wine buyers.Segment analysis Similar to other wine segmentation studies (Johnson et al., 1993; Bruwer et al., 2002), a k-clustering approach to segmentation was adopted. The k-clustering approach is an exploratory approach, which examines the relationships between more than one variable and produces clusters based on these relationships. The benefit of employing a k-clustering approach is that clusters are formed based on actual relationships within the data set, and not prior assumptions by the researcher.Using the k-means algorithm is a hierarchical method of clustering which reassigns cases to segments "whose centroid is closest to the case" (Punj and Stewart, 1983, p. 139). Table I illustrates the final cluster centres of 17 variables for three clusters. Three of the variables relate to buyer behaviour characteristics (i.e. volumes and average spends). An involvement variable is included in the k-clustering analysis. The involvement variable was developed using Lockshin et al. (2001) product measurement instrument. Fourteen variables in the k-cluster analysis were variables quantifying the importance to the sample of fourteen wine characteristics when choosing wine. The final two variables are the type of wine most frequently purchased, and the age category of the sample. Figure 2 compares the age category results for the WDB (2004) and the total sample of this research. As there is a difference in the two sets of results, age category was included in the k-clustering.The researcher chooses how many clusters are desirable before conducting the k-clustering analysis. As highlighted by Bruwer et al. (2002), there can be difficulty in determining how many clusters are appropriate. A trial and error approach was adopted and the k-clustering was conducted four times to firstly produce two clusters, then three, four and finally five clusters. The four sets of results were assessed to determine which set of results yielded the most sizeable and actionable clusters. Of the four sets of results, the k-clustering which yielded three clusters was found to produce three distinctive and substantial clusters, and these clusters were accepted as the three final segments. To develop three segment profiles, the data for the three clusters is cross tabulated. Three profiles of the average wine buyer in each segment are developed below, followed by a commentary on how the findings are of use to wine marketers.Segment profiles Segment 1: Casual wine buyer (52 per cent of the sample) The average wine buyer in this segment is female, aged from 45 to 54 years, buys on average six bottles of wine per month, spends on average EUR10 a bottle, resulting in an average monthly spend of about EUR60. She is moderately involved in the purchase, with the region of origin, price per bottle, and grape variety being the most important criteria when buying wine. This wine buyer chooses wine they like over more concrete attributes such as alcohol level and price. The style of wine and the brand name are also important. She shows a preference for red wine, and buys in the supermarket usually choosing Australian or French wine. As the Segment 1 wine buyer chooses wine according to more ethereal qualities like style and brand, over more concrete wine attributes, this segment possibly has a casual approach to choosing wine, and is therefore labelled a "Casual Wine Buyer".Marketing to the casual wine buyer The casual wine buyer Segment has the greatest the degree of variability of country of origin preference, with the top two wine countries, Australia and France, being from both Old and New Worlds of wine. The segment is therefore suitable as a target market for wine brands from all countries in both the Old and New World, particularly, Australian, France, Chile, South Africa and Spain. For the casual wine buyer, region of origin is the most important product attribute in choosing wine. Emphasising in the brand communication message, the country from which the wine originates, may be an effective method for wine brands to gain membership in the target market's evoked set of alternative brands.The style of wine preferred by the casual wine buyer is red wine (58 per cent of the segment) with a low preference for white wine (13 per cent of the segment). In promotion activities, such as special price offers when buying more than one bottle, marketers targeting the casual wine buyer Segment, should avoid combining both red and white varieties of the same brand, as the casual wine buyer shows a high preference for red wine. In reaching the casual wine buyer, wine marketers must be conscious, that while wine is usually bought in supermarkets, off licences and wine shops are also popular places of purchase. The variety in country of origin and place of purchase in this segment, gives this segment an image of being casual and informal in their approach to wine buying. A possible point of difference for wine brands targeting this segment could be a casual and informal brand identity, to mirror the possible sentiments of its target market.Segment 2: Value seeking wine buyer (35 per cent of the sample) The average wine buyer in this segment is also female, but slightly younger than the casual wine buyer, being in the age category of 25 to 34 years. Similar to the casual wine buyer, she is employed, with secondary or third level education. She spends on average EUR55 a month on wine, on five bottles. She is the least involved in the wine purchase process of the three segments. For the average buyer in Segment 2, wine is chosen in a purchase situation based on price, style of wine, and brand name, alcohol level and grape variety. Price is the most important aspect of the purchase decision, with alcohol level also of importance, which might suggest she looks for measurable indicators of getting value for her money. This wine drinker buys wine in the supermarket or off licence, drinks the majority of wine at home and buys both red and white wine. This segment shows a preference for New World wine, with 80 per cent of the countries of origin of wine bought by this segment being New World countries. The importance of easily comparable indicators of value (price and alcohol level) in choosing wine gives this segment the title "Value Seeking Wine Buyer".Marketing to the value seeking wine buyer The value seeking segment shows greater preference for wine from New World countries, and proves a suitable target market for wine brands from Chile, South Africa, Australia and the USA. Attracting the value seeking segment requires marketers to emphasise product attributes that are easily measurable. As price is the most important product attribute, attracting this segment with a low price is an obvious positioning strategy. The level of alcohol is also one of the top important attributes in choosing wine. Brands that communicate the value for money qualities of the wine, and resonate with the segment's desire to find affordable wines, may succeed in attracting this segment.The value seeking segment is an attractive segment, despite buying less and spending less than either of the other segments. The profile highlights a segment with consistent preferences and behaviours, which might suggest a greater propensity to being brand loyal, making the segment attractive to wine marketers with a brand image of value for money.Segment 3: Wine traditionalist (13 per cent of the sample) Unlike the other two segments, the average wine buyer in Segment 3 is male, aged between 35 to 44 years, and self-employed with third level education. He is greatly involved in his purchases and buys on average15 bottles of wine per month, spending on average EUR16 a bottle, resulting in an expensive average monthly spend of EUR235. He chooses wine according to traditionally perceived indicators of wine quality such as grape variety, style, region, classification and vintage. Price does not appear in his top five criteria for choosing wine. The Segment 3 wine buyer prefers Old World wines, which account for 63 per cent of the segment's country of origin preferences. The top three wines of preference are from France, Spain and Italy. This wine buyer predominantly buys his wine in wine shops. Despite adopting a connoisseur approach to choosing wine, he is similar to buyers in the other segments, in that wine is consumed most often when dining at home, and is usually bought in standard size bottles. Due to the importance of traditional indicators of wine quality (i.e. classification and vintage), the average wine buyer in this segment may be referred to as a "Wine Traditionalist".Marketing to the wine traditionalist While the wine traditionalist segment accounts for just 13 per cent of the market, in the context of average spend per month, the segment is equally large as the casual wine buyer segment. The wine traditionalist segment is a suitable target market, for Old World wine brands, especially French, Italian and Spanish brands. The wine traditionalist segment may also be suitable as a target market for New World wine brands, especially Australian and Chilean wine, provided the brand message communicates specific indicators of quality, namely, the grape variety, classification, vintage, region of origin and the style of wine.Marketers targeting the wine traditionalists should note the segment may also buy wine by the case, and is most likely to buy both red and white wine. Similar to the casual wine buyer segment, the wine traditionalist shows variety in the type of wine bought, which may suggest, gaining brand loyalty may be difficult. However, the large average spend per month (EUR234.03) deems the wine traditionalist segment an attractive target market for wine brands which command a higher price. Figure 3 illustrates the proportion of the sample in each segment. The research examines how the Irish wine market can be effectively segmented to improve brand positioning. The increases in consumption of wine and increases in preference for wine from New World countries are key trends in the Irish wine market. Consumer behaviour, particularly involvement in wine purchases, and the importance of wine attributes, are necessary considerations in a wine market segmentation study. In terms of relevance, substance, and accessibility, a k-clustering segmentation design, with a behavioural basis including an involvement variable, proves to be an appropriate approach to segmenting the Irish wine market. The profiles of the three resulting segments: casual wine buyer, value seeking wine buyer and wine traditionalist, are sizeable, accessible, relevant and actionable.The profiles developed as a result of the primary research, provides wine marketers with an insight in Irish wine consumer behaviour. Specifically, marketers are provided with accessible and sizeable segments, with meaningfully distinctions and similarities drawn between them. Brand positioning can be improved by ensuring the brand communicates and emphasises the product attributes, which the targeted segments values the most when choosing wine. The demographic information and the buyer behaviour data provide marketers with points of access to their target market. The involvement base, when used in conjunction with other behaviour variables, proves effective in producing sizeable, accessible and actionable segments. A limitation of adopting a behavioural basis in conducting the segmentation is the highly descriptive nature of the resulting data. Examining behaviours give an insight into how consumers act, but fails to take into account the underlying motivations and rationale for consumer actions. The use of more complex segmentation bases, such as value systems and lifestyles would wield a richer understanding of the Irish wine consumer. A second suggestion for future research is an empirically tested wine market behavioural segmentation study, to confirm the findings in this research at a national level.In answering the research question, the Irish wine market can be effectively segmented, with a k-cluster design, with a behavioural basis. Effectively segmenting the Irish wine market requires more than the involvement variable, and calls for other behavioural variables, including the importance of product attributes and country of origin preferences to be included in the segmentation process.Opens in a new window.Figure 1 Wine country of origin preferenceOpens in a new window.Figure 2 Age category bar chartOpens in a new window.Figure 3 Final segmentsOpens in a new window.Table I k-cluster analysis: final cluster centres
|
- This research adopts a lifestyle segmentation approach by linking lifestyle values, product attributes and buying and consumption patterns. The primary research is descriptive in design, employing a self-administered questionnaire to collect quantitative data on wine consumer behaviour. Efforts made to ensure a highly representative sample included choosing a large sample size, administering the questionnaire in a range of outlets, and gathering information from wine drinkers with wide ranging involvement levels.
|
[SECTION: Purpose] There is ample empirical evidence in support of the important role that reference prices (both internal and external) play in determining consumers' evaluations of posted prices (for comprehensive reviews of the many operationalizations of reference price, see Briesch et al., 1997; Mazumdar et al., 2005). Collectively, the body of evidence strongly supports the conclusion that consumers rely on (internal and/or external) reference prices to make judgments about perceived value, perceived fairness and perceived expensiveness that, in turn, influence purchase intentions. The inescapable conclusion for practitioners is that they must pay attention to understanding, measuring and managing consumers' reference prices to ensure that their prices are evaluated favorably. However, the literature offers many different operationalizations of reference price. While some definitions of reference price are based on past prices or on consumers' knowledge of past/current market prices, for example highest price, normal price, average price, lowest price and expected price (Jacobson and Obermiller, 1990; Kalwani et al., 1990; Chandrashekaran and Jagpal, 1995), others are related to subjective notions of fairness (Herrmann et al., 2007; Xia et al., 2004). The most commonly accepted view is that reference price is a brand-specific construct that represents an expectation based primarily on some summary of a brand's own past prices (Briesch et al., 1997). Indeed, based on a comprehensive review of the literature, Mazumdar et al. (2005, p. 86) concluded that "the strongest determinant of a consumer's IRP is the prior prices he or she observes". In addition to such a temporal view of reference price, previous research has offered other potential standards against which consumers may evaluate posted prices. For example, Rajendran and Tellis (1994) introduced the idea of a contextual component that is based on an average of current prices of competing brands (see also Rajendran, 2009). Such varied operationalizations underscore that reference price is a complex, multi-faceted construct, and that no single operationalization is likely to capture it in its entirety. In addition, there are likely to be individual differences with respect to both the number and types of references evoked in particular situations. Despite overwhelming experimental and empirical evidence that reference prices play a crucial role in determining perceived value, we know considerably less about the process by which consumers may combine/integrate multiple reference prices. Even less is known about individual differences in consumers' utilization of multiple reference prices. It is imperative that managers not only understand that reference prices influence consumers' evaluations, but they may benefit from knowledge about which specific reference prices affect consumers' purchase intentions and the process by which this influence may occur. Such information could be vital in planning appropriate price communication strategies designed to affect specific outcomes, for example raising the appropriate reference price(s) to enhance consumers' perceptions of offer-value that will, in turn, increase the likelihood of purchase. The motivation for this paper comes from recognizing two main gaps in the literature. First, despite multiple definitions of reference price, prior research has not adequately addressed heterogeneity in the types of reference prices that are evoked and the ways in which consumers might process available price information in relation to multiple reference prices. This paper considers how one specific source of heterogeneity, involvement, influences consumers' utilization of (multiple) reference prices. Research on consumer information processing (see Bettman, 1979) suggests that several individual differences (e.g. involvement with the product and/or the purchase) may influence the type of information considered and the extent to which consumers are willing to elaborate on available information. This research attempts to fill a gap in the literature by investigating the role of involvement in the context of reference prices.Second, there have been very few attempts to examine the inter-relationship between different reference prices and their combined influence on consumers' evaluations. Previous research (Chandrashekaran and Jagpal, 1995; Rajendran and Tellis, 1994; Mayhew and Winer, 1992) has examined the simultaneous and independent influences of internal and external references on consumers' choices. However, questions about the inter-relationships between different types of reference prices remain. This study attempts to extend current thinking to include sequential effects of reference prices on consumers' evaluations. It is generally established in the literature that consumers may utilize some external reference prices to adjust internal reference prices, which in turn affects evaluation. However, some research (e.g. Rajendran and Tellis, 1994) has also shown that consumers may utilize certain external reference prices in conjunction with internal reference price. Hence, there is a need to validate the process empirically. In addition, the interrelationship between specific internal and external references prices along with the role of involvement has not been empirically investigated.In summary, this study proposes to extend the literature by attempting to answer two questions:1. Do all consumers evaluate a given offer against the same set of reference prices?2. Do all consumers utilize internal and external reference prices in the same way (i.e. using the same underlying process)? When consumers evaluate retail prices, they may do so against several reference prices. Of these, some are internal and others are external. The term "internal reference price" (henceforth referred to as IRP) has generally been used to refer to price information held in consumers' long-term memories. For example, Winer (1986) proposed that consumers use past price information to form expectations of what the retail price is likely to be on the next purchase occasion. As described in the literature, these price expectations represent internally held reference prices, and they are based on past price information that must be retrieved from consumers' long-term memories during the purchase occasion. In fact, most empirical studies have based their measures of internal reference price on such a temporally generated construct (Kalwani et al., 1990; Mayhew and Winer, 1992; Rajendran and Tellis, 1994; Briesch et al., 1997). The literature offers compelling empirical evidence that such (memory-based) internal reference prices are crucial in predicting consumers' evaluations (of retail prices) and subsequent choice behavior.In contrast to internal reference price, external reference prices (henceforth referred to as ERP) refer to comparative (competitive) price information that is available in the purchase environment, i.e. prices that consumers are likely to encounter during the search process. Thus, consumers do not have to expend cognitive resources in trying to retrieve information on current market prices from their long terms memories. This information is likely to be obtained during the search process and utilized directly from working memory. Here, researchers have identified several possible references including normal (typically charged) market price (Urbany and Dickson, 1991) and the lowest market price (Biswas and Sherrell, 1993). These and other similar studies have shown that such market-based references are important predictors of consumers' choices.It is important to note the conceptual distinction between IRP and ERP. While IRP refers to price information that is stored in and retrieved from consumers' long term memories, ERPs are derived from price cues that are either present or evoked at the time of purchase. Specifically, ERPs may represent consumers' beliefs about current market prices that may be formed during the search associated with the purchase process. Such beliefs may include highest price, lowest available price and normal/average market price. Note, however, that there is likely to be significant heterogeneity (due in part to differing levels of involvement) in the extent to which consumers search for the lowest/best price. Such selective and differential exposure is likely to result in differences across individuals in their perceptions of ERPs. It is well established that reference price is a complex, multi-faceted construct, and that reference prices are important in explaining consumers' brand choices (Briesch et al., 1997; Mazumdar et al., 2005). In addition, consumers are likely to use more than one type of reference price to evaluate an offer (Mayhew and Winer, 1992; Rajendran and Tellis, 1994; Chandrashekaran and Jagpal, 1995; Shirai, 2003). However, we know considerably less about the process by which consumers may combine/integrate multiple reference prices. Even less is known about individual differences in consumers' utilization of multiple reference prices.Prior research has acknowledged that different consumer groups (segments) are likely to evoke and use different sets of reference prices (Winer, 1986; Mazumdar and Papatla, 2000). Indeed, prior research has demonstrated that several factors lead to differences in the references used to evaluate a posted price. For example, reference price utilization and price perceptions have been linked to frequency of purchase (Thomas and Menon, 2007), product type (Chandrashekaran and Jagpal, 1995; Lowe and Alpert, 2007) and price consciousness (Alford and Biswas, 2002; Palazon and Delgado, 2009). In summary, prior research supports the claim that different segments may evoke and utilize different sets of reference prices, and suggests that it may not be appropriate to assume that all consumers utilize the same types (and number) of reference prices.As emphasized by Lowe and Alpert (2007) and by Mazumdar et al. (2005), despite some research on the topic, we do not know much about how consumers integrate multiple reference prices. More importantly, we need a better understanding of the factors that may determine which type(s) of reference price(s) are evoked and utilized when making price-related judgments. Along those lines, this research proposes to investigate how the level of involvement leads to differences in the number and types of reference prices that are used in the evaluation process. More specifically, it focuses on how consumers combine and utilize internal reference price (operationalized as expected price) and market-based external reference prices (operationalized as subjective knowledge of normal and lowest market prices)[1]. Consistent with established theory, this research acknowledges that consumers may adjust their internal reference prices based on their perceptions of market-based references. However, the conceptualization adopted here (explained later) also allows for consumers to evaluate posted prices directly against market-based (external) references.Regarding the relative impacts of the two types of reference prices on consumers' evaluations, there is some empirical evidence (Mayhew and Winer, 1992) suggesting that external reference prices may be more influential than internal reference prices. However, that conclusion might be premature because previous research has not paid attention to the moderating roles of individual level variables, for example involvement, that are likely to influence how consumers choose to allocate their cognitive resources - this is likely to impact the relative emphasis that consumers place on the two types of reference prices when evaluating a posted price. Consumers' involvement with a product class has been shown to influence consumers' information processing strategy (Bettman, 1979; Bloch et al., 1986) by affecting the extent of search, the type of information sought, the relative importance of different types of information (e.g. price versus product attribute information), and the motivation to process available information. High involvement is generally associated with product knowledge (familiarity) and the motivation to search for and use relevant information. On the other hand, low involvement has been associated with low product/price knowledge (lack of familiarity), and the lack of motivation to engage in detailed processing of information.Although there is substantial research dealing with how involvement affects consumers' use of product-attribute information and on how involvement affects consumers' responses to marketing communications (advertising), research in the (reference) pricing context is relatively limited. Available evidence reveals that such variables as involvement, prior knowledge, and price consciousness moderate consumers' internal reference prices and their acceptance of retail prices. These variables have been found to affect consumers' internal standards (Kosenko and Rahtz, 1988; Lichtenstein et al., 1988; Biswas and Sherrell, 1993), the widths (i.e. upper and lower limits) of their price acceptance regions (Lichtenstein et al., 1988; Rao and Sieben, 1992), and, finally, their confidence in their reference price estimates (Biswas and Sherrell, 1993). However, we know less about how such factors are likely to influence the ways in which consumers may combine/utilize multiple reference prices. The current study focuses on one such factor, i.e. involvement. This research explores the notion that different reference price utilization strategies unfold for consumers with high and low involvement. The idea that involvement affects which decision route consumers are likely to follow is not new. Extant research on consumers' information processing has shown that involvement affects the way consumers acquire, store, retrieve, and use relevant information to make decisions (Bettman, 1979; Bloch et al., 1986). Compared to those who are not involved, involved consumers are more motivated to search for relevant information. They also possess better knowledge about products and prices than consumers who lack involvement. Therefore, these consumers are more likely to possess well-defined internal standards than low involvement consumers (see also Chandrashekaran and Jagpal, 1995). Consequently, they are likely to be more comfortable evaluating retail prices against their internal (memory-based) reference prices. In contrast, less involved consumers are not as motivated as their involved counterparts to engage in extensive search behavior. In addition, they not likely to engage in detailed processing of available (price) information. As a result, their internal standards are not likely to be well defined, and they are not confident about acting on this information (Chandrashekaran and Jagpal, 1995).Consistent with such theorizing, Biswas and Sherrell (1993) found that high-involvement consumers are indeed more confident of their internal reference price estimates than low-involvement consumers. Therefore, there is some empirical evidence to support the expectation that the level of involvement is likely to moderate consumers' utilization of reference prices.Consistent with the notion that consumers are likely to possess reference prices that are brand specific (Briesch et al., 1997), a brand's own previous prices may be considered to be the most relevant information on which to base a reference price for the brand. Thus, it may be reasonable to expect high involvement consumers' evaluations to be most closely associated with a summary of a brand's past prices, i.e. the expected price construct. Briesch et al. (1997) empirically compared five different model formulations from the literature and confirmed that a model using a temporally generated, brand-specific reference price is the best predictor of consumer choice. However, the authors did not examine individual differences in the utilization of past price information.As discussed earlier, low involvement individuals are by definition not equipped with the knowledge to evaluate price on the basis of relevant factors. Recall that these consumers are not expected to internalize (past) price information, and thus cannot be expected to have well-defined (reliable) price expectations. Therefore, these individuals are likely to evaluate a brand's current price on the basis of more readily accessible perceptions they may have formed of current market prices.In summary, high- and low-involvement consumers are expected to follow different processes in evaluating the same objective price information. Highly involved consumers base their evaluations on internally held beliefs. However, because they are exposed to market price information in the course of their search, it is likely that they utilize this information to update (fine-tune) their internal standards to be consistent with current market information. In contrast, low involvement consumers are not motivated to devote cognitive resources to storing price information in their long-term memories and, therefore, are less confident about their knowledge of past prices. Consequently, they are likely to rely on immediately available external cues (i.e., market-based reference prices) to make their evaluations (as opposed to first updating their internal reference prices and, subsequently utilizing them to evaluate posted prices).More formally, it is expected that:H1. High-involvement consumers evaluate retail prices only against their price expectations (i.e. internal reference price).H2. High-involvement consumers utilize market-based external references only to update their internal beliefs (expected prices).H3. Low involvement consumers evaluate retail prices against one or more market-based (external) reference prices.H4. Low involvement consumers do not utilize internal reference prices (price expectations) in evaluating retail prices.Figure 1 presents the hypothesized model on the processes by which market-based and truly internal reference prices affect consumers' evaluations. Two hundred undergraduate business majors attending a large Northeastern state university participated in this study. The data required for this study was collected in four stages, over a two-week period. The entire data collection task was designed in the form of an experiential exercise.Stage one: initial measures In the first stage, subjects provided general information on their familiarity and involvement with the chosen product, i.e. Levi jeans. This product was chosen because of its relevance to the student subjects, and because no major differences across genders and age groups were expected.Product knowledge Prior research (Rao and Sieben, 1992; Biswas and Sherrell, 1993) has shown that consumers' prior knowledge affects the levels and widths of their price acceptance regions. Product knowledge was measured using a single seven-point item on which subjects indicated their familiarity with the product (brand). Generally, multiple scale items measures are preferred. However, in this study, consumer knowledge (expertise) is not central to the hypotheses. Therefore, no major problems are foreseen.Involvement Subjects' involvement with the product was assessed using the personal involvement inventory (PII) developed by Zaichkowsky (1985). Subjects rated the product on 20 bi-polar adjective scales. These individual items were summed to yield an involvement index for each subject. Later, high- and low-involvement groups were created by performing a median split of the distribution of involvement indices.The shopping exercise After completing the task in the first stage, subjects were given an instruction sheet that described an experiential exercise. In the instructions, subjects were informed that this exercise was intended to demonstrate to them the differences in consumers' shopping behaviors for convenience and shopping goods. The experiential exercise required students to pretend that they were in the market for products listed on the sheet. To mask the actual intent of the study, the sheet listed several products (e.g. toothpaste, laundry detergent, and sneakers) in addition to jeans. To preserve realism, students were instructed to behave as if they were shopping for the products listed (including a pair of jeans), and were asked to engage in behaviors that they would normally engage in when actually buying the products listed (a pair of jeans, toothpaste, laundry detergent, sneakers) for themselves. They were especially asked to use only the information they would normally use when making purchase decisions in the product categories listed. Finally, the subjects were informed that they were expected to come back to class the following week to engage in a class discussion about their experiences and about where they would buy the products listed.Stage two: measuring perceptions of market prices A week after the initial measures were taken, subjects were questioned about the outcomes of their search during a class discussion. At this stage, subjects' perceptions of market prices were measured on two items corresponding to their estimates of the lowest price for which they can purchase the product and its normal (average) market price[2].Stage three: measuring internal reference price At the next class meeting, subjects were asked to provide estimates of their internal reference prices (prices they expected to see on the next purchase occasion). The measures of market-based and internal reference prices were separated in time to avoid any contamination of the reference price measures.Stage four: the decision task At the end of the second week of the exercise, each subject was shown a mock retail ad for a pair of Levi jeans, which included a picture of the product, information on the retail store (a popular retail chain), and a selling price ($34.99). To stick with the main theme of understanding how consumers utilize perceptions of market prices and their reference prices to evaluate a posted price, the ad did not identify that the product was "on sale" and, therefore, it did not include an advertised comparison price or any other semantic cues that are normally associated with comparative price claims[3]. The retail price ($34.99) was chosen for two reasons. First, it reflected the actual market prices in the stores that most students reported they are most likely to visit. Second, it was close to the median of the average reference price (=$36.17) calculated as (Lowest+Normal+Expected)/3. Thus, for approximately half the subjects the offered price was likely to have been above their average reference price, and for the remaining subjects it was likely to have been below their average reference price. Each subject evaluated the posted retail price on three seven-point scale items corresponding to perceived offer value, attractiveness of the deal, and willingness to buy the product at the stated price. Similar items have been used in previous research (see, for example, Chandrashekaran and Jagpal, 1995; Chandrashekaran and Grewal, 2006). The first task was to split the sample into high and low groups based on their involvement with the product category (jeans). A median split of the distribution of PII scores (median involvement=103) yielded high and low involvement groups of approximately equal sizes (nhigh=111 and nlow=114). In addition, the mean involvement scores in the two groups (mean=119.26 and SD=10.50 in the high-involvement group versus mean=83.03 and SD=16.87 in the low involvement group) were significantly different (t=19.27, p<0.001). Finally, an examination of the mean knowledge (familiarity) scores in the two involvement groups (mean familiarity=5.99 and 4.75 for high- and low-involvement groups, respectively) confirmed the premise that, in this product category, high-involvement consumers are significantly more knowledgeable than low-involvement consumers (t=7.91, p<0.01).To test the hypotheses in this study empirically, correlation matrices in each group were analyzed using the structural equations methodology (interested readers may refer to Joreskog and Sorbom, 1988). Figure 1 shows the model structure specified in both groups.The conceptual model shown in Figure 1 allows for the possibility that market-based reference prices (lowest price and normal price) may affect evaluations directly (along with expected price), or indirectly by first influencing expected price, which, in turn, affects how the offer is evaluated. Therefore, the model enables us to investigate the process by which consumers evoke, combine and utilize internal and market-based reference prices. Recall that this is an improvement over Rajendran and Tellis (1994), who examined how both temporal and contextual reference prices influence consumers' evaluations simultaneously. In relation to Figure 1, the hypotheses presented here predict that:* g11, g12, and b21> 0; and g21 and g22=0 in the high-involvement group; and* b21, g11, and g12=0; and g21 and g22>0 in the low-involvement group.As discussed earlier, consumers' evaluations were measured using three scale items corresponding to their perceptions of the value of the offer, the attractiveness of the deal, and their willingness to buy the item at the stated price. It could be argued that the concept of perceived value is related to consumers' perceptions of acquisition utility (AU), while their perceptions of the deal are related to transaction utility (TU) as defined by Thaler (1985). According to Thaler, these are conceptually different from purchase intentions (behavioral responses). However, principal components factor analysis resulted in a single factor and did not reveal conceptual differences between the three measures. Table I compares the two involvement groups on their reference price estimates and evaluations of the offer. It is interesting to note that there is no significant difference in reference price estimates across the two involvement groups (this point is also discussed later in the manuscript). Furthermore, there are no major differences between the two groups in their evaluations of the offer, except that high involvement consumers perceive the deal to be moderately more attractive than low involvement consumers (p<=0.10)[4].However, one cannot conclude on the basis of these results alone that high- and low-involvement consumers are homogeneous, thus justifying pooled analyses. Although consumers' final evaluations are important, it is crucial that we examine the process (route) by which high and low involvement consumers arrive at the same destination. In addition, consumers' final evaluations are not central to this study. Rather, the primary objective of this study is to examine heterogeneity in consumers' utilization of market-based and truly internal reference prices to evaluate the same objective price information. Indeed, although consumers do not differ significantly in their final evaluations of the offer, this paper seeks to uncover heterogeneity in reference price utilization, i.e. the means by which high and low involvement consumers reach the end (their final evaluations).Table II shows the estimated models in each group along with several fit indices. It is clear that the specified model structure fits the data well in both groups (kh2 with six df=9.59, p=0.143 in the high involvement group; and kh2 with six df=6.09, p=0.413 in the low-involvement group). In addition, the structural models explain a significant proportion of the variance (R2=0.51 and 0.53 in the high- and low-involvement groups respectively). For both groups, the RMSEA (Steiger, 1985) and other fit indices indicate satisfactory model fit. All indicated paths are of the predicted sign, and are significant at a=0.01. All non-significant effects, i.e. paths that are not significantly different from zero, are indicated as "NS".From Table II, it is clear that consumers who are highly involved with the product evaluate the retail price against a single internal reference price, expected price (b21 is significant at p<=0.01), but do not utilize any market-based references to make evaluations (note that g21 and g22 are not significantly different from zero). These findings strongly support the hypothesis (H1) for consumers' utilization of truly internal reference prices under high involvement.H2 suggests that involved consumers are likely to use market-based references primarily to update their internal references. Consistent with this hypothesis, Table II shows that normal price (a market-based reference price) has a significant positive effect on the internal standards of involved consumers (g12=0.727 is significant at a=0.01). This finding is consistent with the view that involved consumers process price information sequentially, i.e. an estimate of the normal/average market price is used to update (fine-tune) the internal standard, which, in turn, is used to evaluate the retail price. Contrary to expectation, these consumers do not utilize their perceptions of the lowest market price (g11 is not significant). Thus, overall, H2 is partially supported.In contrast to high-involvement consumers, those in the low-involvement group do not use their internal standards to evaluate the retail price (b21=0). As shown in Table II, these consumers base their evaluations of the retail price on market-based references (g21=0.257 and g22=0.211 are both significant at a=0.01). Thus, H3 and H4 are supported. In conjunction, these findings support the conclusion that low levels of involved facilitate simultaneous utilization of several market-based reference prices to evaluate retail prices[5].It is interesting to note that highly involved consumers utilize only normal price, but not their perceptions of lowest price, in formulating their price expectations. Clearly, further research is required to be able to offer a convincing explanation to this observation. One possible explanation may be that these consumers may generally be more aware that the observed lowest price (a single data point) is likely to be a temporary promotional offer that does not reflect the "normal" price for the product/category. Consequently, these consumers may discount this information and rely to a greater extent on more comprehensive information, for example overall estimate of normal/average price. Urbany and Dickson (1991) found that consumers' estimates of normal prices are reasonable surrogates for their reference prices. However, the authors did not investigate individual differences in consumers' perceptions of normal/average prices. It is likely that one segment (e.g. high-involvement consumers) has more accurate estimates of normal prices than another segment (e.g. low-involvement consumers). Hopefully, future research will investigate and shed some light on this and other similar issues dealing with individual differences in price perceptions and evaluation.In summary, the results reveal that although the two types of consumers (high- and low-involvement) may have similar reference prices and may even evaluate an offer similarly, they do so via distinct routes, i.e. by utilizing different types of reference prices. Specifically, whereas highly involved consumers rely solely on their internal standards to evaluate retail prices, those who are less involved with the product category utilize several market-based references to judge the attractiveness of retail prices. A general discussion of the results along with implications and limitations of the study follows. This study intended to draw attention to the heterogeneity in consumers' utilization of multiple reference prices. The results obtained here offer some evidence that involvement affects consumers' utilization of reference prices. Although involvement does not affect the levels of consumers' internal standards or their final evaluations, it plays a significant role in the way consumers utilize these standards to evaluate the same objective price information. Highly involved consumers, who are more knowledgeable about the product class, evaluate offers against a single internal standard (expected price), which supports previous researchers' (e.g. Winer, 1986; Kalwani et al., 1990) in their use of expected price as a reference price to explain consumers' choices. However, low-involvement consumers do not use expected price to evaluate offers. Rather, they use two market-based references (lowest price and normal price) to determine the overall value of an offer. Thus, involvement affects the number and types of reference prices used in the evaluation process.The results are also consistent with information-processing theory (Bettman, 1979), which advocates that high involvement encourages a deeper and more complex processing strategy than low involvement, which supports less complex processing and the use of easily available information. Highly involved consumers retrieve price information stored in long-term memory and fine tune this internal standard (expected price) using current market price information (estimated normal/average market price). Thus, high involvement supports a hierarchical model in which market-based and truly internal reference prices are used sequentially. This conclusion is consistent with the underlying process implied in vast amount of research on reference price formation. In contrast, low-involvement consumers do not possess the motivation to engage in such a complex evaluation process. They simply utilize the more readily available market-based references and evaluate offers in a single step.It is only appropriate to point out that some of the results obtained here are inconsistent with previous research findings and call for further investigation. For example, Lichtenstein et al. (1988) found that consumers' price acceptability level (i.e. internal reference price level) is positively related to involvement. However, no such differences were found here. One possibility is that the results are being moderated by consumers' knowledge in this category. Rao and Sieben (1992) found that the upper and lower limits of consumers' price-acceptance regions increase with knowledge up to a point, and then levels off. It is likely that, given the nature of product category used in this study (jeans), most consumers were familiar with the product. Indeed, an examination of the mean knowledge scores revealed that both high- and low-involvement consumers lie in the upper half of the scale (means for high- and low-involvement groups=5.99 and 4.75 on seven-point scales), which may account for the lack of difference in their reference prices. Furthermore, product knowledge was assessed using a single item that measured subjects' familiarity with the product category. Additional research is needed to investigate this issue further.Another possible limitation of this study is the omission of several internal and market-based reference prices (e.g. fair price and highest price) that have been mentioned in the literature. It might be useful for future research to include more definitions of reference price to investigate whether the premise of this research holds. Finally, it may be useful to examine the roles of other moderating factors (e.g. gender, product experience, price consciousness, etc.).This research has important implications for both marketing research and practice. It is important for researchers to identify other sources of heterogeneity and their impact on consumers' construction and utilization of reference prices. The finding that high- and low-involvement consumers are different in the types of reference prices used and in the way they use this information to evaluate offers strongly supports the need for using different model structures to represent these sub-populations. From a practical standpoint, this study underscores the importance of including individual differences in strategies designed to affect consumers' perceptions of retail prices. For example, marketers might segment the market on the basis of reference price utilization and design different strategies to obtain optimal results. It is hoped that this study will stimulate further research in the area so that we may gain a better understanding of how consumers process and respond to retail pricing strategies. Such knowledge will undoubtedly be useful in designing more effective and efficient marketing strategies. Opens in a new window.Figure 1 Process of reference price effects on evaluations Opens in a new window.Table I Summary statistics Opens in a new window.Table II Estimated models
|
- The purpose of this article is to investigate whether involved consumers utilize the same set of reference prices to evaluate an offer as compared to those who are less involved. Additionally, this study aims to investigate whether the processes employed in the two groups are different.
|
[SECTION: Method] There is ample empirical evidence in support of the important role that reference prices (both internal and external) play in determining consumers' evaluations of posted prices (for comprehensive reviews of the many operationalizations of reference price, see Briesch et al., 1997; Mazumdar et al., 2005). Collectively, the body of evidence strongly supports the conclusion that consumers rely on (internal and/or external) reference prices to make judgments about perceived value, perceived fairness and perceived expensiveness that, in turn, influence purchase intentions. The inescapable conclusion for practitioners is that they must pay attention to understanding, measuring and managing consumers' reference prices to ensure that their prices are evaluated favorably. However, the literature offers many different operationalizations of reference price. While some definitions of reference price are based on past prices or on consumers' knowledge of past/current market prices, for example highest price, normal price, average price, lowest price and expected price (Jacobson and Obermiller, 1990; Kalwani et al., 1990; Chandrashekaran and Jagpal, 1995), others are related to subjective notions of fairness (Herrmann et al., 2007; Xia et al., 2004). The most commonly accepted view is that reference price is a brand-specific construct that represents an expectation based primarily on some summary of a brand's own past prices (Briesch et al., 1997). Indeed, based on a comprehensive review of the literature, Mazumdar et al. (2005, p. 86) concluded that "the strongest determinant of a consumer's IRP is the prior prices he or she observes". In addition to such a temporal view of reference price, previous research has offered other potential standards against which consumers may evaluate posted prices. For example, Rajendran and Tellis (1994) introduced the idea of a contextual component that is based on an average of current prices of competing brands (see also Rajendran, 2009). Such varied operationalizations underscore that reference price is a complex, multi-faceted construct, and that no single operationalization is likely to capture it in its entirety. In addition, there are likely to be individual differences with respect to both the number and types of references evoked in particular situations. Despite overwhelming experimental and empirical evidence that reference prices play a crucial role in determining perceived value, we know considerably less about the process by which consumers may combine/integrate multiple reference prices. Even less is known about individual differences in consumers' utilization of multiple reference prices. It is imperative that managers not only understand that reference prices influence consumers' evaluations, but they may benefit from knowledge about which specific reference prices affect consumers' purchase intentions and the process by which this influence may occur. Such information could be vital in planning appropriate price communication strategies designed to affect specific outcomes, for example raising the appropriate reference price(s) to enhance consumers' perceptions of offer-value that will, in turn, increase the likelihood of purchase. The motivation for this paper comes from recognizing two main gaps in the literature. First, despite multiple definitions of reference price, prior research has not adequately addressed heterogeneity in the types of reference prices that are evoked and the ways in which consumers might process available price information in relation to multiple reference prices. This paper considers how one specific source of heterogeneity, involvement, influences consumers' utilization of (multiple) reference prices. Research on consumer information processing (see Bettman, 1979) suggests that several individual differences (e.g. involvement with the product and/or the purchase) may influence the type of information considered and the extent to which consumers are willing to elaborate on available information. This research attempts to fill a gap in the literature by investigating the role of involvement in the context of reference prices.Second, there have been very few attempts to examine the inter-relationship between different reference prices and their combined influence on consumers' evaluations. Previous research (Chandrashekaran and Jagpal, 1995; Rajendran and Tellis, 1994; Mayhew and Winer, 1992) has examined the simultaneous and independent influences of internal and external references on consumers' choices. However, questions about the inter-relationships between different types of reference prices remain. This study attempts to extend current thinking to include sequential effects of reference prices on consumers' evaluations. It is generally established in the literature that consumers may utilize some external reference prices to adjust internal reference prices, which in turn affects evaluation. However, some research (e.g. Rajendran and Tellis, 1994) has also shown that consumers may utilize certain external reference prices in conjunction with internal reference price. Hence, there is a need to validate the process empirically. In addition, the interrelationship between specific internal and external references prices along with the role of involvement has not been empirically investigated.In summary, this study proposes to extend the literature by attempting to answer two questions:1. Do all consumers evaluate a given offer against the same set of reference prices?2. Do all consumers utilize internal and external reference prices in the same way (i.e. using the same underlying process)? When consumers evaluate retail prices, they may do so against several reference prices. Of these, some are internal and others are external. The term "internal reference price" (henceforth referred to as IRP) has generally been used to refer to price information held in consumers' long-term memories. For example, Winer (1986) proposed that consumers use past price information to form expectations of what the retail price is likely to be on the next purchase occasion. As described in the literature, these price expectations represent internally held reference prices, and they are based on past price information that must be retrieved from consumers' long-term memories during the purchase occasion. In fact, most empirical studies have based their measures of internal reference price on such a temporally generated construct (Kalwani et al., 1990; Mayhew and Winer, 1992; Rajendran and Tellis, 1994; Briesch et al., 1997). The literature offers compelling empirical evidence that such (memory-based) internal reference prices are crucial in predicting consumers' evaluations (of retail prices) and subsequent choice behavior.In contrast to internal reference price, external reference prices (henceforth referred to as ERP) refer to comparative (competitive) price information that is available in the purchase environment, i.e. prices that consumers are likely to encounter during the search process. Thus, consumers do not have to expend cognitive resources in trying to retrieve information on current market prices from their long terms memories. This information is likely to be obtained during the search process and utilized directly from working memory. Here, researchers have identified several possible references including normal (typically charged) market price (Urbany and Dickson, 1991) and the lowest market price (Biswas and Sherrell, 1993). These and other similar studies have shown that such market-based references are important predictors of consumers' choices.It is important to note the conceptual distinction between IRP and ERP. While IRP refers to price information that is stored in and retrieved from consumers' long term memories, ERPs are derived from price cues that are either present or evoked at the time of purchase. Specifically, ERPs may represent consumers' beliefs about current market prices that may be formed during the search associated with the purchase process. Such beliefs may include highest price, lowest available price and normal/average market price. Note, however, that there is likely to be significant heterogeneity (due in part to differing levels of involvement) in the extent to which consumers search for the lowest/best price. Such selective and differential exposure is likely to result in differences across individuals in their perceptions of ERPs. It is well established that reference price is a complex, multi-faceted construct, and that reference prices are important in explaining consumers' brand choices (Briesch et al., 1997; Mazumdar et al., 2005). In addition, consumers are likely to use more than one type of reference price to evaluate an offer (Mayhew and Winer, 1992; Rajendran and Tellis, 1994; Chandrashekaran and Jagpal, 1995; Shirai, 2003). However, we know considerably less about the process by which consumers may combine/integrate multiple reference prices. Even less is known about individual differences in consumers' utilization of multiple reference prices.Prior research has acknowledged that different consumer groups (segments) are likely to evoke and use different sets of reference prices (Winer, 1986; Mazumdar and Papatla, 2000). Indeed, prior research has demonstrated that several factors lead to differences in the references used to evaluate a posted price. For example, reference price utilization and price perceptions have been linked to frequency of purchase (Thomas and Menon, 2007), product type (Chandrashekaran and Jagpal, 1995; Lowe and Alpert, 2007) and price consciousness (Alford and Biswas, 2002; Palazon and Delgado, 2009). In summary, prior research supports the claim that different segments may evoke and utilize different sets of reference prices, and suggests that it may not be appropriate to assume that all consumers utilize the same types (and number) of reference prices.As emphasized by Lowe and Alpert (2007) and by Mazumdar et al. (2005), despite some research on the topic, we do not know much about how consumers integrate multiple reference prices. More importantly, we need a better understanding of the factors that may determine which type(s) of reference price(s) are evoked and utilized when making price-related judgments. Along those lines, this research proposes to investigate how the level of involvement leads to differences in the number and types of reference prices that are used in the evaluation process. More specifically, it focuses on how consumers combine and utilize internal reference price (operationalized as expected price) and market-based external reference prices (operationalized as subjective knowledge of normal and lowest market prices)[1]. Consistent with established theory, this research acknowledges that consumers may adjust their internal reference prices based on their perceptions of market-based references. However, the conceptualization adopted here (explained later) also allows for consumers to evaluate posted prices directly against market-based (external) references.Regarding the relative impacts of the two types of reference prices on consumers' evaluations, there is some empirical evidence (Mayhew and Winer, 1992) suggesting that external reference prices may be more influential than internal reference prices. However, that conclusion might be premature because previous research has not paid attention to the moderating roles of individual level variables, for example involvement, that are likely to influence how consumers choose to allocate their cognitive resources - this is likely to impact the relative emphasis that consumers place on the two types of reference prices when evaluating a posted price. Consumers' involvement with a product class has been shown to influence consumers' information processing strategy (Bettman, 1979; Bloch et al., 1986) by affecting the extent of search, the type of information sought, the relative importance of different types of information (e.g. price versus product attribute information), and the motivation to process available information. High involvement is generally associated with product knowledge (familiarity) and the motivation to search for and use relevant information. On the other hand, low involvement has been associated with low product/price knowledge (lack of familiarity), and the lack of motivation to engage in detailed processing of information.Although there is substantial research dealing with how involvement affects consumers' use of product-attribute information and on how involvement affects consumers' responses to marketing communications (advertising), research in the (reference) pricing context is relatively limited. Available evidence reveals that such variables as involvement, prior knowledge, and price consciousness moderate consumers' internal reference prices and their acceptance of retail prices. These variables have been found to affect consumers' internal standards (Kosenko and Rahtz, 1988; Lichtenstein et al., 1988; Biswas and Sherrell, 1993), the widths (i.e. upper and lower limits) of their price acceptance regions (Lichtenstein et al., 1988; Rao and Sieben, 1992), and, finally, their confidence in their reference price estimates (Biswas and Sherrell, 1993). However, we know less about how such factors are likely to influence the ways in which consumers may combine/utilize multiple reference prices. The current study focuses on one such factor, i.e. involvement. This research explores the notion that different reference price utilization strategies unfold for consumers with high and low involvement. The idea that involvement affects which decision route consumers are likely to follow is not new. Extant research on consumers' information processing has shown that involvement affects the way consumers acquire, store, retrieve, and use relevant information to make decisions (Bettman, 1979; Bloch et al., 1986). Compared to those who are not involved, involved consumers are more motivated to search for relevant information. They also possess better knowledge about products and prices than consumers who lack involvement. Therefore, these consumers are more likely to possess well-defined internal standards than low involvement consumers (see also Chandrashekaran and Jagpal, 1995). Consequently, they are likely to be more comfortable evaluating retail prices against their internal (memory-based) reference prices. In contrast, less involved consumers are not as motivated as their involved counterparts to engage in extensive search behavior. In addition, they not likely to engage in detailed processing of available (price) information. As a result, their internal standards are not likely to be well defined, and they are not confident about acting on this information (Chandrashekaran and Jagpal, 1995).Consistent with such theorizing, Biswas and Sherrell (1993) found that high-involvement consumers are indeed more confident of their internal reference price estimates than low-involvement consumers. Therefore, there is some empirical evidence to support the expectation that the level of involvement is likely to moderate consumers' utilization of reference prices.Consistent with the notion that consumers are likely to possess reference prices that are brand specific (Briesch et al., 1997), a brand's own previous prices may be considered to be the most relevant information on which to base a reference price for the brand. Thus, it may be reasonable to expect high involvement consumers' evaluations to be most closely associated with a summary of a brand's past prices, i.e. the expected price construct. Briesch et al. (1997) empirically compared five different model formulations from the literature and confirmed that a model using a temporally generated, brand-specific reference price is the best predictor of consumer choice. However, the authors did not examine individual differences in the utilization of past price information.As discussed earlier, low involvement individuals are by definition not equipped with the knowledge to evaluate price on the basis of relevant factors. Recall that these consumers are not expected to internalize (past) price information, and thus cannot be expected to have well-defined (reliable) price expectations. Therefore, these individuals are likely to evaluate a brand's current price on the basis of more readily accessible perceptions they may have formed of current market prices.In summary, high- and low-involvement consumers are expected to follow different processes in evaluating the same objective price information. Highly involved consumers base their evaluations on internally held beliefs. However, because they are exposed to market price information in the course of their search, it is likely that they utilize this information to update (fine-tune) their internal standards to be consistent with current market information. In contrast, low involvement consumers are not motivated to devote cognitive resources to storing price information in their long-term memories and, therefore, are less confident about their knowledge of past prices. Consequently, they are likely to rely on immediately available external cues (i.e., market-based reference prices) to make their evaluations (as opposed to first updating their internal reference prices and, subsequently utilizing them to evaluate posted prices).More formally, it is expected that:H1. High-involvement consumers evaluate retail prices only against their price expectations (i.e. internal reference price).H2. High-involvement consumers utilize market-based external references only to update their internal beliefs (expected prices).H3. Low involvement consumers evaluate retail prices against one or more market-based (external) reference prices.H4. Low involvement consumers do not utilize internal reference prices (price expectations) in evaluating retail prices.Figure 1 presents the hypothesized model on the processes by which market-based and truly internal reference prices affect consumers' evaluations. Two hundred undergraduate business majors attending a large Northeastern state university participated in this study. The data required for this study was collected in four stages, over a two-week period. The entire data collection task was designed in the form of an experiential exercise.Stage one: initial measures In the first stage, subjects provided general information on their familiarity and involvement with the chosen product, i.e. Levi jeans. This product was chosen because of its relevance to the student subjects, and because no major differences across genders and age groups were expected.Product knowledge Prior research (Rao and Sieben, 1992; Biswas and Sherrell, 1993) has shown that consumers' prior knowledge affects the levels and widths of their price acceptance regions. Product knowledge was measured using a single seven-point item on which subjects indicated their familiarity with the product (brand). Generally, multiple scale items measures are preferred. However, in this study, consumer knowledge (expertise) is not central to the hypotheses. Therefore, no major problems are foreseen.Involvement Subjects' involvement with the product was assessed using the personal involvement inventory (PII) developed by Zaichkowsky (1985). Subjects rated the product on 20 bi-polar adjective scales. These individual items were summed to yield an involvement index for each subject. Later, high- and low-involvement groups were created by performing a median split of the distribution of involvement indices.The shopping exercise After completing the task in the first stage, subjects were given an instruction sheet that described an experiential exercise. In the instructions, subjects were informed that this exercise was intended to demonstrate to them the differences in consumers' shopping behaviors for convenience and shopping goods. The experiential exercise required students to pretend that they were in the market for products listed on the sheet. To mask the actual intent of the study, the sheet listed several products (e.g. toothpaste, laundry detergent, and sneakers) in addition to jeans. To preserve realism, students were instructed to behave as if they were shopping for the products listed (including a pair of jeans), and were asked to engage in behaviors that they would normally engage in when actually buying the products listed (a pair of jeans, toothpaste, laundry detergent, sneakers) for themselves. They were especially asked to use only the information they would normally use when making purchase decisions in the product categories listed. Finally, the subjects were informed that they were expected to come back to class the following week to engage in a class discussion about their experiences and about where they would buy the products listed.Stage two: measuring perceptions of market prices A week after the initial measures were taken, subjects were questioned about the outcomes of their search during a class discussion. At this stage, subjects' perceptions of market prices were measured on two items corresponding to their estimates of the lowest price for which they can purchase the product and its normal (average) market price[2].Stage three: measuring internal reference price At the next class meeting, subjects were asked to provide estimates of their internal reference prices (prices they expected to see on the next purchase occasion). The measures of market-based and internal reference prices were separated in time to avoid any contamination of the reference price measures.Stage four: the decision task At the end of the second week of the exercise, each subject was shown a mock retail ad for a pair of Levi jeans, which included a picture of the product, information on the retail store (a popular retail chain), and a selling price ($34.99). To stick with the main theme of understanding how consumers utilize perceptions of market prices and their reference prices to evaluate a posted price, the ad did not identify that the product was "on sale" and, therefore, it did not include an advertised comparison price or any other semantic cues that are normally associated with comparative price claims[3]. The retail price ($34.99) was chosen for two reasons. First, it reflected the actual market prices in the stores that most students reported they are most likely to visit. Second, it was close to the median of the average reference price (=$36.17) calculated as (Lowest+Normal+Expected)/3. Thus, for approximately half the subjects the offered price was likely to have been above their average reference price, and for the remaining subjects it was likely to have been below their average reference price. Each subject evaluated the posted retail price on three seven-point scale items corresponding to perceived offer value, attractiveness of the deal, and willingness to buy the product at the stated price. Similar items have been used in previous research (see, for example, Chandrashekaran and Jagpal, 1995; Chandrashekaran and Grewal, 2006). The first task was to split the sample into high and low groups based on their involvement with the product category (jeans). A median split of the distribution of PII scores (median involvement=103) yielded high and low involvement groups of approximately equal sizes (nhigh=111 and nlow=114). In addition, the mean involvement scores in the two groups (mean=119.26 and SD=10.50 in the high-involvement group versus mean=83.03 and SD=16.87 in the low involvement group) were significantly different (t=19.27, p<0.001). Finally, an examination of the mean knowledge (familiarity) scores in the two involvement groups (mean familiarity=5.99 and 4.75 for high- and low-involvement groups, respectively) confirmed the premise that, in this product category, high-involvement consumers are significantly more knowledgeable than low-involvement consumers (t=7.91, p<0.01).To test the hypotheses in this study empirically, correlation matrices in each group were analyzed using the structural equations methodology (interested readers may refer to Joreskog and Sorbom, 1988). Figure 1 shows the model structure specified in both groups.The conceptual model shown in Figure 1 allows for the possibility that market-based reference prices (lowest price and normal price) may affect evaluations directly (along with expected price), or indirectly by first influencing expected price, which, in turn, affects how the offer is evaluated. Therefore, the model enables us to investigate the process by which consumers evoke, combine and utilize internal and market-based reference prices. Recall that this is an improvement over Rajendran and Tellis (1994), who examined how both temporal and contextual reference prices influence consumers' evaluations simultaneously. In relation to Figure 1, the hypotheses presented here predict that:* g11, g12, and b21> 0; and g21 and g22=0 in the high-involvement group; and* b21, g11, and g12=0; and g21 and g22>0 in the low-involvement group.As discussed earlier, consumers' evaluations were measured using three scale items corresponding to their perceptions of the value of the offer, the attractiveness of the deal, and their willingness to buy the item at the stated price. It could be argued that the concept of perceived value is related to consumers' perceptions of acquisition utility (AU), while their perceptions of the deal are related to transaction utility (TU) as defined by Thaler (1985). According to Thaler, these are conceptually different from purchase intentions (behavioral responses). However, principal components factor analysis resulted in a single factor and did not reveal conceptual differences between the three measures. Table I compares the two involvement groups on their reference price estimates and evaluations of the offer. It is interesting to note that there is no significant difference in reference price estimates across the two involvement groups (this point is also discussed later in the manuscript). Furthermore, there are no major differences between the two groups in their evaluations of the offer, except that high involvement consumers perceive the deal to be moderately more attractive than low involvement consumers (p<=0.10)[4].However, one cannot conclude on the basis of these results alone that high- and low-involvement consumers are homogeneous, thus justifying pooled analyses. Although consumers' final evaluations are important, it is crucial that we examine the process (route) by which high and low involvement consumers arrive at the same destination. In addition, consumers' final evaluations are not central to this study. Rather, the primary objective of this study is to examine heterogeneity in consumers' utilization of market-based and truly internal reference prices to evaluate the same objective price information. Indeed, although consumers do not differ significantly in their final evaluations of the offer, this paper seeks to uncover heterogeneity in reference price utilization, i.e. the means by which high and low involvement consumers reach the end (their final evaluations).Table II shows the estimated models in each group along with several fit indices. It is clear that the specified model structure fits the data well in both groups (kh2 with six df=9.59, p=0.143 in the high involvement group; and kh2 with six df=6.09, p=0.413 in the low-involvement group). In addition, the structural models explain a significant proportion of the variance (R2=0.51 and 0.53 in the high- and low-involvement groups respectively). For both groups, the RMSEA (Steiger, 1985) and other fit indices indicate satisfactory model fit. All indicated paths are of the predicted sign, and are significant at a=0.01. All non-significant effects, i.e. paths that are not significantly different from zero, are indicated as "NS".From Table II, it is clear that consumers who are highly involved with the product evaluate the retail price against a single internal reference price, expected price (b21 is significant at p<=0.01), but do not utilize any market-based references to make evaluations (note that g21 and g22 are not significantly different from zero). These findings strongly support the hypothesis (H1) for consumers' utilization of truly internal reference prices under high involvement.H2 suggests that involved consumers are likely to use market-based references primarily to update their internal references. Consistent with this hypothesis, Table II shows that normal price (a market-based reference price) has a significant positive effect on the internal standards of involved consumers (g12=0.727 is significant at a=0.01). This finding is consistent with the view that involved consumers process price information sequentially, i.e. an estimate of the normal/average market price is used to update (fine-tune) the internal standard, which, in turn, is used to evaluate the retail price. Contrary to expectation, these consumers do not utilize their perceptions of the lowest market price (g11 is not significant). Thus, overall, H2 is partially supported.In contrast to high-involvement consumers, those in the low-involvement group do not use their internal standards to evaluate the retail price (b21=0). As shown in Table II, these consumers base their evaluations of the retail price on market-based references (g21=0.257 and g22=0.211 are both significant at a=0.01). Thus, H3 and H4 are supported. In conjunction, these findings support the conclusion that low levels of involved facilitate simultaneous utilization of several market-based reference prices to evaluate retail prices[5].It is interesting to note that highly involved consumers utilize only normal price, but not their perceptions of lowest price, in formulating their price expectations. Clearly, further research is required to be able to offer a convincing explanation to this observation. One possible explanation may be that these consumers may generally be more aware that the observed lowest price (a single data point) is likely to be a temporary promotional offer that does not reflect the "normal" price for the product/category. Consequently, these consumers may discount this information and rely to a greater extent on more comprehensive information, for example overall estimate of normal/average price. Urbany and Dickson (1991) found that consumers' estimates of normal prices are reasonable surrogates for their reference prices. However, the authors did not investigate individual differences in consumers' perceptions of normal/average prices. It is likely that one segment (e.g. high-involvement consumers) has more accurate estimates of normal prices than another segment (e.g. low-involvement consumers). Hopefully, future research will investigate and shed some light on this and other similar issues dealing with individual differences in price perceptions and evaluation.In summary, the results reveal that although the two types of consumers (high- and low-involvement) may have similar reference prices and may even evaluate an offer similarly, they do so via distinct routes, i.e. by utilizing different types of reference prices. Specifically, whereas highly involved consumers rely solely on their internal standards to evaluate retail prices, those who are less involved with the product category utilize several market-based references to judge the attractiveness of retail prices. A general discussion of the results along with implications and limitations of the study follows. This study intended to draw attention to the heterogeneity in consumers' utilization of multiple reference prices. The results obtained here offer some evidence that involvement affects consumers' utilization of reference prices. Although involvement does not affect the levels of consumers' internal standards or their final evaluations, it plays a significant role in the way consumers utilize these standards to evaluate the same objective price information. Highly involved consumers, who are more knowledgeable about the product class, evaluate offers against a single internal standard (expected price), which supports previous researchers' (e.g. Winer, 1986; Kalwani et al., 1990) in their use of expected price as a reference price to explain consumers' choices. However, low-involvement consumers do not use expected price to evaluate offers. Rather, they use two market-based references (lowest price and normal price) to determine the overall value of an offer. Thus, involvement affects the number and types of reference prices used in the evaluation process.The results are also consistent with information-processing theory (Bettman, 1979), which advocates that high involvement encourages a deeper and more complex processing strategy than low involvement, which supports less complex processing and the use of easily available information. Highly involved consumers retrieve price information stored in long-term memory and fine tune this internal standard (expected price) using current market price information (estimated normal/average market price). Thus, high involvement supports a hierarchical model in which market-based and truly internal reference prices are used sequentially. This conclusion is consistent with the underlying process implied in vast amount of research on reference price formation. In contrast, low-involvement consumers do not possess the motivation to engage in such a complex evaluation process. They simply utilize the more readily available market-based references and evaluate offers in a single step.It is only appropriate to point out that some of the results obtained here are inconsistent with previous research findings and call for further investigation. For example, Lichtenstein et al. (1988) found that consumers' price acceptability level (i.e. internal reference price level) is positively related to involvement. However, no such differences were found here. One possibility is that the results are being moderated by consumers' knowledge in this category. Rao and Sieben (1992) found that the upper and lower limits of consumers' price-acceptance regions increase with knowledge up to a point, and then levels off. It is likely that, given the nature of product category used in this study (jeans), most consumers were familiar with the product. Indeed, an examination of the mean knowledge scores revealed that both high- and low-involvement consumers lie in the upper half of the scale (means for high- and low-involvement groups=5.99 and 4.75 on seven-point scales), which may account for the lack of difference in their reference prices. Furthermore, product knowledge was assessed using a single item that measured subjects' familiarity with the product category. Additional research is needed to investigate this issue further.Another possible limitation of this study is the omission of several internal and market-based reference prices (e.g. fair price and highest price) that have been mentioned in the literature. It might be useful for future research to include more definitions of reference price to investigate whether the premise of this research holds. Finally, it may be useful to examine the roles of other moderating factors (e.g. gender, product experience, price consciousness, etc.).This research has important implications for both marketing research and practice. It is important for researchers to identify other sources of heterogeneity and their impact on consumers' construction and utilization of reference prices. The finding that high- and low-involvement consumers are different in the types of reference prices used and in the way they use this information to evaluate offers strongly supports the need for using different model structures to represent these sub-populations. From a practical standpoint, this study underscores the importance of including individual differences in strategies designed to affect consumers' perceptions of retail prices. For example, marketers might segment the market on the basis of reference price utilization and design different strategies to obtain optimal results. It is hoped that this study will stimulate further research in the area so that we may gain a better understanding of how consumers process and respond to retail pricing strategies. Such knowledge will undoubtedly be useful in designing more effective and efficient marketing strategies. Opens in a new window.Figure 1 Process of reference price effects on evaluations Opens in a new window.Table I Summary statistics Opens in a new window.Table II Estimated models
|
- A total of 200 students were enrolled to participate in a realistic shopping experience over a two-week period. In the course of the study, subjects were asked to provide information on their reference prices. They were also asked to evaluate an advertised offer for a pair of jeans. The data were analyzed using a structural equations methodology.
|
[SECTION: Findings] There is ample empirical evidence in support of the important role that reference prices (both internal and external) play in determining consumers' evaluations of posted prices (for comprehensive reviews of the many operationalizations of reference price, see Briesch et al., 1997; Mazumdar et al., 2005). Collectively, the body of evidence strongly supports the conclusion that consumers rely on (internal and/or external) reference prices to make judgments about perceived value, perceived fairness and perceived expensiveness that, in turn, influence purchase intentions. The inescapable conclusion for practitioners is that they must pay attention to understanding, measuring and managing consumers' reference prices to ensure that their prices are evaluated favorably. However, the literature offers many different operationalizations of reference price. While some definitions of reference price are based on past prices or on consumers' knowledge of past/current market prices, for example highest price, normal price, average price, lowest price and expected price (Jacobson and Obermiller, 1990; Kalwani et al., 1990; Chandrashekaran and Jagpal, 1995), others are related to subjective notions of fairness (Herrmann et al., 2007; Xia et al., 2004). The most commonly accepted view is that reference price is a brand-specific construct that represents an expectation based primarily on some summary of a brand's own past prices (Briesch et al., 1997). Indeed, based on a comprehensive review of the literature, Mazumdar et al. (2005, p. 86) concluded that "the strongest determinant of a consumer's IRP is the prior prices he or she observes". In addition to such a temporal view of reference price, previous research has offered other potential standards against which consumers may evaluate posted prices. For example, Rajendran and Tellis (1994) introduced the idea of a contextual component that is based on an average of current prices of competing brands (see also Rajendran, 2009). Such varied operationalizations underscore that reference price is a complex, multi-faceted construct, and that no single operationalization is likely to capture it in its entirety. In addition, there are likely to be individual differences with respect to both the number and types of references evoked in particular situations. Despite overwhelming experimental and empirical evidence that reference prices play a crucial role in determining perceived value, we know considerably less about the process by which consumers may combine/integrate multiple reference prices. Even less is known about individual differences in consumers' utilization of multiple reference prices. It is imperative that managers not only understand that reference prices influence consumers' evaluations, but they may benefit from knowledge about which specific reference prices affect consumers' purchase intentions and the process by which this influence may occur. Such information could be vital in planning appropriate price communication strategies designed to affect specific outcomes, for example raising the appropriate reference price(s) to enhance consumers' perceptions of offer-value that will, in turn, increase the likelihood of purchase. The motivation for this paper comes from recognizing two main gaps in the literature. First, despite multiple definitions of reference price, prior research has not adequately addressed heterogeneity in the types of reference prices that are evoked and the ways in which consumers might process available price information in relation to multiple reference prices. This paper considers how one specific source of heterogeneity, involvement, influences consumers' utilization of (multiple) reference prices. Research on consumer information processing (see Bettman, 1979) suggests that several individual differences (e.g. involvement with the product and/or the purchase) may influence the type of information considered and the extent to which consumers are willing to elaborate on available information. This research attempts to fill a gap in the literature by investigating the role of involvement in the context of reference prices.Second, there have been very few attempts to examine the inter-relationship between different reference prices and their combined influence on consumers' evaluations. Previous research (Chandrashekaran and Jagpal, 1995; Rajendran and Tellis, 1994; Mayhew and Winer, 1992) has examined the simultaneous and independent influences of internal and external references on consumers' choices. However, questions about the inter-relationships between different types of reference prices remain. This study attempts to extend current thinking to include sequential effects of reference prices on consumers' evaluations. It is generally established in the literature that consumers may utilize some external reference prices to adjust internal reference prices, which in turn affects evaluation. However, some research (e.g. Rajendran and Tellis, 1994) has also shown that consumers may utilize certain external reference prices in conjunction with internal reference price. Hence, there is a need to validate the process empirically. In addition, the interrelationship between specific internal and external references prices along with the role of involvement has not been empirically investigated.In summary, this study proposes to extend the literature by attempting to answer two questions:1. Do all consumers evaluate a given offer against the same set of reference prices?2. Do all consumers utilize internal and external reference prices in the same way (i.e. using the same underlying process)? When consumers evaluate retail prices, they may do so against several reference prices. Of these, some are internal and others are external. The term "internal reference price" (henceforth referred to as IRP) has generally been used to refer to price information held in consumers' long-term memories. For example, Winer (1986) proposed that consumers use past price information to form expectations of what the retail price is likely to be on the next purchase occasion. As described in the literature, these price expectations represent internally held reference prices, and they are based on past price information that must be retrieved from consumers' long-term memories during the purchase occasion. In fact, most empirical studies have based their measures of internal reference price on such a temporally generated construct (Kalwani et al., 1990; Mayhew and Winer, 1992; Rajendran and Tellis, 1994; Briesch et al., 1997). The literature offers compelling empirical evidence that such (memory-based) internal reference prices are crucial in predicting consumers' evaluations (of retail prices) and subsequent choice behavior.In contrast to internal reference price, external reference prices (henceforth referred to as ERP) refer to comparative (competitive) price information that is available in the purchase environment, i.e. prices that consumers are likely to encounter during the search process. Thus, consumers do not have to expend cognitive resources in trying to retrieve information on current market prices from their long terms memories. This information is likely to be obtained during the search process and utilized directly from working memory. Here, researchers have identified several possible references including normal (typically charged) market price (Urbany and Dickson, 1991) and the lowest market price (Biswas and Sherrell, 1993). These and other similar studies have shown that such market-based references are important predictors of consumers' choices.It is important to note the conceptual distinction between IRP and ERP. While IRP refers to price information that is stored in and retrieved from consumers' long term memories, ERPs are derived from price cues that are either present or evoked at the time of purchase. Specifically, ERPs may represent consumers' beliefs about current market prices that may be formed during the search associated with the purchase process. Such beliefs may include highest price, lowest available price and normal/average market price. Note, however, that there is likely to be significant heterogeneity (due in part to differing levels of involvement) in the extent to which consumers search for the lowest/best price. Such selective and differential exposure is likely to result in differences across individuals in their perceptions of ERPs. It is well established that reference price is a complex, multi-faceted construct, and that reference prices are important in explaining consumers' brand choices (Briesch et al., 1997; Mazumdar et al., 2005). In addition, consumers are likely to use more than one type of reference price to evaluate an offer (Mayhew and Winer, 1992; Rajendran and Tellis, 1994; Chandrashekaran and Jagpal, 1995; Shirai, 2003). However, we know considerably less about the process by which consumers may combine/integrate multiple reference prices. Even less is known about individual differences in consumers' utilization of multiple reference prices.Prior research has acknowledged that different consumer groups (segments) are likely to evoke and use different sets of reference prices (Winer, 1986; Mazumdar and Papatla, 2000). Indeed, prior research has demonstrated that several factors lead to differences in the references used to evaluate a posted price. For example, reference price utilization and price perceptions have been linked to frequency of purchase (Thomas and Menon, 2007), product type (Chandrashekaran and Jagpal, 1995; Lowe and Alpert, 2007) and price consciousness (Alford and Biswas, 2002; Palazon and Delgado, 2009). In summary, prior research supports the claim that different segments may evoke and utilize different sets of reference prices, and suggests that it may not be appropriate to assume that all consumers utilize the same types (and number) of reference prices.As emphasized by Lowe and Alpert (2007) and by Mazumdar et al. (2005), despite some research on the topic, we do not know much about how consumers integrate multiple reference prices. More importantly, we need a better understanding of the factors that may determine which type(s) of reference price(s) are evoked and utilized when making price-related judgments. Along those lines, this research proposes to investigate how the level of involvement leads to differences in the number and types of reference prices that are used in the evaluation process. More specifically, it focuses on how consumers combine and utilize internal reference price (operationalized as expected price) and market-based external reference prices (operationalized as subjective knowledge of normal and lowest market prices)[1]. Consistent with established theory, this research acknowledges that consumers may adjust their internal reference prices based on their perceptions of market-based references. However, the conceptualization adopted here (explained later) also allows for consumers to evaluate posted prices directly against market-based (external) references.Regarding the relative impacts of the two types of reference prices on consumers' evaluations, there is some empirical evidence (Mayhew and Winer, 1992) suggesting that external reference prices may be more influential than internal reference prices. However, that conclusion might be premature because previous research has not paid attention to the moderating roles of individual level variables, for example involvement, that are likely to influence how consumers choose to allocate their cognitive resources - this is likely to impact the relative emphasis that consumers place on the two types of reference prices when evaluating a posted price. Consumers' involvement with a product class has been shown to influence consumers' information processing strategy (Bettman, 1979; Bloch et al., 1986) by affecting the extent of search, the type of information sought, the relative importance of different types of information (e.g. price versus product attribute information), and the motivation to process available information. High involvement is generally associated with product knowledge (familiarity) and the motivation to search for and use relevant information. On the other hand, low involvement has been associated with low product/price knowledge (lack of familiarity), and the lack of motivation to engage in detailed processing of information.Although there is substantial research dealing with how involvement affects consumers' use of product-attribute information and on how involvement affects consumers' responses to marketing communications (advertising), research in the (reference) pricing context is relatively limited. Available evidence reveals that such variables as involvement, prior knowledge, and price consciousness moderate consumers' internal reference prices and their acceptance of retail prices. These variables have been found to affect consumers' internal standards (Kosenko and Rahtz, 1988; Lichtenstein et al., 1988; Biswas and Sherrell, 1993), the widths (i.e. upper and lower limits) of their price acceptance regions (Lichtenstein et al., 1988; Rao and Sieben, 1992), and, finally, their confidence in their reference price estimates (Biswas and Sherrell, 1993). However, we know less about how such factors are likely to influence the ways in which consumers may combine/utilize multiple reference prices. The current study focuses on one such factor, i.e. involvement. This research explores the notion that different reference price utilization strategies unfold for consumers with high and low involvement. The idea that involvement affects which decision route consumers are likely to follow is not new. Extant research on consumers' information processing has shown that involvement affects the way consumers acquire, store, retrieve, and use relevant information to make decisions (Bettman, 1979; Bloch et al., 1986). Compared to those who are not involved, involved consumers are more motivated to search for relevant information. They also possess better knowledge about products and prices than consumers who lack involvement. Therefore, these consumers are more likely to possess well-defined internal standards than low involvement consumers (see also Chandrashekaran and Jagpal, 1995). Consequently, they are likely to be more comfortable evaluating retail prices against their internal (memory-based) reference prices. In contrast, less involved consumers are not as motivated as their involved counterparts to engage in extensive search behavior. In addition, they not likely to engage in detailed processing of available (price) information. As a result, their internal standards are not likely to be well defined, and they are not confident about acting on this information (Chandrashekaran and Jagpal, 1995).Consistent with such theorizing, Biswas and Sherrell (1993) found that high-involvement consumers are indeed more confident of their internal reference price estimates than low-involvement consumers. Therefore, there is some empirical evidence to support the expectation that the level of involvement is likely to moderate consumers' utilization of reference prices.Consistent with the notion that consumers are likely to possess reference prices that are brand specific (Briesch et al., 1997), a brand's own previous prices may be considered to be the most relevant information on which to base a reference price for the brand. Thus, it may be reasonable to expect high involvement consumers' evaluations to be most closely associated with a summary of a brand's past prices, i.e. the expected price construct. Briesch et al. (1997) empirically compared five different model formulations from the literature and confirmed that a model using a temporally generated, brand-specific reference price is the best predictor of consumer choice. However, the authors did not examine individual differences in the utilization of past price information.As discussed earlier, low involvement individuals are by definition not equipped with the knowledge to evaluate price on the basis of relevant factors. Recall that these consumers are not expected to internalize (past) price information, and thus cannot be expected to have well-defined (reliable) price expectations. Therefore, these individuals are likely to evaluate a brand's current price on the basis of more readily accessible perceptions they may have formed of current market prices.In summary, high- and low-involvement consumers are expected to follow different processes in evaluating the same objective price information. Highly involved consumers base their evaluations on internally held beliefs. However, because they are exposed to market price information in the course of their search, it is likely that they utilize this information to update (fine-tune) their internal standards to be consistent with current market information. In contrast, low involvement consumers are not motivated to devote cognitive resources to storing price information in their long-term memories and, therefore, are less confident about their knowledge of past prices. Consequently, they are likely to rely on immediately available external cues (i.e., market-based reference prices) to make their evaluations (as opposed to first updating their internal reference prices and, subsequently utilizing them to evaluate posted prices).More formally, it is expected that:H1. High-involvement consumers evaluate retail prices only against their price expectations (i.e. internal reference price).H2. High-involvement consumers utilize market-based external references only to update their internal beliefs (expected prices).H3. Low involvement consumers evaluate retail prices against one or more market-based (external) reference prices.H4. Low involvement consumers do not utilize internal reference prices (price expectations) in evaluating retail prices.Figure 1 presents the hypothesized model on the processes by which market-based and truly internal reference prices affect consumers' evaluations. Two hundred undergraduate business majors attending a large Northeastern state university participated in this study. The data required for this study was collected in four stages, over a two-week period. The entire data collection task was designed in the form of an experiential exercise.Stage one: initial measures In the first stage, subjects provided general information on their familiarity and involvement with the chosen product, i.e. Levi jeans. This product was chosen because of its relevance to the student subjects, and because no major differences across genders and age groups were expected.Product knowledge Prior research (Rao and Sieben, 1992; Biswas and Sherrell, 1993) has shown that consumers' prior knowledge affects the levels and widths of their price acceptance regions. Product knowledge was measured using a single seven-point item on which subjects indicated their familiarity with the product (brand). Generally, multiple scale items measures are preferred. However, in this study, consumer knowledge (expertise) is not central to the hypotheses. Therefore, no major problems are foreseen.Involvement Subjects' involvement with the product was assessed using the personal involvement inventory (PII) developed by Zaichkowsky (1985). Subjects rated the product on 20 bi-polar adjective scales. These individual items were summed to yield an involvement index for each subject. Later, high- and low-involvement groups were created by performing a median split of the distribution of involvement indices.The shopping exercise After completing the task in the first stage, subjects were given an instruction sheet that described an experiential exercise. In the instructions, subjects were informed that this exercise was intended to demonstrate to them the differences in consumers' shopping behaviors for convenience and shopping goods. The experiential exercise required students to pretend that they were in the market for products listed on the sheet. To mask the actual intent of the study, the sheet listed several products (e.g. toothpaste, laundry detergent, and sneakers) in addition to jeans. To preserve realism, students were instructed to behave as if they were shopping for the products listed (including a pair of jeans), and were asked to engage in behaviors that they would normally engage in when actually buying the products listed (a pair of jeans, toothpaste, laundry detergent, sneakers) for themselves. They were especially asked to use only the information they would normally use when making purchase decisions in the product categories listed. Finally, the subjects were informed that they were expected to come back to class the following week to engage in a class discussion about their experiences and about where they would buy the products listed.Stage two: measuring perceptions of market prices A week after the initial measures were taken, subjects were questioned about the outcomes of their search during a class discussion. At this stage, subjects' perceptions of market prices were measured on two items corresponding to their estimates of the lowest price for which they can purchase the product and its normal (average) market price[2].Stage three: measuring internal reference price At the next class meeting, subjects were asked to provide estimates of their internal reference prices (prices they expected to see on the next purchase occasion). The measures of market-based and internal reference prices were separated in time to avoid any contamination of the reference price measures.Stage four: the decision task At the end of the second week of the exercise, each subject was shown a mock retail ad for a pair of Levi jeans, which included a picture of the product, information on the retail store (a popular retail chain), and a selling price ($34.99). To stick with the main theme of understanding how consumers utilize perceptions of market prices and their reference prices to evaluate a posted price, the ad did not identify that the product was "on sale" and, therefore, it did not include an advertised comparison price or any other semantic cues that are normally associated with comparative price claims[3]. The retail price ($34.99) was chosen for two reasons. First, it reflected the actual market prices in the stores that most students reported they are most likely to visit. Second, it was close to the median of the average reference price (=$36.17) calculated as (Lowest+Normal+Expected)/3. Thus, for approximately half the subjects the offered price was likely to have been above their average reference price, and for the remaining subjects it was likely to have been below their average reference price. Each subject evaluated the posted retail price on three seven-point scale items corresponding to perceived offer value, attractiveness of the deal, and willingness to buy the product at the stated price. Similar items have been used in previous research (see, for example, Chandrashekaran and Jagpal, 1995; Chandrashekaran and Grewal, 2006). The first task was to split the sample into high and low groups based on their involvement with the product category (jeans). A median split of the distribution of PII scores (median involvement=103) yielded high and low involvement groups of approximately equal sizes (nhigh=111 and nlow=114). In addition, the mean involvement scores in the two groups (mean=119.26 and SD=10.50 in the high-involvement group versus mean=83.03 and SD=16.87 in the low involvement group) were significantly different (t=19.27, p<0.001). Finally, an examination of the mean knowledge (familiarity) scores in the two involvement groups (mean familiarity=5.99 and 4.75 for high- and low-involvement groups, respectively) confirmed the premise that, in this product category, high-involvement consumers are significantly more knowledgeable than low-involvement consumers (t=7.91, p<0.01).To test the hypotheses in this study empirically, correlation matrices in each group were analyzed using the structural equations methodology (interested readers may refer to Joreskog and Sorbom, 1988). Figure 1 shows the model structure specified in both groups.The conceptual model shown in Figure 1 allows for the possibility that market-based reference prices (lowest price and normal price) may affect evaluations directly (along with expected price), or indirectly by first influencing expected price, which, in turn, affects how the offer is evaluated. Therefore, the model enables us to investigate the process by which consumers evoke, combine and utilize internal and market-based reference prices. Recall that this is an improvement over Rajendran and Tellis (1994), who examined how both temporal and contextual reference prices influence consumers' evaluations simultaneously. In relation to Figure 1, the hypotheses presented here predict that:* g11, g12, and b21> 0; and g21 and g22=0 in the high-involvement group; and* b21, g11, and g12=0; and g21 and g22>0 in the low-involvement group.As discussed earlier, consumers' evaluations were measured using three scale items corresponding to their perceptions of the value of the offer, the attractiveness of the deal, and their willingness to buy the item at the stated price. It could be argued that the concept of perceived value is related to consumers' perceptions of acquisition utility (AU), while their perceptions of the deal are related to transaction utility (TU) as defined by Thaler (1985). According to Thaler, these are conceptually different from purchase intentions (behavioral responses). However, principal components factor analysis resulted in a single factor and did not reveal conceptual differences between the three measures. Table I compares the two involvement groups on their reference price estimates and evaluations of the offer. It is interesting to note that there is no significant difference in reference price estimates across the two involvement groups (this point is also discussed later in the manuscript). Furthermore, there are no major differences between the two groups in their evaluations of the offer, except that high involvement consumers perceive the deal to be moderately more attractive than low involvement consumers (p<=0.10)[4].However, one cannot conclude on the basis of these results alone that high- and low-involvement consumers are homogeneous, thus justifying pooled analyses. Although consumers' final evaluations are important, it is crucial that we examine the process (route) by which high and low involvement consumers arrive at the same destination. In addition, consumers' final evaluations are not central to this study. Rather, the primary objective of this study is to examine heterogeneity in consumers' utilization of market-based and truly internal reference prices to evaluate the same objective price information. Indeed, although consumers do not differ significantly in their final evaluations of the offer, this paper seeks to uncover heterogeneity in reference price utilization, i.e. the means by which high and low involvement consumers reach the end (their final evaluations).Table II shows the estimated models in each group along with several fit indices. It is clear that the specified model structure fits the data well in both groups (kh2 with six df=9.59, p=0.143 in the high involvement group; and kh2 with six df=6.09, p=0.413 in the low-involvement group). In addition, the structural models explain a significant proportion of the variance (R2=0.51 and 0.53 in the high- and low-involvement groups respectively). For both groups, the RMSEA (Steiger, 1985) and other fit indices indicate satisfactory model fit. All indicated paths are of the predicted sign, and are significant at a=0.01. All non-significant effects, i.e. paths that are not significantly different from zero, are indicated as "NS".From Table II, it is clear that consumers who are highly involved with the product evaluate the retail price against a single internal reference price, expected price (b21 is significant at p<=0.01), but do not utilize any market-based references to make evaluations (note that g21 and g22 are not significantly different from zero). These findings strongly support the hypothesis (H1) for consumers' utilization of truly internal reference prices under high involvement.H2 suggests that involved consumers are likely to use market-based references primarily to update their internal references. Consistent with this hypothesis, Table II shows that normal price (a market-based reference price) has a significant positive effect on the internal standards of involved consumers (g12=0.727 is significant at a=0.01). This finding is consistent with the view that involved consumers process price information sequentially, i.e. an estimate of the normal/average market price is used to update (fine-tune) the internal standard, which, in turn, is used to evaluate the retail price. Contrary to expectation, these consumers do not utilize their perceptions of the lowest market price (g11 is not significant). Thus, overall, H2 is partially supported.In contrast to high-involvement consumers, those in the low-involvement group do not use their internal standards to evaluate the retail price (b21=0). As shown in Table II, these consumers base their evaluations of the retail price on market-based references (g21=0.257 and g22=0.211 are both significant at a=0.01). Thus, H3 and H4 are supported. In conjunction, these findings support the conclusion that low levels of involved facilitate simultaneous utilization of several market-based reference prices to evaluate retail prices[5].It is interesting to note that highly involved consumers utilize only normal price, but not their perceptions of lowest price, in formulating their price expectations. Clearly, further research is required to be able to offer a convincing explanation to this observation. One possible explanation may be that these consumers may generally be more aware that the observed lowest price (a single data point) is likely to be a temporary promotional offer that does not reflect the "normal" price for the product/category. Consequently, these consumers may discount this information and rely to a greater extent on more comprehensive information, for example overall estimate of normal/average price. Urbany and Dickson (1991) found that consumers' estimates of normal prices are reasonable surrogates for their reference prices. However, the authors did not investigate individual differences in consumers' perceptions of normal/average prices. It is likely that one segment (e.g. high-involvement consumers) has more accurate estimates of normal prices than another segment (e.g. low-involvement consumers). Hopefully, future research will investigate and shed some light on this and other similar issues dealing with individual differences in price perceptions and evaluation.In summary, the results reveal that although the two types of consumers (high- and low-involvement) may have similar reference prices and may even evaluate an offer similarly, they do so via distinct routes, i.e. by utilizing different types of reference prices. Specifically, whereas highly involved consumers rely solely on their internal standards to evaluate retail prices, those who are less involved with the product category utilize several market-based references to judge the attractiveness of retail prices. A general discussion of the results along with implications and limitations of the study follows. This study intended to draw attention to the heterogeneity in consumers' utilization of multiple reference prices. The results obtained here offer some evidence that involvement affects consumers' utilization of reference prices. Although involvement does not affect the levels of consumers' internal standards or their final evaluations, it plays a significant role in the way consumers utilize these standards to evaluate the same objective price information. Highly involved consumers, who are more knowledgeable about the product class, evaluate offers against a single internal standard (expected price), which supports previous researchers' (e.g. Winer, 1986; Kalwani et al., 1990) in their use of expected price as a reference price to explain consumers' choices. However, low-involvement consumers do not use expected price to evaluate offers. Rather, they use two market-based references (lowest price and normal price) to determine the overall value of an offer. Thus, involvement affects the number and types of reference prices used in the evaluation process.The results are also consistent with information-processing theory (Bettman, 1979), which advocates that high involvement encourages a deeper and more complex processing strategy than low involvement, which supports less complex processing and the use of easily available information. Highly involved consumers retrieve price information stored in long-term memory and fine tune this internal standard (expected price) using current market price information (estimated normal/average market price). Thus, high involvement supports a hierarchical model in which market-based and truly internal reference prices are used sequentially. This conclusion is consistent with the underlying process implied in vast amount of research on reference price formation. In contrast, low-involvement consumers do not possess the motivation to engage in such a complex evaluation process. They simply utilize the more readily available market-based references and evaluate offers in a single step.It is only appropriate to point out that some of the results obtained here are inconsistent with previous research findings and call for further investigation. For example, Lichtenstein et al. (1988) found that consumers' price acceptability level (i.e. internal reference price level) is positively related to involvement. However, no such differences were found here. One possibility is that the results are being moderated by consumers' knowledge in this category. Rao and Sieben (1992) found that the upper and lower limits of consumers' price-acceptance regions increase with knowledge up to a point, and then levels off. It is likely that, given the nature of product category used in this study (jeans), most consumers were familiar with the product. Indeed, an examination of the mean knowledge scores revealed that both high- and low-involvement consumers lie in the upper half of the scale (means for high- and low-involvement groups=5.99 and 4.75 on seven-point scales), which may account for the lack of difference in their reference prices. Furthermore, product knowledge was assessed using a single item that measured subjects' familiarity with the product category. Additional research is needed to investigate this issue further.Another possible limitation of this study is the omission of several internal and market-based reference prices (e.g. fair price and highest price) that have been mentioned in the literature. It might be useful for future research to include more definitions of reference price to investigate whether the premise of this research holds. Finally, it may be useful to examine the roles of other moderating factors (e.g. gender, product experience, price consciousness, etc.).This research has important implications for both marketing research and practice. It is important for researchers to identify other sources of heterogeneity and their impact on consumers' construction and utilization of reference prices. The finding that high- and low-involvement consumers are different in the types of reference prices used and in the way they use this information to evaluate offers strongly supports the need for using different model structures to represent these sub-populations. From a practical standpoint, this study underscores the importance of including individual differences in strategies designed to affect consumers' perceptions of retail prices. For example, marketers might segment the market on the basis of reference price utilization and design different strategies to obtain optimal results. It is hoped that this study will stimulate further research in the area so that we may gain a better understanding of how consumers process and respond to retail pricing strategies. Such knowledge will undoubtedly be useful in designing more effective and efficient marketing strategies. Opens in a new window.Figure 1 Process of reference price effects on evaluations Opens in a new window.Table I Summary statistics Opens in a new window.Table II Estimated models
|
- Under high involvement, consumers utilize perceived normal price, an external market-based reference price, to adjust their internal standards, which in turn is used to evaluate retail price. In contrast, low-involvement consumers do not use their internal standards. Rather, they evaluate retail prices directly against external market-based references.
|
[SECTION: Value] There is ample empirical evidence in support of the important role that reference prices (both internal and external) play in determining consumers' evaluations of posted prices (for comprehensive reviews of the many operationalizations of reference price, see Briesch et al., 1997; Mazumdar et al., 2005). Collectively, the body of evidence strongly supports the conclusion that consumers rely on (internal and/or external) reference prices to make judgments about perceived value, perceived fairness and perceived expensiveness that, in turn, influence purchase intentions. The inescapable conclusion for practitioners is that they must pay attention to understanding, measuring and managing consumers' reference prices to ensure that their prices are evaluated favorably. However, the literature offers many different operationalizations of reference price. While some definitions of reference price are based on past prices or on consumers' knowledge of past/current market prices, for example highest price, normal price, average price, lowest price and expected price (Jacobson and Obermiller, 1990; Kalwani et al., 1990; Chandrashekaran and Jagpal, 1995), others are related to subjective notions of fairness (Herrmann et al., 2007; Xia et al., 2004). The most commonly accepted view is that reference price is a brand-specific construct that represents an expectation based primarily on some summary of a brand's own past prices (Briesch et al., 1997). Indeed, based on a comprehensive review of the literature, Mazumdar et al. (2005, p. 86) concluded that "the strongest determinant of a consumer's IRP is the prior prices he or she observes". In addition to such a temporal view of reference price, previous research has offered other potential standards against which consumers may evaluate posted prices. For example, Rajendran and Tellis (1994) introduced the idea of a contextual component that is based on an average of current prices of competing brands (see also Rajendran, 2009). Such varied operationalizations underscore that reference price is a complex, multi-faceted construct, and that no single operationalization is likely to capture it in its entirety. In addition, there are likely to be individual differences with respect to both the number and types of references evoked in particular situations. Despite overwhelming experimental and empirical evidence that reference prices play a crucial role in determining perceived value, we know considerably less about the process by which consumers may combine/integrate multiple reference prices. Even less is known about individual differences in consumers' utilization of multiple reference prices. It is imperative that managers not only understand that reference prices influence consumers' evaluations, but they may benefit from knowledge about which specific reference prices affect consumers' purchase intentions and the process by which this influence may occur. Such information could be vital in planning appropriate price communication strategies designed to affect specific outcomes, for example raising the appropriate reference price(s) to enhance consumers' perceptions of offer-value that will, in turn, increase the likelihood of purchase. The motivation for this paper comes from recognizing two main gaps in the literature. First, despite multiple definitions of reference price, prior research has not adequately addressed heterogeneity in the types of reference prices that are evoked and the ways in which consumers might process available price information in relation to multiple reference prices. This paper considers how one specific source of heterogeneity, involvement, influences consumers' utilization of (multiple) reference prices. Research on consumer information processing (see Bettman, 1979) suggests that several individual differences (e.g. involvement with the product and/or the purchase) may influence the type of information considered and the extent to which consumers are willing to elaborate on available information. This research attempts to fill a gap in the literature by investigating the role of involvement in the context of reference prices.Second, there have been very few attempts to examine the inter-relationship between different reference prices and their combined influence on consumers' evaluations. Previous research (Chandrashekaran and Jagpal, 1995; Rajendran and Tellis, 1994; Mayhew and Winer, 1992) has examined the simultaneous and independent influences of internal and external references on consumers' choices. However, questions about the inter-relationships between different types of reference prices remain. This study attempts to extend current thinking to include sequential effects of reference prices on consumers' evaluations. It is generally established in the literature that consumers may utilize some external reference prices to adjust internal reference prices, which in turn affects evaluation. However, some research (e.g. Rajendran and Tellis, 1994) has also shown that consumers may utilize certain external reference prices in conjunction with internal reference price. Hence, there is a need to validate the process empirically. In addition, the interrelationship between specific internal and external references prices along with the role of involvement has not been empirically investigated.In summary, this study proposes to extend the literature by attempting to answer two questions:1. Do all consumers evaluate a given offer against the same set of reference prices?2. Do all consumers utilize internal and external reference prices in the same way (i.e. using the same underlying process)? When consumers evaluate retail prices, they may do so against several reference prices. Of these, some are internal and others are external. The term "internal reference price" (henceforth referred to as IRP) has generally been used to refer to price information held in consumers' long-term memories. For example, Winer (1986) proposed that consumers use past price information to form expectations of what the retail price is likely to be on the next purchase occasion. As described in the literature, these price expectations represent internally held reference prices, and they are based on past price information that must be retrieved from consumers' long-term memories during the purchase occasion. In fact, most empirical studies have based their measures of internal reference price on such a temporally generated construct (Kalwani et al., 1990; Mayhew and Winer, 1992; Rajendran and Tellis, 1994; Briesch et al., 1997). The literature offers compelling empirical evidence that such (memory-based) internal reference prices are crucial in predicting consumers' evaluations (of retail prices) and subsequent choice behavior.In contrast to internal reference price, external reference prices (henceforth referred to as ERP) refer to comparative (competitive) price information that is available in the purchase environment, i.e. prices that consumers are likely to encounter during the search process. Thus, consumers do not have to expend cognitive resources in trying to retrieve information on current market prices from their long terms memories. This information is likely to be obtained during the search process and utilized directly from working memory. Here, researchers have identified several possible references including normal (typically charged) market price (Urbany and Dickson, 1991) and the lowest market price (Biswas and Sherrell, 1993). These and other similar studies have shown that such market-based references are important predictors of consumers' choices.It is important to note the conceptual distinction between IRP and ERP. While IRP refers to price information that is stored in and retrieved from consumers' long term memories, ERPs are derived from price cues that are either present or evoked at the time of purchase. Specifically, ERPs may represent consumers' beliefs about current market prices that may be formed during the search associated with the purchase process. Such beliefs may include highest price, lowest available price and normal/average market price. Note, however, that there is likely to be significant heterogeneity (due in part to differing levels of involvement) in the extent to which consumers search for the lowest/best price. Such selective and differential exposure is likely to result in differences across individuals in their perceptions of ERPs. It is well established that reference price is a complex, multi-faceted construct, and that reference prices are important in explaining consumers' brand choices (Briesch et al., 1997; Mazumdar et al., 2005). In addition, consumers are likely to use more than one type of reference price to evaluate an offer (Mayhew and Winer, 1992; Rajendran and Tellis, 1994; Chandrashekaran and Jagpal, 1995; Shirai, 2003). However, we know considerably less about the process by which consumers may combine/integrate multiple reference prices. Even less is known about individual differences in consumers' utilization of multiple reference prices.Prior research has acknowledged that different consumer groups (segments) are likely to evoke and use different sets of reference prices (Winer, 1986; Mazumdar and Papatla, 2000). Indeed, prior research has demonstrated that several factors lead to differences in the references used to evaluate a posted price. For example, reference price utilization and price perceptions have been linked to frequency of purchase (Thomas and Menon, 2007), product type (Chandrashekaran and Jagpal, 1995; Lowe and Alpert, 2007) and price consciousness (Alford and Biswas, 2002; Palazon and Delgado, 2009). In summary, prior research supports the claim that different segments may evoke and utilize different sets of reference prices, and suggests that it may not be appropriate to assume that all consumers utilize the same types (and number) of reference prices.As emphasized by Lowe and Alpert (2007) and by Mazumdar et al. (2005), despite some research on the topic, we do not know much about how consumers integrate multiple reference prices. More importantly, we need a better understanding of the factors that may determine which type(s) of reference price(s) are evoked and utilized when making price-related judgments. Along those lines, this research proposes to investigate how the level of involvement leads to differences in the number and types of reference prices that are used in the evaluation process. More specifically, it focuses on how consumers combine and utilize internal reference price (operationalized as expected price) and market-based external reference prices (operationalized as subjective knowledge of normal and lowest market prices)[1]. Consistent with established theory, this research acknowledges that consumers may adjust their internal reference prices based on their perceptions of market-based references. However, the conceptualization adopted here (explained later) also allows for consumers to evaluate posted prices directly against market-based (external) references.Regarding the relative impacts of the two types of reference prices on consumers' evaluations, there is some empirical evidence (Mayhew and Winer, 1992) suggesting that external reference prices may be more influential than internal reference prices. However, that conclusion might be premature because previous research has not paid attention to the moderating roles of individual level variables, for example involvement, that are likely to influence how consumers choose to allocate their cognitive resources - this is likely to impact the relative emphasis that consumers place on the two types of reference prices when evaluating a posted price. Consumers' involvement with a product class has been shown to influence consumers' information processing strategy (Bettman, 1979; Bloch et al., 1986) by affecting the extent of search, the type of information sought, the relative importance of different types of information (e.g. price versus product attribute information), and the motivation to process available information. High involvement is generally associated with product knowledge (familiarity) and the motivation to search for and use relevant information. On the other hand, low involvement has been associated with low product/price knowledge (lack of familiarity), and the lack of motivation to engage in detailed processing of information.Although there is substantial research dealing with how involvement affects consumers' use of product-attribute information and on how involvement affects consumers' responses to marketing communications (advertising), research in the (reference) pricing context is relatively limited. Available evidence reveals that such variables as involvement, prior knowledge, and price consciousness moderate consumers' internal reference prices and their acceptance of retail prices. These variables have been found to affect consumers' internal standards (Kosenko and Rahtz, 1988; Lichtenstein et al., 1988; Biswas and Sherrell, 1993), the widths (i.e. upper and lower limits) of their price acceptance regions (Lichtenstein et al., 1988; Rao and Sieben, 1992), and, finally, their confidence in their reference price estimates (Biswas and Sherrell, 1993). However, we know less about how such factors are likely to influence the ways in which consumers may combine/utilize multiple reference prices. The current study focuses on one such factor, i.e. involvement. This research explores the notion that different reference price utilization strategies unfold for consumers with high and low involvement. The idea that involvement affects which decision route consumers are likely to follow is not new. Extant research on consumers' information processing has shown that involvement affects the way consumers acquire, store, retrieve, and use relevant information to make decisions (Bettman, 1979; Bloch et al., 1986). Compared to those who are not involved, involved consumers are more motivated to search for relevant information. They also possess better knowledge about products and prices than consumers who lack involvement. Therefore, these consumers are more likely to possess well-defined internal standards than low involvement consumers (see also Chandrashekaran and Jagpal, 1995). Consequently, they are likely to be more comfortable evaluating retail prices against their internal (memory-based) reference prices. In contrast, less involved consumers are not as motivated as their involved counterparts to engage in extensive search behavior. In addition, they not likely to engage in detailed processing of available (price) information. As a result, their internal standards are not likely to be well defined, and they are not confident about acting on this information (Chandrashekaran and Jagpal, 1995).Consistent with such theorizing, Biswas and Sherrell (1993) found that high-involvement consumers are indeed more confident of their internal reference price estimates than low-involvement consumers. Therefore, there is some empirical evidence to support the expectation that the level of involvement is likely to moderate consumers' utilization of reference prices.Consistent with the notion that consumers are likely to possess reference prices that are brand specific (Briesch et al., 1997), a brand's own previous prices may be considered to be the most relevant information on which to base a reference price for the brand. Thus, it may be reasonable to expect high involvement consumers' evaluations to be most closely associated with a summary of a brand's past prices, i.e. the expected price construct. Briesch et al. (1997) empirically compared five different model formulations from the literature and confirmed that a model using a temporally generated, brand-specific reference price is the best predictor of consumer choice. However, the authors did not examine individual differences in the utilization of past price information.As discussed earlier, low involvement individuals are by definition not equipped with the knowledge to evaluate price on the basis of relevant factors. Recall that these consumers are not expected to internalize (past) price information, and thus cannot be expected to have well-defined (reliable) price expectations. Therefore, these individuals are likely to evaluate a brand's current price on the basis of more readily accessible perceptions they may have formed of current market prices.In summary, high- and low-involvement consumers are expected to follow different processes in evaluating the same objective price information. Highly involved consumers base their evaluations on internally held beliefs. However, because they are exposed to market price information in the course of their search, it is likely that they utilize this information to update (fine-tune) their internal standards to be consistent with current market information. In contrast, low involvement consumers are not motivated to devote cognitive resources to storing price information in their long-term memories and, therefore, are less confident about their knowledge of past prices. Consequently, they are likely to rely on immediately available external cues (i.e., market-based reference prices) to make their evaluations (as opposed to first updating their internal reference prices and, subsequently utilizing them to evaluate posted prices).More formally, it is expected that:H1. High-involvement consumers evaluate retail prices only against their price expectations (i.e. internal reference price).H2. High-involvement consumers utilize market-based external references only to update their internal beliefs (expected prices).H3. Low involvement consumers evaluate retail prices against one or more market-based (external) reference prices.H4. Low involvement consumers do not utilize internal reference prices (price expectations) in evaluating retail prices.Figure 1 presents the hypothesized model on the processes by which market-based and truly internal reference prices affect consumers' evaluations. Two hundred undergraduate business majors attending a large Northeastern state university participated in this study. The data required for this study was collected in four stages, over a two-week period. The entire data collection task was designed in the form of an experiential exercise.Stage one: initial measures In the first stage, subjects provided general information on their familiarity and involvement with the chosen product, i.e. Levi jeans. This product was chosen because of its relevance to the student subjects, and because no major differences across genders and age groups were expected.Product knowledge Prior research (Rao and Sieben, 1992; Biswas and Sherrell, 1993) has shown that consumers' prior knowledge affects the levels and widths of their price acceptance regions. Product knowledge was measured using a single seven-point item on which subjects indicated their familiarity with the product (brand). Generally, multiple scale items measures are preferred. However, in this study, consumer knowledge (expertise) is not central to the hypotheses. Therefore, no major problems are foreseen.Involvement Subjects' involvement with the product was assessed using the personal involvement inventory (PII) developed by Zaichkowsky (1985). Subjects rated the product on 20 bi-polar adjective scales. These individual items were summed to yield an involvement index for each subject. Later, high- and low-involvement groups were created by performing a median split of the distribution of involvement indices.The shopping exercise After completing the task in the first stage, subjects were given an instruction sheet that described an experiential exercise. In the instructions, subjects were informed that this exercise was intended to demonstrate to them the differences in consumers' shopping behaviors for convenience and shopping goods. The experiential exercise required students to pretend that they were in the market for products listed on the sheet. To mask the actual intent of the study, the sheet listed several products (e.g. toothpaste, laundry detergent, and sneakers) in addition to jeans. To preserve realism, students were instructed to behave as if they were shopping for the products listed (including a pair of jeans), and were asked to engage in behaviors that they would normally engage in when actually buying the products listed (a pair of jeans, toothpaste, laundry detergent, sneakers) for themselves. They were especially asked to use only the information they would normally use when making purchase decisions in the product categories listed. Finally, the subjects were informed that they were expected to come back to class the following week to engage in a class discussion about their experiences and about where they would buy the products listed.Stage two: measuring perceptions of market prices A week after the initial measures were taken, subjects were questioned about the outcomes of their search during a class discussion. At this stage, subjects' perceptions of market prices were measured on two items corresponding to their estimates of the lowest price for which they can purchase the product and its normal (average) market price[2].Stage three: measuring internal reference price At the next class meeting, subjects were asked to provide estimates of their internal reference prices (prices they expected to see on the next purchase occasion). The measures of market-based and internal reference prices were separated in time to avoid any contamination of the reference price measures.Stage four: the decision task At the end of the second week of the exercise, each subject was shown a mock retail ad for a pair of Levi jeans, which included a picture of the product, information on the retail store (a popular retail chain), and a selling price ($34.99). To stick with the main theme of understanding how consumers utilize perceptions of market prices and their reference prices to evaluate a posted price, the ad did not identify that the product was "on sale" and, therefore, it did not include an advertised comparison price or any other semantic cues that are normally associated with comparative price claims[3]. The retail price ($34.99) was chosen for two reasons. First, it reflected the actual market prices in the stores that most students reported they are most likely to visit. Second, it was close to the median of the average reference price (=$36.17) calculated as (Lowest+Normal+Expected)/3. Thus, for approximately half the subjects the offered price was likely to have been above their average reference price, and for the remaining subjects it was likely to have been below their average reference price. Each subject evaluated the posted retail price on three seven-point scale items corresponding to perceived offer value, attractiveness of the deal, and willingness to buy the product at the stated price. Similar items have been used in previous research (see, for example, Chandrashekaran and Jagpal, 1995; Chandrashekaran and Grewal, 2006). The first task was to split the sample into high and low groups based on their involvement with the product category (jeans). A median split of the distribution of PII scores (median involvement=103) yielded high and low involvement groups of approximately equal sizes (nhigh=111 and nlow=114). In addition, the mean involvement scores in the two groups (mean=119.26 and SD=10.50 in the high-involvement group versus mean=83.03 and SD=16.87 in the low involvement group) were significantly different (t=19.27, p<0.001). Finally, an examination of the mean knowledge (familiarity) scores in the two involvement groups (mean familiarity=5.99 and 4.75 for high- and low-involvement groups, respectively) confirmed the premise that, in this product category, high-involvement consumers are significantly more knowledgeable than low-involvement consumers (t=7.91, p<0.01).To test the hypotheses in this study empirically, correlation matrices in each group were analyzed using the structural equations methodology (interested readers may refer to Joreskog and Sorbom, 1988). Figure 1 shows the model structure specified in both groups.The conceptual model shown in Figure 1 allows for the possibility that market-based reference prices (lowest price and normal price) may affect evaluations directly (along with expected price), or indirectly by first influencing expected price, which, in turn, affects how the offer is evaluated. Therefore, the model enables us to investigate the process by which consumers evoke, combine and utilize internal and market-based reference prices. Recall that this is an improvement over Rajendran and Tellis (1994), who examined how both temporal and contextual reference prices influence consumers' evaluations simultaneously. In relation to Figure 1, the hypotheses presented here predict that:* g11, g12, and b21> 0; and g21 and g22=0 in the high-involvement group; and* b21, g11, and g12=0; and g21 and g22>0 in the low-involvement group.As discussed earlier, consumers' evaluations were measured using three scale items corresponding to their perceptions of the value of the offer, the attractiveness of the deal, and their willingness to buy the item at the stated price. It could be argued that the concept of perceived value is related to consumers' perceptions of acquisition utility (AU), while their perceptions of the deal are related to transaction utility (TU) as defined by Thaler (1985). According to Thaler, these are conceptually different from purchase intentions (behavioral responses). However, principal components factor analysis resulted in a single factor and did not reveal conceptual differences between the three measures. Table I compares the two involvement groups on their reference price estimates and evaluations of the offer. It is interesting to note that there is no significant difference in reference price estimates across the two involvement groups (this point is also discussed later in the manuscript). Furthermore, there are no major differences between the two groups in their evaluations of the offer, except that high involvement consumers perceive the deal to be moderately more attractive than low involvement consumers (p<=0.10)[4].However, one cannot conclude on the basis of these results alone that high- and low-involvement consumers are homogeneous, thus justifying pooled analyses. Although consumers' final evaluations are important, it is crucial that we examine the process (route) by which high and low involvement consumers arrive at the same destination. In addition, consumers' final evaluations are not central to this study. Rather, the primary objective of this study is to examine heterogeneity in consumers' utilization of market-based and truly internal reference prices to evaluate the same objective price information. Indeed, although consumers do not differ significantly in their final evaluations of the offer, this paper seeks to uncover heterogeneity in reference price utilization, i.e. the means by which high and low involvement consumers reach the end (their final evaluations).Table II shows the estimated models in each group along with several fit indices. It is clear that the specified model structure fits the data well in both groups (kh2 with six df=9.59, p=0.143 in the high involvement group; and kh2 with six df=6.09, p=0.413 in the low-involvement group). In addition, the structural models explain a significant proportion of the variance (R2=0.51 and 0.53 in the high- and low-involvement groups respectively). For both groups, the RMSEA (Steiger, 1985) and other fit indices indicate satisfactory model fit. All indicated paths are of the predicted sign, and are significant at a=0.01. All non-significant effects, i.e. paths that are not significantly different from zero, are indicated as "NS".From Table II, it is clear that consumers who are highly involved with the product evaluate the retail price against a single internal reference price, expected price (b21 is significant at p<=0.01), but do not utilize any market-based references to make evaluations (note that g21 and g22 are not significantly different from zero). These findings strongly support the hypothesis (H1) for consumers' utilization of truly internal reference prices under high involvement.H2 suggests that involved consumers are likely to use market-based references primarily to update their internal references. Consistent with this hypothesis, Table II shows that normal price (a market-based reference price) has a significant positive effect on the internal standards of involved consumers (g12=0.727 is significant at a=0.01). This finding is consistent with the view that involved consumers process price information sequentially, i.e. an estimate of the normal/average market price is used to update (fine-tune) the internal standard, which, in turn, is used to evaluate the retail price. Contrary to expectation, these consumers do not utilize their perceptions of the lowest market price (g11 is not significant). Thus, overall, H2 is partially supported.In contrast to high-involvement consumers, those in the low-involvement group do not use their internal standards to evaluate the retail price (b21=0). As shown in Table II, these consumers base their evaluations of the retail price on market-based references (g21=0.257 and g22=0.211 are both significant at a=0.01). Thus, H3 and H4 are supported. In conjunction, these findings support the conclusion that low levels of involved facilitate simultaneous utilization of several market-based reference prices to evaluate retail prices[5].It is interesting to note that highly involved consumers utilize only normal price, but not their perceptions of lowest price, in formulating their price expectations. Clearly, further research is required to be able to offer a convincing explanation to this observation. One possible explanation may be that these consumers may generally be more aware that the observed lowest price (a single data point) is likely to be a temporary promotional offer that does not reflect the "normal" price for the product/category. Consequently, these consumers may discount this information and rely to a greater extent on more comprehensive information, for example overall estimate of normal/average price. Urbany and Dickson (1991) found that consumers' estimates of normal prices are reasonable surrogates for their reference prices. However, the authors did not investigate individual differences in consumers' perceptions of normal/average prices. It is likely that one segment (e.g. high-involvement consumers) has more accurate estimates of normal prices than another segment (e.g. low-involvement consumers). Hopefully, future research will investigate and shed some light on this and other similar issues dealing with individual differences in price perceptions and evaluation.In summary, the results reveal that although the two types of consumers (high- and low-involvement) may have similar reference prices and may even evaluate an offer similarly, they do so via distinct routes, i.e. by utilizing different types of reference prices. Specifically, whereas highly involved consumers rely solely on their internal standards to evaluate retail prices, those who are less involved with the product category utilize several market-based references to judge the attractiveness of retail prices. A general discussion of the results along with implications and limitations of the study follows. This study intended to draw attention to the heterogeneity in consumers' utilization of multiple reference prices. The results obtained here offer some evidence that involvement affects consumers' utilization of reference prices. Although involvement does not affect the levels of consumers' internal standards or their final evaluations, it plays a significant role in the way consumers utilize these standards to evaluate the same objective price information. Highly involved consumers, who are more knowledgeable about the product class, evaluate offers against a single internal standard (expected price), which supports previous researchers' (e.g. Winer, 1986; Kalwani et al., 1990) in their use of expected price as a reference price to explain consumers' choices. However, low-involvement consumers do not use expected price to evaluate offers. Rather, they use two market-based references (lowest price and normal price) to determine the overall value of an offer. Thus, involvement affects the number and types of reference prices used in the evaluation process.The results are also consistent with information-processing theory (Bettman, 1979), which advocates that high involvement encourages a deeper and more complex processing strategy than low involvement, which supports less complex processing and the use of easily available information. Highly involved consumers retrieve price information stored in long-term memory and fine tune this internal standard (expected price) using current market price information (estimated normal/average market price). Thus, high involvement supports a hierarchical model in which market-based and truly internal reference prices are used sequentially. This conclusion is consistent with the underlying process implied in vast amount of research on reference price formation. In contrast, low-involvement consumers do not possess the motivation to engage in such a complex evaluation process. They simply utilize the more readily available market-based references and evaluate offers in a single step.It is only appropriate to point out that some of the results obtained here are inconsistent with previous research findings and call for further investigation. For example, Lichtenstein et al. (1988) found that consumers' price acceptability level (i.e. internal reference price level) is positively related to involvement. However, no such differences were found here. One possibility is that the results are being moderated by consumers' knowledge in this category. Rao and Sieben (1992) found that the upper and lower limits of consumers' price-acceptance regions increase with knowledge up to a point, and then levels off. It is likely that, given the nature of product category used in this study (jeans), most consumers were familiar with the product. Indeed, an examination of the mean knowledge scores revealed that both high- and low-involvement consumers lie in the upper half of the scale (means for high- and low-involvement groups=5.99 and 4.75 on seven-point scales), which may account for the lack of difference in their reference prices. Furthermore, product knowledge was assessed using a single item that measured subjects' familiarity with the product category. Additional research is needed to investigate this issue further.Another possible limitation of this study is the omission of several internal and market-based reference prices (e.g. fair price and highest price) that have been mentioned in the literature. It might be useful for future research to include more definitions of reference price to investigate whether the premise of this research holds. Finally, it may be useful to examine the roles of other moderating factors (e.g. gender, product experience, price consciousness, etc.).This research has important implications for both marketing research and practice. It is important for researchers to identify other sources of heterogeneity and their impact on consumers' construction and utilization of reference prices. The finding that high- and low-involvement consumers are different in the types of reference prices used and in the way they use this information to evaluate offers strongly supports the need for using different model structures to represent these sub-populations. From a practical standpoint, this study underscores the importance of including individual differences in strategies designed to affect consumers' perceptions of retail prices. For example, marketers might segment the market on the basis of reference price utilization and design different strategies to obtain optimal results. It is hoped that this study will stimulate further research in the area so that we may gain a better understanding of how consumers process and respond to retail pricing strategies. Such knowledge will undoubtedly be useful in designing more effective and efficient marketing strategies. Opens in a new window.Figure 1 Process of reference price effects on evaluations Opens in a new window.Table I Summary statistics Opens in a new window.Table II Estimated models
|
- Despite decades of research on the role of reference prices, the moderating role of involvement on reference price utilization has not been researched adequately. The findings reported here add to existing knowledge in the field and shed additional light on the process by which consumers evaluate posted prices. The findings also emphasize the need to segment the market on the basis of reference price utilization and to design appropriate communication strategies for each.
|
[SECTION: Purpose] In South Korea, restrictions on political speech surrounding elections are more stringent than in many other countries. The Public Official Election Act (hereinafter POEA, enacted in March 1994 as a result of integration of four different election laws corresponding to different layers of public administration) contains a number of provisions prohibiting campaign activities that would be standard practice in most democratic countries. According to Article 59, for example, official campaigns are allowed only during a period from the day following the closing date of candidate registration to the eve of the election. This amounts to 23 days in the case of presidential elections, and 14 days for elections of legislators and local governors. Even within this brief designated period, campaigns are subject to tight protocols prescribed in Articles 58 ("Definition of election campaign"), 60 ("Persons barred from election campaign"), 68 ("Campaign sashes"), 92 ("Prohibition of election campaign using motion pictures, etc."), 98 ("Restriction on use of broadcast for election campaign"), 99 ("Prohibition of election campaign by internal broadcast, etc."), and 100 ("Prohibition of use of recorders, etc."), to name a few.Having a more direct effect on individual voters are Article 90 and Article 93, which ban the display or distribution of election-related paraphernalia in the 180 days prior to an election. Section 1 of Article 93 states that "no one shall distribute or display advertisements, letters, posters, photographs, documents, drawings, printed materials, audiotapes, videotapes or the like that convey endorsement of or opposition to a candidate or a party" during the 180 days.The already overarching scope of Article 93(1) has been further stretched since the National Election Commission (NEC) and the Public Prosecutors' Office started to apply it to the online context in 1996. Tracking down the author of an online source and enforcing this provision is practically easy in Korean cyberspace, in that all Korean users are law-bound to verify their real identities (by providing their resident ID numbers) when joining major online services. Thereby, numerous bulletin board entries, blog posts, viewer comments on news sites, and user-generated content on Web 2.0 platforms have resulted in legal ramifications ranging from fines to imprisonment.The 2007 presidential election The latest presidential election in 2007 epitomized this long-standing conflict between citizens and institutions in the Korean electoral environment and how the global internet affected ways in which this conflict played out. The 2007 election saw Korea's cyberspace surprisingly silent, at least on the surface (Park and Lee, 2008). The victor seemed predetermined; Lee Myung-Bak, the conservative Grand National Party[1] (GNP) candidate, maintained a huge lead in the opinion polls since his nomination. However, the event took an unexpected sharp turn when, shortly before the day of the election, the rival United New Democratic Party[2] (UNDP) released two video clips implicating Lee in a financial scandal.The first clip showed Lee, in an interview by a journalist[3] in 2000, stating that he was directly involved in the establishment of BBK, a financial firm that was investigated in 2001 and found to be a scam. This contradicted Lee's campaign speeches denying any connection to BBK, as well as state prosecutors' conclusion clearing him of any involvement. The video clip was then rapidly circulated among the voters, even though the NEC restricted the sharing of such information in cyberspace, based on Article 93(1) of the POEA. Voters opposing Lee abandoned domestic social networking sites subject to the country's online real-name verification system and moved to foreign websites to further discuss the candidacy of Lee. The clip, which was posted on YouTube on December 5, 2007, generated more than 900,000 views in two weeks. It was viewed more often than any other user-generated video content available in Korean cyberspace during the election (Chang and Park, 2009, p. 93). As a result, this phenomenon popularized the term "cyber-exile" in the media (The Hankyoreh, 2009), and the term has become increasingly in use since. No news reports mentioned the term between 2000 and 2007, but between 2008 and 2010, it was mentioned 52 times[4].The second video clip was posted on the website of the Hankyoreh, a liberal daily, three days before the election. This 150-second clip showed Lee giving a speech at a local university in 2000, once again introducing himself as the founder and successful entrepreneur of BBK. According to The Hankyoreh's web statistics, this exclusive scoop attracted 1.5 million viewers by the time of the election. The clip also appeared on Daum, a major South Korean web portal, receiving more than 40,000 comments.Lee overcame this scandal and went on to win the election, but these video clips posed a serious threat to his presidential bid and has been a recurring issue throughout his term in office. In this study, we will examine how the two video clips were shared and discussed among the Korean online users, in order to explore the causes and consequences of the "cyber-exile" phenomenon, as well as its political implications for Korea in addition to the global community. Given the mainstream popularity of Web 2.0 services in recent years, current literature offers extensive discussion regarding the political implications of the increasing practice of digitally mediated social networking and content sharing. Specific interest has been shown in the question of whether such practice increases voter turnout or other forms of political participation in the traditional sense of the term. Research findings have, however, been mixed so far; while some consider that the use of Facebook and other social networking sites (SNSs) for political purposes is a significant predictor of general political participation (e.g. Vitak et al., 2011), others suggest that reliance on SNSs is not necessarily related to the increase in political participation (e.g. Baumgartner and Morris, 2010; Zhang et al., 2010).This broad discussion itself is inconclusive and is in need of further exploration, but for the purposes of this paper, we decided to focus on the political activity on and through YouTube. This commercially successful video-sharing site often features among SNSs in journalistic and scholarly accounts (boyd and Ellison, 2007), and this fact appropriately captures some, but not all, of its characteristics. It indeed provides a range of social networking features, such as "Comments", "Favorite", "Share" and "Subscriptions". However, it also stands apart from other SNSs, due to it being a unique hybrid form of social media and mass media. Despite social networks that underlie or emerge from its use (Lange, 2007b; Rotman and Golbeck, 2011; Sifman, 2011), YouTube originally has a character of "a media distribution platform, kind of like television" (Burgess and Green, 2009, p. 3). Burgess and Green (2009, p. 3) argue that "YouTube's ascendancy has occurred amid a fog of uncertainty and contradiction around what it is actually for." It is, so the two authors' argument goes, a new breed of businesses, or an example of what Weinberger (2007) calls "meta businesses", where old media heavyweights and amateur content creators form a curious cohabitation (Burgess and Green, 2009, pp. 4-5).Given that its format induces user participation, YouTube along with other Web 2.0 models revitalized the early internet idealism for better democracy (Bruns, 2008; Jenkins, 2006; Weinberger, 2007). As Marwick (2007) also pointed out based on a content analysis of news coverage, YouTube has been portrayed in a rather celebratory tone in the mass media for its democratic (or democratizing) potential. This kind of portrayal has been supported by a handful of anecdotal cases. During the 2006 midterm elections in the USA, a YouTube-publicized gaffe, later called the "macaca moment", cost Republican Senator George Allen (Virginia) his re-election bid, despite his huge initial lead in the opinion polls over the Democratic challenger, Jim Webb. This result was attributed to a video clip showing Allen making racially discriminatory remarks, targeting one of Webb's volunteers, during a campaign tour. The incident quickly became a widely publicized issue once the footage was posted on YouTube, which effectively led to Allen's defeat (Sidarth, 2006). Since then, the term "macaca moment" has been used to refer to "high-profile candidate gaffes that are captured on YouTube, receive a cascade of citizen views and contribute to some substantial political impact" (Karpf, 2010). A further example of such an incident took place in the 2007 Finnish national elections (Carlson and Strandberg, 2008, p. 171).The discussion of the political implications of YouTube intensified around the 2008 US presidential election because that was when Web 2.0 applications started to be incorporated into the mainstream campaign repertoire in the USA. Then-candidate Barack Obama received considerable media attention for his online presence, which outshone that of the rival candidate Hillary Clinton during the primaries and later that of his Republican opponent John McCain. Obama's campaign made heavy use of Facebook and YouTube to engage the attention of younger voters (Young, 2008), securing record campaign contributions mainly through online donations (Cooper, 2008; Carpenter, 2010).The 2008 election is widely considered as having been a pivotal moment in the US campaign history. Carpenter (2010) and Ricke (2010), for example, pointed out that YouTube, particularly its joint projects with PBS ("Video Your Vote") and with CNN ("The CNN/YouTube Debates"), afforded more room than ever for ordinary voters to participate in the campaign process and consequently served as "an instrument of 'checks and balances'" (Carpenter, 2010, 223). Some other scholars focused on how individual candidates made use of YouTube to deliver their messages in this particular election. Church (2010) examined leadership discourse by analyzing YouTube clips, featuring 16 candidates in the race, and suggested that given the emergence of what he termed "the postmodern constituency" (Church, 2010, p. 138), as well as the unfiltered nature of the medium (Church, 2010, p. 139), voters' focus shifted from candidates' political experience to their character. Duman and Locher (2008) examined Barack Obama and Hilary Clinton's YouTube campaigns and highlighted how the two presidential hopefuls attempted to create and uphold a conversational allusion through their videos.In regards to YouTube, another growing body of literature is concerned with the patterns of user interaction in this "viral marketing wonderland" (Burgess and Green, 2009, p. 3), and the methodological implications of exploring them. Three themes were identified for our research design. First, some studies have focused on what motivates users to share their videos on YouTube in the first place, ranging from perceived usefulness to interpersonal norms (e.g. Yang et al., 2010). Second, from a similar yet more specific perspective, some have established that YouTube is not only a source for information or entertainment for individual purposes, but also a new terrain for social interaction involving acts of "co-viewing" (Haridakis and Hanson, 2009), video responses (Adami, 2009), peer comments (Lange, 2007a; Jones and Schieffelin, 2009), and links to Facebook Walls (Robertson et al., 2010). In this sense, Chu (2009) went further to argue that YouTube plays a role as a "cultural public sphere".That said, others have taken a cautionary stance regarding YouTube's capacity to function as a public sphere. For example, Hess (2009) suggested there are certain obstacles to public deliberation on YouTube, such as the platform's dismissive and playful atmosphere. Moreover, Carpentier (2009) analyzed 16plus, a YouTube-like online platform provided by the north Belgian public broadcaster VRT, and suggested that the users may not be as appreciative of additional means of participation as expected. Lange (2007a) found that video-based communication on YouTube is no less hostile than faceless text-based communication. Blitvich (2010) added that impolite comments, although sometimes strategically employed, are likely to result in polarization.Chadwick (2009) argued that deliberative processes and "thick" citizenship cannot be the sole yardstick for the assessment of the functioning of e-democracy in the Web 2.0 era. However, if YouTube proves to be an unviable site for discussing serious political issues, it is important to ask why that is so. Answers to this question would advance our understanding of its political potential and how to harness it. As established in the previous section, YouTube is a uniquely interesting environment for political communication. What merits further attention is how this global platform intersects with local political and cultural dynamics. This is an underexplored line of inquiry, especially in the existing literature, which has been predominantly based on incidents from Western polities. With the aim of contributing to filling this lacuna, we address in the present paper the following three questions.1. What were the salient features of the discussion surrounding the YouTube clip, and how did that discussion develop during the campaign period?2. What were the salient features of the discussion surrounding the Daum clip, and how did that discussion develop during the campaign period?3. To what extent and how did the two discussions differ, and what lessons can be drawn from the differences (if any) in terms of harnessing the political energy of Web 2.0? Data collection The two video clips observably changed the course of the campaign, and the present study was conducted in "real time" as the event unfolded. As soon as we identified the research need to examine how the discussions developed in connection with the first clip posted on YouTube on December 5, 2007, we started to collect different types of data to have as detailed understanding as possible regarding how the clip was shared and discussed among the Korean voters.First, we monitored YouTube users' reaction manifested in terms of view counts, number of comments, and number of Favorites.Second, we collected all user comments added to the clip. By the election day of December 19, a total of 436 comments were manually retrieved.Third, we gathered the URLs of webpages which either sent a hyperlink to the YouTube clip or quoted its address www.youtube.com/watch?v=8ZFAywhGTgk[5] in the text. The URLs were retrieved from search engine indices on December 11, 14, and 17, 2007. This third dataset was particularly important to our study, as we aimed to investigate where YouTube was situated in a broader context of the Korean presidential election in question. For this facet of data collection, we used three search engines - Yahoo, Google, and Naver (the most popular search engine in Korea) - because we expected that these search engines would complement one another in terms of coverage. It turned out that Yahoo was better at locating webpages that sent out links to this particular clip, and Google was better at retrieving webpages quoting the URL in the body of the content, than Naver. Nonetheless, we included Naver search results in our data corpus because of Naver's dominance over Yahoo and Google in the Korean search engine market (Lee, 2009, p. 312).Once the second video clip was disclosed through the online edition of the local daily Hankyoreh, the scope of our data collection expanded. We collected all user comments left under the clip posted on the web portal Daum. Founded as early as 1995, Daum is one of the three largest portals in South Korea (along with Naver and Nate) and offers a wide range of services such as email, online communities (Daum Cafe), public forums (Daum Agora), blogs, web searches, and news aggregation. The Hankyoreh site also provides some space for reader comments, but we decided to collect the data from Daum because domestic web portals play a uniquely influential role in Koreans' news consumption patterns (Lim, 2005; Kim and Park, 2008). More relevant to this study, Daum users are known to be particularly active in political mobilization (Yun, 2009; Chang and Park, 2012).As a result, we manually collected a total of 43,493 comments under 7,641 unique screen IDs by the election day. Most comments (39,120 out of 43,493) were made on the eve of the election.Data analysis After the establishment of data corpus, we employed a combination of interaction network analysis, semantic network analysis, and hyperlink analysis.In more detail, first, for the interaction network analysis, the user comments, collected from both YouTube and Daum, were coded in terms of the attitudes conveyed in each entry and were sorted by the user IDs. The coding scheme was straightforward: each comment was manually classified into one of the Positive, Negative, and Neutral categories, depending on its contributor's attitude toward candidate Lee. Positive comments included messages supporting Lee, such as those arguing that the video clip was either quoted out of context or doctored, or arguing that despite the scandal Lee's strengths outweighed his shortcomings as a potential president. On the other hand, negative comments included those criticizing Lee and/or the prosecution team that vindicated him. Neutral comments included messages from users who appeared to be unaware of or ambivalent about the BBK scandal. Irrelevant or nonsensical comments were classified as "Others."Next, we conducted a content and discourse analysis of the collected comments from both YouTube (420 comments) and Daum (400 comments, chosen by a systematic sampling of every tenth entry) to identify any semantic patterns emerging from the respective discussions. For this, we conducted a keyword-in-context (KWIC) analysis, a methodological strategy that focuses on the correlation between a given word and its surroundings within the text. KWIC refers to a set of methods that process large bodies of messages, through the identification key words or phrases within a text (Biddix et al., 2009). For the purpose of this study, we used KrKwic (Korean Key Word In Context), a free software package that was developed by Park and Leydesdorff (2004) for KWIC analysis in the Korean language.We also charted 100 critical words from the 400 Daum comments, according to KWIC results, and conducted a CONCOR (convergence of iterated correlations) analysis of those 100 words to identify any issue clusters. CONCOR is a procedure that partition words into positions based on the structural equivalence (Wasserman and Faust, 1994, pp. 375-381). This partition generates subsets of the original set of words. Each of these subsets contains words that have similar connections to and from other words in the text. KrKwic and CONCOR have been used as means of forming content analysis of the Korean texts. For example, Chung and Park (2010) have recently used both KrKwic and CONCOR to compare the ideological stances of two Korean presidents, using their inaugural addresses, concluding that two methods greatly assisted in categorizing Korean words using the semantic links between words grounded in the text.Finally, we sorted the webpages that were linked to the YouTube clip according to their URLs and visited them manually to identify the nature of their content and link. As discussed in the following section, this qualitative approach allowed for an examination of Korean users' innovative efforts to circumvent Article 93(1) of the POEA.Table I summarizes the techniques employed in this study. We conducted the CONCOR analysis only for the Daum clip because the YouTube comments provided little text. On the other hand, the Daum clip required no hyperlink analysis because Korean users are known for their tendency to 'flock' to the semi-fenced playground of web portals, and particularly that of Daum, for political debates (Yun, 2009; Chang and Park, 2012). Patterns of interactions among users As shown in Table II, YouTube users paid considerable attention to the "BBK interview" footage. Most of the comments leveled criticism at Lee, as summarized in Table III. Likewise, the footage of Lee's guest speech also attracted a huge number of viewer comments on Daum, mostly critical of him.The data indicated a power-law distribution of participation on both YouTube and Daum (although any power-law interpretation should be made with caution because our datasets were relatively small). In other words, only a handful of users accounted for a large part of the commentary. For example, nine out of 13 comments on YouTube by JapanEmpire (a screen name) supported Lee, including the following:GNP candidate and former Seoul mayor Lee Myung-bak is a wealthy, experienced, and sincere Christian, I'm sure he doesn't lie like others. Also, he will be the second best president after Park Chung-hee of the ROK.Korea needs a great president with strong leadership who will really improve the economy and possibly eliminate corruption (quoted verbatim).On the other hand, Seebuddy9 left 12 comments that consistently opposed Lee, including the following:Hey Mr Lee! You are not qualified as Korean president. You know why? Korean people saw your bottom of immorality. What about your criminal record in Korean police too. Now, Korean people fighting corruption against Mr Lee and grand national party. The president of Korea should required transparency with strong accountability. That is reason you are not qualified. Withdraw your candidacy immediately! (quoted verbatim).Our observation was that ordinary users were more likely to respond to messages put up by a few highly active contributors than to initiate a thread themselves. Except for those highly active contributors (in the case of YouTube, danfrphils, JapanEmpire, Youngkuk, and seebuddy9, among others, occupying the central position in the network in Figure 1), the number of new comments decreased rapidly over time. However, this does not necessarily imply that voters' interest in the issue was short-lived. Instead, the discussion became dispersed, as duplicates of the video clip started to appear on YouTube.The data also indicated the preferential bonding among like-minded users. In other words, users tended to interact more with others sharing the same opinion. Figure 1 visualizes these interaction patterns. The shape represents the coded attitude: square means "Positive"; inverted triangle, "Negative"; triangle, "Neutral"; and circle, "Others." The size of each node is proportional to that of the number of "replies" exchanged. Isolators (i.e. those who did not send nor receive a reply) were excluded from the visualization.The use of this visualization technique allowed us to explore the variant ways in which YouTube users appropriated the comment facility and conversed among themselves. Inverted triangles show some affinity with other inverted triangles. Square nodes show the same pattern. In particular, those opposing Lee communicated with one another most frequently. This is consistent with the findings of Park and his colleagues (Park and Jankowski, 2008; Park and Thelwall, 2008); they found that SNS users who belong to or sympathize with the progressive camp in Korea are more connected among themselves than SNS users who support the conservative Grand National Party.Having said that, those interaction patterns we observed were only suggestive, rather implying that during the campaign period, YouTube users tended to express their opinions and leave it at that, instead of engaging in further discussions or debates with others (see also Vergeer and Hermans, 2008). This finding is generally consistent with the findings of Mislove et al. (2007), who suggested that YouTube users are less likely to form interest groups than users of other SNSs (e.g. Live Journal, Flickr, and Orkut).Semantic and discursive patterns of comments The proportion of irrelevant comments was interestingly high on YouTube. Comments in the "Others" category, in Table III, included those containing unintelligible phrases and, more notably, those engaging in "cyber-flaming" (i.e. directing unmotivated anonymous abuse toward other users). The discussion on YouTube was occasionally derailed by racist attacks. For example, the very first comment about the video clip was:Stupid Korean want to make crook to the president ... haha (itman21c, quoted verbatim).This was immediately followed byWho's the president you talking about? Do you even know what you're saying, itman21c? Just shut up, retarded singapole, singapore whatever! (Youngkuk, quoted verbatim).Given this observation, we applied the same coding scheme to the YouTube comments in collection but focused on the country, not Lee (as seen in Table III), for exploratory purposes. The results are summarized in Table IV, pointing to the high proportion of irrelevant comments in English.Comments added to the Daum clip took playful forms, specific to the Korean online culture, such as emoticons (e.g. 238 instances of "^^," which represents smiling), internet slang or jargon (e.g. 5,470 instances of "Graphic 1," a Korean consonant representing giggling in this context), and sarcastic nicknames for those allegedly involved in the BBK scandal (e.g. 220 instances of "tteok-chal," one given to the prosecution team, implying corruption) (for a detailed explanation of these terms, see Park and Lee, 2009). Despite these playful uses of language, the discussion on Daum stayed closely relevant to the topic.We also identified issue clusters from this rather intense discussion, on the basis of the CONCOR results. After reviewing user comments and considering their relevance to the topic, we labeled each cluster. As shown in Table V, there were one major cluster and seven minor clusters. The major cluster indicates commonly expressed negative opinions about Lee. Frequent keywords in this category included "Graphic 2," "Lee Myung-bak," "negative," "crook," "withdrawal of endorsement," and "Chung Dong-young" (a major opposition candidate), among others. A closer examination of the clusters revealed that the discussion surrounding the Daum video grew to cover other controversies associated with Lee (e.g. some clusters representing Lee's false residence registration and real estate speculation).Hyperlink networks around the YouTube clip Finally, Table VI summarizes the hyperlink data we collected, which were in turn supplemented and extended by a qualitative analysis of the actual content of each webpage.The video clip spread rapidly in Korean cyberspace through personal blogs, the Naver Q&A service, online communities, and public bulletin boards on web portals/newspaper sites, in that order. In these cases, the URL was simply presented as text without displaying the actual video content and expressing opinions about it, which allowed for the circumvention of Article 93(1) of the POEA. Q1. What were the salient features of the discussion surrounding the YouTube clip? In terms of the first research question, the YouTube clip targeted an audience of Korean voters exclusively. This was self-evident because the clip was in Korean with no subtitles, and the original contributor's message, explicitly addressed to the Korean public, was also in Korean with no translation. The contributor made his motive clear vis-a-vis the media attention he unwittingly caused. In a 1,150-word message next to the clip, he explained that he could not accept the state prosecutors' conclusion that Lee was not guilty. He also added that he first considered posting the clip on the bulletin board on the website of the Public Prosecutors' Office, as a gesture of protest, but decided to post it on YouTube instead because of (in his own words) the "bother" that he would otherwise have received from the NEC. He also emphasized that contrary to the mass media's speculation, he was neither a member of any party nor a campaigner for any candidate. This incident illustrates how Korean users appropriated a global social networking space for discussion of domestic political issues, specifically to circumvent their country's election law.It is noteworthy that the diffusion of information and the political discussion with respect to the YouTube clip stayed within the Korean cyberspace, bounded by the Korean language. According to the YouTube statistics, most of the "referrals" were from Korean news media websites. As of December 11, Kyunghyang (a daily newspaper) accounted for the largest number of views. Kyunghyang's news story (http://news.khan.co.kr/kh_news/khan_art_view.html?artid=200712061653151 &code=940301) about the video generated 13,385 views, followed by the Hankyoreh's news story (www.hani.co.kr/arti/politics/politics_general/255686.html), which generated 10,699 views. Both progressive newspapers had the YouTube clip embedded in the news story. The results based on this study's collection of search engine indices indicate that, despite Article 93(1) of the POEA, a number of individual voters shared the clip by relaying its URL without providing detailed information or comments regarding its content.YouTube is not a "walled garden" (i.e. a closed or exclusive virtual environment) and is thus susceptible to irrelevant comments. Consequently, its comment facility did not serve well as a space for further discussion in this case. The racially motivated attacks were initiated by those who were apparently not directly concerned with Korean politics. Our results were in lines with Gueorguieva's (2008) findings, which suggested that YouTube and other popular SNSs have posed a new set of challenges to campaigns, particularly the risk of reduced control over the images and messages of the campaigns.Q2. What were the salient features of the discussion surrounding the Daum clip? Beyond the abundance of playful emoticons and internet jargon, as shown in Table V, the discussion on Daum was quantitatively intense, as evidenced by the number of comments, and stayed relevant to the video clip. The irony is that if it had not been for the Hankyoreh and Daum, the vibrant discussion would not have been possible because of the NEC's enforcement of the POEA.Q3. To what extent did the two discussions differ and what do the differences suggest? The difference between the two discussions illustrated above can be explained by the characteristics of their respective spaces. As discussed earlier, YouTube is a complex media environment that defies an easy definition. Its characteristic openness made it relatively easy for Korean voters to employ it as an alternative channel for their discussions and debates outside election laws. At the same time, however, YouTube is more likely to be composed of uninterested parties than domestic sites, and therefore YouTube users might have distracted Korean voters' discussions and debates away from issues surrounding Korean politics. The address of the YouTube clip was widely circulated in Korean cyberspace, through blogs and online forums. However, no in-depth discussion took place because this "sphere" was fragmented. The dialectics between global media and local contexts have indeed been well documented (e.g. Turner, 2005), and the internet has complicated the matter even further. Contrary to the general description of the internet as a supranational network of computers, what we have is, in boyd's (2006) words, "all sorts of local cultures connected through a global network, resulting in all sorts of ugly tensions."What is unique about the case studied in this paper is, however, that users turned to the global space in order to circumvent a local political conflict, not the other way around - hence the neologism "cyber-exile". During the 2007 presidential election, Korean voters used YouTube to share election-related information because the laws of their country prohibited them from doing so on domestic websites. YouTube provided them with a higher level of anonymity than Korean sites. This incident of cyber-exile illustrates the tension between internet-mediated grassroots political activity and the authorities' restrictive interpretation and application of existing laws - the POEA in this case - to curtail the activity (see also Lee, 2011).However, the discussion surrounding the YouTube video clip was brought "back" to the Korean cyberspace. Korean users are generally known for having a strong preference for local services (Lee, 2009, 312), but more importantly, YouTube could not provide a suitable environment for this particular discussion. Dialogs within YouTube's comment facility, among Koreans as well as non-Korean users, were often off the topic, and in some instances, unexpectedly turned into "racist flaming".Zuckerman (2010) analyzed various tools for circumventing the internet censorship worldwide and suggested that circumvention cannot be a long-term solution. His argument is focused largely on technical aspects, but it still leads to important questions of how, and to what extent, we then should go "beyond" circumvention.We, the authors of this paper, do not intend to advocate the "walled garden" model for online forums, as opposed to the wild of YouTube, nor do we wish to suggest there should be zero government intervention. Our findings indicate that users can circumvent local regulations via the internet and other digital communication technologies (and probably more will do so), but subsequent discussions are likely to become fragmented as a result. Despite the specifics of the case studied, the significance of the present paper lies in the fact that epitomizes the tension between "old" laws and "new" media. Moreover, our findings clearly demonstrate that an innovative circumvention attempt on the user's part is not enough to harness the potential of online discussion for measured, sustained discourse of the issue at hand.This study has an important limitation. We conducted the analysis on a real-time basis as the event unfolded, and therefore the need to compare YouTube and Daum arose only during the later stages of data collection. As a result, a point-to-point comparison was not feasible, which invites further research.Noteworthy is that on December 29, 2011, the Constitutional Court of South Korea declared the unconstitutionality of the NEC's extended application of Article 93(1) of the POEA to social media, particularly Twitter. With the next presidential election scheduled for December 2012, future research should examine the impact of this court decision on the long-awaited relaxation of restrictions on election-related online communication in the country. An earlier version of this paper was presented at the Oxford Internet Institute's 2010 conference, and some of the findings (answering different research questions) were published in a Korean-language journal. The authors are grateful to Ae-Jie Bae for her assistance in data collection. Opens in a new window.Graphic 1 Opens in a new window.Graphic 2 Opens in a new window.Graphic 3 Opens in a new window.Graphic 4 Opens in a new window.Graphic 5 Opens in a new window.Figure 1 Interaction network of YouTube users commenting on Lee Opens in a new window.Table I Summary of methodological techniques Opens in a new window.Table II YouTube users' reaction to the "BBK Interview" video Opens in a new window.Table III Attitudes toward Lee Opens in a new window.Table IV Attitudes toward Korea Opens in a new window.Table V Frequent keywords for eight CONCOR clusters Opens in a new window.Table VI Number of webpages directing users to the YouTube video clip
|
- South Korea imposes more stringent restrictions on political speeches during elections than many other democratic countries. The purpose of this paper is to examine the long-standing conflict between citizens and institutions in the Korean electoral environment and the effects of the internet on this conflict.
|
[SECTION: Method] In South Korea, restrictions on political speech surrounding elections are more stringent than in many other countries. The Public Official Election Act (hereinafter POEA, enacted in March 1994 as a result of integration of four different election laws corresponding to different layers of public administration) contains a number of provisions prohibiting campaign activities that would be standard practice in most democratic countries. According to Article 59, for example, official campaigns are allowed only during a period from the day following the closing date of candidate registration to the eve of the election. This amounts to 23 days in the case of presidential elections, and 14 days for elections of legislators and local governors. Even within this brief designated period, campaigns are subject to tight protocols prescribed in Articles 58 ("Definition of election campaign"), 60 ("Persons barred from election campaign"), 68 ("Campaign sashes"), 92 ("Prohibition of election campaign using motion pictures, etc."), 98 ("Restriction on use of broadcast for election campaign"), 99 ("Prohibition of election campaign by internal broadcast, etc."), and 100 ("Prohibition of use of recorders, etc."), to name a few.Having a more direct effect on individual voters are Article 90 and Article 93, which ban the display or distribution of election-related paraphernalia in the 180 days prior to an election. Section 1 of Article 93 states that "no one shall distribute or display advertisements, letters, posters, photographs, documents, drawings, printed materials, audiotapes, videotapes or the like that convey endorsement of or opposition to a candidate or a party" during the 180 days.The already overarching scope of Article 93(1) has been further stretched since the National Election Commission (NEC) and the Public Prosecutors' Office started to apply it to the online context in 1996. Tracking down the author of an online source and enforcing this provision is practically easy in Korean cyberspace, in that all Korean users are law-bound to verify their real identities (by providing their resident ID numbers) when joining major online services. Thereby, numerous bulletin board entries, blog posts, viewer comments on news sites, and user-generated content on Web 2.0 platforms have resulted in legal ramifications ranging from fines to imprisonment.The 2007 presidential election The latest presidential election in 2007 epitomized this long-standing conflict between citizens and institutions in the Korean electoral environment and how the global internet affected ways in which this conflict played out. The 2007 election saw Korea's cyberspace surprisingly silent, at least on the surface (Park and Lee, 2008). The victor seemed predetermined; Lee Myung-Bak, the conservative Grand National Party[1] (GNP) candidate, maintained a huge lead in the opinion polls since his nomination. However, the event took an unexpected sharp turn when, shortly before the day of the election, the rival United New Democratic Party[2] (UNDP) released two video clips implicating Lee in a financial scandal.The first clip showed Lee, in an interview by a journalist[3] in 2000, stating that he was directly involved in the establishment of BBK, a financial firm that was investigated in 2001 and found to be a scam. This contradicted Lee's campaign speeches denying any connection to BBK, as well as state prosecutors' conclusion clearing him of any involvement. The video clip was then rapidly circulated among the voters, even though the NEC restricted the sharing of such information in cyberspace, based on Article 93(1) of the POEA. Voters opposing Lee abandoned domestic social networking sites subject to the country's online real-name verification system and moved to foreign websites to further discuss the candidacy of Lee. The clip, which was posted on YouTube on December 5, 2007, generated more than 900,000 views in two weeks. It was viewed more often than any other user-generated video content available in Korean cyberspace during the election (Chang and Park, 2009, p. 93). As a result, this phenomenon popularized the term "cyber-exile" in the media (The Hankyoreh, 2009), and the term has become increasingly in use since. No news reports mentioned the term between 2000 and 2007, but between 2008 and 2010, it was mentioned 52 times[4].The second video clip was posted on the website of the Hankyoreh, a liberal daily, three days before the election. This 150-second clip showed Lee giving a speech at a local university in 2000, once again introducing himself as the founder and successful entrepreneur of BBK. According to The Hankyoreh's web statistics, this exclusive scoop attracted 1.5 million viewers by the time of the election. The clip also appeared on Daum, a major South Korean web portal, receiving more than 40,000 comments.Lee overcame this scandal and went on to win the election, but these video clips posed a serious threat to his presidential bid and has been a recurring issue throughout his term in office. In this study, we will examine how the two video clips were shared and discussed among the Korean online users, in order to explore the causes and consequences of the "cyber-exile" phenomenon, as well as its political implications for Korea in addition to the global community. Given the mainstream popularity of Web 2.0 services in recent years, current literature offers extensive discussion regarding the political implications of the increasing practice of digitally mediated social networking and content sharing. Specific interest has been shown in the question of whether such practice increases voter turnout or other forms of political participation in the traditional sense of the term. Research findings have, however, been mixed so far; while some consider that the use of Facebook and other social networking sites (SNSs) for political purposes is a significant predictor of general political participation (e.g. Vitak et al., 2011), others suggest that reliance on SNSs is not necessarily related to the increase in political participation (e.g. Baumgartner and Morris, 2010; Zhang et al., 2010).This broad discussion itself is inconclusive and is in need of further exploration, but for the purposes of this paper, we decided to focus on the political activity on and through YouTube. This commercially successful video-sharing site often features among SNSs in journalistic and scholarly accounts (boyd and Ellison, 2007), and this fact appropriately captures some, but not all, of its characteristics. It indeed provides a range of social networking features, such as "Comments", "Favorite", "Share" and "Subscriptions". However, it also stands apart from other SNSs, due to it being a unique hybrid form of social media and mass media. Despite social networks that underlie or emerge from its use (Lange, 2007b; Rotman and Golbeck, 2011; Sifman, 2011), YouTube originally has a character of "a media distribution platform, kind of like television" (Burgess and Green, 2009, p. 3). Burgess and Green (2009, p. 3) argue that "YouTube's ascendancy has occurred amid a fog of uncertainty and contradiction around what it is actually for." It is, so the two authors' argument goes, a new breed of businesses, or an example of what Weinberger (2007) calls "meta businesses", where old media heavyweights and amateur content creators form a curious cohabitation (Burgess and Green, 2009, pp. 4-5).Given that its format induces user participation, YouTube along with other Web 2.0 models revitalized the early internet idealism for better democracy (Bruns, 2008; Jenkins, 2006; Weinberger, 2007). As Marwick (2007) also pointed out based on a content analysis of news coverage, YouTube has been portrayed in a rather celebratory tone in the mass media for its democratic (or democratizing) potential. This kind of portrayal has been supported by a handful of anecdotal cases. During the 2006 midterm elections in the USA, a YouTube-publicized gaffe, later called the "macaca moment", cost Republican Senator George Allen (Virginia) his re-election bid, despite his huge initial lead in the opinion polls over the Democratic challenger, Jim Webb. This result was attributed to a video clip showing Allen making racially discriminatory remarks, targeting one of Webb's volunteers, during a campaign tour. The incident quickly became a widely publicized issue once the footage was posted on YouTube, which effectively led to Allen's defeat (Sidarth, 2006). Since then, the term "macaca moment" has been used to refer to "high-profile candidate gaffes that are captured on YouTube, receive a cascade of citizen views and contribute to some substantial political impact" (Karpf, 2010). A further example of such an incident took place in the 2007 Finnish national elections (Carlson and Strandberg, 2008, p. 171).The discussion of the political implications of YouTube intensified around the 2008 US presidential election because that was when Web 2.0 applications started to be incorporated into the mainstream campaign repertoire in the USA. Then-candidate Barack Obama received considerable media attention for his online presence, which outshone that of the rival candidate Hillary Clinton during the primaries and later that of his Republican opponent John McCain. Obama's campaign made heavy use of Facebook and YouTube to engage the attention of younger voters (Young, 2008), securing record campaign contributions mainly through online donations (Cooper, 2008; Carpenter, 2010).The 2008 election is widely considered as having been a pivotal moment in the US campaign history. Carpenter (2010) and Ricke (2010), for example, pointed out that YouTube, particularly its joint projects with PBS ("Video Your Vote") and with CNN ("The CNN/YouTube Debates"), afforded more room than ever for ordinary voters to participate in the campaign process and consequently served as "an instrument of 'checks and balances'" (Carpenter, 2010, 223). Some other scholars focused on how individual candidates made use of YouTube to deliver their messages in this particular election. Church (2010) examined leadership discourse by analyzing YouTube clips, featuring 16 candidates in the race, and suggested that given the emergence of what he termed "the postmodern constituency" (Church, 2010, p. 138), as well as the unfiltered nature of the medium (Church, 2010, p. 139), voters' focus shifted from candidates' political experience to their character. Duman and Locher (2008) examined Barack Obama and Hilary Clinton's YouTube campaigns and highlighted how the two presidential hopefuls attempted to create and uphold a conversational allusion through their videos.In regards to YouTube, another growing body of literature is concerned with the patterns of user interaction in this "viral marketing wonderland" (Burgess and Green, 2009, p. 3), and the methodological implications of exploring them. Three themes were identified for our research design. First, some studies have focused on what motivates users to share their videos on YouTube in the first place, ranging from perceived usefulness to interpersonal norms (e.g. Yang et al., 2010). Second, from a similar yet more specific perspective, some have established that YouTube is not only a source for information or entertainment for individual purposes, but also a new terrain for social interaction involving acts of "co-viewing" (Haridakis and Hanson, 2009), video responses (Adami, 2009), peer comments (Lange, 2007a; Jones and Schieffelin, 2009), and links to Facebook Walls (Robertson et al., 2010). In this sense, Chu (2009) went further to argue that YouTube plays a role as a "cultural public sphere".That said, others have taken a cautionary stance regarding YouTube's capacity to function as a public sphere. For example, Hess (2009) suggested there are certain obstacles to public deliberation on YouTube, such as the platform's dismissive and playful atmosphere. Moreover, Carpentier (2009) analyzed 16plus, a YouTube-like online platform provided by the north Belgian public broadcaster VRT, and suggested that the users may not be as appreciative of additional means of participation as expected. Lange (2007a) found that video-based communication on YouTube is no less hostile than faceless text-based communication. Blitvich (2010) added that impolite comments, although sometimes strategically employed, are likely to result in polarization.Chadwick (2009) argued that deliberative processes and "thick" citizenship cannot be the sole yardstick for the assessment of the functioning of e-democracy in the Web 2.0 era. However, if YouTube proves to be an unviable site for discussing serious political issues, it is important to ask why that is so. Answers to this question would advance our understanding of its political potential and how to harness it. As established in the previous section, YouTube is a uniquely interesting environment for political communication. What merits further attention is how this global platform intersects with local political and cultural dynamics. This is an underexplored line of inquiry, especially in the existing literature, which has been predominantly based on incidents from Western polities. With the aim of contributing to filling this lacuna, we address in the present paper the following three questions.1. What were the salient features of the discussion surrounding the YouTube clip, and how did that discussion develop during the campaign period?2. What were the salient features of the discussion surrounding the Daum clip, and how did that discussion develop during the campaign period?3. To what extent and how did the two discussions differ, and what lessons can be drawn from the differences (if any) in terms of harnessing the political energy of Web 2.0? Data collection The two video clips observably changed the course of the campaign, and the present study was conducted in "real time" as the event unfolded. As soon as we identified the research need to examine how the discussions developed in connection with the first clip posted on YouTube on December 5, 2007, we started to collect different types of data to have as detailed understanding as possible regarding how the clip was shared and discussed among the Korean voters.First, we monitored YouTube users' reaction manifested in terms of view counts, number of comments, and number of Favorites.Second, we collected all user comments added to the clip. By the election day of December 19, a total of 436 comments were manually retrieved.Third, we gathered the URLs of webpages which either sent a hyperlink to the YouTube clip or quoted its address www.youtube.com/watch?v=8ZFAywhGTgk[5] in the text. The URLs were retrieved from search engine indices on December 11, 14, and 17, 2007. This third dataset was particularly important to our study, as we aimed to investigate where YouTube was situated in a broader context of the Korean presidential election in question. For this facet of data collection, we used three search engines - Yahoo, Google, and Naver (the most popular search engine in Korea) - because we expected that these search engines would complement one another in terms of coverage. It turned out that Yahoo was better at locating webpages that sent out links to this particular clip, and Google was better at retrieving webpages quoting the URL in the body of the content, than Naver. Nonetheless, we included Naver search results in our data corpus because of Naver's dominance over Yahoo and Google in the Korean search engine market (Lee, 2009, p. 312).Once the second video clip was disclosed through the online edition of the local daily Hankyoreh, the scope of our data collection expanded. We collected all user comments left under the clip posted on the web portal Daum. Founded as early as 1995, Daum is one of the three largest portals in South Korea (along with Naver and Nate) and offers a wide range of services such as email, online communities (Daum Cafe), public forums (Daum Agora), blogs, web searches, and news aggregation. The Hankyoreh site also provides some space for reader comments, but we decided to collect the data from Daum because domestic web portals play a uniquely influential role in Koreans' news consumption patterns (Lim, 2005; Kim and Park, 2008). More relevant to this study, Daum users are known to be particularly active in political mobilization (Yun, 2009; Chang and Park, 2012).As a result, we manually collected a total of 43,493 comments under 7,641 unique screen IDs by the election day. Most comments (39,120 out of 43,493) were made on the eve of the election.Data analysis After the establishment of data corpus, we employed a combination of interaction network analysis, semantic network analysis, and hyperlink analysis.In more detail, first, for the interaction network analysis, the user comments, collected from both YouTube and Daum, were coded in terms of the attitudes conveyed in each entry and were sorted by the user IDs. The coding scheme was straightforward: each comment was manually classified into one of the Positive, Negative, and Neutral categories, depending on its contributor's attitude toward candidate Lee. Positive comments included messages supporting Lee, such as those arguing that the video clip was either quoted out of context or doctored, or arguing that despite the scandal Lee's strengths outweighed his shortcomings as a potential president. On the other hand, negative comments included those criticizing Lee and/or the prosecution team that vindicated him. Neutral comments included messages from users who appeared to be unaware of or ambivalent about the BBK scandal. Irrelevant or nonsensical comments were classified as "Others."Next, we conducted a content and discourse analysis of the collected comments from both YouTube (420 comments) and Daum (400 comments, chosen by a systematic sampling of every tenth entry) to identify any semantic patterns emerging from the respective discussions. For this, we conducted a keyword-in-context (KWIC) analysis, a methodological strategy that focuses on the correlation between a given word and its surroundings within the text. KWIC refers to a set of methods that process large bodies of messages, through the identification key words or phrases within a text (Biddix et al., 2009). For the purpose of this study, we used KrKwic (Korean Key Word In Context), a free software package that was developed by Park and Leydesdorff (2004) for KWIC analysis in the Korean language.We also charted 100 critical words from the 400 Daum comments, according to KWIC results, and conducted a CONCOR (convergence of iterated correlations) analysis of those 100 words to identify any issue clusters. CONCOR is a procedure that partition words into positions based on the structural equivalence (Wasserman and Faust, 1994, pp. 375-381). This partition generates subsets of the original set of words. Each of these subsets contains words that have similar connections to and from other words in the text. KrKwic and CONCOR have been used as means of forming content analysis of the Korean texts. For example, Chung and Park (2010) have recently used both KrKwic and CONCOR to compare the ideological stances of two Korean presidents, using their inaugural addresses, concluding that two methods greatly assisted in categorizing Korean words using the semantic links between words grounded in the text.Finally, we sorted the webpages that were linked to the YouTube clip according to their URLs and visited them manually to identify the nature of their content and link. As discussed in the following section, this qualitative approach allowed for an examination of Korean users' innovative efforts to circumvent Article 93(1) of the POEA.Table I summarizes the techniques employed in this study. We conducted the CONCOR analysis only for the Daum clip because the YouTube comments provided little text. On the other hand, the Daum clip required no hyperlink analysis because Korean users are known for their tendency to 'flock' to the semi-fenced playground of web portals, and particularly that of Daum, for political debates (Yun, 2009; Chang and Park, 2012). Patterns of interactions among users As shown in Table II, YouTube users paid considerable attention to the "BBK interview" footage. Most of the comments leveled criticism at Lee, as summarized in Table III. Likewise, the footage of Lee's guest speech also attracted a huge number of viewer comments on Daum, mostly critical of him.The data indicated a power-law distribution of participation on both YouTube and Daum (although any power-law interpretation should be made with caution because our datasets were relatively small). In other words, only a handful of users accounted for a large part of the commentary. For example, nine out of 13 comments on YouTube by JapanEmpire (a screen name) supported Lee, including the following:GNP candidate and former Seoul mayor Lee Myung-bak is a wealthy, experienced, and sincere Christian, I'm sure he doesn't lie like others. Also, he will be the second best president after Park Chung-hee of the ROK.Korea needs a great president with strong leadership who will really improve the economy and possibly eliminate corruption (quoted verbatim).On the other hand, Seebuddy9 left 12 comments that consistently opposed Lee, including the following:Hey Mr Lee! You are not qualified as Korean president. You know why? Korean people saw your bottom of immorality. What about your criminal record in Korean police too. Now, Korean people fighting corruption against Mr Lee and grand national party. The president of Korea should required transparency with strong accountability. That is reason you are not qualified. Withdraw your candidacy immediately! (quoted verbatim).Our observation was that ordinary users were more likely to respond to messages put up by a few highly active contributors than to initiate a thread themselves. Except for those highly active contributors (in the case of YouTube, danfrphils, JapanEmpire, Youngkuk, and seebuddy9, among others, occupying the central position in the network in Figure 1), the number of new comments decreased rapidly over time. However, this does not necessarily imply that voters' interest in the issue was short-lived. Instead, the discussion became dispersed, as duplicates of the video clip started to appear on YouTube.The data also indicated the preferential bonding among like-minded users. In other words, users tended to interact more with others sharing the same opinion. Figure 1 visualizes these interaction patterns. The shape represents the coded attitude: square means "Positive"; inverted triangle, "Negative"; triangle, "Neutral"; and circle, "Others." The size of each node is proportional to that of the number of "replies" exchanged. Isolators (i.e. those who did not send nor receive a reply) were excluded from the visualization.The use of this visualization technique allowed us to explore the variant ways in which YouTube users appropriated the comment facility and conversed among themselves. Inverted triangles show some affinity with other inverted triangles. Square nodes show the same pattern. In particular, those opposing Lee communicated with one another most frequently. This is consistent with the findings of Park and his colleagues (Park and Jankowski, 2008; Park and Thelwall, 2008); they found that SNS users who belong to or sympathize with the progressive camp in Korea are more connected among themselves than SNS users who support the conservative Grand National Party.Having said that, those interaction patterns we observed were only suggestive, rather implying that during the campaign period, YouTube users tended to express their opinions and leave it at that, instead of engaging in further discussions or debates with others (see also Vergeer and Hermans, 2008). This finding is generally consistent with the findings of Mislove et al. (2007), who suggested that YouTube users are less likely to form interest groups than users of other SNSs (e.g. Live Journal, Flickr, and Orkut).Semantic and discursive patterns of comments The proportion of irrelevant comments was interestingly high on YouTube. Comments in the "Others" category, in Table III, included those containing unintelligible phrases and, more notably, those engaging in "cyber-flaming" (i.e. directing unmotivated anonymous abuse toward other users). The discussion on YouTube was occasionally derailed by racist attacks. For example, the very first comment about the video clip was:Stupid Korean want to make crook to the president ... haha (itman21c, quoted verbatim).This was immediately followed byWho's the president you talking about? Do you even know what you're saying, itman21c? Just shut up, retarded singapole, singapore whatever! (Youngkuk, quoted verbatim).Given this observation, we applied the same coding scheme to the YouTube comments in collection but focused on the country, not Lee (as seen in Table III), for exploratory purposes. The results are summarized in Table IV, pointing to the high proportion of irrelevant comments in English.Comments added to the Daum clip took playful forms, specific to the Korean online culture, such as emoticons (e.g. 238 instances of "^^," which represents smiling), internet slang or jargon (e.g. 5,470 instances of "Graphic 1," a Korean consonant representing giggling in this context), and sarcastic nicknames for those allegedly involved in the BBK scandal (e.g. 220 instances of "tteok-chal," one given to the prosecution team, implying corruption) (for a detailed explanation of these terms, see Park and Lee, 2009). Despite these playful uses of language, the discussion on Daum stayed closely relevant to the topic.We also identified issue clusters from this rather intense discussion, on the basis of the CONCOR results. After reviewing user comments and considering their relevance to the topic, we labeled each cluster. As shown in Table V, there were one major cluster and seven minor clusters. The major cluster indicates commonly expressed negative opinions about Lee. Frequent keywords in this category included "Graphic 2," "Lee Myung-bak," "negative," "crook," "withdrawal of endorsement," and "Chung Dong-young" (a major opposition candidate), among others. A closer examination of the clusters revealed that the discussion surrounding the Daum video grew to cover other controversies associated with Lee (e.g. some clusters representing Lee's false residence registration and real estate speculation).Hyperlink networks around the YouTube clip Finally, Table VI summarizes the hyperlink data we collected, which were in turn supplemented and extended by a qualitative analysis of the actual content of each webpage.The video clip spread rapidly in Korean cyberspace through personal blogs, the Naver Q&A service, online communities, and public bulletin boards on web portals/newspaper sites, in that order. In these cases, the URL was simply presented as text without displaying the actual video content and expressing opinions about it, which allowed for the circumvention of Article 93(1) of the POEA. Q1. What were the salient features of the discussion surrounding the YouTube clip? In terms of the first research question, the YouTube clip targeted an audience of Korean voters exclusively. This was self-evident because the clip was in Korean with no subtitles, and the original contributor's message, explicitly addressed to the Korean public, was also in Korean with no translation. The contributor made his motive clear vis-a-vis the media attention he unwittingly caused. In a 1,150-word message next to the clip, he explained that he could not accept the state prosecutors' conclusion that Lee was not guilty. He also added that he first considered posting the clip on the bulletin board on the website of the Public Prosecutors' Office, as a gesture of protest, but decided to post it on YouTube instead because of (in his own words) the "bother" that he would otherwise have received from the NEC. He also emphasized that contrary to the mass media's speculation, he was neither a member of any party nor a campaigner for any candidate. This incident illustrates how Korean users appropriated a global social networking space for discussion of domestic political issues, specifically to circumvent their country's election law.It is noteworthy that the diffusion of information and the political discussion with respect to the YouTube clip stayed within the Korean cyberspace, bounded by the Korean language. According to the YouTube statistics, most of the "referrals" were from Korean news media websites. As of December 11, Kyunghyang (a daily newspaper) accounted for the largest number of views. Kyunghyang's news story (http://news.khan.co.kr/kh_news/khan_art_view.html?artid=200712061653151 &code=940301) about the video generated 13,385 views, followed by the Hankyoreh's news story (www.hani.co.kr/arti/politics/politics_general/255686.html), which generated 10,699 views. Both progressive newspapers had the YouTube clip embedded in the news story. The results based on this study's collection of search engine indices indicate that, despite Article 93(1) of the POEA, a number of individual voters shared the clip by relaying its URL without providing detailed information or comments regarding its content.YouTube is not a "walled garden" (i.e. a closed or exclusive virtual environment) and is thus susceptible to irrelevant comments. Consequently, its comment facility did not serve well as a space for further discussion in this case. The racially motivated attacks were initiated by those who were apparently not directly concerned with Korean politics. Our results were in lines with Gueorguieva's (2008) findings, which suggested that YouTube and other popular SNSs have posed a new set of challenges to campaigns, particularly the risk of reduced control over the images and messages of the campaigns.Q2. What were the salient features of the discussion surrounding the Daum clip? Beyond the abundance of playful emoticons and internet jargon, as shown in Table V, the discussion on Daum was quantitatively intense, as evidenced by the number of comments, and stayed relevant to the video clip. The irony is that if it had not been for the Hankyoreh and Daum, the vibrant discussion would not have been possible because of the NEC's enforcement of the POEA.Q3. To what extent did the two discussions differ and what do the differences suggest? The difference between the two discussions illustrated above can be explained by the characteristics of their respective spaces. As discussed earlier, YouTube is a complex media environment that defies an easy definition. Its characteristic openness made it relatively easy for Korean voters to employ it as an alternative channel for their discussions and debates outside election laws. At the same time, however, YouTube is more likely to be composed of uninterested parties than domestic sites, and therefore YouTube users might have distracted Korean voters' discussions and debates away from issues surrounding Korean politics. The address of the YouTube clip was widely circulated in Korean cyberspace, through blogs and online forums. However, no in-depth discussion took place because this "sphere" was fragmented. The dialectics between global media and local contexts have indeed been well documented (e.g. Turner, 2005), and the internet has complicated the matter even further. Contrary to the general description of the internet as a supranational network of computers, what we have is, in boyd's (2006) words, "all sorts of local cultures connected through a global network, resulting in all sorts of ugly tensions."What is unique about the case studied in this paper is, however, that users turned to the global space in order to circumvent a local political conflict, not the other way around - hence the neologism "cyber-exile". During the 2007 presidential election, Korean voters used YouTube to share election-related information because the laws of their country prohibited them from doing so on domestic websites. YouTube provided them with a higher level of anonymity than Korean sites. This incident of cyber-exile illustrates the tension between internet-mediated grassroots political activity and the authorities' restrictive interpretation and application of existing laws - the POEA in this case - to curtail the activity (see also Lee, 2011).However, the discussion surrounding the YouTube video clip was brought "back" to the Korean cyberspace. Korean users are generally known for having a strong preference for local services (Lee, 2009, 312), but more importantly, YouTube could not provide a suitable environment for this particular discussion. Dialogs within YouTube's comment facility, among Koreans as well as non-Korean users, were often off the topic, and in some instances, unexpectedly turned into "racist flaming".Zuckerman (2010) analyzed various tools for circumventing the internet censorship worldwide and suggested that circumvention cannot be a long-term solution. His argument is focused largely on technical aspects, but it still leads to important questions of how, and to what extent, we then should go "beyond" circumvention.We, the authors of this paper, do not intend to advocate the "walled garden" model for online forums, as opposed to the wild of YouTube, nor do we wish to suggest there should be zero government intervention. Our findings indicate that users can circumvent local regulations via the internet and other digital communication technologies (and probably more will do so), but subsequent discussions are likely to become fragmented as a result. Despite the specifics of the case studied, the significance of the present paper lies in the fact that epitomizes the tension between "old" laws and "new" media. Moreover, our findings clearly demonstrate that an innovative circumvention attempt on the user's part is not enough to harness the potential of online discussion for measured, sustained discourse of the issue at hand.This study has an important limitation. We conducted the analysis on a real-time basis as the event unfolded, and therefore the need to compare YouTube and Daum arose only during the later stages of data collection. As a result, a point-to-point comparison was not feasible, which invites further research.Noteworthy is that on December 29, 2011, the Constitutional Court of South Korea declared the unconstitutionality of the NEC's extended application of Article 93(1) of the POEA to social media, particularly Twitter. With the next presidential election scheduled for December 2012, future research should examine the impact of this court decision on the long-awaited relaxation of restrictions on election-related online communication in the country. An earlier version of this paper was presented at the Oxford Internet Institute's 2010 conference, and some of the findings (answering different research questions) were published in a Korean-language journal. The authors are grateful to Ae-Jie Bae for her assistance in data collection. Opens in a new window.Graphic 1 Opens in a new window.Graphic 2 Opens in a new window.Graphic 3 Opens in a new window.Graphic 4 Opens in a new window.Graphic 5 Opens in a new window.Figure 1 Interaction network of YouTube users commenting on Lee Opens in a new window.Table I Summary of methodological techniques Opens in a new window.Table II YouTube users' reaction to the "BBK Interview" video Opens in a new window.Table III Attitudes toward Lee Opens in a new window.Table IV Attitudes toward Korea Opens in a new window.Table V Frequent keywords for eight CONCOR clusters Opens in a new window.Table VI Number of webpages directing users to the YouTube video clip
|
- The paper provides a case study of the 2007 presidential election in Korea. During the campaign period, two video clips (one on YouTube and the other on Daum, a major domestic web portal) emerged and implicated the then-leading candidate in a financial scandal. The paper investigates how these video clips were shared and discussed among Korean voters, even though the country's election laws restricted the sharing of such information in cyberspace. The paper employs a combination of network analysis techniques, including hyperlink analysis, interaction network analysis, and semantic network analysis.
|
[SECTION: Findings] In South Korea, restrictions on political speech surrounding elections are more stringent than in many other countries. The Public Official Election Act (hereinafter POEA, enacted in March 1994 as a result of integration of four different election laws corresponding to different layers of public administration) contains a number of provisions prohibiting campaign activities that would be standard practice in most democratic countries. According to Article 59, for example, official campaigns are allowed only during a period from the day following the closing date of candidate registration to the eve of the election. This amounts to 23 days in the case of presidential elections, and 14 days for elections of legislators and local governors. Even within this brief designated period, campaigns are subject to tight protocols prescribed in Articles 58 ("Definition of election campaign"), 60 ("Persons barred from election campaign"), 68 ("Campaign sashes"), 92 ("Prohibition of election campaign using motion pictures, etc."), 98 ("Restriction on use of broadcast for election campaign"), 99 ("Prohibition of election campaign by internal broadcast, etc."), and 100 ("Prohibition of use of recorders, etc."), to name a few.Having a more direct effect on individual voters are Article 90 and Article 93, which ban the display or distribution of election-related paraphernalia in the 180 days prior to an election. Section 1 of Article 93 states that "no one shall distribute or display advertisements, letters, posters, photographs, documents, drawings, printed materials, audiotapes, videotapes or the like that convey endorsement of or opposition to a candidate or a party" during the 180 days.The already overarching scope of Article 93(1) has been further stretched since the National Election Commission (NEC) and the Public Prosecutors' Office started to apply it to the online context in 1996. Tracking down the author of an online source and enforcing this provision is practically easy in Korean cyberspace, in that all Korean users are law-bound to verify their real identities (by providing their resident ID numbers) when joining major online services. Thereby, numerous bulletin board entries, blog posts, viewer comments on news sites, and user-generated content on Web 2.0 platforms have resulted in legal ramifications ranging from fines to imprisonment.The 2007 presidential election The latest presidential election in 2007 epitomized this long-standing conflict between citizens and institutions in the Korean electoral environment and how the global internet affected ways in which this conflict played out. The 2007 election saw Korea's cyberspace surprisingly silent, at least on the surface (Park and Lee, 2008). The victor seemed predetermined; Lee Myung-Bak, the conservative Grand National Party[1] (GNP) candidate, maintained a huge lead in the opinion polls since his nomination. However, the event took an unexpected sharp turn when, shortly before the day of the election, the rival United New Democratic Party[2] (UNDP) released two video clips implicating Lee in a financial scandal.The first clip showed Lee, in an interview by a journalist[3] in 2000, stating that he was directly involved in the establishment of BBK, a financial firm that was investigated in 2001 and found to be a scam. This contradicted Lee's campaign speeches denying any connection to BBK, as well as state prosecutors' conclusion clearing him of any involvement. The video clip was then rapidly circulated among the voters, even though the NEC restricted the sharing of such information in cyberspace, based on Article 93(1) of the POEA. Voters opposing Lee abandoned domestic social networking sites subject to the country's online real-name verification system and moved to foreign websites to further discuss the candidacy of Lee. The clip, which was posted on YouTube on December 5, 2007, generated more than 900,000 views in two weeks. It was viewed more often than any other user-generated video content available in Korean cyberspace during the election (Chang and Park, 2009, p. 93). As a result, this phenomenon popularized the term "cyber-exile" in the media (The Hankyoreh, 2009), and the term has become increasingly in use since. No news reports mentioned the term between 2000 and 2007, but between 2008 and 2010, it was mentioned 52 times[4].The second video clip was posted on the website of the Hankyoreh, a liberal daily, three days before the election. This 150-second clip showed Lee giving a speech at a local university in 2000, once again introducing himself as the founder and successful entrepreneur of BBK. According to The Hankyoreh's web statistics, this exclusive scoop attracted 1.5 million viewers by the time of the election. The clip also appeared on Daum, a major South Korean web portal, receiving more than 40,000 comments.Lee overcame this scandal and went on to win the election, but these video clips posed a serious threat to his presidential bid and has been a recurring issue throughout his term in office. In this study, we will examine how the two video clips were shared and discussed among the Korean online users, in order to explore the causes and consequences of the "cyber-exile" phenomenon, as well as its political implications for Korea in addition to the global community. Given the mainstream popularity of Web 2.0 services in recent years, current literature offers extensive discussion regarding the political implications of the increasing practice of digitally mediated social networking and content sharing. Specific interest has been shown in the question of whether such practice increases voter turnout or other forms of political participation in the traditional sense of the term. Research findings have, however, been mixed so far; while some consider that the use of Facebook and other social networking sites (SNSs) for political purposes is a significant predictor of general political participation (e.g. Vitak et al., 2011), others suggest that reliance on SNSs is not necessarily related to the increase in political participation (e.g. Baumgartner and Morris, 2010; Zhang et al., 2010).This broad discussion itself is inconclusive and is in need of further exploration, but for the purposes of this paper, we decided to focus on the political activity on and through YouTube. This commercially successful video-sharing site often features among SNSs in journalistic and scholarly accounts (boyd and Ellison, 2007), and this fact appropriately captures some, but not all, of its characteristics. It indeed provides a range of social networking features, such as "Comments", "Favorite", "Share" and "Subscriptions". However, it also stands apart from other SNSs, due to it being a unique hybrid form of social media and mass media. Despite social networks that underlie or emerge from its use (Lange, 2007b; Rotman and Golbeck, 2011; Sifman, 2011), YouTube originally has a character of "a media distribution platform, kind of like television" (Burgess and Green, 2009, p. 3). Burgess and Green (2009, p. 3) argue that "YouTube's ascendancy has occurred amid a fog of uncertainty and contradiction around what it is actually for." It is, so the two authors' argument goes, a new breed of businesses, or an example of what Weinberger (2007) calls "meta businesses", where old media heavyweights and amateur content creators form a curious cohabitation (Burgess and Green, 2009, pp. 4-5).Given that its format induces user participation, YouTube along with other Web 2.0 models revitalized the early internet idealism for better democracy (Bruns, 2008; Jenkins, 2006; Weinberger, 2007). As Marwick (2007) also pointed out based on a content analysis of news coverage, YouTube has been portrayed in a rather celebratory tone in the mass media for its democratic (or democratizing) potential. This kind of portrayal has been supported by a handful of anecdotal cases. During the 2006 midterm elections in the USA, a YouTube-publicized gaffe, later called the "macaca moment", cost Republican Senator George Allen (Virginia) his re-election bid, despite his huge initial lead in the opinion polls over the Democratic challenger, Jim Webb. This result was attributed to a video clip showing Allen making racially discriminatory remarks, targeting one of Webb's volunteers, during a campaign tour. The incident quickly became a widely publicized issue once the footage was posted on YouTube, which effectively led to Allen's defeat (Sidarth, 2006). Since then, the term "macaca moment" has been used to refer to "high-profile candidate gaffes that are captured on YouTube, receive a cascade of citizen views and contribute to some substantial political impact" (Karpf, 2010). A further example of such an incident took place in the 2007 Finnish national elections (Carlson and Strandberg, 2008, p. 171).The discussion of the political implications of YouTube intensified around the 2008 US presidential election because that was when Web 2.0 applications started to be incorporated into the mainstream campaign repertoire in the USA. Then-candidate Barack Obama received considerable media attention for his online presence, which outshone that of the rival candidate Hillary Clinton during the primaries and later that of his Republican opponent John McCain. Obama's campaign made heavy use of Facebook and YouTube to engage the attention of younger voters (Young, 2008), securing record campaign contributions mainly through online donations (Cooper, 2008; Carpenter, 2010).The 2008 election is widely considered as having been a pivotal moment in the US campaign history. Carpenter (2010) and Ricke (2010), for example, pointed out that YouTube, particularly its joint projects with PBS ("Video Your Vote") and with CNN ("The CNN/YouTube Debates"), afforded more room than ever for ordinary voters to participate in the campaign process and consequently served as "an instrument of 'checks and balances'" (Carpenter, 2010, 223). Some other scholars focused on how individual candidates made use of YouTube to deliver their messages in this particular election. Church (2010) examined leadership discourse by analyzing YouTube clips, featuring 16 candidates in the race, and suggested that given the emergence of what he termed "the postmodern constituency" (Church, 2010, p. 138), as well as the unfiltered nature of the medium (Church, 2010, p. 139), voters' focus shifted from candidates' political experience to their character. Duman and Locher (2008) examined Barack Obama and Hilary Clinton's YouTube campaigns and highlighted how the two presidential hopefuls attempted to create and uphold a conversational allusion through their videos.In regards to YouTube, another growing body of literature is concerned with the patterns of user interaction in this "viral marketing wonderland" (Burgess and Green, 2009, p. 3), and the methodological implications of exploring them. Three themes were identified for our research design. First, some studies have focused on what motivates users to share their videos on YouTube in the first place, ranging from perceived usefulness to interpersonal norms (e.g. Yang et al., 2010). Second, from a similar yet more specific perspective, some have established that YouTube is not only a source for information or entertainment for individual purposes, but also a new terrain for social interaction involving acts of "co-viewing" (Haridakis and Hanson, 2009), video responses (Adami, 2009), peer comments (Lange, 2007a; Jones and Schieffelin, 2009), and links to Facebook Walls (Robertson et al., 2010). In this sense, Chu (2009) went further to argue that YouTube plays a role as a "cultural public sphere".That said, others have taken a cautionary stance regarding YouTube's capacity to function as a public sphere. For example, Hess (2009) suggested there are certain obstacles to public deliberation on YouTube, such as the platform's dismissive and playful atmosphere. Moreover, Carpentier (2009) analyzed 16plus, a YouTube-like online platform provided by the north Belgian public broadcaster VRT, and suggested that the users may not be as appreciative of additional means of participation as expected. Lange (2007a) found that video-based communication on YouTube is no less hostile than faceless text-based communication. Blitvich (2010) added that impolite comments, although sometimes strategically employed, are likely to result in polarization.Chadwick (2009) argued that deliberative processes and "thick" citizenship cannot be the sole yardstick for the assessment of the functioning of e-democracy in the Web 2.0 era. However, if YouTube proves to be an unviable site for discussing serious political issues, it is important to ask why that is so. Answers to this question would advance our understanding of its political potential and how to harness it. As established in the previous section, YouTube is a uniquely interesting environment for political communication. What merits further attention is how this global platform intersects with local political and cultural dynamics. This is an underexplored line of inquiry, especially in the existing literature, which has been predominantly based on incidents from Western polities. With the aim of contributing to filling this lacuna, we address in the present paper the following three questions.1. What were the salient features of the discussion surrounding the YouTube clip, and how did that discussion develop during the campaign period?2. What were the salient features of the discussion surrounding the Daum clip, and how did that discussion develop during the campaign period?3. To what extent and how did the two discussions differ, and what lessons can be drawn from the differences (if any) in terms of harnessing the political energy of Web 2.0? Data collection The two video clips observably changed the course of the campaign, and the present study was conducted in "real time" as the event unfolded. As soon as we identified the research need to examine how the discussions developed in connection with the first clip posted on YouTube on December 5, 2007, we started to collect different types of data to have as detailed understanding as possible regarding how the clip was shared and discussed among the Korean voters.First, we monitored YouTube users' reaction manifested in terms of view counts, number of comments, and number of Favorites.Second, we collected all user comments added to the clip. By the election day of December 19, a total of 436 comments were manually retrieved.Third, we gathered the URLs of webpages which either sent a hyperlink to the YouTube clip or quoted its address www.youtube.com/watch?v=8ZFAywhGTgk[5] in the text. The URLs were retrieved from search engine indices on December 11, 14, and 17, 2007. This third dataset was particularly important to our study, as we aimed to investigate where YouTube was situated in a broader context of the Korean presidential election in question. For this facet of data collection, we used three search engines - Yahoo, Google, and Naver (the most popular search engine in Korea) - because we expected that these search engines would complement one another in terms of coverage. It turned out that Yahoo was better at locating webpages that sent out links to this particular clip, and Google was better at retrieving webpages quoting the URL in the body of the content, than Naver. Nonetheless, we included Naver search results in our data corpus because of Naver's dominance over Yahoo and Google in the Korean search engine market (Lee, 2009, p. 312).Once the second video clip was disclosed through the online edition of the local daily Hankyoreh, the scope of our data collection expanded. We collected all user comments left under the clip posted on the web portal Daum. Founded as early as 1995, Daum is one of the three largest portals in South Korea (along with Naver and Nate) and offers a wide range of services such as email, online communities (Daum Cafe), public forums (Daum Agora), blogs, web searches, and news aggregation. The Hankyoreh site also provides some space for reader comments, but we decided to collect the data from Daum because domestic web portals play a uniquely influential role in Koreans' news consumption patterns (Lim, 2005; Kim and Park, 2008). More relevant to this study, Daum users are known to be particularly active in political mobilization (Yun, 2009; Chang and Park, 2012).As a result, we manually collected a total of 43,493 comments under 7,641 unique screen IDs by the election day. Most comments (39,120 out of 43,493) were made on the eve of the election.Data analysis After the establishment of data corpus, we employed a combination of interaction network analysis, semantic network analysis, and hyperlink analysis.In more detail, first, for the interaction network analysis, the user comments, collected from both YouTube and Daum, were coded in terms of the attitudes conveyed in each entry and were sorted by the user IDs. The coding scheme was straightforward: each comment was manually classified into one of the Positive, Negative, and Neutral categories, depending on its contributor's attitude toward candidate Lee. Positive comments included messages supporting Lee, such as those arguing that the video clip was either quoted out of context or doctored, or arguing that despite the scandal Lee's strengths outweighed his shortcomings as a potential president. On the other hand, negative comments included those criticizing Lee and/or the prosecution team that vindicated him. Neutral comments included messages from users who appeared to be unaware of or ambivalent about the BBK scandal. Irrelevant or nonsensical comments were classified as "Others."Next, we conducted a content and discourse analysis of the collected comments from both YouTube (420 comments) and Daum (400 comments, chosen by a systematic sampling of every tenth entry) to identify any semantic patterns emerging from the respective discussions. For this, we conducted a keyword-in-context (KWIC) analysis, a methodological strategy that focuses on the correlation between a given word and its surroundings within the text. KWIC refers to a set of methods that process large bodies of messages, through the identification key words or phrases within a text (Biddix et al., 2009). For the purpose of this study, we used KrKwic (Korean Key Word In Context), a free software package that was developed by Park and Leydesdorff (2004) for KWIC analysis in the Korean language.We also charted 100 critical words from the 400 Daum comments, according to KWIC results, and conducted a CONCOR (convergence of iterated correlations) analysis of those 100 words to identify any issue clusters. CONCOR is a procedure that partition words into positions based on the structural equivalence (Wasserman and Faust, 1994, pp. 375-381). This partition generates subsets of the original set of words. Each of these subsets contains words that have similar connections to and from other words in the text. KrKwic and CONCOR have been used as means of forming content analysis of the Korean texts. For example, Chung and Park (2010) have recently used both KrKwic and CONCOR to compare the ideological stances of two Korean presidents, using their inaugural addresses, concluding that two methods greatly assisted in categorizing Korean words using the semantic links between words grounded in the text.Finally, we sorted the webpages that were linked to the YouTube clip according to their URLs and visited them manually to identify the nature of their content and link. As discussed in the following section, this qualitative approach allowed for an examination of Korean users' innovative efforts to circumvent Article 93(1) of the POEA.Table I summarizes the techniques employed in this study. We conducted the CONCOR analysis only for the Daum clip because the YouTube comments provided little text. On the other hand, the Daum clip required no hyperlink analysis because Korean users are known for their tendency to 'flock' to the semi-fenced playground of web portals, and particularly that of Daum, for political debates (Yun, 2009; Chang and Park, 2012). Patterns of interactions among users As shown in Table II, YouTube users paid considerable attention to the "BBK interview" footage. Most of the comments leveled criticism at Lee, as summarized in Table III. Likewise, the footage of Lee's guest speech also attracted a huge number of viewer comments on Daum, mostly critical of him.The data indicated a power-law distribution of participation on both YouTube and Daum (although any power-law interpretation should be made with caution because our datasets were relatively small). In other words, only a handful of users accounted for a large part of the commentary. For example, nine out of 13 comments on YouTube by JapanEmpire (a screen name) supported Lee, including the following:GNP candidate and former Seoul mayor Lee Myung-bak is a wealthy, experienced, and sincere Christian, I'm sure he doesn't lie like others. Also, he will be the second best president after Park Chung-hee of the ROK.Korea needs a great president with strong leadership who will really improve the economy and possibly eliminate corruption (quoted verbatim).On the other hand, Seebuddy9 left 12 comments that consistently opposed Lee, including the following:Hey Mr Lee! You are not qualified as Korean president. You know why? Korean people saw your bottom of immorality. What about your criminal record in Korean police too. Now, Korean people fighting corruption against Mr Lee and grand national party. The president of Korea should required transparency with strong accountability. That is reason you are not qualified. Withdraw your candidacy immediately! (quoted verbatim).Our observation was that ordinary users were more likely to respond to messages put up by a few highly active contributors than to initiate a thread themselves. Except for those highly active contributors (in the case of YouTube, danfrphils, JapanEmpire, Youngkuk, and seebuddy9, among others, occupying the central position in the network in Figure 1), the number of new comments decreased rapidly over time. However, this does not necessarily imply that voters' interest in the issue was short-lived. Instead, the discussion became dispersed, as duplicates of the video clip started to appear on YouTube.The data also indicated the preferential bonding among like-minded users. In other words, users tended to interact more with others sharing the same opinion. Figure 1 visualizes these interaction patterns. The shape represents the coded attitude: square means "Positive"; inverted triangle, "Negative"; triangle, "Neutral"; and circle, "Others." The size of each node is proportional to that of the number of "replies" exchanged. Isolators (i.e. those who did not send nor receive a reply) were excluded from the visualization.The use of this visualization technique allowed us to explore the variant ways in which YouTube users appropriated the comment facility and conversed among themselves. Inverted triangles show some affinity with other inverted triangles. Square nodes show the same pattern. In particular, those opposing Lee communicated with one another most frequently. This is consistent with the findings of Park and his colleagues (Park and Jankowski, 2008; Park and Thelwall, 2008); they found that SNS users who belong to or sympathize with the progressive camp in Korea are more connected among themselves than SNS users who support the conservative Grand National Party.Having said that, those interaction patterns we observed were only suggestive, rather implying that during the campaign period, YouTube users tended to express their opinions and leave it at that, instead of engaging in further discussions or debates with others (see also Vergeer and Hermans, 2008). This finding is generally consistent with the findings of Mislove et al. (2007), who suggested that YouTube users are less likely to form interest groups than users of other SNSs (e.g. Live Journal, Flickr, and Orkut).Semantic and discursive patterns of comments The proportion of irrelevant comments was interestingly high on YouTube. Comments in the "Others" category, in Table III, included those containing unintelligible phrases and, more notably, those engaging in "cyber-flaming" (i.e. directing unmotivated anonymous abuse toward other users). The discussion on YouTube was occasionally derailed by racist attacks. For example, the very first comment about the video clip was:Stupid Korean want to make crook to the president ... haha (itman21c, quoted verbatim).This was immediately followed byWho's the president you talking about? Do you even know what you're saying, itman21c? Just shut up, retarded singapole, singapore whatever! (Youngkuk, quoted verbatim).Given this observation, we applied the same coding scheme to the YouTube comments in collection but focused on the country, not Lee (as seen in Table III), for exploratory purposes. The results are summarized in Table IV, pointing to the high proportion of irrelevant comments in English.Comments added to the Daum clip took playful forms, specific to the Korean online culture, such as emoticons (e.g. 238 instances of "^^," which represents smiling), internet slang or jargon (e.g. 5,470 instances of "Graphic 1," a Korean consonant representing giggling in this context), and sarcastic nicknames for those allegedly involved in the BBK scandal (e.g. 220 instances of "tteok-chal," one given to the prosecution team, implying corruption) (for a detailed explanation of these terms, see Park and Lee, 2009). Despite these playful uses of language, the discussion on Daum stayed closely relevant to the topic.We also identified issue clusters from this rather intense discussion, on the basis of the CONCOR results. After reviewing user comments and considering their relevance to the topic, we labeled each cluster. As shown in Table V, there were one major cluster and seven minor clusters. The major cluster indicates commonly expressed negative opinions about Lee. Frequent keywords in this category included "Graphic 2," "Lee Myung-bak," "negative," "crook," "withdrawal of endorsement," and "Chung Dong-young" (a major opposition candidate), among others. A closer examination of the clusters revealed that the discussion surrounding the Daum video grew to cover other controversies associated with Lee (e.g. some clusters representing Lee's false residence registration and real estate speculation).Hyperlink networks around the YouTube clip Finally, Table VI summarizes the hyperlink data we collected, which were in turn supplemented and extended by a qualitative analysis of the actual content of each webpage.The video clip spread rapidly in Korean cyberspace through personal blogs, the Naver Q&A service, online communities, and public bulletin boards on web portals/newspaper sites, in that order. In these cases, the URL was simply presented as text without displaying the actual video content and expressing opinions about it, which allowed for the circumvention of Article 93(1) of the POEA. Q1. What were the salient features of the discussion surrounding the YouTube clip? In terms of the first research question, the YouTube clip targeted an audience of Korean voters exclusively. This was self-evident because the clip was in Korean with no subtitles, and the original contributor's message, explicitly addressed to the Korean public, was also in Korean with no translation. The contributor made his motive clear vis-a-vis the media attention he unwittingly caused. In a 1,150-word message next to the clip, he explained that he could not accept the state prosecutors' conclusion that Lee was not guilty. He also added that he first considered posting the clip on the bulletin board on the website of the Public Prosecutors' Office, as a gesture of protest, but decided to post it on YouTube instead because of (in his own words) the "bother" that he would otherwise have received from the NEC. He also emphasized that contrary to the mass media's speculation, he was neither a member of any party nor a campaigner for any candidate. This incident illustrates how Korean users appropriated a global social networking space for discussion of domestic political issues, specifically to circumvent their country's election law.It is noteworthy that the diffusion of information and the political discussion with respect to the YouTube clip stayed within the Korean cyberspace, bounded by the Korean language. According to the YouTube statistics, most of the "referrals" were from Korean news media websites. As of December 11, Kyunghyang (a daily newspaper) accounted for the largest number of views. Kyunghyang's news story (http://news.khan.co.kr/kh_news/khan_art_view.html?artid=200712061653151 &code=940301) about the video generated 13,385 views, followed by the Hankyoreh's news story (www.hani.co.kr/arti/politics/politics_general/255686.html), which generated 10,699 views. Both progressive newspapers had the YouTube clip embedded in the news story. The results based on this study's collection of search engine indices indicate that, despite Article 93(1) of the POEA, a number of individual voters shared the clip by relaying its URL without providing detailed information or comments regarding its content.YouTube is not a "walled garden" (i.e. a closed or exclusive virtual environment) and is thus susceptible to irrelevant comments. Consequently, its comment facility did not serve well as a space for further discussion in this case. The racially motivated attacks were initiated by those who were apparently not directly concerned with Korean politics. Our results were in lines with Gueorguieva's (2008) findings, which suggested that YouTube and other popular SNSs have posed a new set of challenges to campaigns, particularly the risk of reduced control over the images and messages of the campaigns.Q2. What were the salient features of the discussion surrounding the Daum clip? Beyond the abundance of playful emoticons and internet jargon, as shown in Table V, the discussion on Daum was quantitatively intense, as evidenced by the number of comments, and stayed relevant to the video clip. The irony is that if it had not been for the Hankyoreh and Daum, the vibrant discussion would not have been possible because of the NEC's enforcement of the POEA.Q3. To what extent did the two discussions differ and what do the differences suggest? The difference between the two discussions illustrated above can be explained by the characteristics of their respective spaces. As discussed earlier, YouTube is a complex media environment that defies an easy definition. Its characteristic openness made it relatively easy for Korean voters to employ it as an alternative channel for their discussions and debates outside election laws. At the same time, however, YouTube is more likely to be composed of uninterested parties than domestic sites, and therefore YouTube users might have distracted Korean voters' discussions and debates away from issues surrounding Korean politics. The address of the YouTube clip was widely circulated in Korean cyberspace, through blogs and online forums. However, no in-depth discussion took place because this "sphere" was fragmented. The dialectics between global media and local contexts have indeed been well documented (e.g. Turner, 2005), and the internet has complicated the matter even further. Contrary to the general description of the internet as a supranational network of computers, what we have is, in boyd's (2006) words, "all sorts of local cultures connected through a global network, resulting in all sorts of ugly tensions."What is unique about the case studied in this paper is, however, that users turned to the global space in order to circumvent a local political conflict, not the other way around - hence the neologism "cyber-exile". During the 2007 presidential election, Korean voters used YouTube to share election-related information because the laws of their country prohibited them from doing so on domestic websites. YouTube provided them with a higher level of anonymity than Korean sites. This incident of cyber-exile illustrates the tension between internet-mediated grassroots political activity and the authorities' restrictive interpretation and application of existing laws - the POEA in this case - to curtail the activity (see also Lee, 2011).However, the discussion surrounding the YouTube video clip was brought "back" to the Korean cyberspace. Korean users are generally known for having a strong preference for local services (Lee, 2009, 312), but more importantly, YouTube could not provide a suitable environment for this particular discussion. Dialogs within YouTube's comment facility, among Koreans as well as non-Korean users, were often off the topic, and in some instances, unexpectedly turned into "racist flaming".Zuckerman (2010) analyzed various tools for circumventing the internet censorship worldwide and suggested that circumvention cannot be a long-term solution. His argument is focused largely on technical aspects, but it still leads to important questions of how, and to what extent, we then should go "beyond" circumvention.We, the authors of this paper, do not intend to advocate the "walled garden" model for online forums, as opposed to the wild of YouTube, nor do we wish to suggest there should be zero government intervention. Our findings indicate that users can circumvent local regulations via the internet and other digital communication technologies (and probably more will do so), but subsequent discussions are likely to become fragmented as a result. Despite the specifics of the case studied, the significance of the present paper lies in the fact that epitomizes the tension between "old" laws and "new" media. Moreover, our findings clearly demonstrate that an innovative circumvention attempt on the user's part is not enough to harness the potential of online discussion for measured, sustained discourse of the issue at hand.This study has an important limitation. We conducted the analysis on a real-time basis as the event unfolded, and therefore the need to compare YouTube and Daum arose only during the later stages of data collection. As a result, a point-to-point comparison was not feasible, which invites further research.Noteworthy is that on December 29, 2011, the Constitutional Court of South Korea declared the unconstitutionality of the NEC's extended application of Article 93(1) of the POEA to social media, particularly Twitter. With the next presidential election scheduled for December 2012, future research should examine the impact of this court decision on the long-awaited relaxation of restrictions on election-related online communication in the country. An earlier version of this paper was presented at the Oxford Internet Institute's 2010 conference, and some of the findings (answering different research questions) were published in a Korean-language journal. The authors are grateful to Ae-Jie Bae for her assistance in data collection. Opens in a new window.Graphic 1 Opens in a new window.Graphic 2 Opens in a new window.Graphic 3 Opens in a new window.Graphic 4 Opens in a new window.Graphic 5 Opens in a new window.Figure 1 Interaction network of YouTube users commenting on Lee Opens in a new window.Table I Summary of methodological techniques Opens in a new window.Table II YouTube users' reaction to the "BBK Interview" video Opens in a new window.Table III Attitudes toward Lee Opens in a new window.Table IV Attitudes toward Korea Opens in a new window.Table V Frequent keywords for eight CONCOR clusters Opens in a new window.Table VI Number of webpages directing users to the YouTube video clip
|
- YouTube served as a medium for Korean voters to circumvent local electoral regulations, thus implying the neologism "cyber-exile". However, unlike Daum, YouTube failed to facilitate discussions on the posted video clip. The discussion through its comment feature was often derailed by irrelevant comments from seemingly uninterested parties. The address of the video clip was shared through personal blogs and online bulletin boards in Korean cyberspace, but these efforts led only to a fragmented sphere.
|
[SECTION: Value] In South Korea, restrictions on political speech surrounding elections are more stringent than in many other countries. The Public Official Election Act (hereinafter POEA, enacted in March 1994 as a result of integration of four different election laws corresponding to different layers of public administration) contains a number of provisions prohibiting campaign activities that would be standard practice in most democratic countries. According to Article 59, for example, official campaigns are allowed only during a period from the day following the closing date of candidate registration to the eve of the election. This amounts to 23 days in the case of presidential elections, and 14 days for elections of legislators and local governors. Even within this brief designated period, campaigns are subject to tight protocols prescribed in Articles 58 ("Definition of election campaign"), 60 ("Persons barred from election campaign"), 68 ("Campaign sashes"), 92 ("Prohibition of election campaign using motion pictures, etc."), 98 ("Restriction on use of broadcast for election campaign"), 99 ("Prohibition of election campaign by internal broadcast, etc."), and 100 ("Prohibition of use of recorders, etc."), to name a few.Having a more direct effect on individual voters are Article 90 and Article 93, which ban the display or distribution of election-related paraphernalia in the 180 days prior to an election. Section 1 of Article 93 states that "no one shall distribute or display advertisements, letters, posters, photographs, documents, drawings, printed materials, audiotapes, videotapes or the like that convey endorsement of or opposition to a candidate or a party" during the 180 days.The already overarching scope of Article 93(1) has been further stretched since the National Election Commission (NEC) and the Public Prosecutors' Office started to apply it to the online context in 1996. Tracking down the author of an online source and enforcing this provision is practically easy in Korean cyberspace, in that all Korean users are law-bound to verify their real identities (by providing their resident ID numbers) when joining major online services. Thereby, numerous bulletin board entries, blog posts, viewer comments on news sites, and user-generated content on Web 2.0 platforms have resulted in legal ramifications ranging from fines to imprisonment.The 2007 presidential election The latest presidential election in 2007 epitomized this long-standing conflict between citizens and institutions in the Korean electoral environment and how the global internet affected ways in which this conflict played out. The 2007 election saw Korea's cyberspace surprisingly silent, at least on the surface (Park and Lee, 2008). The victor seemed predetermined; Lee Myung-Bak, the conservative Grand National Party[1] (GNP) candidate, maintained a huge lead in the opinion polls since his nomination. However, the event took an unexpected sharp turn when, shortly before the day of the election, the rival United New Democratic Party[2] (UNDP) released two video clips implicating Lee in a financial scandal.The first clip showed Lee, in an interview by a journalist[3] in 2000, stating that he was directly involved in the establishment of BBK, a financial firm that was investigated in 2001 and found to be a scam. This contradicted Lee's campaign speeches denying any connection to BBK, as well as state prosecutors' conclusion clearing him of any involvement. The video clip was then rapidly circulated among the voters, even though the NEC restricted the sharing of such information in cyberspace, based on Article 93(1) of the POEA. Voters opposing Lee abandoned domestic social networking sites subject to the country's online real-name verification system and moved to foreign websites to further discuss the candidacy of Lee. The clip, which was posted on YouTube on December 5, 2007, generated more than 900,000 views in two weeks. It was viewed more often than any other user-generated video content available in Korean cyberspace during the election (Chang and Park, 2009, p. 93). As a result, this phenomenon popularized the term "cyber-exile" in the media (The Hankyoreh, 2009), and the term has become increasingly in use since. No news reports mentioned the term between 2000 and 2007, but between 2008 and 2010, it was mentioned 52 times[4].The second video clip was posted on the website of the Hankyoreh, a liberal daily, three days before the election. This 150-second clip showed Lee giving a speech at a local university in 2000, once again introducing himself as the founder and successful entrepreneur of BBK. According to The Hankyoreh's web statistics, this exclusive scoop attracted 1.5 million viewers by the time of the election. The clip also appeared on Daum, a major South Korean web portal, receiving more than 40,000 comments.Lee overcame this scandal and went on to win the election, but these video clips posed a serious threat to his presidential bid and has been a recurring issue throughout his term in office. In this study, we will examine how the two video clips were shared and discussed among the Korean online users, in order to explore the causes and consequences of the "cyber-exile" phenomenon, as well as its political implications for Korea in addition to the global community. Given the mainstream popularity of Web 2.0 services in recent years, current literature offers extensive discussion regarding the political implications of the increasing practice of digitally mediated social networking and content sharing. Specific interest has been shown in the question of whether such practice increases voter turnout or other forms of political participation in the traditional sense of the term. Research findings have, however, been mixed so far; while some consider that the use of Facebook and other social networking sites (SNSs) for political purposes is a significant predictor of general political participation (e.g. Vitak et al., 2011), others suggest that reliance on SNSs is not necessarily related to the increase in political participation (e.g. Baumgartner and Morris, 2010; Zhang et al., 2010).This broad discussion itself is inconclusive and is in need of further exploration, but for the purposes of this paper, we decided to focus on the political activity on and through YouTube. This commercially successful video-sharing site often features among SNSs in journalistic and scholarly accounts (boyd and Ellison, 2007), and this fact appropriately captures some, but not all, of its characteristics. It indeed provides a range of social networking features, such as "Comments", "Favorite", "Share" and "Subscriptions". However, it also stands apart from other SNSs, due to it being a unique hybrid form of social media and mass media. Despite social networks that underlie or emerge from its use (Lange, 2007b; Rotman and Golbeck, 2011; Sifman, 2011), YouTube originally has a character of "a media distribution platform, kind of like television" (Burgess and Green, 2009, p. 3). Burgess and Green (2009, p. 3) argue that "YouTube's ascendancy has occurred amid a fog of uncertainty and contradiction around what it is actually for." It is, so the two authors' argument goes, a new breed of businesses, or an example of what Weinberger (2007) calls "meta businesses", where old media heavyweights and amateur content creators form a curious cohabitation (Burgess and Green, 2009, pp. 4-5).Given that its format induces user participation, YouTube along with other Web 2.0 models revitalized the early internet idealism for better democracy (Bruns, 2008; Jenkins, 2006; Weinberger, 2007). As Marwick (2007) also pointed out based on a content analysis of news coverage, YouTube has been portrayed in a rather celebratory tone in the mass media for its democratic (or democratizing) potential. This kind of portrayal has been supported by a handful of anecdotal cases. During the 2006 midterm elections in the USA, a YouTube-publicized gaffe, later called the "macaca moment", cost Republican Senator George Allen (Virginia) his re-election bid, despite his huge initial lead in the opinion polls over the Democratic challenger, Jim Webb. This result was attributed to a video clip showing Allen making racially discriminatory remarks, targeting one of Webb's volunteers, during a campaign tour. The incident quickly became a widely publicized issue once the footage was posted on YouTube, which effectively led to Allen's defeat (Sidarth, 2006). Since then, the term "macaca moment" has been used to refer to "high-profile candidate gaffes that are captured on YouTube, receive a cascade of citizen views and contribute to some substantial political impact" (Karpf, 2010). A further example of such an incident took place in the 2007 Finnish national elections (Carlson and Strandberg, 2008, p. 171).The discussion of the political implications of YouTube intensified around the 2008 US presidential election because that was when Web 2.0 applications started to be incorporated into the mainstream campaign repertoire in the USA. Then-candidate Barack Obama received considerable media attention for his online presence, which outshone that of the rival candidate Hillary Clinton during the primaries and later that of his Republican opponent John McCain. Obama's campaign made heavy use of Facebook and YouTube to engage the attention of younger voters (Young, 2008), securing record campaign contributions mainly through online donations (Cooper, 2008; Carpenter, 2010).The 2008 election is widely considered as having been a pivotal moment in the US campaign history. Carpenter (2010) and Ricke (2010), for example, pointed out that YouTube, particularly its joint projects with PBS ("Video Your Vote") and with CNN ("The CNN/YouTube Debates"), afforded more room than ever for ordinary voters to participate in the campaign process and consequently served as "an instrument of 'checks and balances'" (Carpenter, 2010, 223). Some other scholars focused on how individual candidates made use of YouTube to deliver their messages in this particular election. Church (2010) examined leadership discourse by analyzing YouTube clips, featuring 16 candidates in the race, and suggested that given the emergence of what he termed "the postmodern constituency" (Church, 2010, p. 138), as well as the unfiltered nature of the medium (Church, 2010, p. 139), voters' focus shifted from candidates' political experience to their character. Duman and Locher (2008) examined Barack Obama and Hilary Clinton's YouTube campaigns and highlighted how the two presidential hopefuls attempted to create and uphold a conversational allusion through their videos.In regards to YouTube, another growing body of literature is concerned with the patterns of user interaction in this "viral marketing wonderland" (Burgess and Green, 2009, p. 3), and the methodological implications of exploring them. Three themes were identified for our research design. First, some studies have focused on what motivates users to share their videos on YouTube in the first place, ranging from perceived usefulness to interpersonal norms (e.g. Yang et al., 2010). Second, from a similar yet more specific perspective, some have established that YouTube is not only a source for information or entertainment for individual purposes, but also a new terrain for social interaction involving acts of "co-viewing" (Haridakis and Hanson, 2009), video responses (Adami, 2009), peer comments (Lange, 2007a; Jones and Schieffelin, 2009), and links to Facebook Walls (Robertson et al., 2010). In this sense, Chu (2009) went further to argue that YouTube plays a role as a "cultural public sphere".That said, others have taken a cautionary stance regarding YouTube's capacity to function as a public sphere. For example, Hess (2009) suggested there are certain obstacles to public deliberation on YouTube, such as the platform's dismissive and playful atmosphere. Moreover, Carpentier (2009) analyzed 16plus, a YouTube-like online platform provided by the north Belgian public broadcaster VRT, and suggested that the users may not be as appreciative of additional means of participation as expected. Lange (2007a) found that video-based communication on YouTube is no less hostile than faceless text-based communication. Blitvich (2010) added that impolite comments, although sometimes strategically employed, are likely to result in polarization.Chadwick (2009) argued that deliberative processes and "thick" citizenship cannot be the sole yardstick for the assessment of the functioning of e-democracy in the Web 2.0 era. However, if YouTube proves to be an unviable site for discussing serious political issues, it is important to ask why that is so. Answers to this question would advance our understanding of its political potential and how to harness it. As established in the previous section, YouTube is a uniquely interesting environment for political communication. What merits further attention is how this global platform intersects with local political and cultural dynamics. This is an underexplored line of inquiry, especially in the existing literature, which has been predominantly based on incidents from Western polities. With the aim of contributing to filling this lacuna, we address in the present paper the following three questions.1. What were the salient features of the discussion surrounding the YouTube clip, and how did that discussion develop during the campaign period?2. What were the salient features of the discussion surrounding the Daum clip, and how did that discussion develop during the campaign period?3. To what extent and how did the two discussions differ, and what lessons can be drawn from the differences (if any) in terms of harnessing the political energy of Web 2.0? Data collection The two video clips observably changed the course of the campaign, and the present study was conducted in "real time" as the event unfolded. As soon as we identified the research need to examine how the discussions developed in connection with the first clip posted on YouTube on December 5, 2007, we started to collect different types of data to have as detailed understanding as possible regarding how the clip was shared and discussed among the Korean voters.First, we monitored YouTube users' reaction manifested in terms of view counts, number of comments, and number of Favorites.Second, we collected all user comments added to the clip. By the election day of December 19, a total of 436 comments were manually retrieved.Third, we gathered the URLs of webpages which either sent a hyperlink to the YouTube clip or quoted its address www.youtube.com/watch?v=8ZFAywhGTgk[5] in the text. The URLs were retrieved from search engine indices on December 11, 14, and 17, 2007. This third dataset was particularly important to our study, as we aimed to investigate where YouTube was situated in a broader context of the Korean presidential election in question. For this facet of data collection, we used three search engines - Yahoo, Google, and Naver (the most popular search engine in Korea) - because we expected that these search engines would complement one another in terms of coverage. It turned out that Yahoo was better at locating webpages that sent out links to this particular clip, and Google was better at retrieving webpages quoting the URL in the body of the content, than Naver. Nonetheless, we included Naver search results in our data corpus because of Naver's dominance over Yahoo and Google in the Korean search engine market (Lee, 2009, p. 312).Once the second video clip was disclosed through the online edition of the local daily Hankyoreh, the scope of our data collection expanded. We collected all user comments left under the clip posted on the web portal Daum. Founded as early as 1995, Daum is one of the three largest portals in South Korea (along with Naver and Nate) and offers a wide range of services such as email, online communities (Daum Cafe), public forums (Daum Agora), blogs, web searches, and news aggregation. The Hankyoreh site also provides some space for reader comments, but we decided to collect the data from Daum because domestic web portals play a uniquely influential role in Koreans' news consumption patterns (Lim, 2005; Kim and Park, 2008). More relevant to this study, Daum users are known to be particularly active in political mobilization (Yun, 2009; Chang and Park, 2012).As a result, we manually collected a total of 43,493 comments under 7,641 unique screen IDs by the election day. Most comments (39,120 out of 43,493) were made on the eve of the election.Data analysis After the establishment of data corpus, we employed a combination of interaction network analysis, semantic network analysis, and hyperlink analysis.In more detail, first, for the interaction network analysis, the user comments, collected from both YouTube and Daum, were coded in terms of the attitudes conveyed in each entry and were sorted by the user IDs. The coding scheme was straightforward: each comment was manually classified into one of the Positive, Negative, and Neutral categories, depending on its contributor's attitude toward candidate Lee. Positive comments included messages supporting Lee, such as those arguing that the video clip was either quoted out of context or doctored, or arguing that despite the scandal Lee's strengths outweighed his shortcomings as a potential president. On the other hand, negative comments included those criticizing Lee and/or the prosecution team that vindicated him. Neutral comments included messages from users who appeared to be unaware of or ambivalent about the BBK scandal. Irrelevant or nonsensical comments were classified as "Others."Next, we conducted a content and discourse analysis of the collected comments from both YouTube (420 comments) and Daum (400 comments, chosen by a systematic sampling of every tenth entry) to identify any semantic patterns emerging from the respective discussions. For this, we conducted a keyword-in-context (KWIC) analysis, a methodological strategy that focuses on the correlation between a given word and its surroundings within the text. KWIC refers to a set of methods that process large bodies of messages, through the identification key words or phrases within a text (Biddix et al., 2009). For the purpose of this study, we used KrKwic (Korean Key Word In Context), a free software package that was developed by Park and Leydesdorff (2004) for KWIC analysis in the Korean language.We also charted 100 critical words from the 400 Daum comments, according to KWIC results, and conducted a CONCOR (convergence of iterated correlations) analysis of those 100 words to identify any issue clusters. CONCOR is a procedure that partition words into positions based on the structural equivalence (Wasserman and Faust, 1994, pp. 375-381). This partition generates subsets of the original set of words. Each of these subsets contains words that have similar connections to and from other words in the text. KrKwic and CONCOR have been used as means of forming content analysis of the Korean texts. For example, Chung and Park (2010) have recently used both KrKwic and CONCOR to compare the ideological stances of two Korean presidents, using their inaugural addresses, concluding that two methods greatly assisted in categorizing Korean words using the semantic links between words grounded in the text.Finally, we sorted the webpages that were linked to the YouTube clip according to their URLs and visited them manually to identify the nature of their content and link. As discussed in the following section, this qualitative approach allowed for an examination of Korean users' innovative efforts to circumvent Article 93(1) of the POEA.Table I summarizes the techniques employed in this study. We conducted the CONCOR analysis only for the Daum clip because the YouTube comments provided little text. On the other hand, the Daum clip required no hyperlink analysis because Korean users are known for their tendency to 'flock' to the semi-fenced playground of web portals, and particularly that of Daum, for political debates (Yun, 2009; Chang and Park, 2012). Patterns of interactions among users As shown in Table II, YouTube users paid considerable attention to the "BBK interview" footage. Most of the comments leveled criticism at Lee, as summarized in Table III. Likewise, the footage of Lee's guest speech also attracted a huge number of viewer comments on Daum, mostly critical of him.The data indicated a power-law distribution of participation on both YouTube and Daum (although any power-law interpretation should be made with caution because our datasets were relatively small). In other words, only a handful of users accounted for a large part of the commentary. For example, nine out of 13 comments on YouTube by JapanEmpire (a screen name) supported Lee, including the following:GNP candidate and former Seoul mayor Lee Myung-bak is a wealthy, experienced, and sincere Christian, I'm sure he doesn't lie like others. Also, he will be the second best president after Park Chung-hee of the ROK.Korea needs a great president with strong leadership who will really improve the economy and possibly eliminate corruption (quoted verbatim).On the other hand, Seebuddy9 left 12 comments that consistently opposed Lee, including the following:Hey Mr Lee! You are not qualified as Korean president. You know why? Korean people saw your bottom of immorality. What about your criminal record in Korean police too. Now, Korean people fighting corruption against Mr Lee and grand national party. The president of Korea should required transparency with strong accountability. That is reason you are not qualified. Withdraw your candidacy immediately! (quoted verbatim).Our observation was that ordinary users were more likely to respond to messages put up by a few highly active contributors than to initiate a thread themselves. Except for those highly active contributors (in the case of YouTube, danfrphils, JapanEmpire, Youngkuk, and seebuddy9, among others, occupying the central position in the network in Figure 1), the number of new comments decreased rapidly over time. However, this does not necessarily imply that voters' interest in the issue was short-lived. Instead, the discussion became dispersed, as duplicates of the video clip started to appear on YouTube.The data also indicated the preferential bonding among like-minded users. In other words, users tended to interact more with others sharing the same opinion. Figure 1 visualizes these interaction patterns. The shape represents the coded attitude: square means "Positive"; inverted triangle, "Negative"; triangle, "Neutral"; and circle, "Others." The size of each node is proportional to that of the number of "replies" exchanged. Isolators (i.e. those who did not send nor receive a reply) were excluded from the visualization.The use of this visualization technique allowed us to explore the variant ways in which YouTube users appropriated the comment facility and conversed among themselves. Inverted triangles show some affinity with other inverted triangles. Square nodes show the same pattern. In particular, those opposing Lee communicated with one another most frequently. This is consistent with the findings of Park and his colleagues (Park and Jankowski, 2008; Park and Thelwall, 2008); they found that SNS users who belong to or sympathize with the progressive camp in Korea are more connected among themselves than SNS users who support the conservative Grand National Party.Having said that, those interaction patterns we observed were only suggestive, rather implying that during the campaign period, YouTube users tended to express their opinions and leave it at that, instead of engaging in further discussions or debates with others (see also Vergeer and Hermans, 2008). This finding is generally consistent with the findings of Mislove et al. (2007), who suggested that YouTube users are less likely to form interest groups than users of other SNSs (e.g. Live Journal, Flickr, and Orkut).Semantic and discursive patterns of comments The proportion of irrelevant comments was interestingly high on YouTube. Comments in the "Others" category, in Table III, included those containing unintelligible phrases and, more notably, those engaging in "cyber-flaming" (i.e. directing unmotivated anonymous abuse toward other users). The discussion on YouTube was occasionally derailed by racist attacks. For example, the very first comment about the video clip was:Stupid Korean want to make crook to the president ... haha (itman21c, quoted verbatim).This was immediately followed byWho's the president you talking about? Do you even know what you're saying, itman21c? Just shut up, retarded singapole, singapore whatever! (Youngkuk, quoted verbatim).Given this observation, we applied the same coding scheme to the YouTube comments in collection but focused on the country, not Lee (as seen in Table III), for exploratory purposes. The results are summarized in Table IV, pointing to the high proportion of irrelevant comments in English.Comments added to the Daum clip took playful forms, specific to the Korean online culture, such as emoticons (e.g. 238 instances of "^^," which represents smiling), internet slang or jargon (e.g. 5,470 instances of "Graphic 1," a Korean consonant representing giggling in this context), and sarcastic nicknames for those allegedly involved in the BBK scandal (e.g. 220 instances of "tteok-chal," one given to the prosecution team, implying corruption) (for a detailed explanation of these terms, see Park and Lee, 2009). Despite these playful uses of language, the discussion on Daum stayed closely relevant to the topic.We also identified issue clusters from this rather intense discussion, on the basis of the CONCOR results. After reviewing user comments and considering their relevance to the topic, we labeled each cluster. As shown in Table V, there were one major cluster and seven minor clusters. The major cluster indicates commonly expressed negative opinions about Lee. Frequent keywords in this category included "Graphic 2," "Lee Myung-bak," "negative," "crook," "withdrawal of endorsement," and "Chung Dong-young" (a major opposition candidate), among others. A closer examination of the clusters revealed that the discussion surrounding the Daum video grew to cover other controversies associated with Lee (e.g. some clusters representing Lee's false residence registration and real estate speculation).Hyperlink networks around the YouTube clip Finally, Table VI summarizes the hyperlink data we collected, which were in turn supplemented and extended by a qualitative analysis of the actual content of each webpage.The video clip spread rapidly in Korean cyberspace through personal blogs, the Naver Q&A service, online communities, and public bulletin boards on web portals/newspaper sites, in that order. In these cases, the URL was simply presented as text without displaying the actual video content and expressing opinions about it, which allowed for the circumvention of Article 93(1) of the POEA. Q1. What were the salient features of the discussion surrounding the YouTube clip? In terms of the first research question, the YouTube clip targeted an audience of Korean voters exclusively. This was self-evident because the clip was in Korean with no subtitles, and the original contributor's message, explicitly addressed to the Korean public, was also in Korean with no translation. The contributor made his motive clear vis-a-vis the media attention he unwittingly caused. In a 1,150-word message next to the clip, he explained that he could not accept the state prosecutors' conclusion that Lee was not guilty. He also added that he first considered posting the clip on the bulletin board on the website of the Public Prosecutors' Office, as a gesture of protest, but decided to post it on YouTube instead because of (in his own words) the "bother" that he would otherwise have received from the NEC. He also emphasized that contrary to the mass media's speculation, he was neither a member of any party nor a campaigner for any candidate. This incident illustrates how Korean users appropriated a global social networking space for discussion of domestic political issues, specifically to circumvent their country's election law.It is noteworthy that the diffusion of information and the political discussion with respect to the YouTube clip stayed within the Korean cyberspace, bounded by the Korean language. According to the YouTube statistics, most of the "referrals" were from Korean news media websites. As of December 11, Kyunghyang (a daily newspaper) accounted for the largest number of views. Kyunghyang's news story (http://news.khan.co.kr/kh_news/khan_art_view.html?artid=200712061653151 &code=940301) about the video generated 13,385 views, followed by the Hankyoreh's news story (www.hani.co.kr/arti/politics/politics_general/255686.html), which generated 10,699 views. Both progressive newspapers had the YouTube clip embedded in the news story. The results based on this study's collection of search engine indices indicate that, despite Article 93(1) of the POEA, a number of individual voters shared the clip by relaying its URL without providing detailed information or comments regarding its content.YouTube is not a "walled garden" (i.e. a closed or exclusive virtual environment) and is thus susceptible to irrelevant comments. Consequently, its comment facility did not serve well as a space for further discussion in this case. The racially motivated attacks were initiated by those who were apparently not directly concerned with Korean politics. Our results were in lines with Gueorguieva's (2008) findings, which suggested that YouTube and other popular SNSs have posed a new set of challenges to campaigns, particularly the risk of reduced control over the images and messages of the campaigns.Q2. What were the salient features of the discussion surrounding the Daum clip? Beyond the abundance of playful emoticons and internet jargon, as shown in Table V, the discussion on Daum was quantitatively intense, as evidenced by the number of comments, and stayed relevant to the video clip. The irony is that if it had not been for the Hankyoreh and Daum, the vibrant discussion would not have been possible because of the NEC's enforcement of the POEA.Q3. To what extent did the two discussions differ and what do the differences suggest? The difference between the two discussions illustrated above can be explained by the characteristics of their respective spaces. As discussed earlier, YouTube is a complex media environment that defies an easy definition. Its characteristic openness made it relatively easy for Korean voters to employ it as an alternative channel for their discussions and debates outside election laws. At the same time, however, YouTube is more likely to be composed of uninterested parties than domestic sites, and therefore YouTube users might have distracted Korean voters' discussions and debates away from issues surrounding Korean politics. The address of the YouTube clip was widely circulated in Korean cyberspace, through blogs and online forums. However, no in-depth discussion took place because this "sphere" was fragmented. The dialectics between global media and local contexts have indeed been well documented (e.g. Turner, 2005), and the internet has complicated the matter even further. Contrary to the general description of the internet as a supranational network of computers, what we have is, in boyd's (2006) words, "all sorts of local cultures connected through a global network, resulting in all sorts of ugly tensions."What is unique about the case studied in this paper is, however, that users turned to the global space in order to circumvent a local political conflict, not the other way around - hence the neologism "cyber-exile". During the 2007 presidential election, Korean voters used YouTube to share election-related information because the laws of their country prohibited them from doing so on domestic websites. YouTube provided them with a higher level of anonymity than Korean sites. This incident of cyber-exile illustrates the tension between internet-mediated grassroots political activity and the authorities' restrictive interpretation and application of existing laws - the POEA in this case - to curtail the activity (see also Lee, 2011).However, the discussion surrounding the YouTube video clip was brought "back" to the Korean cyberspace. Korean users are generally known for having a strong preference for local services (Lee, 2009, 312), but more importantly, YouTube could not provide a suitable environment for this particular discussion. Dialogs within YouTube's comment facility, among Koreans as well as non-Korean users, were often off the topic, and in some instances, unexpectedly turned into "racist flaming".Zuckerman (2010) analyzed various tools for circumventing the internet censorship worldwide and suggested that circumvention cannot be a long-term solution. His argument is focused largely on technical aspects, but it still leads to important questions of how, and to what extent, we then should go "beyond" circumvention.We, the authors of this paper, do not intend to advocate the "walled garden" model for online forums, as opposed to the wild of YouTube, nor do we wish to suggest there should be zero government intervention. Our findings indicate that users can circumvent local regulations via the internet and other digital communication technologies (and probably more will do so), but subsequent discussions are likely to become fragmented as a result. Despite the specifics of the case studied, the significance of the present paper lies in the fact that epitomizes the tension between "old" laws and "new" media. Moreover, our findings clearly demonstrate that an innovative circumvention attempt on the user's part is not enough to harness the potential of online discussion for measured, sustained discourse of the issue at hand.This study has an important limitation. We conducted the analysis on a real-time basis as the event unfolded, and therefore the need to compare YouTube and Daum arose only during the later stages of data collection. As a result, a point-to-point comparison was not feasible, which invites further research.Noteworthy is that on December 29, 2011, the Constitutional Court of South Korea declared the unconstitutionality of the NEC's extended application of Article 93(1) of the POEA to social media, particularly Twitter. With the next presidential election scheduled for December 2012, future research should examine the impact of this court decision on the long-awaited relaxation of restrictions on election-related online communication in the country. An earlier version of this paper was presented at the Oxford Internet Institute's 2010 conference, and some of the findings (answering different research questions) were published in a Korean-language journal. The authors are grateful to Ae-Jie Bae for her assistance in data collection. Opens in a new window.Graphic 1 Opens in a new window.Graphic 2 Opens in a new window.Graphic 3 Opens in a new window.Graphic 4 Opens in a new window.Graphic 5 Opens in a new window.Figure 1 Interaction network of YouTube users commenting on Lee Opens in a new window.Table I Summary of methodological techniques Opens in a new window.Table II YouTube users' reaction to the "BBK Interview" video Opens in a new window.Table III Attitudes toward Lee Opens in a new window.Table IV Attitudes toward Korea Opens in a new window.Table V Frequent keywords for eight CONCOR clusters Opens in a new window.Table VI Number of webpages directing users to the YouTube video clip
|
- This paper highlights how pre-internet institutions shape its members' political activity on the internet. In addition, the results clearly demonstrate that an innovative effort to circumvent barriers on the part of internet users is not enough to harness the potential of online discussions for a measured and sustained discourse on the issue at hand.
|
[SECTION: Purpose] In this study, I present the findings from a qualitative investigation that explored the other side of the issue of integration of migrants; that is, the views and perceptions of information intermediaries working with migrants in Israel about the integration process. The notion of integration has been the focus of academic debate in recent years and has been defined in different ways (Ager and Strang, 2008; Cheung and Phillimore, 2014; Farach et al., 2015; Gilmartin and Migge, 2015; Harder et al., 2018; Phillimore, 2012). In its broadest sense, integration means the process or transition by which people who are relatively new to a country become part of society (Rudiger and Spencer, 2003). Integration is achieved when "people, whatever their background, live, work, learn and socialise together, based on shared rights, responsibilities and opportunities" (Ndofor-Tah et al., 2019) while keeping a measure of their original cultural identity (Threadgold and Court, 2005). Integration is being characterized today as a multidimensional and multidirectional (Harder et al., 2018) process that encompasses "access to resources and opportunities as well as social mixing" involving adjustments by everyone in society (Ndofor-Tah et al., 2019). Integration is the desired outcome of migration which has been described in the literature as a transition (Bronstein, 2019; Meleis, 2015) - a disruptive process that fractures migrants' information landscapes (Lloyd, 2017) separates them from their traditional social networks, and renders their known information sources and practices inadequate (Giralt, 2015; Lloyd et al., 2013). To integrate into their host society, migrants need to find ways to redefine themselves, redevelop self-agency (Kralik et al., 2006) and learn news "ways of knowing" (Lloyd, 2017; Martzoukou and Burnett, 2018). Past studies that focused on this population have investigated the challenges and obstacles migrants face in the process of social inclusion and integration into their new society (Atiso et al., 2018; Allard and Caidi, 2018; Bronstein, 2017; Clark et al., 2014; Khoir et al., 2015; Lloyd, 2015). Migrants are one of the most marginalized population groups worldwide (OECD, 2018). To adequately design and provide information and services for these populations, LIS professionals need to understand how marginalization works. This study draws from Gibson and Martin's (2019) notion of information marginalization "that describes the systematic, interactive socio-technical processes that can push and hold certain groups of people at social 'margins,' where their needs are persistently ignored or overlooked" (p. 1). Information marginalization shifts the focus away from the behavior of the information poor who mistrust their surroundings (Chatman, 1996) and often lack the necessary skills or abilities to effectively seek and retrieve information (Britz, 2004) toward the institutions and places in which information poverty is deeply embedded. In my investigation into the views and perceptions of intermediaries working to provide information and services to the migrant community, I realized that the obstacles and difficulties experienced by migrants when trying to access and use the information they needed in their everyday lives originated in both the migrants themselves and in the systems and institutions designed to assist them. Hence, this paper proposes a new perspective of Gibson and Martin's (2019) concept that addresses the cultural and social gaps existing between the migrants and the individuals and organizations working with them. The second concept I propose is information resistance based on Dervin's (1999) sense-making concept of resistance. Dervin (1999) asserted that sense-making "assumes that articulation of one's lived experience including its struggles and resistances as well as alignments with given order is in itself a Sense-Making journey" (p. 742). I would argue that the social, cultural and language gaps that exist between migrants and intermediaries render the information and services offered often inadequate or irrelevant resulting in migrants resisting or rejecting the information offered to them that ultimately results in a situation of information marginalization that hinders migrants' process of integration. This study contributes to the literature on the role of information in the integration of migrants in three ways: first, it explores a different perspective about the role that information plays in the integration process of migrant workers by exploring the views and opinions of individuals and organizations that work with these communities daily. Most LIS studies considering migrants worldwide have focused on their information needs and behavior (Allard and Caidi, 2018; Bronstein, 2017, 2019; Caidi and Allard, 2005; Fisher et al., 2004; Lloyd, 2015) and their use of information technologies (Baron et al., 2014; Dekker and Engbersen, 2014; Suh and Hsieh, 2019). Their intent was to facilitate their integration into their host society. Second, instead of a top-down approach to integration that quantifies social, economic and educational indicators and reviews integration programs and policies as outcomes of the process of social inclusion (Cheung and Phillimore, 2017; Lichtenstein and Puma, 2018; Puma et al., 2018; Schibel et al., 2002; Schneider and Crul, 2010), this study considers the issue from a qualitative perspective that allows participants to talk about their experiences and opinions. This study provides a different perspective about an issue that is often hard to measure because it relies on "surveys of attitudes, feelings and perceptions" (OECD, 2018, p. 25). Third, the study proposes a new perspective to Ager and Strang's (2008) framework of integration by looking at its different elements through the perspective provided by Gibson and Martin's (2019) concept of information marginalization and Dervin's (1999) sense-making notion of resistance. Social integration of migrants This study adopted the United Nations Migration Agency (IOM)'s (United Nations, n.d.) general definition of migrant as any person who is moving or has moved across an international border or within a State away from his/her habitual place of residence, regardless of the person's legal status; whether the movement is voluntary or involuntary; what the causes for the movement are; or what the length of the stay is. The study draws from Ager and Strang (2008) framework of indicators of integration that define the core domains of integration of migrants by describing the different information needs (markers and means), information sources (social connections) and facilitators that can support and foster the integration process. I would argue that behind this framework lies the need for information (Wilson, 1981) as one of the factors that foster or hinder the process of social integration of migrants. The framework includes the following elements: Markers and means: to live and to be integrated into a new society, migrants need information to fulfill their immediate needs (what Ager and Strang (2008) termed markers and means): employment, housing, education and health. Social connections: Ager and Strang (2008) described these as a "connective tissue" (p. 177) that represents the defining feature of an integrated community. The authors recalled three forms of social connections according to Putnam (1995): social bonds (with family and co-ethnic, co-national, co-religious or other forms of group), social bridges (with other communities) and social links (with the structures of the state). These social connections serve as sources of informal and formal information and support. The current study investigates the perceptions of the third kind of social connections of migrants, social links with governmental and voluntary organizations. Facilitators: the process of integration requires facilitators that foster social inclusion. Ager and Strang (2008) outlined two major facilitators, language and knowledge acquisition that are key areas of cultural competence, safety and stability that allow migrants to feel at home in their host society. Foundation: notions of citizenship and rights are discussed as an essential prerequisite for integration. This element did not appear in the findings of the study. Ager and Strang's (2008) framework of integration as served as the basis for a number of studies dealing with integration of different migrant communities (Erdal, 2013; Hainmueller et al., 2017; Platts-Fowler and Robinson, 2015; Spencer and Charsley, 2016). Recently, the UK Home office published a new report on indicators of integration (Ndofor-Tah et al., 2019). This new framework is structured around Ager and Strang's (2008) domains adding leisure to the marker and means domains and digital skills as a new facilitator. In addition, the new framework relates to language and communication as separate domains from culture as well as safety and stability. There is a considerable body of work dealing with integration of migrants that can only be briefly summarized here. Prior research has discussed the difficulty in conceptualizing integration (Ager and Strang, 2008; Da Lomba, 2010; Robinson, 1998). This is why Castles et al. (2002) suggested that integration should be recognized as an umbrella term that encompasses a wide range of criteria that help researchers "assess the migrants' own motivations, strategies and networks." Da Lomba (2010) asserted that integration has two broad dimensions: public and private. The public dimension of integration refers to the legal and social environment (i.e. housing, education and health) in which refugees and immigrants are located. The private dimension of integration discusses to personal situations and demographic status (i.e. gender, education, dependence on social benefits and employment prospects) as well as a migrant's social connections (e.g. religious or ethnic identity, familial connections). Past studies can be divided into two main groups based on the methodological perspective chosen to conceptualize, examine and measure integration. The first group of studies used different quantitative criteria to measure integration. In an early study, Kuhlman (1991) proposed a complex model of integration that took into account a range of different indicators of integration that included spatial integration, economic integration, social integration, political integration, legal integration and psychological integration. He further stated that long-term outcomes may be influenced by early experiences that could result in disadvantage and marginalization. Along this line of thought, Hainmueller et al.'s (2016) longitudinal study found that receiving the host country citizenship early in the process had a lasting impact on enhancing social integration. Cheung and Phillimore (2017) examined gender differences of integration in the complex relationships between language proficiency, social networks and three key social policy areas: employment/education, health and housing. Their findings showed clear genders differences for these criteria. Saggar et al. (2012) focused on social cohesion as an integration indicator using the following measures: national identity measures (feelings of belonging); integration by group measures (outcomes in education, jobs, housing and so on); Cohesion measures in local neighborhoods (feelings of safety, well-being and belonging in the local community). Following this line of research, Torunczyk-Ruiz and Brunarska (2018) examined access to emotional and instrumental social capital in the host country as a strong indicator that enables migrants to develop a stronger sense and shape settlement intentions of temporary migrants. Kearns and Whitley (2015) found a positive correlation between time and place and self-reported indicators of social integration such as trust, reliance and safety, social relations and a sense of community. Hahn et al.'s (2019) study reported that an increased incremental validity of personality traits and cognitive factors on the integration process of migrants even with respect to contextual and sociodemographic factors. A number of studies have identified mental health as a significant factor of integration that is negatively impacted by other indicators such as precarious employment and lack of social connections, trust and self-identity (Brydsten et al., 2019; Lin et al., 2016). A second group of studies have investigated different aspects of integration qualitatively using semi-structured interviews to explore a more personal perspective on different aspects of integration (Grzymala-Kazlowska, 2018; Vancea and Boso, 2015). Gilmartin and Migge (2015) explored the pathways of integration of European migrants to Ireland. They found that cultural and social pathways (i.e. language acquisition, education, and building social relationships) are central to integration while economic factors such as employment were perceived as enablers of social pathways. Phillimore et al. (2014) interviewed migrants in the UK on what integration meant to them. Participants talked about "integration as a process and that the passage of time allows people to become more integrated" (p. 8). They related how integration was strongly associated with language proficiency, understanding British culture and developing strong social connections with close social environments and institutional settings. The significance of social capital has been recognized in other studies that have focused on the role that close and distant social networks play in the integration process and in creating a sense of belonging (Cheung and Phillimore, 2013; Grzymala-Kazlowska, 2018; Jorgensen, 2017; Phillimore, 2012; Wessendorf, 2019). In a recent study, Kyeremeh et al. (2019) investigated the notion of successful integration among African migrants in Canada who felt that integration should be a gradual and ongoing process and explained that the host society should allow migrants with ample time to acquaint and adjust to the new life. Information marginalization and information resistance Social exclusion or marginalization has traditionally been attributed to cultural, economic and social factors, but a number of studies in the LIS field have also shown that information provision is a key component of social exclusion. Communities that lack access to information sources are oftentimes socially excluded (Britz, 2004; Caidi and Allard, 2005; Lloyd, 2015; Yu, 2012). Recent studies have stated that migrants' social integration or inclusion "depends on their preparedness to trust new information and information sources and their understanding of information flow in their new environment" (Lloyd et al., 2017). The theoretical framework of information poverty has been used in the past to explain the social exclusion of marginalized groups in society (Bronstein, 2017; Chatman, 1996, 1999; Lingel and Boyd, 2013). These studies have asserted that people living in an information impoverished world: experience social isolation and mistrust of their surroundings, see themselves as devoid of information sources, live in economic poverty or experience social exclusion or stigma, and engage in defensive behaviors (Britz, 2004; Chatman, 1996; Lingel and Boyd, 2013) that hinder their access and use of information. However, Haider and Bawden (2007) asserted that this economic and technological deterministic view that dictates "there is a 'right kind' as well as a 'right amount' of information" (p. 3) has transformed information into a commodity and has caused the literature on the subject to neglect "the social and communicative aspects of information and in particular of information use that would allow for a serious engagement with different types of informational practices by considering the various social conditions" (p. 535). Following this line of thought, the current study focuses on the social and communicative aspects of information. This is done by analyzing different elements that comprise the gaps that exist between migrants and their host society that hinder their access and use of information, and ultimately their social integration. The social approach to information behavior (Burnett et al., 2001; Chatman, 1991, 1996) proposed a theoretical framework that suggested a link between social marginalization, isolation, limited information access and particular self-protective information-related practices as well as selective acceptance or avoidance of new information. The current study reframes the notion of selective acceptance or avoidance found in past studies (Chatman, 1999; Lingel and Boyd, 2013) within the sense-making framework and proposes the concept of information resistance. Dervin (1992) defined the sense-making approach as "a set of assumptions and propositions, and a set of methods which have been developed to study the making of sense that people do in their everyday experiences" (p. 61). Making sense is a process, and sense is the product of this process, that "includes 'knowledge,' but also a host of intuitions, opinions, hunches effective responses, evaluations, questions, etc." (Savolainen, 1993, p. 16). At the foundation of this framework, sense-making assumes that reality is neither complete nor constant, but rather it is filled with fundamental and pervasive discontinuities or gaps (Dervin, 1983) that bring humans to constantly seek continuity as the "ultimate goal of life in general and human condition in particular" (Savolainen, 1993, p. 16). Discontinuity is the negation of this state of desired continuity that occurs when the individual faces a challenge or a problematic situation and needs to find a new sense in her new reality. That is, when a new reality blocks the movement through time-space, a discontinuity is created. This discontinuity or gap can be resolved through information seeking and sense-making can take place. However, humans face many barriers and constraints that block the sense-making process and bring them to resist or challenge the information encountered (Dervin, 1992). I would argue that during the integration process, migrants face discontinuities or gaps in all aspects of their lives and need to bridge these gaps by acquiring sociocultural and language knowledge of their host society. Information marginalization occurs when migrants experience information resistance that renders them unable or unwilling to bridge these gaps because of sociocultural, political, and economic barriers and a sense of insecurity and instability that hinders their advancement toward integration (Dervin, 1992). The information role of intermediaries for marginalized groups Prior research has shown that people rely on networks of personal contacts to access and acquire information (Buchanan et al., 2019; Agada, 1999; Lu, 2007; Palsdottir, 2012; Sabelli, 2016). People who seek and disseminate information for others have been referred to in the LIS literature in different ways: gatekeepers (Sturges, 2001), social mediators (Sabelli, 2012) or lay information mediaries (Abrahamson and Fisher, 2007) who "through filtering and transmitting information subtly shape and inform our social reality and knowledge worlds" (Lu, 2007, p. 115). This study adopted Buchanan et al.'s (2019) notion of information intermediaries who "play a key role in recognizing, understanding, and progressing information needs in the problematic context" (p. 127). Participants in the current study are "socially positioned" (p. 113) gatekeepers working at institutions such as government agencies, NGOs and health organizations that have a strategic advantage that endow them with ready access to resources and allow them to seek and disseminate information. A number of past studies have examined the information role that intermediaries have for marginalized or disadvantaged populations (Buchanan et al., 2019; Sabelli, 2016; Schilderman, 2002). In an early study, Kurtz (1966) investigated the acculturation process of a Spanish-surnamed population in the USA who migrated from rural areas to an urban environment, and for whom gatekeepers held a strategic position for providing ready access to resources needed to solve problems. Agada (1999) investigated the "information use environments," of residents of an African-American inner-city neighborhood and found that gatekeepers were those individuals within the community who were able to move between cultures and link their communities with resources, and who were perceived as trustworthy and credible sources of information. Durrance et al. (2006) examined everyday information behaviors for local community-based citizen's groups in the USA and found that they play an important gatekeeping role because they make information more relevant for their constituents by distilling, tailoring and vetting the information according to the residents' needs. Schilderman (2002) reported on the role that intermediaries have as sources of information for the urban poor in Sri Lanka, Peru and Zimbabwe who prefer gatekeepers from social networks based on kinship, proximity or friendship rather than institutional gatekeepers (government agencies, NGOs and health organizations) who might be better informed but who have been unsuccessful in providing the urban poor with information in personalized formats - and in perceiving the urban poor as partners who are an important source of indigenous knowledge. Contrary to Schilderman (2002), two other studies described the successful role that institutional intermediaries have in advancing the information needs of marginalized populations. Sabelli (2012) found that for young women in vulnerable settings in Uruguay, institutional mediators represent an "essential bridge" in "the process of appropriation of information and knowledge for both personal and collective development." Using Granovetter's (1983) weak/strong ties terminology, Bronstein (2018) examined the information behavior of migrant workers in Israel and found that as weak ties, institutional intermediaries can extend access to information sources and skills not available through the strong ties within the migrant community. Moreover, because the information provided comes from people with whom migrant workers do not have a reciprocal or intimate bond, this information is perceived as social support that fosters a sense of belonging. Past studies have examined different elements of the work being done at humanitarian and public organizations assisting migrant communities in different countries (Mathews and Jiayi, 2017). These studies have examined the experiences of health practitioners (Broom et al., 2019; Kelly, 2019; Teunissen et al., 2016), privacy concerns in humanitarian organizations (Vannini et al., 2019), immediate refugee support (Cederquist, 2019), settings providing social support (Stock, 2019) and motivations for volunteering for refugees (Artero, 2019). Population of the study Ten people were interviewed for the study. Table I shows details about the participants. All participants worked directly with migrants at NGOs or public institutions. These organizations assist all the migrant community in Israel, that is, no differentiation was made between different groups of migrants at an organizational or a personal level. In other words, in the interviews, participants talked about migrants in general without referring to their legal status (i.e. refugees, migrant workers, undocumented migrants). Characteristics of study participants are presented in Table I. Data collection To gain access to the field, I volunteered at a workers' hotline named Kav La'oved, as a worker's counselor for domestic migrant workers for four years. Through my work at Kav La'oved, I developed social connections with social workers and other volunteers at different NGOs who participated in the study and referred me to additional participants. Participants were interviewed at their places of work using a semi-structured questionnaire (see Appendix). The interviews lasted 60-90 min were conducted in Hebrew, recorded and fully transcribed. The textual data were translated using the double or back translation method to secure the reliability of the data translated. Using this method, I first translated the data from Hebrew to English (forward translation) and then the data were translated to Hebrew by an expert in both languages (back translation) at which time both version of the data were checked for accuracy (Dhamani and Richter, 2011). Data analysis The study used a hybrid approach of inductive and deductive coding. In the first phase of the analysis, I used deductive content analysis by which existing data were analyzed by testing categories, concepts, models or hypotheses based on earlier theories, models, mind maps and literature reviews (Elo and Kyngas, 2008). That is, I first looked for elements of two theoretical frameworks, information poverty and sense-making in the data. When it became clear that these two frameworks did not fully reflect the complexities, I encountered in the data I adopted an inductive analysis in which I identified new ideas or themes derived from the data (Elo and Kyngas, 2008). A categorization matrix (Table II) was developed after the first phase of the analysis and was augmented with the categories that resulted in the second phase. The content analysis of the interviews revealed three major themes: information marginalization includes data that describe the different factors that keep migrants at the social margins and the ways that this marginalization is reflected in their information seeking of everyday life information; information resistance includes data that describe the ways by which migrants hold off or rebuff accessing and receiving information; overcoming resistance includes data that reveal the ways by which migrants and social mediators try to overcome information resistance and ultimately information marginalization (see Table II). This theme has two main categories: elements of information marginalization that hinder migrants' social integration, and Ager and Strang's (2008) markers and means: three indicators of integration viewed through the perspective of information marginalization. The categories comprised in this theme related to Ager and Strang's (2008) facilitators: language and cultural knowledge and safety and security. Findings from the content analysis showed that participants did not view these elements as facilitators of integration, but rather as factors in the migrants' lives that hinder their integration into Israeli society. This theme is comprised of four subcategories: lack of cultural knowledge, lack of language proficiency, living in an unsafe and insecure environment, and discrimination. Lack of cultural knowledge Ager and Strang's (2008) framework of integration relates to the acquisition of knowledge as a facilitator of integration that has been described as "a prerequisite for social inclusion is knowledge about the social, economic and community dimensions through which any society is constituted" (Lloyd et al., 2013, p. 122). Findings from the content analysis of the interviews revealed that participants in the study viewed the lack of cultural knowledge that is the gaps or discontinuities that migrants experience as one of the main obstacles in their process of integration. For example, Sandra, who works at an NGO that organizes educational training workshops, described how the lack of cultural knowledge makes it difficult for migrants to make sense of their host society, and access the information they need in their everyday life:There are a lot of people who really have nowhere to turn for all sorts of things. There is a lack of understanding of the system, lack of understanding of the law, they do not understand the system, do not understand the law, here [in Israel] we have a different culture and language. This [lack of cultural knowledge] creates a lot of emotional stress, they talk a lot about this emotional stress. Sofia, a worker at an NGO dedicated to protect the migrants' civil rights, described how the lack of cultural knowledge as an obstacle for integration is especially salient for migrants who come from rural societies:A lack of understanding of the systems, a lack of understanding of the culture in which you live, a lack of understanding of bureaucracy, it is much more obvious in people who come from a rural society, and suddenly they have to deal with bureaucracy and forms, which is terrible. So it is precisely there that the shock is more noticeable, how to get along at all, with all the rules, with all the procedures. It's very difficult for them. Information marginalization does not exist solely on the users' side, it is reinforced by the system's inability to reach its users and understand their needs. Natalia, a social worker at a prominent municipal NGO, describes the gap that exists between her organization and its clients:I feel it all the time. I do not know how to reach them [the migrants], to talk to them and help them, we need to speak their cultural language, their real language so we can communicate with them. Lack of language proficiency Language acquisition has been described as one of the first stepping stones of social integration into a new society (Aspinall, 2007; Clark et al., 2014) and, alongside cultural knowledge, it is considered by Ager and Strang (2008) as a facilitator of integration. Sandra related the importance of language acquisition to the process of integration:Obviously, the first thing is language. Language is the first obstacle you will encounter in a foreign country. Language has also a cognitive element; how am I supposed to function in a society if I don't speak the language? Language is also related to trust. Natalia commented on how information that is available in the migrants' native language will be more accessible and will be perceived as more reliable:I think that we miss the point in terms of transmitting messages when it is not done in their [the migrants'] mother language in terms of trust, they find it easier to trust the information if it's in their own language. Sara explained how talking to migrants in their own language helps open lines of communication:My Spanish is completely broken, but when I sit down and talk to her [a Latin American migrant] in Spanish, she will open a lot more, she will feel much more relaxed. It also brings a lot of humor into the conversation, and then she opens up and feels more relaxed. Living in an unsafe and unstable environment Ager and Strang (2008) stressed the importance of achieving a sense of safety and stability as a facilitator of integration. Undocumented migrants who experience an unsafe and unstable environment remain on the outskirts of society and cannot access the services and information they need even in extreme cases. Sara described how the fear of deportation affects the migrants' sense of safety and stability, and how this fear directly impacts their health:This is the situation. They [migrants] live in fear of deportation, they somehow manage to live under the radar, not knowing their rights, and in the context of health, it results in neglect and fear to seek treatment until it's sometimes too late. Emily, a social worker at a local NGO related a case in which a woman was experiencing a clear danger to her safety and well-being:There was one very serious case of physical violence and she did not know where to turn. Who is supposed to take care of them? If you are undocumented, you do not have the protection of the state, so if you find yourself in a difficult situation you have no one who can help you. Discrimination Many times, it is the system that discriminates and reacts with intolerance toward the migrants' needs. Participants described a number of discriminatory incidents experienced by migrants in their interactions with the healthcare system, as Sara explained:[Bureaucracy in the health system] is also difficult for Israelis to understand, so it is much harder for migrants. Very often they face intolerance from the administration of the HMO, it is very difficult. I mean, many times they encounter hostility and discrimination. For example, I tell them to call me from the clinic and I will talk to the clerk and I will explain the situation to her. But then, when they get to the clinic the clerks will not talk to us, they are not willing to help them. Sofia noted how migrants find it hard to deal with long-term bureaucratic processes because of intolerance and discrimination:When they have to ask for a disability pension or a rehabilitation allowance, these long processes are very difficult to manage for them because they will need to meet again and again the unfriendly face and the crooked eye and all the other things that make it more difficult for them as a person who lives on the margins of society to get a little bit closer to the mainstream of that society. Means and markers Ager and Strang (2008) proposed four key areas that are indicative of a successful integration: housing, education, health and employment. The following excerpts from the interviews show how information marginalization was part of three of these aspects in the migrants' lives in the following three subcategories: health, education and employment. Health Lack of understanding of how the health system works is one of the main obstacles to migrant integration. Natalia explained how migrants struggle to understand the concept of health insurance:They have a real problem understanding the concept of health insurance. We often receive complaints of the following kind: "I paid health insurance for two years and I was not sick and they would not pay me my money back." I say it with half a smile but it's really not funny for the people who feel like they have wasted their money. Then what happens is that they sometimes cancel their insurance and have no health coverage when they get sick. Dealing with emergency rooms in big urban hospitals can be a trying experience for migrants. Sara described how they are afraid to sign the paperwork and pay the admission fee and get sent away:When they go to the emergency room they are not willing to sign the paperwork, they are very scared of it and then they come to us and say they did not want to treat me. In other words, they refused to treat me and wanted me to pay. They say, "pay or you will not receive treatment". There is some misunderstanding around the procedures in the emergency rooms. Education Ager and Strang (2008) referred to education as one of the means and markers of integration. Content analysis of the interviews revealed that migrants face socioeconomic barriers and personal obstacles such as a lack of language proficiency that impedes their access to education:The issue of education, migrants want it very much, they are eager to learn, but it feels impossible for them. This is something lacking in their lives, it's too hard for them, it's not accessible to them. Employment The impact that information marginalization has on the integration process is especially noticeable in the area of employment. According to participants, migrants find it hard to understand different notions related to employment such as taxes and work visas. Sofia related her experiences in this area:They get a culture shock when people tell them that to work you need a visa. "Why do I need a visa to work?" "I just work and you pay me." On a very basic level, what they know is that if you work for someone in the fields, then he pays you for the days you work. For migrant populations, the issue of taxation can be difficult to understand:The whole issue of taxes is very strange to them, they constantly say: why do we need it? Is always very difficult for them to understand and assimilate this [the issue of taxes]. There is a lot to explain, to try to explain the reason behind it [the taxes] they do not understand why there are so many laws. So we had to explain this issue to them all the time. Migrants' employment opportunities are also affected by their legal status and by the information sources accessible to them. Sofia mentions the obstacles that migrants encounter when they are looking for a job:As an Israeli, how do I look for a job? In our world, for the last few years, it is very normal to go online and look for work; we send resumes to online job sites. But how can a migrant worker send his resume online? They have to rely on friends or relatives who will help them find a job. If someone has a job, maybe he can ask if they need more workers, so their employment possibilities are very limited. The second theme in this study is information resistance. Content analysis of the interviews revealed that migrants resist information as a defensive behavior in response to the unstable and unsafe situation they face in their host country and to their lack of cultural knowledge. This theme includes two subcategories: secrecy and disinformation. Secrecy The notion of information resistance is related to Chatman's (1991) principle of secrecy. In this study, secrecy refers to the need that migrants have to safeguard themselves from unwanted exposure and purposively conceal their need for information as a defensive behavior. The purpose of secrecy is to guard against disclosure, which results in people not being receptive to advice or receiving information. Natalia described how undocumented migrants that live in a state of secrecy are oftentimes not receptive to the services provided by their organizations for fear of exposure and deportation:Sometimes they are reluctant to get help from us, or families that will try to disappear under the radar so as not to accept our intervention. And then it's a real courtship that sometimes lasts for months until we find the family. It's really hard to find people who are undocumented, nobody knows their address. I have no way of locating them, except for the help of activists, pastors, priests; we work with churches a lot. Sara conveyed how fear and the need for secrecy prevent migrants from getting the medical help they need. Concurring with Chatman's (1996) assertion that "it became clear that an influence on one's decision to remain secretive about a concern was fear" (p. 199):Health is a big issue for them [the migrants]. Being here without a visa, is a constant threat. Things can turn bad if they have a child, they don't know what to do, so they don't ask for help if they become ill. To be isolated like that, to be away from your family and bear all the responsibility all by yourself for the family's livelihood. Even without being a doctor I can see that all of these things affect them mentally. Disinformation Resisting or rejecting information that comes from outsiders (Chatman, 1996) and trusting insiders' unreliable sources of information that appear trustworthy is one of the outcomes of information resistance. Participants relate how disinformation negatively impacts migrants' lives:They have a strong will to live here that it is very empowering, but there are a lot of anxieties, a lot. A lot of times these anxieties come from rumours that grow bigger and bigger into something really monstrous and catastrophic, that ultimately does a lot of damage. Disinformation can also preclude migrants from reaching out for help. Natalia described a situation in which misinformation and a lack of trust stopped a family from getting the help they needed:We had such a case in which a girl had missed a lot of school days, and no one had been able to find her mother. Eventually, we found out that her mother was always afraid to contact us or reveal details about them as a family because she was afraid we would take her child away from her or pick her up and deport her. All sorts of things like that that don't actually happen. Migrants are also affected by the disinformation about them in Israeli society and directly affect their employment prospects:Many Israelis believe that if they employ an undocumented migrant they will be fined or prosecuted. This obstacle is certainly a structural obstacle that comes from campaigns disseminating a false message. They say, whoever employs an undocumented foreign worker is considered a criminal. This theme consists of two categories that describe the efforts of social mediators to find new, relevant ways to communicate with migrants as well as the role that social connections play in overcoming information resistance and marginalization. Situational relevance How can information resistance be overcome? What constitutes a trustworthy and relevant information source? People perceive an information source as relevant and accessible if they can make sense and trust the information provided by the source (Bronstein, 2014; Ruthven et al., 2018). That is, "even if a source is perceived as potentially useful, it will not do much good to the individual if that source is not legitimized by contextual others" (Chatman, 1996, p. 202). Yona, a social worker at an NGO that provides health services to migrants and other marginalized populations, explained the importance of trust in overcoming resistance:To migrants, there are those who can be trusted and those who cannot, so they won't trust information if it's written and I show them that it's written in a law book, but they will not believe it. They will not believe it. It does not matter if I show them that there is a law now and the law says it this way, it will not give them the confidence; they will only trust the information coming from people they trust, then they will believe it. Sandra describes the difficulties she encountered when trying to market a training workshop and how she solves the problem by marketing the information through information sources that the community trusted:We tried marketing the workshop on Facebook, advertised it in all kinds of groups that we thought were relevant, but it didn't work so well. I felt that in this community, I had to act differently, not using normal marketing channels, but really personal. So I called people. I called someone who was a community leader and said, "Listen, it's [the workshop] really worth it to people. Help them, tell people to come." As time passed I realized that this is the way. Find the people who have connections, gatekeepers of the community, let them advertise it, let them push it and do it in cooperation with them because they bring the people. This way the workshop was a success. Emily talked about another project that helped migrant workers with small children register their children in public kindergarten:We [at her organization] want as many children as possible from the age of 3 to go to kindergarten. But registration for the kindergarten opens in December. These communities don't think long term, so a lot of people miss the registration period, so we created videos, in several languages, uploaded them to Facebook announcing that the registration for kindergarten was open, and described what needs to be done and where to go [in the videos]. You see a person from the community who does the whole registration process from the beginning. He shows all the registration steps, shows them what kind of paperwork they need. It was a success; three years ago, only 40% of migrant children were registered on time and today 95% of migrant children are registered. Social connections Ager and Strang (2008) considered social connections as facilitators that remove barriers that help migrants make sense in their host society and advance their social integration. The content analysis of the data identified two subcategories: social bonds and social links. Social bonds Ager and Strang (2008) asserted that social bonds created with the migrants' close social network play an important role helping them feel settled and integrated. The intermediaries who participated in the study talked about different initiatives that provided migrants with information that made sense to them and fulfilled some of their needs. Natalia talked about one of these initiatives that supported communal leadership:The aim at the beginning was to form a communal network. We tried to create a social network of people who help each other. We organized all sorts of committees. The Filipino and the South American communities have groups of mothers who support each other, for example. When one of them is ill, the others take care of her children. They visit her at the hospital, take care of her, a real network of mutual support. Social linking Ager and Strang (2008) described social links as the connection between individuals and structures of the state. Sherry talked about the important role that NGOs and other institutions have as information sources:A lot of word of mouth [...] many times they will come to us because we are considered a relatively general organization, we direct them to other places. We use Facebook and other organizations also use Facebook and distribute information there, also Whatsapp, and all sorts of mailing lists. Tamara, a social worker at a municipal NGO, described the important role that churches play in the lives of migrants as social links to the host society:Churches are a place to seek information, they are places that give first aid. In other words, there are churches that will turn the world around and find a solution for someone living in the street. Whether they give him money or give him a room in the church for a couple of days, they are a really necessary support resource for communities. Sara explained how they organize talks with different communities through churches and other community centers to talk to migrants about the healthcare system and health insurance:We have done informational evenings in churches on Friday or Sunday evenings, where the pastor gave us half an hour to talk about how emergency rooms work or health insurance, but there are still gaps. We need to go out more into the communities, reach out especially in small communities where it's a bit more complicated. This study presented a new and more nuanced understanding of the process of integration of migrants by re-examining Ager and Strang's (2008) framework from an informational perspective told by the intermediaries who work with these populations. Findings from the content analysis revealed that the process of integration is shaped by the inability of both migrants and the institutions in the host society to close the cultural and social gaps that ultimately result in information marginalization. Two main themes were revealed in the content analysis: information marginalization and information resistance. The first theme contained two main categories that reflected the elements of marginalization and how it is reflected in Ager and Strang's (2008) means and markers. Although the framework related to cultural and language knowledge as facilitators of integration, participants in this study perceived the lack of these indicators as obstacles that precluded migrants from getting their basic needs covered, reinforcing their marginalization. Echoing this finding, Martzoukou and Burnett (2018) observed that "the direct relationship between social inclusion and knowledge of the socioeconomic and community information-related constructs," and found "that there were fundamental gaps in Syrian Scots knowledge domains in relation to everyday life matters, which would be considered as 'routine' knowledge in mainstream Scottish society and beyond" (p. 1126). Just as participants in this study related the migrants' lack of understanding of admission procedures at emergency rooms, Martzoukou and Burnett (2018) described an instance in which a participant in their study missed a doctor's appointment when she could not understand how to ride the bus. The second element of information marginalization revealed in the content analysis is the lack of language proficiency that represents a salient theme in the literature on migration (Bronstein, 2019; Ikafa, 2015; Leith, 2009; Lloyd et al., 2013; Martzoukou and Burnett, 2018; Wessendorf and Phillimore, 2019). Grzymala-Kazlowska (2018) described language as an anchor and "a foothold that allows individuals to acquire socio-psychological stability and security and function effectively in a new or substantially changed life" (p. 235). Findings from this study reinforced this assertion and showed that language proficiency was directly related to the ability of migrants to solve everyday problems and meet basic needs, as well as to develop trust, make sense of their reality and be receptive to outsiders and to the information they provide. Information resistance occurred when information in their native language was perceived as the only information that made sense and could be trusted (Chatman, 1996). Two other elements of information marginalization were: living in an unsafe and unstable environment, and discrimination. These elements directly relate to the migrants' ability to understand their new reality and lessen the resistance they experience as a safeguard against a hostile environment bringing migrants "to engage in information-related practices, such as information sharing, which would enable them to understand how the various instrumental environments (health, education, employment) operate" (Lloyd et al., 2017) and allowing people to safely engage with services and with other people to integrate (Ndofor-Tah et al., 2019). Supporting Ager and Strang's (2008) assertion that to feel at home, migrants need to feel safe, participants in this study related how the fear of deportation and discrimination precluded them from seeking information and how that often negatively affected their well-being. Past studies have noted the significant role that a safe and stable environment plays in the process of integration (Ager and Strang, 2008; Ikafa, 2015; Saggar et al., 2012) and represented a crucial foothold and point of reference which enabled migrants to adapt to their new life and feel settled (Grzymala-Kazlowska, 2018). Findings showed that accessing and understanding information related to Ager and Strang's (2008) markers and means is directly related to the migrants' ability to bridge the gaps in their reality, make sense of their surroundings and advance in their integration into Israeli society (Baird, 2012; Ikafa, 2015). That is, the sociocultural gap that prevents migrants from understanding how the education and the healthcare systems work, and how to find and retain proper employment directly contributes to their marginality and does not allow them "to move beyond the liminal zones of society and become included" (Lloyd et al., 2017). The second theme revealed by content analysis refers to the information resistance experienced by migrants when they resist information that comes from outsiders and renders them unable or unwilling to bridge the cultural gap that will help them make sense of their reality. This resistance is oftentimes compounded by inability of the institutions in the host country to help them. This theme contains two categories: secrecy and disinformation. Echoing Chatman's (1996) concept of secrecy as a safeguarding behavior, participants described how migrants, "cease to be receptive to advice and information" (p. 195) and resisted the information they needed in an attempt to protect themselves from an environment they perceived as unsafe. Participants related how resistance to information perceived as not trustworthy (i.e. coming from outsiders) negatively impacted migrants' lives and resulted in disinformation that ultimately hindered their integration process. The third theme revealed in the study presents the findings related to the migrants' and the intermediaries' attempts to overcome information resistance. This theme contains two categories: situational relevance and social connections. The intermediaries participating in the study described different strategies they used to reach migrants and present information in formats that made sense to them. Supporting Chatman's (1996) assertion that "things that make sense are relevant" (p. 202), participants tried to make information relevant to migrants by creating videos to explain the kindergarten registration process or by having trusted gatekeepers share information about training programs or community services. That is, participants understood the importance of legitimizing the information through contextual others (Chatman, 1996). The legitimation of information by contextualization is related to the creation of social connections. The significant role that social connections play in the integration process of migrants is a salient feature in the literature (Bronstein, 2019; Cheung and Phillimore, 2013; Gilmartin and Migge, 2015; Grzymala-Kazlowska, 2018; OECD, 2018). Two of Ager and Strang's (2008) social connections were revealed in the content analysis: social bonds as close social relations and social linking as the relationships built with the intermediaries. Echoing past studies findings showed that creating social bonds within their community helped migrants resolve problems, feel more secure and stable (Grzymala-Kazlowska, 2018) and reinforce their settlement intentions (Torunczyk-Ruiz and Brunarska, 2018). Interestingly, the two new indicators that were added in the 2019 UK Home Office indicators (Ndofor-Tah et al., 2019), leisure and digital skills did not appeared in the data analysis. Further research should investigate the role of these two elements in the integration process of migrants. Phillimore (2012) wrote, "Integration implies the development of a sense of belonging in the host community, with some renegotiation of identity by both newcomers and hosts" (p. 3). Findings from the study showed that it is in this oftentimes failed renegotiation that information marginalization emerges. By allowing intermediaries to articulate their views and opinions, I was able to understand how personal, cultural and social factors impacted the lives of migrants, and distanced them from the sources of information and support they need to feel at home in their new country. Findings showed that for integration to be successful, it should be the result of an effort by both migrants and local institutions/structures to reach a middle ground of understanding and compromise. This study extending Gibson and Martin's (2019) notion of information marginalization to encompass both sides of the equation of marginalization and situating it in the social, economic and contextual conditions that create information poverty. By examining Ager and Strang's (2008) framework of integration from an information perspective, this study highlighted the role that information plays as a facilitating or an impeding element in the integration process, as well as a vehicle to help migrants make sense of their reality, regain social capital, perceive their environment as safe and develop a sense of belonging to their new home.
|
The purpose of this paper is to explore a different perspective about the role that information plays in the integration process of migrant workers by exploring the views and opinions of individuals and organizations that work with these communities on a daily basis. The study proposes a new perspective of Ager and Strang's framework of integration by looking at its different elements through the perspective provided by Gibson and Martin's (2019) concept of information marginalization and Dervin's sense-making notion of resistance.
|
[SECTION: Method] In this study, I present the findings from a qualitative investigation that explored the other side of the issue of integration of migrants; that is, the views and perceptions of information intermediaries working with migrants in Israel about the integration process. The notion of integration has been the focus of academic debate in recent years and has been defined in different ways (Ager and Strang, 2008; Cheung and Phillimore, 2014; Farach et al., 2015; Gilmartin and Migge, 2015; Harder et al., 2018; Phillimore, 2012). In its broadest sense, integration means the process or transition by which people who are relatively new to a country become part of society (Rudiger and Spencer, 2003). Integration is achieved when "people, whatever their background, live, work, learn and socialise together, based on shared rights, responsibilities and opportunities" (Ndofor-Tah et al., 2019) while keeping a measure of their original cultural identity (Threadgold and Court, 2005). Integration is being characterized today as a multidimensional and multidirectional (Harder et al., 2018) process that encompasses "access to resources and opportunities as well as social mixing" involving adjustments by everyone in society (Ndofor-Tah et al., 2019). Integration is the desired outcome of migration which has been described in the literature as a transition (Bronstein, 2019; Meleis, 2015) - a disruptive process that fractures migrants' information landscapes (Lloyd, 2017) separates them from their traditional social networks, and renders their known information sources and practices inadequate (Giralt, 2015; Lloyd et al., 2013). To integrate into their host society, migrants need to find ways to redefine themselves, redevelop self-agency (Kralik et al., 2006) and learn news "ways of knowing" (Lloyd, 2017; Martzoukou and Burnett, 2018). Past studies that focused on this population have investigated the challenges and obstacles migrants face in the process of social inclusion and integration into their new society (Atiso et al., 2018; Allard and Caidi, 2018; Bronstein, 2017; Clark et al., 2014; Khoir et al., 2015; Lloyd, 2015). Migrants are one of the most marginalized population groups worldwide (OECD, 2018). To adequately design and provide information and services for these populations, LIS professionals need to understand how marginalization works. This study draws from Gibson and Martin's (2019) notion of information marginalization "that describes the systematic, interactive socio-technical processes that can push and hold certain groups of people at social 'margins,' where their needs are persistently ignored or overlooked" (p. 1). Information marginalization shifts the focus away from the behavior of the information poor who mistrust their surroundings (Chatman, 1996) and often lack the necessary skills or abilities to effectively seek and retrieve information (Britz, 2004) toward the institutions and places in which information poverty is deeply embedded. In my investigation into the views and perceptions of intermediaries working to provide information and services to the migrant community, I realized that the obstacles and difficulties experienced by migrants when trying to access and use the information they needed in their everyday lives originated in both the migrants themselves and in the systems and institutions designed to assist them. Hence, this paper proposes a new perspective of Gibson and Martin's (2019) concept that addresses the cultural and social gaps existing between the migrants and the individuals and organizations working with them. The second concept I propose is information resistance based on Dervin's (1999) sense-making concept of resistance. Dervin (1999) asserted that sense-making "assumes that articulation of one's lived experience including its struggles and resistances as well as alignments with given order is in itself a Sense-Making journey" (p. 742). I would argue that the social, cultural and language gaps that exist between migrants and intermediaries render the information and services offered often inadequate or irrelevant resulting in migrants resisting or rejecting the information offered to them that ultimately results in a situation of information marginalization that hinders migrants' process of integration. This study contributes to the literature on the role of information in the integration of migrants in three ways: first, it explores a different perspective about the role that information plays in the integration process of migrant workers by exploring the views and opinions of individuals and organizations that work with these communities daily. Most LIS studies considering migrants worldwide have focused on their information needs and behavior (Allard and Caidi, 2018; Bronstein, 2017, 2019; Caidi and Allard, 2005; Fisher et al., 2004; Lloyd, 2015) and their use of information technologies (Baron et al., 2014; Dekker and Engbersen, 2014; Suh and Hsieh, 2019). Their intent was to facilitate their integration into their host society. Second, instead of a top-down approach to integration that quantifies social, economic and educational indicators and reviews integration programs and policies as outcomes of the process of social inclusion (Cheung and Phillimore, 2017; Lichtenstein and Puma, 2018; Puma et al., 2018; Schibel et al., 2002; Schneider and Crul, 2010), this study considers the issue from a qualitative perspective that allows participants to talk about their experiences and opinions. This study provides a different perspective about an issue that is often hard to measure because it relies on "surveys of attitudes, feelings and perceptions" (OECD, 2018, p. 25). Third, the study proposes a new perspective to Ager and Strang's (2008) framework of integration by looking at its different elements through the perspective provided by Gibson and Martin's (2019) concept of information marginalization and Dervin's (1999) sense-making notion of resistance. Social integration of migrants This study adopted the United Nations Migration Agency (IOM)'s (United Nations, n.d.) general definition of migrant as any person who is moving or has moved across an international border or within a State away from his/her habitual place of residence, regardless of the person's legal status; whether the movement is voluntary or involuntary; what the causes for the movement are; or what the length of the stay is. The study draws from Ager and Strang (2008) framework of indicators of integration that define the core domains of integration of migrants by describing the different information needs (markers and means), information sources (social connections) and facilitators that can support and foster the integration process. I would argue that behind this framework lies the need for information (Wilson, 1981) as one of the factors that foster or hinder the process of social integration of migrants. The framework includes the following elements: Markers and means: to live and to be integrated into a new society, migrants need information to fulfill their immediate needs (what Ager and Strang (2008) termed markers and means): employment, housing, education and health. Social connections: Ager and Strang (2008) described these as a "connective tissue" (p. 177) that represents the defining feature of an integrated community. The authors recalled three forms of social connections according to Putnam (1995): social bonds (with family and co-ethnic, co-national, co-religious or other forms of group), social bridges (with other communities) and social links (with the structures of the state). These social connections serve as sources of informal and formal information and support. The current study investigates the perceptions of the third kind of social connections of migrants, social links with governmental and voluntary organizations. Facilitators: the process of integration requires facilitators that foster social inclusion. Ager and Strang (2008) outlined two major facilitators, language and knowledge acquisition that are key areas of cultural competence, safety and stability that allow migrants to feel at home in their host society. Foundation: notions of citizenship and rights are discussed as an essential prerequisite for integration. This element did not appear in the findings of the study. Ager and Strang's (2008) framework of integration as served as the basis for a number of studies dealing with integration of different migrant communities (Erdal, 2013; Hainmueller et al., 2017; Platts-Fowler and Robinson, 2015; Spencer and Charsley, 2016). Recently, the UK Home office published a new report on indicators of integration (Ndofor-Tah et al., 2019). This new framework is structured around Ager and Strang's (2008) domains adding leisure to the marker and means domains and digital skills as a new facilitator. In addition, the new framework relates to language and communication as separate domains from culture as well as safety and stability. There is a considerable body of work dealing with integration of migrants that can only be briefly summarized here. Prior research has discussed the difficulty in conceptualizing integration (Ager and Strang, 2008; Da Lomba, 2010; Robinson, 1998). This is why Castles et al. (2002) suggested that integration should be recognized as an umbrella term that encompasses a wide range of criteria that help researchers "assess the migrants' own motivations, strategies and networks." Da Lomba (2010) asserted that integration has two broad dimensions: public and private. The public dimension of integration refers to the legal and social environment (i.e. housing, education and health) in which refugees and immigrants are located. The private dimension of integration discusses to personal situations and demographic status (i.e. gender, education, dependence on social benefits and employment prospects) as well as a migrant's social connections (e.g. religious or ethnic identity, familial connections). Past studies can be divided into two main groups based on the methodological perspective chosen to conceptualize, examine and measure integration. The first group of studies used different quantitative criteria to measure integration. In an early study, Kuhlman (1991) proposed a complex model of integration that took into account a range of different indicators of integration that included spatial integration, economic integration, social integration, political integration, legal integration and psychological integration. He further stated that long-term outcomes may be influenced by early experiences that could result in disadvantage and marginalization. Along this line of thought, Hainmueller et al.'s (2016) longitudinal study found that receiving the host country citizenship early in the process had a lasting impact on enhancing social integration. Cheung and Phillimore (2017) examined gender differences of integration in the complex relationships between language proficiency, social networks and three key social policy areas: employment/education, health and housing. Their findings showed clear genders differences for these criteria. Saggar et al. (2012) focused on social cohesion as an integration indicator using the following measures: national identity measures (feelings of belonging); integration by group measures (outcomes in education, jobs, housing and so on); Cohesion measures in local neighborhoods (feelings of safety, well-being and belonging in the local community). Following this line of research, Torunczyk-Ruiz and Brunarska (2018) examined access to emotional and instrumental social capital in the host country as a strong indicator that enables migrants to develop a stronger sense and shape settlement intentions of temporary migrants. Kearns and Whitley (2015) found a positive correlation between time and place and self-reported indicators of social integration such as trust, reliance and safety, social relations and a sense of community. Hahn et al.'s (2019) study reported that an increased incremental validity of personality traits and cognitive factors on the integration process of migrants even with respect to contextual and sociodemographic factors. A number of studies have identified mental health as a significant factor of integration that is negatively impacted by other indicators such as precarious employment and lack of social connections, trust and self-identity (Brydsten et al., 2019; Lin et al., 2016). A second group of studies have investigated different aspects of integration qualitatively using semi-structured interviews to explore a more personal perspective on different aspects of integration (Grzymala-Kazlowska, 2018; Vancea and Boso, 2015). Gilmartin and Migge (2015) explored the pathways of integration of European migrants to Ireland. They found that cultural and social pathways (i.e. language acquisition, education, and building social relationships) are central to integration while economic factors such as employment were perceived as enablers of social pathways. Phillimore et al. (2014) interviewed migrants in the UK on what integration meant to them. Participants talked about "integration as a process and that the passage of time allows people to become more integrated" (p. 8). They related how integration was strongly associated with language proficiency, understanding British culture and developing strong social connections with close social environments and institutional settings. The significance of social capital has been recognized in other studies that have focused on the role that close and distant social networks play in the integration process and in creating a sense of belonging (Cheung and Phillimore, 2013; Grzymala-Kazlowska, 2018; Jorgensen, 2017; Phillimore, 2012; Wessendorf, 2019). In a recent study, Kyeremeh et al. (2019) investigated the notion of successful integration among African migrants in Canada who felt that integration should be a gradual and ongoing process and explained that the host society should allow migrants with ample time to acquaint and adjust to the new life. Information marginalization and information resistance Social exclusion or marginalization has traditionally been attributed to cultural, economic and social factors, but a number of studies in the LIS field have also shown that information provision is a key component of social exclusion. Communities that lack access to information sources are oftentimes socially excluded (Britz, 2004; Caidi and Allard, 2005; Lloyd, 2015; Yu, 2012). Recent studies have stated that migrants' social integration or inclusion "depends on their preparedness to trust new information and information sources and their understanding of information flow in their new environment" (Lloyd et al., 2017). The theoretical framework of information poverty has been used in the past to explain the social exclusion of marginalized groups in society (Bronstein, 2017; Chatman, 1996, 1999; Lingel and Boyd, 2013). These studies have asserted that people living in an information impoverished world: experience social isolation and mistrust of their surroundings, see themselves as devoid of information sources, live in economic poverty or experience social exclusion or stigma, and engage in defensive behaviors (Britz, 2004; Chatman, 1996; Lingel and Boyd, 2013) that hinder their access and use of information. However, Haider and Bawden (2007) asserted that this economic and technological deterministic view that dictates "there is a 'right kind' as well as a 'right amount' of information" (p. 3) has transformed information into a commodity and has caused the literature on the subject to neglect "the social and communicative aspects of information and in particular of information use that would allow for a serious engagement with different types of informational practices by considering the various social conditions" (p. 535). Following this line of thought, the current study focuses on the social and communicative aspects of information. This is done by analyzing different elements that comprise the gaps that exist between migrants and their host society that hinder their access and use of information, and ultimately their social integration. The social approach to information behavior (Burnett et al., 2001; Chatman, 1991, 1996) proposed a theoretical framework that suggested a link between social marginalization, isolation, limited information access and particular self-protective information-related practices as well as selective acceptance or avoidance of new information. The current study reframes the notion of selective acceptance or avoidance found in past studies (Chatman, 1999; Lingel and Boyd, 2013) within the sense-making framework and proposes the concept of information resistance. Dervin (1992) defined the sense-making approach as "a set of assumptions and propositions, and a set of methods which have been developed to study the making of sense that people do in their everyday experiences" (p. 61). Making sense is a process, and sense is the product of this process, that "includes 'knowledge,' but also a host of intuitions, opinions, hunches effective responses, evaluations, questions, etc." (Savolainen, 1993, p. 16). At the foundation of this framework, sense-making assumes that reality is neither complete nor constant, but rather it is filled with fundamental and pervasive discontinuities or gaps (Dervin, 1983) that bring humans to constantly seek continuity as the "ultimate goal of life in general and human condition in particular" (Savolainen, 1993, p. 16). Discontinuity is the negation of this state of desired continuity that occurs when the individual faces a challenge or a problematic situation and needs to find a new sense in her new reality. That is, when a new reality blocks the movement through time-space, a discontinuity is created. This discontinuity or gap can be resolved through information seeking and sense-making can take place. However, humans face many barriers and constraints that block the sense-making process and bring them to resist or challenge the information encountered (Dervin, 1992). I would argue that during the integration process, migrants face discontinuities or gaps in all aspects of their lives and need to bridge these gaps by acquiring sociocultural and language knowledge of their host society. Information marginalization occurs when migrants experience information resistance that renders them unable or unwilling to bridge these gaps because of sociocultural, political, and economic barriers and a sense of insecurity and instability that hinders their advancement toward integration (Dervin, 1992). The information role of intermediaries for marginalized groups Prior research has shown that people rely on networks of personal contacts to access and acquire information (Buchanan et al., 2019; Agada, 1999; Lu, 2007; Palsdottir, 2012; Sabelli, 2016). People who seek and disseminate information for others have been referred to in the LIS literature in different ways: gatekeepers (Sturges, 2001), social mediators (Sabelli, 2012) or lay information mediaries (Abrahamson and Fisher, 2007) who "through filtering and transmitting information subtly shape and inform our social reality and knowledge worlds" (Lu, 2007, p. 115). This study adopted Buchanan et al.'s (2019) notion of information intermediaries who "play a key role in recognizing, understanding, and progressing information needs in the problematic context" (p. 127). Participants in the current study are "socially positioned" (p. 113) gatekeepers working at institutions such as government agencies, NGOs and health organizations that have a strategic advantage that endow them with ready access to resources and allow them to seek and disseminate information. A number of past studies have examined the information role that intermediaries have for marginalized or disadvantaged populations (Buchanan et al., 2019; Sabelli, 2016; Schilderman, 2002). In an early study, Kurtz (1966) investigated the acculturation process of a Spanish-surnamed population in the USA who migrated from rural areas to an urban environment, and for whom gatekeepers held a strategic position for providing ready access to resources needed to solve problems. Agada (1999) investigated the "information use environments," of residents of an African-American inner-city neighborhood and found that gatekeepers were those individuals within the community who were able to move between cultures and link their communities with resources, and who were perceived as trustworthy and credible sources of information. Durrance et al. (2006) examined everyday information behaviors for local community-based citizen's groups in the USA and found that they play an important gatekeeping role because they make information more relevant for their constituents by distilling, tailoring and vetting the information according to the residents' needs. Schilderman (2002) reported on the role that intermediaries have as sources of information for the urban poor in Sri Lanka, Peru and Zimbabwe who prefer gatekeepers from social networks based on kinship, proximity or friendship rather than institutional gatekeepers (government agencies, NGOs and health organizations) who might be better informed but who have been unsuccessful in providing the urban poor with information in personalized formats - and in perceiving the urban poor as partners who are an important source of indigenous knowledge. Contrary to Schilderman (2002), two other studies described the successful role that institutional intermediaries have in advancing the information needs of marginalized populations. Sabelli (2012) found that for young women in vulnerable settings in Uruguay, institutional mediators represent an "essential bridge" in "the process of appropriation of information and knowledge for both personal and collective development." Using Granovetter's (1983) weak/strong ties terminology, Bronstein (2018) examined the information behavior of migrant workers in Israel and found that as weak ties, institutional intermediaries can extend access to information sources and skills not available through the strong ties within the migrant community. Moreover, because the information provided comes from people with whom migrant workers do not have a reciprocal or intimate bond, this information is perceived as social support that fosters a sense of belonging. Past studies have examined different elements of the work being done at humanitarian and public organizations assisting migrant communities in different countries (Mathews and Jiayi, 2017). These studies have examined the experiences of health practitioners (Broom et al., 2019; Kelly, 2019; Teunissen et al., 2016), privacy concerns in humanitarian organizations (Vannini et al., 2019), immediate refugee support (Cederquist, 2019), settings providing social support (Stock, 2019) and motivations for volunteering for refugees (Artero, 2019). Population of the study Ten people were interviewed for the study. Table I shows details about the participants. All participants worked directly with migrants at NGOs or public institutions. These organizations assist all the migrant community in Israel, that is, no differentiation was made between different groups of migrants at an organizational or a personal level. In other words, in the interviews, participants talked about migrants in general without referring to their legal status (i.e. refugees, migrant workers, undocumented migrants). Characteristics of study participants are presented in Table I. Data collection To gain access to the field, I volunteered at a workers' hotline named Kav La'oved, as a worker's counselor for domestic migrant workers for four years. Through my work at Kav La'oved, I developed social connections with social workers and other volunteers at different NGOs who participated in the study and referred me to additional participants. Participants were interviewed at their places of work using a semi-structured questionnaire (see Appendix). The interviews lasted 60-90 min were conducted in Hebrew, recorded and fully transcribed. The textual data were translated using the double or back translation method to secure the reliability of the data translated. Using this method, I first translated the data from Hebrew to English (forward translation) and then the data were translated to Hebrew by an expert in both languages (back translation) at which time both version of the data were checked for accuracy (Dhamani and Richter, 2011). Data analysis The study used a hybrid approach of inductive and deductive coding. In the first phase of the analysis, I used deductive content analysis by which existing data were analyzed by testing categories, concepts, models or hypotheses based on earlier theories, models, mind maps and literature reviews (Elo and Kyngas, 2008). That is, I first looked for elements of two theoretical frameworks, information poverty and sense-making in the data. When it became clear that these two frameworks did not fully reflect the complexities, I encountered in the data I adopted an inductive analysis in which I identified new ideas or themes derived from the data (Elo and Kyngas, 2008). A categorization matrix (Table II) was developed after the first phase of the analysis and was augmented with the categories that resulted in the second phase. The content analysis of the interviews revealed three major themes: information marginalization includes data that describe the different factors that keep migrants at the social margins and the ways that this marginalization is reflected in their information seeking of everyday life information; information resistance includes data that describe the ways by which migrants hold off or rebuff accessing and receiving information; overcoming resistance includes data that reveal the ways by which migrants and social mediators try to overcome information resistance and ultimately information marginalization (see Table II). This theme has two main categories: elements of information marginalization that hinder migrants' social integration, and Ager and Strang's (2008) markers and means: three indicators of integration viewed through the perspective of information marginalization. The categories comprised in this theme related to Ager and Strang's (2008) facilitators: language and cultural knowledge and safety and security. Findings from the content analysis showed that participants did not view these elements as facilitators of integration, but rather as factors in the migrants' lives that hinder their integration into Israeli society. This theme is comprised of four subcategories: lack of cultural knowledge, lack of language proficiency, living in an unsafe and insecure environment, and discrimination. Lack of cultural knowledge Ager and Strang's (2008) framework of integration relates to the acquisition of knowledge as a facilitator of integration that has been described as "a prerequisite for social inclusion is knowledge about the social, economic and community dimensions through which any society is constituted" (Lloyd et al., 2013, p. 122). Findings from the content analysis of the interviews revealed that participants in the study viewed the lack of cultural knowledge that is the gaps or discontinuities that migrants experience as one of the main obstacles in their process of integration. For example, Sandra, who works at an NGO that organizes educational training workshops, described how the lack of cultural knowledge makes it difficult for migrants to make sense of their host society, and access the information they need in their everyday life:There are a lot of people who really have nowhere to turn for all sorts of things. There is a lack of understanding of the system, lack of understanding of the law, they do not understand the system, do not understand the law, here [in Israel] we have a different culture and language. This [lack of cultural knowledge] creates a lot of emotional stress, they talk a lot about this emotional stress. Sofia, a worker at an NGO dedicated to protect the migrants' civil rights, described how the lack of cultural knowledge as an obstacle for integration is especially salient for migrants who come from rural societies:A lack of understanding of the systems, a lack of understanding of the culture in which you live, a lack of understanding of bureaucracy, it is much more obvious in people who come from a rural society, and suddenly they have to deal with bureaucracy and forms, which is terrible. So it is precisely there that the shock is more noticeable, how to get along at all, with all the rules, with all the procedures. It's very difficult for them. Information marginalization does not exist solely on the users' side, it is reinforced by the system's inability to reach its users and understand their needs. Natalia, a social worker at a prominent municipal NGO, describes the gap that exists between her organization and its clients:I feel it all the time. I do not know how to reach them [the migrants], to talk to them and help them, we need to speak their cultural language, their real language so we can communicate with them. Lack of language proficiency Language acquisition has been described as one of the first stepping stones of social integration into a new society (Aspinall, 2007; Clark et al., 2014) and, alongside cultural knowledge, it is considered by Ager and Strang (2008) as a facilitator of integration. Sandra related the importance of language acquisition to the process of integration:Obviously, the first thing is language. Language is the first obstacle you will encounter in a foreign country. Language has also a cognitive element; how am I supposed to function in a society if I don't speak the language? Language is also related to trust. Natalia commented on how information that is available in the migrants' native language will be more accessible and will be perceived as more reliable:I think that we miss the point in terms of transmitting messages when it is not done in their [the migrants'] mother language in terms of trust, they find it easier to trust the information if it's in their own language. Sara explained how talking to migrants in their own language helps open lines of communication:My Spanish is completely broken, but when I sit down and talk to her [a Latin American migrant] in Spanish, she will open a lot more, she will feel much more relaxed. It also brings a lot of humor into the conversation, and then she opens up and feels more relaxed. Living in an unsafe and unstable environment Ager and Strang (2008) stressed the importance of achieving a sense of safety and stability as a facilitator of integration. Undocumented migrants who experience an unsafe and unstable environment remain on the outskirts of society and cannot access the services and information they need even in extreme cases. Sara described how the fear of deportation affects the migrants' sense of safety and stability, and how this fear directly impacts their health:This is the situation. They [migrants] live in fear of deportation, they somehow manage to live under the radar, not knowing their rights, and in the context of health, it results in neglect and fear to seek treatment until it's sometimes too late. Emily, a social worker at a local NGO related a case in which a woman was experiencing a clear danger to her safety and well-being:There was one very serious case of physical violence and she did not know where to turn. Who is supposed to take care of them? If you are undocumented, you do not have the protection of the state, so if you find yourself in a difficult situation you have no one who can help you. Discrimination Many times, it is the system that discriminates and reacts with intolerance toward the migrants' needs. Participants described a number of discriminatory incidents experienced by migrants in their interactions with the healthcare system, as Sara explained:[Bureaucracy in the health system] is also difficult for Israelis to understand, so it is much harder for migrants. Very often they face intolerance from the administration of the HMO, it is very difficult. I mean, many times they encounter hostility and discrimination. For example, I tell them to call me from the clinic and I will talk to the clerk and I will explain the situation to her. But then, when they get to the clinic the clerks will not talk to us, they are not willing to help them. Sofia noted how migrants find it hard to deal with long-term bureaucratic processes because of intolerance and discrimination:When they have to ask for a disability pension or a rehabilitation allowance, these long processes are very difficult to manage for them because they will need to meet again and again the unfriendly face and the crooked eye and all the other things that make it more difficult for them as a person who lives on the margins of society to get a little bit closer to the mainstream of that society. Means and markers Ager and Strang (2008) proposed four key areas that are indicative of a successful integration: housing, education, health and employment. The following excerpts from the interviews show how information marginalization was part of three of these aspects in the migrants' lives in the following three subcategories: health, education and employment. Health Lack of understanding of how the health system works is one of the main obstacles to migrant integration. Natalia explained how migrants struggle to understand the concept of health insurance:They have a real problem understanding the concept of health insurance. We often receive complaints of the following kind: "I paid health insurance for two years and I was not sick and they would not pay me my money back." I say it with half a smile but it's really not funny for the people who feel like they have wasted their money. Then what happens is that they sometimes cancel their insurance and have no health coverage when they get sick. Dealing with emergency rooms in big urban hospitals can be a trying experience for migrants. Sara described how they are afraid to sign the paperwork and pay the admission fee and get sent away:When they go to the emergency room they are not willing to sign the paperwork, they are very scared of it and then they come to us and say they did not want to treat me. In other words, they refused to treat me and wanted me to pay. They say, "pay or you will not receive treatment". There is some misunderstanding around the procedures in the emergency rooms. Education Ager and Strang (2008) referred to education as one of the means and markers of integration. Content analysis of the interviews revealed that migrants face socioeconomic barriers and personal obstacles such as a lack of language proficiency that impedes their access to education:The issue of education, migrants want it very much, they are eager to learn, but it feels impossible for them. This is something lacking in their lives, it's too hard for them, it's not accessible to them. Employment The impact that information marginalization has on the integration process is especially noticeable in the area of employment. According to participants, migrants find it hard to understand different notions related to employment such as taxes and work visas. Sofia related her experiences in this area:They get a culture shock when people tell them that to work you need a visa. "Why do I need a visa to work?" "I just work and you pay me." On a very basic level, what they know is that if you work for someone in the fields, then he pays you for the days you work. For migrant populations, the issue of taxation can be difficult to understand:The whole issue of taxes is very strange to them, they constantly say: why do we need it? Is always very difficult for them to understand and assimilate this [the issue of taxes]. There is a lot to explain, to try to explain the reason behind it [the taxes] they do not understand why there are so many laws. So we had to explain this issue to them all the time. Migrants' employment opportunities are also affected by their legal status and by the information sources accessible to them. Sofia mentions the obstacles that migrants encounter when they are looking for a job:As an Israeli, how do I look for a job? In our world, for the last few years, it is very normal to go online and look for work; we send resumes to online job sites. But how can a migrant worker send his resume online? They have to rely on friends or relatives who will help them find a job. If someone has a job, maybe he can ask if they need more workers, so their employment possibilities are very limited. The second theme in this study is information resistance. Content analysis of the interviews revealed that migrants resist information as a defensive behavior in response to the unstable and unsafe situation they face in their host country and to their lack of cultural knowledge. This theme includes two subcategories: secrecy and disinformation. Secrecy The notion of information resistance is related to Chatman's (1991) principle of secrecy. In this study, secrecy refers to the need that migrants have to safeguard themselves from unwanted exposure and purposively conceal their need for information as a defensive behavior. The purpose of secrecy is to guard against disclosure, which results in people not being receptive to advice or receiving information. Natalia described how undocumented migrants that live in a state of secrecy are oftentimes not receptive to the services provided by their organizations for fear of exposure and deportation:Sometimes they are reluctant to get help from us, or families that will try to disappear under the radar so as not to accept our intervention. And then it's a real courtship that sometimes lasts for months until we find the family. It's really hard to find people who are undocumented, nobody knows their address. I have no way of locating them, except for the help of activists, pastors, priests; we work with churches a lot. Sara conveyed how fear and the need for secrecy prevent migrants from getting the medical help they need. Concurring with Chatman's (1996) assertion that "it became clear that an influence on one's decision to remain secretive about a concern was fear" (p. 199):Health is a big issue for them [the migrants]. Being here without a visa, is a constant threat. Things can turn bad if they have a child, they don't know what to do, so they don't ask for help if they become ill. To be isolated like that, to be away from your family and bear all the responsibility all by yourself for the family's livelihood. Even without being a doctor I can see that all of these things affect them mentally. Disinformation Resisting or rejecting information that comes from outsiders (Chatman, 1996) and trusting insiders' unreliable sources of information that appear trustworthy is one of the outcomes of information resistance. Participants relate how disinformation negatively impacts migrants' lives:They have a strong will to live here that it is very empowering, but there are a lot of anxieties, a lot. A lot of times these anxieties come from rumours that grow bigger and bigger into something really monstrous and catastrophic, that ultimately does a lot of damage. Disinformation can also preclude migrants from reaching out for help. Natalia described a situation in which misinformation and a lack of trust stopped a family from getting the help they needed:We had such a case in which a girl had missed a lot of school days, and no one had been able to find her mother. Eventually, we found out that her mother was always afraid to contact us or reveal details about them as a family because she was afraid we would take her child away from her or pick her up and deport her. All sorts of things like that that don't actually happen. Migrants are also affected by the disinformation about them in Israeli society and directly affect their employment prospects:Many Israelis believe that if they employ an undocumented migrant they will be fined or prosecuted. This obstacle is certainly a structural obstacle that comes from campaigns disseminating a false message. They say, whoever employs an undocumented foreign worker is considered a criminal. This theme consists of two categories that describe the efforts of social mediators to find new, relevant ways to communicate with migrants as well as the role that social connections play in overcoming information resistance and marginalization. Situational relevance How can information resistance be overcome? What constitutes a trustworthy and relevant information source? People perceive an information source as relevant and accessible if they can make sense and trust the information provided by the source (Bronstein, 2014; Ruthven et al., 2018). That is, "even if a source is perceived as potentially useful, it will not do much good to the individual if that source is not legitimized by contextual others" (Chatman, 1996, p. 202). Yona, a social worker at an NGO that provides health services to migrants and other marginalized populations, explained the importance of trust in overcoming resistance:To migrants, there are those who can be trusted and those who cannot, so they won't trust information if it's written and I show them that it's written in a law book, but they will not believe it. They will not believe it. It does not matter if I show them that there is a law now and the law says it this way, it will not give them the confidence; they will only trust the information coming from people they trust, then they will believe it. Sandra describes the difficulties she encountered when trying to market a training workshop and how she solves the problem by marketing the information through information sources that the community trusted:We tried marketing the workshop on Facebook, advertised it in all kinds of groups that we thought were relevant, but it didn't work so well. I felt that in this community, I had to act differently, not using normal marketing channels, but really personal. So I called people. I called someone who was a community leader and said, "Listen, it's [the workshop] really worth it to people. Help them, tell people to come." As time passed I realized that this is the way. Find the people who have connections, gatekeepers of the community, let them advertise it, let them push it and do it in cooperation with them because they bring the people. This way the workshop was a success. Emily talked about another project that helped migrant workers with small children register their children in public kindergarten:We [at her organization] want as many children as possible from the age of 3 to go to kindergarten. But registration for the kindergarten opens in December. These communities don't think long term, so a lot of people miss the registration period, so we created videos, in several languages, uploaded them to Facebook announcing that the registration for kindergarten was open, and described what needs to be done and where to go [in the videos]. You see a person from the community who does the whole registration process from the beginning. He shows all the registration steps, shows them what kind of paperwork they need. It was a success; three years ago, only 40% of migrant children were registered on time and today 95% of migrant children are registered. Social connections Ager and Strang (2008) considered social connections as facilitators that remove barriers that help migrants make sense in their host society and advance their social integration. The content analysis of the data identified two subcategories: social bonds and social links. Social bonds Ager and Strang (2008) asserted that social bonds created with the migrants' close social network play an important role helping them feel settled and integrated. The intermediaries who participated in the study talked about different initiatives that provided migrants with information that made sense to them and fulfilled some of their needs. Natalia talked about one of these initiatives that supported communal leadership:The aim at the beginning was to form a communal network. We tried to create a social network of people who help each other. We organized all sorts of committees. The Filipino and the South American communities have groups of mothers who support each other, for example. When one of them is ill, the others take care of her children. They visit her at the hospital, take care of her, a real network of mutual support. Social linking Ager and Strang (2008) described social links as the connection between individuals and structures of the state. Sherry talked about the important role that NGOs and other institutions have as information sources:A lot of word of mouth [...] many times they will come to us because we are considered a relatively general organization, we direct them to other places. We use Facebook and other organizations also use Facebook and distribute information there, also Whatsapp, and all sorts of mailing lists. Tamara, a social worker at a municipal NGO, described the important role that churches play in the lives of migrants as social links to the host society:Churches are a place to seek information, they are places that give first aid. In other words, there are churches that will turn the world around and find a solution for someone living in the street. Whether they give him money or give him a room in the church for a couple of days, they are a really necessary support resource for communities. Sara explained how they organize talks with different communities through churches and other community centers to talk to migrants about the healthcare system and health insurance:We have done informational evenings in churches on Friday or Sunday evenings, where the pastor gave us half an hour to talk about how emergency rooms work or health insurance, but there are still gaps. We need to go out more into the communities, reach out especially in small communities where it's a bit more complicated. This study presented a new and more nuanced understanding of the process of integration of migrants by re-examining Ager and Strang's (2008) framework from an informational perspective told by the intermediaries who work with these populations. Findings from the content analysis revealed that the process of integration is shaped by the inability of both migrants and the institutions in the host society to close the cultural and social gaps that ultimately result in information marginalization. Two main themes were revealed in the content analysis: information marginalization and information resistance. The first theme contained two main categories that reflected the elements of marginalization and how it is reflected in Ager and Strang's (2008) means and markers. Although the framework related to cultural and language knowledge as facilitators of integration, participants in this study perceived the lack of these indicators as obstacles that precluded migrants from getting their basic needs covered, reinforcing their marginalization. Echoing this finding, Martzoukou and Burnett (2018) observed that "the direct relationship between social inclusion and knowledge of the socioeconomic and community information-related constructs," and found "that there were fundamental gaps in Syrian Scots knowledge domains in relation to everyday life matters, which would be considered as 'routine' knowledge in mainstream Scottish society and beyond" (p. 1126). Just as participants in this study related the migrants' lack of understanding of admission procedures at emergency rooms, Martzoukou and Burnett (2018) described an instance in which a participant in their study missed a doctor's appointment when she could not understand how to ride the bus. The second element of information marginalization revealed in the content analysis is the lack of language proficiency that represents a salient theme in the literature on migration (Bronstein, 2019; Ikafa, 2015; Leith, 2009; Lloyd et al., 2013; Martzoukou and Burnett, 2018; Wessendorf and Phillimore, 2019). Grzymala-Kazlowska (2018) described language as an anchor and "a foothold that allows individuals to acquire socio-psychological stability and security and function effectively in a new or substantially changed life" (p. 235). Findings from this study reinforced this assertion and showed that language proficiency was directly related to the ability of migrants to solve everyday problems and meet basic needs, as well as to develop trust, make sense of their reality and be receptive to outsiders and to the information they provide. Information resistance occurred when information in their native language was perceived as the only information that made sense and could be trusted (Chatman, 1996). Two other elements of information marginalization were: living in an unsafe and unstable environment, and discrimination. These elements directly relate to the migrants' ability to understand their new reality and lessen the resistance they experience as a safeguard against a hostile environment bringing migrants "to engage in information-related practices, such as information sharing, which would enable them to understand how the various instrumental environments (health, education, employment) operate" (Lloyd et al., 2017) and allowing people to safely engage with services and with other people to integrate (Ndofor-Tah et al., 2019). Supporting Ager and Strang's (2008) assertion that to feel at home, migrants need to feel safe, participants in this study related how the fear of deportation and discrimination precluded them from seeking information and how that often negatively affected their well-being. Past studies have noted the significant role that a safe and stable environment plays in the process of integration (Ager and Strang, 2008; Ikafa, 2015; Saggar et al., 2012) and represented a crucial foothold and point of reference which enabled migrants to adapt to their new life and feel settled (Grzymala-Kazlowska, 2018). Findings showed that accessing and understanding information related to Ager and Strang's (2008) markers and means is directly related to the migrants' ability to bridge the gaps in their reality, make sense of their surroundings and advance in their integration into Israeli society (Baird, 2012; Ikafa, 2015). That is, the sociocultural gap that prevents migrants from understanding how the education and the healthcare systems work, and how to find and retain proper employment directly contributes to their marginality and does not allow them "to move beyond the liminal zones of society and become included" (Lloyd et al., 2017). The second theme revealed by content analysis refers to the information resistance experienced by migrants when they resist information that comes from outsiders and renders them unable or unwilling to bridge the cultural gap that will help them make sense of their reality. This resistance is oftentimes compounded by inability of the institutions in the host country to help them. This theme contains two categories: secrecy and disinformation. Echoing Chatman's (1996) concept of secrecy as a safeguarding behavior, participants described how migrants, "cease to be receptive to advice and information" (p. 195) and resisted the information they needed in an attempt to protect themselves from an environment they perceived as unsafe. Participants related how resistance to information perceived as not trustworthy (i.e. coming from outsiders) negatively impacted migrants' lives and resulted in disinformation that ultimately hindered their integration process. The third theme revealed in the study presents the findings related to the migrants' and the intermediaries' attempts to overcome information resistance. This theme contains two categories: situational relevance and social connections. The intermediaries participating in the study described different strategies they used to reach migrants and present information in formats that made sense to them. Supporting Chatman's (1996) assertion that "things that make sense are relevant" (p. 202), participants tried to make information relevant to migrants by creating videos to explain the kindergarten registration process or by having trusted gatekeepers share information about training programs or community services. That is, participants understood the importance of legitimizing the information through contextual others (Chatman, 1996). The legitimation of information by contextualization is related to the creation of social connections. The significant role that social connections play in the integration process of migrants is a salient feature in the literature (Bronstein, 2019; Cheung and Phillimore, 2013; Gilmartin and Migge, 2015; Grzymala-Kazlowska, 2018; OECD, 2018). Two of Ager and Strang's (2008) social connections were revealed in the content analysis: social bonds as close social relations and social linking as the relationships built with the intermediaries. Echoing past studies findings showed that creating social bonds within their community helped migrants resolve problems, feel more secure and stable (Grzymala-Kazlowska, 2018) and reinforce their settlement intentions (Torunczyk-Ruiz and Brunarska, 2018). Interestingly, the two new indicators that were added in the 2019 UK Home Office indicators (Ndofor-Tah et al., 2019), leisure and digital skills did not appeared in the data analysis. Further research should investigate the role of these two elements in the integration process of migrants. Phillimore (2012) wrote, "Integration implies the development of a sense of belonging in the host community, with some renegotiation of identity by both newcomers and hosts" (p. 3). Findings from the study showed that it is in this oftentimes failed renegotiation that information marginalization emerges. By allowing intermediaries to articulate their views and opinions, I was able to understand how personal, cultural and social factors impacted the lives of migrants, and distanced them from the sources of information and support they need to feel at home in their new country. Findings showed that for integration to be successful, it should be the result of an effort by both migrants and local institutions/structures to reach a middle ground of understanding and compromise. This study extending Gibson and Martin's (2019) notion of information marginalization to encompass both sides of the equation of marginalization and situating it in the social, economic and contextual conditions that create information poverty. By examining Ager and Strang's (2008) framework of integration from an information perspective, this study highlighted the role that information plays as a facilitating or an impeding element in the integration process, as well as a vehicle to help migrants make sense of their reality, regain social capital, perceive their environment as safe and develop a sense of belonging to their new home.
|
Ten intermediaries working with migrants were interviewed using semi-structure interviews. They were analyzed using an integrative approach of deductive and inductive content analysis and rendered categories drawn from the theoretical frameworks and categories that emerged from the data.
|
[SECTION: Findings] In this study, I present the findings from a qualitative investigation that explored the other side of the issue of integration of migrants; that is, the views and perceptions of information intermediaries working with migrants in Israel about the integration process. The notion of integration has been the focus of academic debate in recent years and has been defined in different ways (Ager and Strang, 2008; Cheung and Phillimore, 2014; Farach et al., 2015; Gilmartin and Migge, 2015; Harder et al., 2018; Phillimore, 2012). In its broadest sense, integration means the process or transition by which people who are relatively new to a country become part of society (Rudiger and Spencer, 2003). Integration is achieved when "people, whatever their background, live, work, learn and socialise together, based on shared rights, responsibilities and opportunities" (Ndofor-Tah et al., 2019) while keeping a measure of their original cultural identity (Threadgold and Court, 2005). Integration is being characterized today as a multidimensional and multidirectional (Harder et al., 2018) process that encompasses "access to resources and opportunities as well as social mixing" involving adjustments by everyone in society (Ndofor-Tah et al., 2019). Integration is the desired outcome of migration which has been described in the literature as a transition (Bronstein, 2019; Meleis, 2015) - a disruptive process that fractures migrants' information landscapes (Lloyd, 2017) separates them from their traditional social networks, and renders their known information sources and practices inadequate (Giralt, 2015; Lloyd et al., 2013). To integrate into their host society, migrants need to find ways to redefine themselves, redevelop self-agency (Kralik et al., 2006) and learn news "ways of knowing" (Lloyd, 2017; Martzoukou and Burnett, 2018). Past studies that focused on this population have investigated the challenges and obstacles migrants face in the process of social inclusion and integration into their new society (Atiso et al., 2018; Allard and Caidi, 2018; Bronstein, 2017; Clark et al., 2014; Khoir et al., 2015; Lloyd, 2015). Migrants are one of the most marginalized population groups worldwide (OECD, 2018). To adequately design and provide information and services for these populations, LIS professionals need to understand how marginalization works. This study draws from Gibson and Martin's (2019) notion of information marginalization "that describes the systematic, interactive socio-technical processes that can push and hold certain groups of people at social 'margins,' where their needs are persistently ignored or overlooked" (p. 1). Information marginalization shifts the focus away from the behavior of the information poor who mistrust their surroundings (Chatman, 1996) and often lack the necessary skills or abilities to effectively seek and retrieve information (Britz, 2004) toward the institutions and places in which information poverty is deeply embedded. In my investigation into the views and perceptions of intermediaries working to provide information and services to the migrant community, I realized that the obstacles and difficulties experienced by migrants when trying to access and use the information they needed in their everyday lives originated in both the migrants themselves and in the systems and institutions designed to assist them. Hence, this paper proposes a new perspective of Gibson and Martin's (2019) concept that addresses the cultural and social gaps existing between the migrants and the individuals and organizations working with them. The second concept I propose is information resistance based on Dervin's (1999) sense-making concept of resistance. Dervin (1999) asserted that sense-making "assumes that articulation of one's lived experience including its struggles and resistances as well as alignments with given order is in itself a Sense-Making journey" (p. 742). I would argue that the social, cultural and language gaps that exist between migrants and intermediaries render the information and services offered often inadequate or irrelevant resulting in migrants resisting or rejecting the information offered to them that ultimately results in a situation of information marginalization that hinders migrants' process of integration. This study contributes to the literature on the role of information in the integration of migrants in three ways: first, it explores a different perspective about the role that information plays in the integration process of migrant workers by exploring the views and opinions of individuals and organizations that work with these communities daily. Most LIS studies considering migrants worldwide have focused on their information needs and behavior (Allard and Caidi, 2018; Bronstein, 2017, 2019; Caidi and Allard, 2005; Fisher et al., 2004; Lloyd, 2015) and their use of information technologies (Baron et al., 2014; Dekker and Engbersen, 2014; Suh and Hsieh, 2019). Their intent was to facilitate their integration into their host society. Second, instead of a top-down approach to integration that quantifies social, economic and educational indicators and reviews integration programs and policies as outcomes of the process of social inclusion (Cheung and Phillimore, 2017; Lichtenstein and Puma, 2018; Puma et al., 2018; Schibel et al., 2002; Schneider and Crul, 2010), this study considers the issue from a qualitative perspective that allows participants to talk about their experiences and opinions. This study provides a different perspective about an issue that is often hard to measure because it relies on "surveys of attitudes, feelings and perceptions" (OECD, 2018, p. 25). Third, the study proposes a new perspective to Ager and Strang's (2008) framework of integration by looking at its different elements through the perspective provided by Gibson and Martin's (2019) concept of information marginalization and Dervin's (1999) sense-making notion of resistance. Social integration of migrants This study adopted the United Nations Migration Agency (IOM)'s (United Nations, n.d.) general definition of migrant as any person who is moving or has moved across an international border or within a State away from his/her habitual place of residence, regardless of the person's legal status; whether the movement is voluntary or involuntary; what the causes for the movement are; or what the length of the stay is. The study draws from Ager and Strang (2008) framework of indicators of integration that define the core domains of integration of migrants by describing the different information needs (markers and means), information sources (social connections) and facilitators that can support and foster the integration process. I would argue that behind this framework lies the need for information (Wilson, 1981) as one of the factors that foster or hinder the process of social integration of migrants. The framework includes the following elements: Markers and means: to live and to be integrated into a new society, migrants need information to fulfill their immediate needs (what Ager and Strang (2008) termed markers and means): employment, housing, education and health. Social connections: Ager and Strang (2008) described these as a "connective tissue" (p. 177) that represents the defining feature of an integrated community. The authors recalled three forms of social connections according to Putnam (1995): social bonds (with family and co-ethnic, co-national, co-religious or other forms of group), social bridges (with other communities) and social links (with the structures of the state). These social connections serve as sources of informal and formal information and support. The current study investigates the perceptions of the third kind of social connections of migrants, social links with governmental and voluntary organizations. Facilitators: the process of integration requires facilitators that foster social inclusion. Ager and Strang (2008) outlined two major facilitators, language and knowledge acquisition that are key areas of cultural competence, safety and stability that allow migrants to feel at home in their host society. Foundation: notions of citizenship and rights are discussed as an essential prerequisite for integration. This element did not appear in the findings of the study. Ager and Strang's (2008) framework of integration as served as the basis for a number of studies dealing with integration of different migrant communities (Erdal, 2013; Hainmueller et al., 2017; Platts-Fowler and Robinson, 2015; Spencer and Charsley, 2016). Recently, the UK Home office published a new report on indicators of integration (Ndofor-Tah et al., 2019). This new framework is structured around Ager and Strang's (2008) domains adding leisure to the marker and means domains and digital skills as a new facilitator. In addition, the new framework relates to language and communication as separate domains from culture as well as safety and stability. There is a considerable body of work dealing with integration of migrants that can only be briefly summarized here. Prior research has discussed the difficulty in conceptualizing integration (Ager and Strang, 2008; Da Lomba, 2010; Robinson, 1998). This is why Castles et al. (2002) suggested that integration should be recognized as an umbrella term that encompasses a wide range of criteria that help researchers "assess the migrants' own motivations, strategies and networks." Da Lomba (2010) asserted that integration has two broad dimensions: public and private. The public dimension of integration refers to the legal and social environment (i.e. housing, education and health) in which refugees and immigrants are located. The private dimension of integration discusses to personal situations and demographic status (i.e. gender, education, dependence on social benefits and employment prospects) as well as a migrant's social connections (e.g. religious or ethnic identity, familial connections). Past studies can be divided into two main groups based on the methodological perspective chosen to conceptualize, examine and measure integration. The first group of studies used different quantitative criteria to measure integration. In an early study, Kuhlman (1991) proposed a complex model of integration that took into account a range of different indicators of integration that included spatial integration, economic integration, social integration, political integration, legal integration and psychological integration. He further stated that long-term outcomes may be influenced by early experiences that could result in disadvantage and marginalization. Along this line of thought, Hainmueller et al.'s (2016) longitudinal study found that receiving the host country citizenship early in the process had a lasting impact on enhancing social integration. Cheung and Phillimore (2017) examined gender differences of integration in the complex relationships between language proficiency, social networks and three key social policy areas: employment/education, health and housing. Their findings showed clear genders differences for these criteria. Saggar et al. (2012) focused on social cohesion as an integration indicator using the following measures: national identity measures (feelings of belonging); integration by group measures (outcomes in education, jobs, housing and so on); Cohesion measures in local neighborhoods (feelings of safety, well-being and belonging in the local community). Following this line of research, Torunczyk-Ruiz and Brunarska (2018) examined access to emotional and instrumental social capital in the host country as a strong indicator that enables migrants to develop a stronger sense and shape settlement intentions of temporary migrants. Kearns and Whitley (2015) found a positive correlation between time and place and self-reported indicators of social integration such as trust, reliance and safety, social relations and a sense of community. Hahn et al.'s (2019) study reported that an increased incremental validity of personality traits and cognitive factors on the integration process of migrants even with respect to contextual and sociodemographic factors. A number of studies have identified mental health as a significant factor of integration that is negatively impacted by other indicators such as precarious employment and lack of social connections, trust and self-identity (Brydsten et al., 2019; Lin et al., 2016). A second group of studies have investigated different aspects of integration qualitatively using semi-structured interviews to explore a more personal perspective on different aspects of integration (Grzymala-Kazlowska, 2018; Vancea and Boso, 2015). Gilmartin and Migge (2015) explored the pathways of integration of European migrants to Ireland. They found that cultural and social pathways (i.e. language acquisition, education, and building social relationships) are central to integration while economic factors such as employment were perceived as enablers of social pathways. Phillimore et al. (2014) interviewed migrants in the UK on what integration meant to them. Participants talked about "integration as a process and that the passage of time allows people to become more integrated" (p. 8). They related how integration was strongly associated with language proficiency, understanding British culture and developing strong social connections with close social environments and institutional settings. The significance of social capital has been recognized in other studies that have focused on the role that close and distant social networks play in the integration process and in creating a sense of belonging (Cheung and Phillimore, 2013; Grzymala-Kazlowska, 2018; Jorgensen, 2017; Phillimore, 2012; Wessendorf, 2019). In a recent study, Kyeremeh et al. (2019) investigated the notion of successful integration among African migrants in Canada who felt that integration should be a gradual and ongoing process and explained that the host society should allow migrants with ample time to acquaint and adjust to the new life. Information marginalization and information resistance Social exclusion or marginalization has traditionally been attributed to cultural, economic and social factors, but a number of studies in the LIS field have also shown that information provision is a key component of social exclusion. Communities that lack access to information sources are oftentimes socially excluded (Britz, 2004; Caidi and Allard, 2005; Lloyd, 2015; Yu, 2012). Recent studies have stated that migrants' social integration or inclusion "depends on their preparedness to trust new information and information sources and their understanding of information flow in their new environment" (Lloyd et al., 2017). The theoretical framework of information poverty has been used in the past to explain the social exclusion of marginalized groups in society (Bronstein, 2017; Chatman, 1996, 1999; Lingel and Boyd, 2013). These studies have asserted that people living in an information impoverished world: experience social isolation and mistrust of their surroundings, see themselves as devoid of information sources, live in economic poverty or experience social exclusion or stigma, and engage in defensive behaviors (Britz, 2004; Chatman, 1996; Lingel and Boyd, 2013) that hinder their access and use of information. However, Haider and Bawden (2007) asserted that this economic and technological deterministic view that dictates "there is a 'right kind' as well as a 'right amount' of information" (p. 3) has transformed information into a commodity and has caused the literature on the subject to neglect "the social and communicative aspects of information and in particular of information use that would allow for a serious engagement with different types of informational practices by considering the various social conditions" (p. 535). Following this line of thought, the current study focuses on the social and communicative aspects of information. This is done by analyzing different elements that comprise the gaps that exist between migrants and their host society that hinder their access and use of information, and ultimately their social integration. The social approach to information behavior (Burnett et al., 2001; Chatman, 1991, 1996) proposed a theoretical framework that suggested a link between social marginalization, isolation, limited information access and particular self-protective information-related practices as well as selective acceptance or avoidance of new information. The current study reframes the notion of selective acceptance or avoidance found in past studies (Chatman, 1999; Lingel and Boyd, 2013) within the sense-making framework and proposes the concept of information resistance. Dervin (1992) defined the sense-making approach as "a set of assumptions and propositions, and a set of methods which have been developed to study the making of sense that people do in their everyday experiences" (p. 61). Making sense is a process, and sense is the product of this process, that "includes 'knowledge,' but also a host of intuitions, opinions, hunches effective responses, evaluations, questions, etc." (Savolainen, 1993, p. 16). At the foundation of this framework, sense-making assumes that reality is neither complete nor constant, but rather it is filled with fundamental and pervasive discontinuities or gaps (Dervin, 1983) that bring humans to constantly seek continuity as the "ultimate goal of life in general and human condition in particular" (Savolainen, 1993, p. 16). Discontinuity is the negation of this state of desired continuity that occurs when the individual faces a challenge or a problematic situation and needs to find a new sense in her new reality. That is, when a new reality blocks the movement through time-space, a discontinuity is created. This discontinuity or gap can be resolved through information seeking and sense-making can take place. However, humans face many barriers and constraints that block the sense-making process and bring them to resist or challenge the information encountered (Dervin, 1992). I would argue that during the integration process, migrants face discontinuities or gaps in all aspects of their lives and need to bridge these gaps by acquiring sociocultural and language knowledge of their host society. Information marginalization occurs when migrants experience information resistance that renders them unable or unwilling to bridge these gaps because of sociocultural, political, and economic barriers and a sense of insecurity and instability that hinders their advancement toward integration (Dervin, 1992). The information role of intermediaries for marginalized groups Prior research has shown that people rely on networks of personal contacts to access and acquire information (Buchanan et al., 2019; Agada, 1999; Lu, 2007; Palsdottir, 2012; Sabelli, 2016). People who seek and disseminate information for others have been referred to in the LIS literature in different ways: gatekeepers (Sturges, 2001), social mediators (Sabelli, 2012) or lay information mediaries (Abrahamson and Fisher, 2007) who "through filtering and transmitting information subtly shape and inform our social reality and knowledge worlds" (Lu, 2007, p. 115). This study adopted Buchanan et al.'s (2019) notion of information intermediaries who "play a key role in recognizing, understanding, and progressing information needs in the problematic context" (p. 127). Participants in the current study are "socially positioned" (p. 113) gatekeepers working at institutions such as government agencies, NGOs and health organizations that have a strategic advantage that endow them with ready access to resources and allow them to seek and disseminate information. A number of past studies have examined the information role that intermediaries have for marginalized or disadvantaged populations (Buchanan et al., 2019; Sabelli, 2016; Schilderman, 2002). In an early study, Kurtz (1966) investigated the acculturation process of a Spanish-surnamed population in the USA who migrated from rural areas to an urban environment, and for whom gatekeepers held a strategic position for providing ready access to resources needed to solve problems. Agada (1999) investigated the "information use environments," of residents of an African-American inner-city neighborhood and found that gatekeepers were those individuals within the community who were able to move between cultures and link their communities with resources, and who were perceived as trustworthy and credible sources of information. Durrance et al. (2006) examined everyday information behaviors for local community-based citizen's groups in the USA and found that they play an important gatekeeping role because they make information more relevant for their constituents by distilling, tailoring and vetting the information according to the residents' needs. Schilderman (2002) reported on the role that intermediaries have as sources of information for the urban poor in Sri Lanka, Peru and Zimbabwe who prefer gatekeepers from social networks based on kinship, proximity or friendship rather than institutional gatekeepers (government agencies, NGOs and health organizations) who might be better informed but who have been unsuccessful in providing the urban poor with information in personalized formats - and in perceiving the urban poor as partners who are an important source of indigenous knowledge. Contrary to Schilderman (2002), two other studies described the successful role that institutional intermediaries have in advancing the information needs of marginalized populations. Sabelli (2012) found that for young women in vulnerable settings in Uruguay, institutional mediators represent an "essential bridge" in "the process of appropriation of information and knowledge for both personal and collective development." Using Granovetter's (1983) weak/strong ties terminology, Bronstein (2018) examined the information behavior of migrant workers in Israel and found that as weak ties, institutional intermediaries can extend access to information sources and skills not available through the strong ties within the migrant community. Moreover, because the information provided comes from people with whom migrant workers do not have a reciprocal or intimate bond, this information is perceived as social support that fosters a sense of belonging. Past studies have examined different elements of the work being done at humanitarian and public organizations assisting migrant communities in different countries (Mathews and Jiayi, 2017). These studies have examined the experiences of health practitioners (Broom et al., 2019; Kelly, 2019; Teunissen et al., 2016), privacy concerns in humanitarian organizations (Vannini et al., 2019), immediate refugee support (Cederquist, 2019), settings providing social support (Stock, 2019) and motivations for volunteering for refugees (Artero, 2019). Population of the study Ten people were interviewed for the study. Table I shows details about the participants. All participants worked directly with migrants at NGOs or public institutions. These organizations assist all the migrant community in Israel, that is, no differentiation was made between different groups of migrants at an organizational or a personal level. In other words, in the interviews, participants talked about migrants in general without referring to their legal status (i.e. refugees, migrant workers, undocumented migrants). Characteristics of study participants are presented in Table I. Data collection To gain access to the field, I volunteered at a workers' hotline named Kav La'oved, as a worker's counselor for domestic migrant workers for four years. Through my work at Kav La'oved, I developed social connections with social workers and other volunteers at different NGOs who participated in the study and referred me to additional participants. Participants were interviewed at their places of work using a semi-structured questionnaire (see Appendix). The interviews lasted 60-90 min were conducted in Hebrew, recorded and fully transcribed. The textual data were translated using the double or back translation method to secure the reliability of the data translated. Using this method, I first translated the data from Hebrew to English (forward translation) and then the data were translated to Hebrew by an expert in both languages (back translation) at which time both version of the data were checked for accuracy (Dhamani and Richter, 2011). Data analysis The study used a hybrid approach of inductive and deductive coding. In the first phase of the analysis, I used deductive content analysis by which existing data were analyzed by testing categories, concepts, models or hypotheses based on earlier theories, models, mind maps and literature reviews (Elo and Kyngas, 2008). That is, I first looked for elements of two theoretical frameworks, information poverty and sense-making in the data. When it became clear that these two frameworks did not fully reflect the complexities, I encountered in the data I adopted an inductive analysis in which I identified new ideas or themes derived from the data (Elo and Kyngas, 2008). A categorization matrix (Table II) was developed after the first phase of the analysis and was augmented with the categories that resulted in the second phase. The content analysis of the interviews revealed three major themes: information marginalization includes data that describe the different factors that keep migrants at the social margins and the ways that this marginalization is reflected in their information seeking of everyday life information; information resistance includes data that describe the ways by which migrants hold off or rebuff accessing and receiving information; overcoming resistance includes data that reveal the ways by which migrants and social mediators try to overcome information resistance and ultimately information marginalization (see Table II). This theme has two main categories: elements of information marginalization that hinder migrants' social integration, and Ager and Strang's (2008) markers and means: three indicators of integration viewed through the perspective of information marginalization. The categories comprised in this theme related to Ager and Strang's (2008) facilitators: language and cultural knowledge and safety and security. Findings from the content analysis showed that participants did not view these elements as facilitators of integration, but rather as factors in the migrants' lives that hinder their integration into Israeli society. This theme is comprised of four subcategories: lack of cultural knowledge, lack of language proficiency, living in an unsafe and insecure environment, and discrimination. Lack of cultural knowledge Ager and Strang's (2008) framework of integration relates to the acquisition of knowledge as a facilitator of integration that has been described as "a prerequisite for social inclusion is knowledge about the social, economic and community dimensions through which any society is constituted" (Lloyd et al., 2013, p. 122). Findings from the content analysis of the interviews revealed that participants in the study viewed the lack of cultural knowledge that is the gaps or discontinuities that migrants experience as one of the main obstacles in their process of integration. For example, Sandra, who works at an NGO that organizes educational training workshops, described how the lack of cultural knowledge makes it difficult for migrants to make sense of their host society, and access the information they need in their everyday life:There are a lot of people who really have nowhere to turn for all sorts of things. There is a lack of understanding of the system, lack of understanding of the law, they do not understand the system, do not understand the law, here [in Israel] we have a different culture and language. This [lack of cultural knowledge] creates a lot of emotional stress, they talk a lot about this emotional stress. Sofia, a worker at an NGO dedicated to protect the migrants' civil rights, described how the lack of cultural knowledge as an obstacle for integration is especially salient for migrants who come from rural societies:A lack of understanding of the systems, a lack of understanding of the culture in which you live, a lack of understanding of bureaucracy, it is much more obvious in people who come from a rural society, and suddenly they have to deal with bureaucracy and forms, which is terrible. So it is precisely there that the shock is more noticeable, how to get along at all, with all the rules, with all the procedures. It's very difficult for them. Information marginalization does not exist solely on the users' side, it is reinforced by the system's inability to reach its users and understand their needs. Natalia, a social worker at a prominent municipal NGO, describes the gap that exists between her organization and its clients:I feel it all the time. I do not know how to reach them [the migrants], to talk to them and help them, we need to speak their cultural language, their real language so we can communicate with them. Lack of language proficiency Language acquisition has been described as one of the first stepping stones of social integration into a new society (Aspinall, 2007; Clark et al., 2014) and, alongside cultural knowledge, it is considered by Ager and Strang (2008) as a facilitator of integration. Sandra related the importance of language acquisition to the process of integration:Obviously, the first thing is language. Language is the first obstacle you will encounter in a foreign country. Language has also a cognitive element; how am I supposed to function in a society if I don't speak the language? Language is also related to trust. Natalia commented on how information that is available in the migrants' native language will be more accessible and will be perceived as more reliable:I think that we miss the point in terms of transmitting messages when it is not done in their [the migrants'] mother language in terms of trust, they find it easier to trust the information if it's in their own language. Sara explained how talking to migrants in their own language helps open lines of communication:My Spanish is completely broken, but when I sit down and talk to her [a Latin American migrant] in Spanish, she will open a lot more, she will feel much more relaxed. It also brings a lot of humor into the conversation, and then she opens up and feels more relaxed. Living in an unsafe and unstable environment Ager and Strang (2008) stressed the importance of achieving a sense of safety and stability as a facilitator of integration. Undocumented migrants who experience an unsafe and unstable environment remain on the outskirts of society and cannot access the services and information they need even in extreme cases. Sara described how the fear of deportation affects the migrants' sense of safety and stability, and how this fear directly impacts their health:This is the situation. They [migrants] live in fear of deportation, they somehow manage to live under the radar, not knowing their rights, and in the context of health, it results in neglect and fear to seek treatment until it's sometimes too late. Emily, a social worker at a local NGO related a case in which a woman was experiencing a clear danger to her safety and well-being:There was one very serious case of physical violence and she did not know where to turn. Who is supposed to take care of them? If you are undocumented, you do not have the protection of the state, so if you find yourself in a difficult situation you have no one who can help you. Discrimination Many times, it is the system that discriminates and reacts with intolerance toward the migrants' needs. Participants described a number of discriminatory incidents experienced by migrants in their interactions with the healthcare system, as Sara explained:[Bureaucracy in the health system] is also difficult for Israelis to understand, so it is much harder for migrants. Very often they face intolerance from the administration of the HMO, it is very difficult. I mean, many times they encounter hostility and discrimination. For example, I tell them to call me from the clinic and I will talk to the clerk and I will explain the situation to her. But then, when they get to the clinic the clerks will not talk to us, they are not willing to help them. Sofia noted how migrants find it hard to deal with long-term bureaucratic processes because of intolerance and discrimination:When they have to ask for a disability pension or a rehabilitation allowance, these long processes are very difficult to manage for them because they will need to meet again and again the unfriendly face and the crooked eye and all the other things that make it more difficult for them as a person who lives on the margins of society to get a little bit closer to the mainstream of that society. Means and markers Ager and Strang (2008) proposed four key areas that are indicative of a successful integration: housing, education, health and employment. The following excerpts from the interviews show how information marginalization was part of three of these aspects in the migrants' lives in the following three subcategories: health, education and employment. Health Lack of understanding of how the health system works is one of the main obstacles to migrant integration. Natalia explained how migrants struggle to understand the concept of health insurance:They have a real problem understanding the concept of health insurance. We often receive complaints of the following kind: "I paid health insurance for two years and I was not sick and they would not pay me my money back." I say it with half a smile but it's really not funny for the people who feel like they have wasted their money. Then what happens is that they sometimes cancel their insurance and have no health coverage when they get sick. Dealing with emergency rooms in big urban hospitals can be a trying experience for migrants. Sara described how they are afraid to sign the paperwork and pay the admission fee and get sent away:When they go to the emergency room they are not willing to sign the paperwork, they are very scared of it and then they come to us and say they did not want to treat me. In other words, they refused to treat me and wanted me to pay. They say, "pay or you will not receive treatment". There is some misunderstanding around the procedures in the emergency rooms. Education Ager and Strang (2008) referred to education as one of the means and markers of integration. Content analysis of the interviews revealed that migrants face socioeconomic barriers and personal obstacles such as a lack of language proficiency that impedes their access to education:The issue of education, migrants want it very much, they are eager to learn, but it feels impossible for them. This is something lacking in their lives, it's too hard for them, it's not accessible to them. Employment The impact that information marginalization has on the integration process is especially noticeable in the area of employment. According to participants, migrants find it hard to understand different notions related to employment such as taxes and work visas. Sofia related her experiences in this area:They get a culture shock when people tell them that to work you need a visa. "Why do I need a visa to work?" "I just work and you pay me." On a very basic level, what they know is that if you work for someone in the fields, then he pays you for the days you work. For migrant populations, the issue of taxation can be difficult to understand:The whole issue of taxes is very strange to them, they constantly say: why do we need it? Is always very difficult for them to understand and assimilate this [the issue of taxes]. There is a lot to explain, to try to explain the reason behind it [the taxes] they do not understand why there are so many laws. So we had to explain this issue to them all the time. Migrants' employment opportunities are also affected by their legal status and by the information sources accessible to them. Sofia mentions the obstacles that migrants encounter when they are looking for a job:As an Israeli, how do I look for a job? In our world, for the last few years, it is very normal to go online and look for work; we send resumes to online job sites. But how can a migrant worker send his resume online? They have to rely on friends or relatives who will help them find a job. If someone has a job, maybe he can ask if they need more workers, so their employment possibilities are very limited. The second theme in this study is information resistance. Content analysis of the interviews revealed that migrants resist information as a defensive behavior in response to the unstable and unsafe situation they face in their host country and to their lack of cultural knowledge. This theme includes two subcategories: secrecy and disinformation. Secrecy The notion of information resistance is related to Chatman's (1991) principle of secrecy. In this study, secrecy refers to the need that migrants have to safeguard themselves from unwanted exposure and purposively conceal their need for information as a defensive behavior. The purpose of secrecy is to guard against disclosure, which results in people not being receptive to advice or receiving information. Natalia described how undocumented migrants that live in a state of secrecy are oftentimes not receptive to the services provided by their organizations for fear of exposure and deportation:Sometimes they are reluctant to get help from us, or families that will try to disappear under the radar so as not to accept our intervention. And then it's a real courtship that sometimes lasts for months until we find the family. It's really hard to find people who are undocumented, nobody knows their address. I have no way of locating them, except for the help of activists, pastors, priests; we work with churches a lot. Sara conveyed how fear and the need for secrecy prevent migrants from getting the medical help they need. Concurring with Chatman's (1996) assertion that "it became clear that an influence on one's decision to remain secretive about a concern was fear" (p. 199):Health is a big issue for them [the migrants]. Being here without a visa, is a constant threat. Things can turn bad if they have a child, they don't know what to do, so they don't ask for help if they become ill. To be isolated like that, to be away from your family and bear all the responsibility all by yourself for the family's livelihood. Even without being a doctor I can see that all of these things affect them mentally. Disinformation Resisting or rejecting information that comes from outsiders (Chatman, 1996) and trusting insiders' unreliable sources of information that appear trustworthy is one of the outcomes of information resistance. Participants relate how disinformation negatively impacts migrants' lives:They have a strong will to live here that it is very empowering, but there are a lot of anxieties, a lot. A lot of times these anxieties come from rumours that grow bigger and bigger into something really monstrous and catastrophic, that ultimately does a lot of damage. Disinformation can also preclude migrants from reaching out for help. Natalia described a situation in which misinformation and a lack of trust stopped a family from getting the help they needed:We had such a case in which a girl had missed a lot of school days, and no one had been able to find her mother. Eventually, we found out that her mother was always afraid to contact us or reveal details about them as a family because she was afraid we would take her child away from her or pick her up and deport her. All sorts of things like that that don't actually happen. Migrants are also affected by the disinformation about them in Israeli society and directly affect their employment prospects:Many Israelis believe that if they employ an undocumented migrant they will be fined or prosecuted. This obstacle is certainly a structural obstacle that comes from campaigns disseminating a false message. They say, whoever employs an undocumented foreign worker is considered a criminal. This theme consists of two categories that describe the efforts of social mediators to find new, relevant ways to communicate with migrants as well as the role that social connections play in overcoming information resistance and marginalization. Situational relevance How can information resistance be overcome? What constitutes a trustworthy and relevant information source? People perceive an information source as relevant and accessible if they can make sense and trust the information provided by the source (Bronstein, 2014; Ruthven et al., 2018). That is, "even if a source is perceived as potentially useful, it will not do much good to the individual if that source is not legitimized by contextual others" (Chatman, 1996, p. 202). Yona, a social worker at an NGO that provides health services to migrants and other marginalized populations, explained the importance of trust in overcoming resistance:To migrants, there are those who can be trusted and those who cannot, so they won't trust information if it's written and I show them that it's written in a law book, but they will not believe it. They will not believe it. It does not matter if I show them that there is a law now and the law says it this way, it will not give them the confidence; they will only trust the information coming from people they trust, then they will believe it. Sandra describes the difficulties she encountered when trying to market a training workshop and how she solves the problem by marketing the information through information sources that the community trusted:We tried marketing the workshop on Facebook, advertised it in all kinds of groups that we thought were relevant, but it didn't work so well. I felt that in this community, I had to act differently, not using normal marketing channels, but really personal. So I called people. I called someone who was a community leader and said, "Listen, it's [the workshop] really worth it to people. Help them, tell people to come." As time passed I realized that this is the way. Find the people who have connections, gatekeepers of the community, let them advertise it, let them push it and do it in cooperation with them because they bring the people. This way the workshop was a success. Emily talked about another project that helped migrant workers with small children register their children in public kindergarten:We [at her organization] want as many children as possible from the age of 3 to go to kindergarten. But registration for the kindergarten opens in December. These communities don't think long term, so a lot of people miss the registration period, so we created videos, in several languages, uploaded them to Facebook announcing that the registration for kindergarten was open, and described what needs to be done and where to go [in the videos]. You see a person from the community who does the whole registration process from the beginning. He shows all the registration steps, shows them what kind of paperwork they need. It was a success; three years ago, only 40% of migrant children were registered on time and today 95% of migrant children are registered. Social connections Ager and Strang (2008) considered social connections as facilitators that remove barriers that help migrants make sense in their host society and advance their social integration. The content analysis of the data identified two subcategories: social bonds and social links. Social bonds Ager and Strang (2008) asserted that social bonds created with the migrants' close social network play an important role helping them feel settled and integrated. The intermediaries who participated in the study talked about different initiatives that provided migrants with information that made sense to them and fulfilled some of their needs. Natalia talked about one of these initiatives that supported communal leadership:The aim at the beginning was to form a communal network. We tried to create a social network of people who help each other. We organized all sorts of committees. The Filipino and the South American communities have groups of mothers who support each other, for example. When one of them is ill, the others take care of her children. They visit her at the hospital, take care of her, a real network of mutual support. Social linking Ager and Strang (2008) described social links as the connection between individuals and structures of the state. Sherry talked about the important role that NGOs and other institutions have as information sources:A lot of word of mouth [...] many times they will come to us because we are considered a relatively general organization, we direct them to other places. We use Facebook and other organizations also use Facebook and distribute information there, also Whatsapp, and all sorts of mailing lists. Tamara, a social worker at a municipal NGO, described the important role that churches play in the lives of migrants as social links to the host society:Churches are a place to seek information, they are places that give first aid. In other words, there are churches that will turn the world around and find a solution for someone living in the street. Whether they give him money or give him a room in the church for a couple of days, they are a really necessary support resource for communities. Sara explained how they organize talks with different communities through churches and other community centers to talk to migrants about the healthcare system and health insurance:We have done informational evenings in churches on Friday or Sunday evenings, where the pastor gave us half an hour to talk about how emergency rooms work or health insurance, but there are still gaps. We need to go out more into the communities, reach out especially in small communities where it's a bit more complicated. This study presented a new and more nuanced understanding of the process of integration of migrants by re-examining Ager and Strang's (2008) framework from an informational perspective told by the intermediaries who work with these populations. Findings from the content analysis revealed that the process of integration is shaped by the inability of both migrants and the institutions in the host society to close the cultural and social gaps that ultimately result in information marginalization. Two main themes were revealed in the content analysis: information marginalization and information resistance. The first theme contained two main categories that reflected the elements of marginalization and how it is reflected in Ager and Strang's (2008) means and markers. Although the framework related to cultural and language knowledge as facilitators of integration, participants in this study perceived the lack of these indicators as obstacles that precluded migrants from getting their basic needs covered, reinforcing their marginalization. Echoing this finding, Martzoukou and Burnett (2018) observed that "the direct relationship between social inclusion and knowledge of the socioeconomic and community information-related constructs," and found "that there were fundamental gaps in Syrian Scots knowledge domains in relation to everyday life matters, which would be considered as 'routine' knowledge in mainstream Scottish society and beyond" (p. 1126). Just as participants in this study related the migrants' lack of understanding of admission procedures at emergency rooms, Martzoukou and Burnett (2018) described an instance in which a participant in their study missed a doctor's appointment when she could not understand how to ride the bus. The second element of information marginalization revealed in the content analysis is the lack of language proficiency that represents a salient theme in the literature on migration (Bronstein, 2019; Ikafa, 2015; Leith, 2009; Lloyd et al., 2013; Martzoukou and Burnett, 2018; Wessendorf and Phillimore, 2019). Grzymala-Kazlowska (2018) described language as an anchor and "a foothold that allows individuals to acquire socio-psychological stability and security and function effectively in a new or substantially changed life" (p. 235). Findings from this study reinforced this assertion and showed that language proficiency was directly related to the ability of migrants to solve everyday problems and meet basic needs, as well as to develop trust, make sense of their reality and be receptive to outsiders and to the information they provide. Information resistance occurred when information in their native language was perceived as the only information that made sense and could be trusted (Chatman, 1996). Two other elements of information marginalization were: living in an unsafe and unstable environment, and discrimination. These elements directly relate to the migrants' ability to understand their new reality and lessen the resistance they experience as a safeguard against a hostile environment bringing migrants "to engage in information-related practices, such as information sharing, which would enable them to understand how the various instrumental environments (health, education, employment) operate" (Lloyd et al., 2017) and allowing people to safely engage with services and with other people to integrate (Ndofor-Tah et al., 2019). Supporting Ager and Strang's (2008) assertion that to feel at home, migrants need to feel safe, participants in this study related how the fear of deportation and discrimination precluded them from seeking information and how that often negatively affected their well-being. Past studies have noted the significant role that a safe and stable environment plays in the process of integration (Ager and Strang, 2008; Ikafa, 2015; Saggar et al., 2012) and represented a crucial foothold and point of reference which enabled migrants to adapt to their new life and feel settled (Grzymala-Kazlowska, 2018). Findings showed that accessing and understanding information related to Ager and Strang's (2008) markers and means is directly related to the migrants' ability to bridge the gaps in their reality, make sense of their surroundings and advance in their integration into Israeli society (Baird, 2012; Ikafa, 2015). That is, the sociocultural gap that prevents migrants from understanding how the education and the healthcare systems work, and how to find and retain proper employment directly contributes to their marginality and does not allow them "to move beyond the liminal zones of society and become included" (Lloyd et al., 2017). The second theme revealed by content analysis refers to the information resistance experienced by migrants when they resist information that comes from outsiders and renders them unable or unwilling to bridge the cultural gap that will help them make sense of their reality. This resistance is oftentimes compounded by inability of the institutions in the host country to help them. This theme contains two categories: secrecy and disinformation. Echoing Chatman's (1996) concept of secrecy as a safeguarding behavior, participants described how migrants, "cease to be receptive to advice and information" (p. 195) and resisted the information they needed in an attempt to protect themselves from an environment they perceived as unsafe. Participants related how resistance to information perceived as not trustworthy (i.e. coming from outsiders) negatively impacted migrants' lives and resulted in disinformation that ultimately hindered their integration process. The third theme revealed in the study presents the findings related to the migrants' and the intermediaries' attempts to overcome information resistance. This theme contains two categories: situational relevance and social connections. The intermediaries participating in the study described different strategies they used to reach migrants and present information in formats that made sense to them. Supporting Chatman's (1996) assertion that "things that make sense are relevant" (p. 202), participants tried to make information relevant to migrants by creating videos to explain the kindergarten registration process or by having trusted gatekeepers share information about training programs or community services. That is, participants understood the importance of legitimizing the information through contextual others (Chatman, 1996). The legitimation of information by contextualization is related to the creation of social connections. The significant role that social connections play in the integration process of migrants is a salient feature in the literature (Bronstein, 2019; Cheung and Phillimore, 2013; Gilmartin and Migge, 2015; Grzymala-Kazlowska, 2018; OECD, 2018). Two of Ager and Strang's (2008) social connections were revealed in the content analysis: social bonds as close social relations and social linking as the relationships built with the intermediaries. Echoing past studies findings showed that creating social bonds within their community helped migrants resolve problems, feel more secure and stable (Grzymala-Kazlowska, 2018) and reinforce their settlement intentions (Torunczyk-Ruiz and Brunarska, 2018). Interestingly, the two new indicators that were added in the 2019 UK Home Office indicators (Ndofor-Tah et al., 2019), leisure and digital skills did not appeared in the data analysis. Further research should investigate the role of these two elements in the integration process of migrants. Phillimore (2012) wrote, "Integration implies the development of a sense of belonging in the host community, with some renegotiation of identity by both newcomers and hosts" (p. 3). Findings from the study showed that it is in this oftentimes failed renegotiation that information marginalization emerges. By allowing intermediaries to articulate their views and opinions, I was able to understand how personal, cultural and social factors impacted the lives of migrants, and distanced them from the sources of information and support they need to feel at home in their new country. Findings showed that for integration to be successful, it should be the result of an effort by both migrants and local institutions/structures to reach a middle ground of understanding and compromise. This study extending Gibson and Martin's (2019) notion of information marginalization to encompass both sides of the equation of marginalization and situating it in the social, economic and contextual conditions that create information poverty. By examining Ager and Strang's (2008) framework of integration from an information perspective, this study highlighted the role that information plays as a facilitating or an impeding element in the integration process, as well as a vehicle to help migrants make sense of their reality, regain social capital, perceive their environment as safe and develop a sense of belonging to their new home.
|
The content analysis of the data revealed that information marginalization is characterized by the lack of cultural knowledge and lack of language proficiency that impact the migrants abilities to fulfill their everyday needs, experience a safe and stable environment. Information marginalization results in migrants experiencing self-protective behaviors such as secrecy and an inability to trust information sources that are not contextualized by insiders. Findings show that information resistance can be overcome by making information available in relevant formats and distributed through trusted sources.
|
[SECTION: Value] In this study, I present the findings from a qualitative investigation that explored the other side of the issue of integration of migrants; that is, the views and perceptions of information intermediaries working with migrants in Israel about the integration process. The notion of integration has been the focus of academic debate in recent years and has been defined in different ways (Ager and Strang, 2008; Cheung and Phillimore, 2014; Farach et al., 2015; Gilmartin and Migge, 2015; Harder et al., 2018; Phillimore, 2012). In its broadest sense, integration means the process or transition by which people who are relatively new to a country become part of society (Rudiger and Spencer, 2003). Integration is achieved when "people, whatever their background, live, work, learn and socialise together, based on shared rights, responsibilities and opportunities" (Ndofor-Tah et al., 2019) while keeping a measure of their original cultural identity (Threadgold and Court, 2005). Integration is being characterized today as a multidimensional and multidirectional (Harder et al., 2018) process that encompasses "access to resources and opportunities as well as social mixing" involving adjustments by everyone in society (Ndofor-Tah et al., 2019). Integration is the desired outcome of migration which has been described in the literature as a transition (Bronstein, 2019; Meleis, 2015) - a disruptive process that fractures migrants' information landscapes (Lloyd, 2017) separates them from their traditional social networks, and renders their known information sources and practices inadequate (Giralt, 2015; Lloyd et al., 2013). To integrate into their host society, migrants need to find ways to redefine themselves, redevelop self-agency (Kralik et al., 2006) and learn news "ways of knowing" (Lloyd, 2017; Martzoukou and Burnett, 2018). Past studies that focused on this population have investigated the challenges and obstacles migrants face in the process of social inclusion and integration into their new society (Atiso et al., 2018; Allard and Caidi, 2018; Bronstein, 2017; Clark et al., 2014; Khoir et al., 2015; Lloyd, 2015). Migrants are one of the most marginalized population groups worldwide (OECD, 2018). To adequately design and provide information and services for these populations, LIS professionals need to understand how marginalization works. This study draws from Gibson and Martin's (2019) notion of information marginalization "that describes the systematic, interactive socio-technical processes that can push and hold certain groups of people at social 'margins,' where their needs are persistently ignored or overlooked" (p. 1). Information marginalization shifts the focus away from the behavior of the information poor who mistrust their surroundings (Chatman, 1996) and often lack the necessary skills or abilities to effectively seek and retrieve information (Britz, 2004) toward the institutions and places in which information poverty is deeply embedded. In my investigation into the views and perceptions of intermediaries working to provide information and services to the migrant community, I realized that the obstacles and difficulties experienced by migrants when trying to access and use the information they needed in their everyday lives originated in both the migrants themselves and in the systems and institutions designed to assist them. Hence, this paper proposes a new perspective of Gibson and Martin's (2019) concept that addresses the cultural and social gaps existing between the migrants and the individuals and organizations working with them. The second concept I propose is information resistance based on Dervin's (1999) sense-making concept of resistance. Dervin (1999) asserted that sense-making "assumes that articulation of one's lived experience including its struggles and resistances as well as alignments with given order is in itself a Sense-Making journey" (p. 742). I would argue that the social, cultural and language gaps that exist between migrants and intermediaries render the information and services offered often inadequate or irrelevant resulting in migrants resisting or rejecting the information offered to them that ultimately results in a situation of information marginalization that hinders migrants' process of integration. This study contributes to the literature on the role of information in the integration of migrants in three ways: first, it explores a different perspective about the role that information plays in the integration process of migrant workers by exploring the views and opinions of individuals and organizations that work with these communities daily. Most LIS studies considering migrants worldwide have focused on their information needs and behavior (Allard and Caidi, 2018; Bronstein, 2017, 2019; Caidi and Allard, 2005; Fisher et al., 2004; Lloyd, 2015) and their use of information technologies (Baron et al., 2014; Dekker and Engbersen, 2014; Suh and Hsieh, 2019). Their intent was to facilitate their integration into their host society. Second, instead of a top-down approach to integration that quantifies social, economic and educational indicators and reviews integration programs and policies as outcomes of the process of social inclusion (Cheung and Phillimore, 2017; Lichtenstein and Puma, 2018; Puma et al., 2018; Schibel et al., 2002; Schneider and Crul, 2010), this study considers the issue from a qualitative perspective that allows participants to talk about their experiences and opinions. This study provides a different perspective about an issue that is often hard to measure because it relies on "surveys of attitudes, feelings and perceptions" (OECD, 2018, p. 25). Third, the study proposes a new perspective to Ager and Strang's (2008) framework of integration by looking at its different elements through the perspective provided by Gibson and Martin's (2019) concept of information marginalization and Dervin's (1999) sense-making notion of resistance. Social integration of migrants This study adopted the United Nations Migration Agency (IOM)'s (United Nations, n.d.) general definition of migrant as any person who is moving or has moved across an international border or within a State away from his/her habitual place of residence, regardless of the person's legal status; whether the movement is voluntary or involuntary; what the causes for the movement are; or what the length of the stay is. The study draws from Ager and Strang (2008) framework of indicators of integration that define the core domains of integration of migrants by describing the different information needs (markers and means), information sources (social connections) and facilitators that can support and foster the integration process. I would argue that behind this framework lies the need for information (Wilson, 1981) as one of the factors that foster or hinder the process of social integration of migrants. The framework includes the following elements: Markers and means: to live and to be integrated into a new society, migrants need information to fulfill their immediate needs (what Ager and Strang (2008) termed markers and means): employment, housing, education and health. Social connections: Ager and Strang (2008) described these as a "connective tissue" (p. 177) that represents the defining feature of an integrated community. The authors recalled three forms of social connections according to Putnam (1995): social bonds (with family and co-ethnic, co-national, co-religious or other forms of group), social bridges (with other communities) and social links (with the structures of the state). These social connections serve as sources of informal and formal information and support. The current study investigates the perceptions of the third kind of social connections of migrants, social links with governmental and voluntary organizations. Facilitators: the process of integration requires facilitators that foster social inclusion. Ager and Strang (2008) outlined two major facilitators, language and knowledge acquisition that are key areas of cultural competence, safety and stability that allow migrants to feel at home in their host society. Foundation: notions of citizenship and rights are discussed as an essential prerequisite for integration. This element did not appear in the findings of the study. Ager and Strang's (2008) framework of integration as served as the basis for a number of studies dealing with integration of different migrant communities (Erdal, 2013; Hainmueller et al., 2017; Platts-Fowler and Robinson, 2015; Spencer and Charsley, 2016). Recently, the UK Home office published a new report on indicators of integration (Ndofor-Tah et al., 2019). This new framework is structured around Ager and Strang's (2008) domains adding leisure to the marker and means domains and digital skills as a new facilitator. In addition, the new framework relates to language and communication as separate domains from culture as well as safety and stability. There is a considerable body of work dealing with integration of migrants that can only be briefly summarized here. Prior research has discussed the difficulty in conceptualizing integration (Ager and Strang, 2008; Da Lomba, 2010; Robinson, 1998). This is why Castles et al. (2002) suggested that integration should be recognized as an umbrella term that encompasses a wide range of criteria that help researchers "assess the migrants' own motivations, strategies and networks." Da Lomba (2010) asserted that integration has two broad dimensions: public and private. The public dimension of integration refers to the legal and social environment (i.e. housing, education and health) in which refugees and immigrants are located. The private dimension of integration discusses to personal situations and demographic status (i.e. gender, education, dependence on social benefits and employment prospects) as well as a migrant's social connections (e.g. religious or ethnic identity, familial connections). Past studies can be divided into two main groups based on the methodological perspective chosen to conceptualize, examine and measure integration. The first group of studies used different quantitative criteria to measure integration. In an early study, Kuhlman (1991) proposed a complex model of integration that took into account a range of different indicators of integration that included spatial integration, economic integration, social integration, political integration, legal integration and psychological integration. He further stated that long-term outcomes may be influenced by early experiences that could result in disadvantage and marginalization. Along this line of thought, Hainmueller et al.'s (2016) longitudinal study found that receiving the host country citizenship early in the process had a lasting impact on enhancing social integration. Cheung and Phillimore (2017) examined gender differences of integration in the complex relationships between language proficiency, social networks and three key social policy areas: employment/education, health and housing. Their findings showed clear genders differences for these criteria. Saggar et al. (2012) focused on social cohesion as an integration indicator using the following measures: national identity measures (feelings of belonging); integration by group measures (outcomes in education, jobs, housing and so on); Cohesion measures in local neighborhoods (feelings of safety, well-being and belonging in the local community). Following this line of research, Torunczyk-Ruiz and Brunarska (2018) examined access to emotional and instrumental social capital in the host country as a strong indicator that enables migrants to develop a stronger sense and shape settlement intentions of temporary migrants. Kearns and Whitley (2015) found a positive correlation between time and place and self-reported indicators of social integration such as trust, reliance and safety, social relations and a sense of community. Hahn et al.'s (2019) study reported that an increased incremental validity of personality traits and cognitive factors on the integration process of migrants even with respect to contextual and sociodemographic factors. A number of studies have identified mental health as a significant factor of integration that is negatively impacted by other indicators such as precarious employment and lack of social connections, trust and self-identity (Brydsten et al., 2019; Lin et al., 2016). A second group of studies have investigated different aspects of integration qualitatively using semi-structured interviews to explore a more personal perspective on different aspects of integration (Grzymala-Kazlowska, 2018; Vancea and Boso, 2015). Gilmartin and Migge (2015) explored the pathways of integration of European migrants to Ireland. They found that cultural and social pathways (i.e. language acquisition, education, and building social relationships) are central to integration while economic factors such as employment were perceived as enablers of social pathways. Phillimore et al. (2014) interviewed migrants in the UK on what integration meant to them. Participants talked about "integration as a process and that the passage of time allows people to become more integrated" (p. 8). They related how integration was strongly associated with language proficiency, understanding British culture and developing strong social connections with close social environments and institutional settings. The significance of social capital has been recognized in other studies that have focused on the role that close and distant social networks play in the integration process and in creating a sense of belonging (Cheung and Phillimore, 2013; Grzymala-Kazlowska, 2018; Jorgensen, 2017; Phillimore, 2012; Wessendorf, 2019). In a recent study, Kyeremeh et al. (2019) investigated the notion of successful integration among African migrants in Canada who felt that integration should be a gradual and ongoing process and explained that the host society should allow migrants with ample time to acquaint and adjust to the new life. Information marginalization and information resistance Social exclusion or marginalization has traditionally been attributed to cultural, economic and social factors, but a number of studies in the LIS field have also shown that information provision is a key component of social exclusion. Communities that lack access to information sources are oftentimes socially excluded (Britz, 2004; Caidi and Allard, 2005; Lloyd, 2015; Yu, 2012). Recent studies have stated that migrants' social integration or inclusion "depends on their preparedness to trust new information and information sources and their understanding of information flow in their new environment" (Lloyd et al., 2017). The theoretical framework of information poverty has been used in the past to explain the social exclusion of marginalized groups in society (Bronstein, 2017; Chatman, 1996, 1999; Lingel and Boyd, 2013). These studies have asserted that people living in an information impoverished world: experience social isolation and mistrust of their surroundings, see themselves as devoid of information sources, live in economic poverty or experience social exclusion or stigma, and engage in defensive behaviors (Britz, 2004; Chatman, 1996; Lingel and Boyd, 2013) that hinder their access and use of information. However, Haider and Bawden (2007) asserted that this economic and technological deterministic view that dictates "there is a 'right kind' as well as a 'right amount' of information" (p. 3) has transformed information into a commodity and has caused the literature on the subject to neglect "the social and communicative aspects of information and in particular of information use that would allow for a serious engagement with different types of informational practices by considering the various social conditions" (p. 535). Following this line of thought, the current study focuses on the social and communicative aspects of information. This is done by analyzing different elements that comprise the gaps that exist between migrants and their host society that hinder their access and use of information, and ultimately their social integration. The social approach to information behavior (Burnett et al., 2001; Chatman, 1991, 1996) proposed a theoretical framework that suggested a link between social marginalization, isolation, limited information access and particular self-protective information-related practices as well as selective acceptance or avoidance of new information. The current study reframes the notion of selective acceptance or avoidance found in past studies (Chatman, 1999; Lingel and Boyd, 2013) within the sense-making framework and proposes the concept of information resistance. Dervin (1992) defined the sense-making approach as "a set of assumptions and propositions, and a set of methods which have been developed to study the making of sense that people do in their everyday experiences" (p. 61). Making sense is a process, and sense is the product of this process, that "includes 'knowledge,' but also a host of intuitions, opinions, hunches effective responses, evaluations, questions, etc." (Savolainen, 1993, p. 16). At the foundation of this framework, sense-making assumes that reality is neither complete nor constant, but rather it is filled with fundamental and pervasive discontinuities or gaps (Dervin, 1983) that bring humans to constantly seek continuity as the "ultimate goal of life in general and human condition in particular" (Savolainen, 1993, p. 16). Discontinuity is the negation of this state of desired continuity that occurs when the individual faces a challenge or a problematic situation and needs to find a new sense in her new reality. That is, when a new reality blocks the movement through time-space, a discontinuity is created. This discontinuity or gap can be resolved through information seeking and sense-making can take place. However, humans face many barriers and constraints that block the sense-making process and bring them to resist or challenge the information encountered (Dervin, 1992). I would argue that during the integration process, migrants face discontinuities or gaps in all aspects of their lives and need to bridge these gaps by acquiring sociocultural and language knowledge of their host society. Information marginalization occurs when migrants experience information resistance that renders them unable or unwilling to bridge these gaps because of sociocultural, political, and economic barriers and a sense of insecurity and instability that hinders their advancement toward integration (Dervin, 1992). The information role of intermediaries for marginalized groups Prior research has shown that people rely on networks of personal contacts to access and acquire information (Buchanan et al., 2019; Agada, 1999; Lu, 2007; Palsdottir, 2012; Sabelli, 2016). People who seek and disseminate information for others have been referred to in the LIS literature in different ways: gatekeepers (Sturges, 2001), social mediators (Sabelli, 2012) or lay information mediaries (Abrahamson and Fisher, 2007) who "through filtering and transmitting information subtly shape and inform our social reality and knowledge worlds" (Lu, 2007, p. 115). This study adopted Buchanan et al.'s (2019) notion of information intermediaries who "play a key role in recognizing, understanding, and progressing information needs in the problematic context" (p. 127). Participants in the current study are "socially positioned" (p. 113) gatekeepers working at institutions such as government agencies, NGOs and health organizations that have a strategic advantage that endow them with ready access to resources and allow them to seek and disseminate information. A number of past studies have examined the information role that intermediaries have for marginalized or disadvantaged populations (Buchanan et al., 2019; Sabelli, 2016; Schilderman, 2002). In an early study, Kurtz (1966) investigated the acculturation process of a Spanish-surnamed population in the USA who migrated from rural areas to an urban environment, and for whom gatekeepers held a strategic position for providing ready access to resources needed to solve problems. Agada (1999) investigated the "information use environments," of residents of an African-American inner-city neighborhood and found that gatekeepers were those individuals within the community who were able to move between cultures and link their communities with resources, and who were perceived as trustworthy and credible sources of information. Durrance et al. (2006) examined everyday information behaviors for local community-based citizen's groups in the USA and found that they play an important gatekeeping role because they make information more relevant for their constituents by distilling, tailoring and vetting the information according to the residents' needs. Schilderman (2002) reported on the role that intermediaries have as sources of information for the urban poor in Sri Lanka, Peru and Zimbabwe who prefer gatekeepers from social networks based on kinship, proximity or friendship rather than institutional gatekeepers (government agencies, NGOs and health organizations) who might be better informed but who have been unsuccessful in providing the urban poor with information in personalized formats - and in perceiving the urban poor as partners who are an important source of indigenous knowledge. Contrary to Schilderman (2002), two other studies described the successful role that institutional intermediaries have in advancing the information needs of marginalized populations. Sabelli (2012) found that for young women in vulnerable settings in Uruguay, institutional mediators represent an "essential bridge" in "the process of appropriation of information and knowledge for both personal and collective development." Using Granovetter's (1983) weak/strong ties terminology, Bronstein (2018) examined the information behavior of migrant workers in Israel and found that as weak ties, institutional intermediaries can extend access to information sources and skills not available through the strong ties within the migrant community. Moreover, because the information provided comes from people with whom migrant workers do not have a reciprocal or intimate bond, this information is perceived as social support that fosters a sense of belonging. Past studies have examined different elements of the work being done at humanitarian and public organizations assisting migrant communities in different countries (Mathews and Jiayi, 2017). These studies have examined the experiences of health practitioners (Broom et al., 2019; Kelly, 2019; Teunissen et al., 2016), privacy concerns in humanitarian organizations (Vannini et al., 2019), immediate refugee support (Cederquist, 2019), settings providing social support (Stock, 2019) and motivations for volunteering for refugees (Artero, 2019). Population of the study Ten people were interviewed for the study. Table I shows details about the participants. All participants worked directly with migrants at NGOs or public institutions. These organizations assist all the migrant community in Israel, that is, no differentiation was made between different groups of migrants at an organizational or a personal level. In other words, in the interviews, participants talked about migrants in general without referring to their legal status (i.e. refugees, migrant workers, undocumented migrants). Characteristics of study participants are presented in Table I. Data collection To gain access to the field, I volunteered at a workers' hotline named Kav La'oved, as a worker's counselor for domestic migrant workers for four years. Through my work at Kav La'oved, I developed social connections with social workers and other volunteers at different NGOs who participated in the study and referred me to additional participants. Participants were interviewed at their places of work using a semi-structured questionnaire (see Appendix). The interviews lasted 60-90 min were conducted in Hebrew, recorded and fully transcribed. The textual data were translated using the double or back translation method to secure the reliability of the data translated. Using this method, I first translated the data from Hebrew to English (forward translation) and then the data were translated to Hebrew by an expert in both languages (back translation) at which time both version of the data were checked for accuracy (Dhamani and Richter, 2011). Data analysis The study used a hybrid approach of inductive and deductive coding. In the first phase of the analysis, I used deductive content analysis by which existing data were analyzed by testing categories, concepts, models or hypotheses based on earlier theories, models, mind maps and literature reviews (Elo and Kyngas, 2008). That is, I first looked for elements of two theoretical frameworks, information poverty and sense-making in the data. When it became clear that these two frameworks did not fully reflect the complexities, I encountered in the data I adopted an inductive analysis in which I identified new ideas or themes derived from the data (Elo and Kyngas, 2008). A categorization matrix (Table II) was developed after the first phase of the analysis and was augmented with the categories that resulted in the second phase. The content analysis of the interviews revealed three major themes: information marginalization includes data that describe the different factors that keep migrants at the social margins and the ways that this marginalization is reflected in their information seeking of everyday life information; information resistance includes data that describe the ways by which migrants hold off or rebuff accessing and receiving information; overcoming resistance includes data that reveal the ways by which migrants and social mediators try to overcome information resistance and ultimately information marginalization (see Table II). This theme has two main categories: elements of information marginalization that hinder migrants' social integration, and Ager and Strang's (2008) markers and means: three indicators of integration viewed through the perspective of information marginalization. The categories comprised in this theme related to Ager and Strang's (2008) facilitators: language and cultural knowledge and safety and security. Findings from the content analysis showed that participants did not view these elements as facilitators of integration, but rather as factors in the migrants' lives that hinder their integration into Israeli society. This theme is comprised of four subcategories: lack of cultural knowledge, lack of language proficiency, living in an unsafe and insecure environment, and discrimination. Lack of cultural knowledge Ager and Strang's (2008) framework of integration relates to the acquisition of knowledge as a facilitator of integration that has been described as "a prerequisite for social inclusion is knowledge about the social, economic and community dimensions through which any society is constituted" (Lloyd et al., 2013, p. 122). Findings from the content analysis of the interviews revealed that participants in the study viewed the lack of cultural knowledge that is the gaps or discontinuities that migrants experience as one of the main obstacles in their process of integration. For example, Sandra, who works at an NGO that organizes educational training workshops, described how the lack of cultural knowledge makes it difficult for migrants to make sense of their host society, and access the information they need in their everyday life:There are a lot of people who really have nowhere to turn for all sorts of things. There is a lack of understanding of the system, lack of understanding of the law, they do not understand the system, do not understand the law, here [in Israel] we have a different culture and language. This [lack of cultural knowledge] creates a lot of emotional stress, they talk a lot about this emotional stress. Sofia, a worker at an NGO dedicated to protect the migrants' civil rights, described how the lack of cultural knowledge as an obstacle for integration is especially salient for migrants who come from rural societies:A lack of understanding of the systems, a lack of understanding of the culture in which you live, a lack of understanding of bureaucracy, it is much more obvious in people who come from a rural society, and suddenly they have to deal with bureaucracy and forms, which is terrible. So it is precisely there that the shock is more noticeable, how to get along at all, with all the rules, with all the procedures. It's very difficult for them. Information marginalization does not exist solely on the users' side, it is reinforced by the system's inability to reach its users and understand their needs. Natalia, a social worker at a prominent municipal NGO, describes the gap that exists between her organization and its clients:I feel it all the time. I do not know how to reach them [the migrants], to talk to them and help them, we need to speak their cultural language, their real language so we can communicate with them. Lack of language proficiency Language acquisition has been described as one of the first stepping stones of social integration into a new society (Aspinall, 2007; Clark et al., 2014) and, alongside cultural knowledge, it is considered by Ager and Strang (2008) as a facilitator of integration. Sandra related the importance of language acquisition to the process of integration:Obviously, the first thing is language. Language is the first obstacle you will encounter in a foreign country. Language has also a cognitive element; how am I supposed to function in a society if I don't speak the language? Language is also related to trust. Natalia commented on how information that is available in the migrants' native language will be more accessible and will be perceived as more reliable:I think that we miss the point in terms of transmitting messages when it is not done in their [the migrants'] mother language in terms of trust, they find it easier to trust the information if it's in their own language. Sara explained how talking to migrants in their own language helps open lines of communication:My Spanish is completely broken, but when I sit down and talk to her [a Latin American migrant] in Spanish, she will open a lot more, she will feel much more relaxed. It also brings a lot of humor into the conversation, and then she opens up and feels more relaxed. Living in an unsafe and unstable environment Ager and Strang (2008) stressed the importance of achieving a sense of safety and stability as a facilitator of integration. Undocumented migrants who experience an unsafe and unstable environment remain on the outskirts of society and cannot access the services and information they need even in extreme cases. Sara described how the fear of deportation affects the migrants' sense of safety and stability, and how this fear directly impacts their health:This is the situation. They [migrants] live in fear of deportation, they somehow manage to live under the radar, not knowing their rights, and in the context of health, it results in neglect and fear to seek treatment until it's sometimes too late. Emily, a social worker at a local NGO related a case in which a woman was experiencing a clear danger to her safety and well-being:There was one very serious case of physical violence and she did not know where to turn. Who is supposed to take care of them? If you are undocumented, you do not have the protection of the state, so if you find yourself in a difficult situation you have no one who can help you. Discrimination Many times, it is the system that discriminates and reacts with intolerance toward the migrants' needs. Participants described a number of discriminatory incidents experienced by migrants in their interactions with the healthcare system, as Sara explained:[Bureaucracy in the health system] is also difficult for Israelis to understand, so it is much harder for migrants. Very often they face intolerance from the administration of the HMO, it is very difficult. I mean, many times they encounter hostility and discrimination. For example, I tell them to call me from the clinic and I will talk to the clerk and I will explain the situation to her. But then, when they get to the clinic the clerks will not talk to us, they are not willing to help them. Sofia noted how migrants find it hard to deal with long-term bureaucratic processes because of intolerance and discrimination:When they have to ask for a disability pension or a rehabilitation allowance, these long processes are very difficult to manage for them because they will need to meet again and again the unfriendly face and the crooked eye and all the other things that make it more difficult for them as a person who lives on the margins of society to get a little bit closer to the mainstream of that society. Means and markers Ager and Strang (2008) proposed four key areas that are indicative of a successful integration: housing, education, health and employment. The following excerpts from the interviews show how information marginalization was part of three of these aspects in the migrants' lives in the following three subcategories: health, education and employment. Health Lack of understanding of how the health system works is one of the main obstacles to migrant integration. Natalia explained how migrants struggle to understand the concept of health insurance:They have a real problem understanding the concept of health insurance. We often receive complaints of the following kind: "I paid health insurance for two years and I was not sick and they would not pay me my money back." I say it with half a smile but it's really not funny for the people who feel like they have wasted their money. Then what happens is that they sometimes cancel their insurance and have no health coverage when they get sick. Dealing with emergency rooms in big urban hospitals can be a trying experience for migrants. Sara described how they are afraid to sign the paperwork and pay the admission fee and get sent away:When they go to the emergency room they are not willing to sign the paperwork, they are very scared of it and then they come to us and say they did not want to treat me. In other words, they refused to treat me and wanted me to pay. They say, "pay or you will not receive treatment". There is some misunderstanding around the procedures in the emergency rooms. Education Ager and Strang (2008) referred to education as one of the means and markers of integration. Content analysis of the interviews revealed that migrants face socioeconomic barriers and personal obstacles such as a lack of language proficiency that impedes their access to education:The issue of education, migrants want it very much, they are eager to learn, but it feels impossible for them. This is something lacking in their lives, it's too hard for them, it's not accessible to them. Employment The impact that information marginalization has on the integration process is especially noticeable in the area of employment. According to participants, migrants find it hard to understand different notions related to employment such as taxes and work visas. Sofia related her experiences in this area:They get a culture shock when people tell them that to work you need a visa. "Why do I need a visa to work?" "I just work and you pay me." On a very basic level, what they know is that if you work for someone in the fields, then he pays you for the days you work. For migrant populations, the issue of taxation can be difficult to understand:The whole issue of taxes is very strange to them, they constantly say: why do we need it? Is always very difficult for them to understand and assimilate this [the issue of taxes]. There is a lot to explain, to try to explain the reason behind it [the taxes] they do not understand why there are so many laws. So we had to explain this issue to them all the time. Migrants' employment opportunities are also affected by their legal status and by the information sources accessible to them. Sofia mentions the obstacles that migrants encounter when they are looking for a job:As an Israeli, how do I look for a job? In our world, for the last few years, it is very normal to go online and look for work; we send resumes to online job sites. But how can a migrant worker send his resume online? They have to rely on friends or relatives who will help them find a job. If someone has a job, maybe he can ask if they need more workers, so their employment possibilities are very limited. The second theme in this study is information resistance. Content analysis of the interviews revealed that migrants resist information as a defensive behavior in response to the unstable and unsafe situation they face in their host country and to their lack of cultural knowledge. This theme includes two subcategories: secrecy and disinformation. Secrecy The notion of information resistance is related to Chatman's (1991) principle of secrecy. In this study, secrecy refers to the need that migrants have to safeguard themselves from unwanted exposure and purposively conceal their need for information as a defensive behavior. The purpose of secrecy is to guard against disclosure, which results in people not being receptive to advice or receiving information. Natalia described how undocumented migrants that live in a state of secrecy are oftentimes not receptive to the services provided by their organizations for fear of exposure and deportation:Sometimes they are reluctant to get help from us, or families that will try to disappear under the radar so as not to accept our intervention. And then it's a real courtship that sometimes lasts for months until we find the family. It's really hard to find people who are undocumented, nobody knows their address. I have no way of locating them, except for the help of activists, pastors, priests; we work with churches a lot. Sara conveyed how fear and the need for secrecy prevent migrants from getting the medical help they need. Concurring with Chatman's (1996) assertion that "it became clear that an influence on one's decision to remain secretive about a concern was fear" (p. 199):Health is a big issue for them [the migrants]. Being here without a visa, is a constant threat. Things can turn bad if they have a child, they don't know what to do, so they don't ask for help if they become ill. To be isolated like that, to be away from your family and bear all the responsibility all by yourself for the family's livelihood. Even without being a doctor I can see that all of these things affect them mentally. Disinformation Resisting or rejecting information that comes from outsiders (Chatman, 1996) and trusting insiders' unreliable sources of information that appear trustworthy is one of the outcomes of information resistance. Participants relate how disinformation negatively impacts migrants' lives:They have a strong will to live here that it is very empowering, but there are a lot of anxieties, a lot. A lot of times these anxieties come from rumours that grow bigger and bigger into something really monstrous and catastrophic, that ultimately does a lot of damage. Disinformation can also preclude migrants from reaching out for help. Natalia described a situation in which misinformation and a lack of trust stopped a family from getting the help they needed:We had such a case in which a girl had missed a lot of school days, and no one had been able to find her mother. Eventually, we found out that her mother was always afraid to contact us or reveal details about them as a family because she was afraid we would take her child away from her or pick her up and deport her. All sorts of things like that that don't actually happen. Migrants are also affected by the disinformation about them in Israeli society and directly affect their employment prospects:Many Israelis believe that if they employ an undocumented migrant they will be fined or prosecuted. This obstacle is certainly a structural obstacle that comes from campaigns disseminating a false message. They say, whoever employs an undocumented foreign worker is considered a criminal. This theme consists of two categories that describe the efforts of social mediators to find new, relevant ways to communicate with migrants as well as the role that social connections play in overcoming information resistance and marginalization. Situational relevance How can information resistance be overcome? What constitutes a trustworthy and relevant information source? People perceive an information source as relevant and accessible if they can make sense and trust the information provided by the source (Bronstein, 2014; Ruthven et al., 2018). That is, "even if a source is perceived as potentially useful, it will not do much good to the individual if that source is not legitimized by contextual others" (Chatman, 1996, p. 202). Yona, a social worker at an NGO that provides health services to migrants and other marginalized populations, explained the importance of trust in overcoming resistance:To migrants, there are those who can be trusted and those who cannot, so they won't trust information if it's written and I show them that it's written in a law book, but they will not believe it. They will not believe it. It does not matter if I show them that there is a law now and the law says it this way, it will not give them the confidence; they will only trust the information coming from people they trust, then they will believe it. Sandra describes the difficulties she encountered when trying to market a training workshop and how she solves the problem by marketing the information through information sources that the community trusted:We tried marketing the workshop on Facebook, advertised it in all kinds of groups that we thought were relevant, but it didn't work so well. I felt that in this community, I had to act differently, not using normal marketing channels, but really personal. So I called people. I called someone who was a community leader and said, "Listen, it's [the workshop] really worth it to people. Help them, tell people to come." As time passed I realized that this is the way. Find the people who have connections, gatekeepers of the community, let them advertise it, let them push it and do it in cooperation with them because they bring the people. This way the workshop was a success. Emily talked about another project that helped migrant workers with small children register their children in public kindergarten:We [at her organization] want as many children as possible from the age of 3 to go to kindergarten. But registration for the kindergarten opens in December. These communities don't think long term, so a lot of people miss the registration period, so we created videos, in several languages, uploaded them to Facebook announcing that the registration for kindergarten was open, and described what needs to be done and where to go [in the videos]. You see a person from the community who does the whole registration process from the beginning. He shows all the registration steps, shows them what kind of paperwork they need. It was a success; three years ago, only 40% of migrant children were registered on time and today 95% of migrant children are registered. Social connections Ager and Strang (2008) considered social connections as facilitators that remove barriers that help migrants make sense in their host society and advance their social integration. The content analysis of the data identified two subcategories: social bonds and social links. Social bonds Ager and Strang (2008) asserted that social bonds created with the migrants' close social network play an important role helping them feel settled and integrated. The intermediaries who participated in the study talked about different initiatives that provided migrants with information that made sense to them and fulfilled some of their needs. Natalia talked about one of these initiatives that supported communal leadership:The aim at the beginning was to form a communal network. We tried to create a social network of people who help each other. We organized all sorts of committees. The Filipino and the South American communities have groups of mothers who support each other, for example. When one of them is ill, the others take care of her children. They visit her at the hospital, take care of her, a real network of mutual support. Social linking Ager and Strang (2008) described social links as the connection between individuals and structures of the state. Sherry talked about the important role that NGOs and other institutions have as information sources:A lot of word of mouth [...] many times they will come to us because we are considered a relatively general organization, we direct them to other places. We use Facebook and other organizations also use Facebook and distribute information there, also Whatsapp, and all sorts of mailing lists. Tamara, a social worker at a municipal NGO, described the important role that churches play in the lives of migrants as social links to the host society:Churches are a place to seek information, they are places that give first aid. In other words, there are churches that will turn the world around and find a solution for someone living in the street. Whether they give him money or give him a room in the church for a couple of days, they are a really necessary support resource for communities. Sara explained how they organize talks with different communities through churches and other community centers to talk to migrants about the healthcare system and health insurance:We have done informational evenings in churches on Friday or Sunday evenings, where the pastor gave us half an hour to talk about how emergency rooms work or health insurance, but there are still gaps. We need to go out more into the communities, reach out especially in small communities where it's a bit more complicated. This study presented a new and more nuanced understanding of the process of integration of migrants by re-examining Ager and Strang's (2008) framework from an informational perspective told by the intermediaries who work with these populations. Findings from the content analysis revealed that the process of integration is shaped by the inability of both migrants and the institutions in the host society to close the cultural and social gaps that ultimately result in information marginalization. Two main themes were revealed in the content analysis: information marginalization and information resistance. The first theme contained two main categories that reflected the elements of marginalization and how it is reflected in Ager and Strang's (2008) means and markers. Although the framework related to cultural and language knowledge as facilitators of integration, participants in this study perceived the lack of these indicators as obstacles that precluded migrants from getting their basic needs covered, reinforcing their marginalization. Echoing this finding, Martzoukou and Burnett (2018) observed that "the direct relationship between social inclusion and knowledge of the socioeconomic and community information-related constructs," and found "that there were fundamental gaps in Syrian Scots knowledge domains in relation to everyday life matters, which would be considered as 'routine' knowledge in mainstream Scottish society and beyond" (p. 1126). Just as participants in this study related the migrants' lack of understanding of admission procedures at emergency rooms, Martzoukou and Burnett (2018) described an instance in which a participant in their study missed a doctor's appointment when she could not understand how to ride the bus. The second element of information marginalization revealed in the content analysis is the lack of language proficiency that represents a salient theme in the literature on migration (Bronstein, 2019; Ikafa, 2015; Leith, 2009; Lloyd et al., 2013; Martzoukou and Burnett, 2018; Wessendorf and Phillimore, 2019). Grzymala-Kazlowska (2018) described language as an anchor and "a foothold that allows individuals to acquire socio-psychological stability and security and function effectively in a new or substantially changed life" (p. 235). Findings from this study reinforced this assertion and showed that language proficiency was directly related to the ability of migrants to solve everyday problems and meet basic needs, as well as to develop trust, make sense of their reality and be receptive to outsiders and to the information they provide. Information resistance occurred when information in their native language was perceived as the only information that made sense and could be trusted (Chatman, 1996). Two other elements of information marginalization were: living in an unsafe and unstable environment, and discrimination. These elements directly relate to the migrants' ability to understand their new reality and lessen the resistance they experience as a safeguard against a hostile environment bringing migrants "to engage in information-related practices, such as information sharing, which would enable them to understand how the various instrumental environments (health, education, employment) operate" (Lloyd et al., 2017) and allowing people to safely engage with services and with other people to integrate (Ndofor-Tah et al., 2019). Supporting Ager and Strang's (2008) assertion that to feel at home, migrants need to feel safe, participants in this study related how the fear of deportation and discrimination precluded them from seeking information and how that often negatively affected their well-being. Past studies have noted the significant role that a safe and stable environment plays in the process of integration (Ager and Strang, 2008; Ikafa, 2015; Saggar et al., 2012) and represented a crucial foothold and point of reference which enabled migrants to adapt to their new life and feel settled (Grzymala-Kazlowska, 2018). Findings showed that accessing and understanding information related to Ager and Strang's (2008) markers and means is directly related to the migrants' ability to bridge the gaps in their reality, make sense of their surroundings and advance in their integration into Israeli society (Baird, 2012; Ikafa, 2015). That is, the sociocultural gap that prevents migrants from understanding how the education and the healthcare systems work, and how to find and retain proper employment directly contributes to their marginality and does not allow them "to move beyond the liminal zones of society and become included" (Lloyd et al., 2017). The second theme revealed by content analysis refers to the information resistance experienced by migrants when they resist information that comes from outsiders and renders them unable or unwilling to bridge the cultural gap that will help them make sense of their reality. This resistance is oftentimes compounded by inability of the institutions in the host country to help them. This theme contains two categories: secrecy and disinformation. Echoing Chatman's (1996) concept of secrecy as a safeguarding behavior, participants described how migrants, "cease to be receptive to advice and information" (p. 195) and resisted the information they needed in an attempt to protect themselves from an environment they perceived as unsafe. Participants related how resistance to information perceived as not trustworthy (i.e. coming from outsiders) negatively impacted migrants' lives and resulted in disinformation that ultimately hindered their integration process. The third theme revealed in the study presents the findings related to the migrants' and the intermediaries' attempts to overcome information resistance. This theme contains two categories: situational relevance and social connections. The intermediaries participating in the study described different strategies they used to reach migrants and present information in formats that made sense to them. Supporting Chatman's (1996) assertion that "things that make sense are relevant" (p. 202), participants tried to make information relevant to migrants by creating videos to explain the kindergarten registration process or by having trusted gatekeepers share information about training programs or community services. That is, participants understood the importance of legitimizing the information through contextual others (Chatman, 1996). The legitimation of information by contextualization is related to the creation of social connections. The significant role that social connections play in the integration process of migrants is a salient feature in the literature (Bronstein, 2019; Cheung and Phillimore, 2013; Gilmartin and Migge, 2015; Grzymala-Kazlowska, 2018; OECD, 2018). Two of Ager and Strang's (2008) social connections were revealed in the content analysis: social bonds as close social relations and social linking as the relationships built with the intermediaries. Echoing past studies findings showed that creating social bonds within their community helped migrants resolve problems, feel more secure and stable (Grzymala-Kazlowska, 2018) and reinforce their settlement intentions (Torunczyk-Ruiz and Brunarska, 2018). Interestingly, the two new indicators that were added in the 2019 UK Home Office indicators (Ndofor-Tah et al., 2019), leisure and digital skills did not appeared in the data analysis. Further research should investigate the role of these two elements in the integration process of migrants. Phillimore (2012) wrote, "Integration implies the development of a sense of belonging in the host community, with some renegotiation of identity by both newcomers and hosts" (p. 3). Findings from the study showed that it is in this oftentimes failed renegotiation that information marginalization emerges. By allowing intermediaries to articulate their views and opinions, I was able to understand how personal, cultural and social factors impacted the lives of migrants, and distanced them from the sources of information and support they need to feel at home in their new country. Findings showed that for integration to be successful, it should be the result of an effort by both migrants and local institutions/structures to reach a middle ground of understanding and compromise. This study extending Gibson and Martin's (2019) notion of information marginalization to encompass both sides of the equation of marginalization and situating it in the social, economic and contextual conditions that create information poverty. By examining Ager and Strang's (2008) framework of integration from an information perspective, this study highlighted the role that information plays as a facilitating or an impeding element in the integration process, as well as a vehicle to help migrants make sense of their reality, regain social capital, perceive their environment as safe and develop a sense of belonging to their new home.
|
The study presents a different perspective of the role of information in the integration process of migrants by examining the views and opinions of intermediaries working with these populations. Also, the study reframes existing notions of information marginalization and resistance by addressing both sides of the cultural and social gap embodies marginalization.
|
[SECTION: Purpose] I see they've been killing cats again [...] "Better bring mine in, then. Sighs, clear tea-cups after social, go home.This case study explores older people's experience of ageing in deprived neighbourhoods. Most cannot afford to move and may not wish to: "This is my home - it's where my heart is - it's been a lifetime".Prior to policy reviews and between 2002 and 2007, Age Concern Liverpool (ACL)[1] commissioned from Liverpool John Moores University (LJMU) consultations with 602 residents aged 50 years or older, in five of Liverpool's poorest wards. Other contributors included: Liverpool City Council (LCC) and Liverpool Primary Care Trust (PCT)[2], three housing associations and 15 voluntary sector bodies. Participation in health and social activities, opportunity and exclusion, needs and aspirations were discussed with the older people and key informants. Consultation reflected the active ageing policy of the World Health Organisation (WHO) and local and national initiatives to facilitate older people's autonomy, independence and participation in communities (WHO, 2002; JRF, 2004; Smith and Gay, 2005). The "general" national and supranational contexts of ageing will be linked to the "particular" (Zeilig, 2011, p. 8), "lived experiences" (Powell, 2006, p. 65) of older people living in deprivation in an advanced economy (Walker, 2009; Dorling, 2011; Zaidi, 2011). The study adds to a growing body of literature relating to deprivation and barriers to active ageing.The article starts with an introduction to supranational and UK active ageing policy. We then focus on the particularities of the study areas as sites of multiple deprivations. The account of study methods is followed by analysis of active ageing in the study localities, focussing on participation in social and health activities. We conclude with a summary of the policy implications of the research findings. Demographic change has stimulated reviews of concepts of ageing, with "active" ageing emerging as an important focus (WHO, 2002; Walker, 2010; DESA, 2011). Activity in older age may be limited by ageism, poverty in its many guises including ill-health (Howse et al., 2011) and policy and resource constraints which inadequately support older people's welfare (Dean, 2012; Lymbery, 2012). Disparities between poverty and wealth prompt examination of social justice and fairness (Harvey, 1973; Rawls, 2001; Barry, 2005; Dorling, 2011). A minority of older people are wealthy; others rely solely on the state retirement pension; others have occupational pensions and some savings, but these may be eroding relative to living costs (HC, 2009, 2012). Internationally and nationally, the elderly are likely to be among societies' poorest (Price, 2006; Walker, 2009; McKee, 2010; DWP, 2012; DESA, 2011).Phillipson (1998) explores changing ideology and social policy relating to ageing post-1945 with ageing's demands becoming individualised in a postmodern setting and generational interdependence being questioned. The complexities of work-retirement transitions, age discrimination and the changing roles of older people in work and society are meriting critical attention (Phillipson, 2004c, a; Macnicol, 2005). Coleman et al. (1993) illustrate negative calculations of age-related costs and changing elder support social networks, including family. Factors converge: global recession; neo-liberal policies of out-sourcing welfare to the private sector and voluntarism; and post-modern discourses of choice and consumerism (Walker, 2005; Coote, 2009). In different cultural settings, older people may find such socio-political change difficult (Moffatt et al., 2011; Miles, 2009).Concepts of ageing Global population change (Europa, 2009; WHO, 2008; ONS, 2012) challenges existing disciplinary concepts (Kohli, 1988; Walker and Naegele, 1999; Marcuse et al., 2009; Buffel et al., 2012; Plouffe and Kalache, 2010) and stimulates new study areas. Dissonance between traditional urban environments and the needs of vulnerable people, including elders (Antonietti and Garrett, 2012) have stimulated proposals for "age-friendly" urban environments (Phillipson, 2004b; WHO, 2007; Cinderby, 2010; ILC, 2011). Demographic change prompts re-examination of age and geography (Walker, 2010); population profiles, worker/dependent ratios (WHO, 2002; Kihara, 2009; ONS, 2012); and intergenerational relationships (Europa, 2009).Overviews of "ageing" (Bowling et al., 2005; Deeming, 2009; Stenner et al., 2011) illustrate conceptual strength and weaknesses. Boudiny (2012, p. 6) rejects "middle-aged perspectives". Societal constructions of age, including the imposition of limitations such as age-related retirement, are explored by Townsend (1989), Torres and Hammarstrom (2009) and Walker (2006a, b). Deeming (2009, p. 93) commends a conceptual and policy focus on "healthy life expectancy", while recognising the relationship between health and social class.Kalachea and Kickbusch (1997) contributed importantly to a focus on "active" ageing (WHO, 2002) as a positive experience combining "physical, social and mental well-being" with active involvement in "social, economic, cultural, spiritual and civic affairs". But "active" ageing is not without criticism (Bowling, 2008; Stenner et al., 2011; Estes et al., 2010). Concerns include the exclusionary potential of interpretations of "active" ageing relating to physical activity and continuing employment, activities specifically rejected in early conceptual formulations (WHO, 2002).Active ageing and supranational policies During deepening recession, active ageing policy directions, from the United Nations (UN), WHO, the Organisation for Economic Co-operation and Development (OECD) and the European Union (EU), differ most clearly with regard to employment. Both the EU (DGCOMM, 2012) and WHO (2011) reject ageing as a burden, commenting upon older people's potentially greater contributions to society if they are allowed to remain active, recognition of age's social construction. Policy and attitudinal changes are essential if older people are to continue to contribute, however (WHO, 2011). The Europe 2020 agenda focuses strongly on sustaining activity and health in order to benefit productivity and competitiveness, limit early retirement and reduce healthcare costs (CEC, 2011). OECD policy includes encouragement of older people's employment (OECD, 2004). Walker (2010, p. 586), however, considers that such policies could undermine older people's status outside the workforce, encourage ageism, elder dependency and social and economic distancing. Eurobarometer's exploration of solidarity between generations, including employment patterns (Europa, 2009), indicates national and generational differences. The mix of altruism and generational self-interest suggests policy-making complexity for the new demography.Ageing and UK policy There are contrasts and similarities between Labour policy on ageing (2002-2012) and Coalition policy (2010-2012) and between constituent UK countries. Work by the Centre for Policy on Ageing (CPA, 2012), Acts of Parliament, legislation in preparation, other official policy documents, and voluntary and campaigning bodies' reports inform this summary.Most UK social policy is decided at Westminster, and devolved powers to Wales, Scotland and Northern Ireland are constrained (Birrell, 2009). Nevertheless, divergencies exist, as in Scotland's approach to care costs (Scott and Mooney, 2009). A Welsh Assembly Government innovation in the UK was to appoint a Commissioner for Older People in Wales to focus on developing coherent age-related strategy (WAG, 2010). Throughout the UK, there are continuities in inequalities, past policy and public sector financial constraint, but devolution is expected to deliver change (Age Concern Northern Ireland, 2007). Further, divergence from Westminster may mean reduced emphasis on neo-liberal consumerism, possibly resulting in greater equity of outcomes (Moffatt et al., 2011; Mooney and Scott, 2005; Mooney and Wright, 2009; Scottish Parliament, 2012).UK policy on ageing refers, for example, to "Well-being" (DH, 2005), "Well" (DWP, 2010), "Healthy" (DH, 2010c), but specific reference to "active" ageing' is rare which, given the international focus on active ageing, appears surprising. An instance of reference to "active ageing" occurs within Labour's Opportunity Age (DWP, 2005), which proposes extending working lives, supporting "active ageing" in the community, extending choice and independence, and combating ageism.Health and adult social care feature strongly in Labour and Coalition policy between 2002 and 2012. Labour policies included proposals for health (DH, 2005); independence (HMG, 2008; DH, 2005); independence and choice, with individual care budgets (HMG, 2008). Potentially, these measures could have produced some equity in care and health, especially if accompanied by two measures: the implementation of the Marmot Review proposals (Marmot, 2010) for a "fair society" with reduction of socially influenced health inequity; and a national care and support system. Both were proposed by Labour (HMG, 2010). The Coalition's election in 2010, however, brought increased emphasis on "choice", "liberating" the NHS (DH, 2010a, b) through the Health and Social Care Act (HMG, 2012). Policy directions within the Act include privatisation, devolved powers and budgets to a range of providers, possibly in association with "Big Society" voluntary contributions (CLG, 2010a), all of which may be at odds with aims of cohesion in health and social care. Both are key areas in sustaining activity in later life. A major concern is the coincidence of recession with promotion of a neo-liberal, consumer-led agenda at the expense of society's most vulnerable, including older people (Walker, 2005).In a recent Treasury Committee, the National Association of Pension Funds and Saga, an organisation specialising in products and services for older people, criticised "loose monetary policy", arguing that "quantitative easing" (QE) was ineffective and particularly disadvantaged poor people, including some elders (HC, 2012, pp. 26-27). Representations in the Scottish Parliament (2012) made similar claims. Monetary policy, however, is likely to support "even further" QE (IFS, 2012). Moreover, pensions and benefits have been aligned with the Consumer Price Index (CPI), rather than the Retail Price Index, a policy described by the Royal Statistical Society (RSS) as "flawed" and unfair to poorer people since CPI fails to reflect their spending patterns and rising costs (Inman, 2012). Additionally, interest rates below inflation (Wilson, 2012) could deflate pension incomes for a lifetime (HC, 2012). AgeUK (2012a, b) lists further issues: treatment of age-related allowances; the position of supporting people (a programme directed to the most vulnerable) following cuts in local government budgets; delays in implementing the Equality Act (2012); and delays, also, in implementing the proposals for social care payment (Dilnot, 2011) which emerged during the labour government (HMG, 2009). Policy divergence has appeared within the Coalition. Most recently, it appears that further, major welfare cuts may be in preparation (Carrell and Watt, 2012), but the Liberal Democrats debate taxation of assets to support welfare (Watt, 2012). Overall, it is questionable how far current social policy directions, in England particularly, will support age equity, with active and healthy ageing. Most participants were primarily dependent on the state retirement pension. Additionally, ageing in poor neighbourhoods confers multiple disadvantages (Scharf et al., 2003) and the study areas are among the most deprived in England (CLG, 2010b; LCC, 2004-2011). During the study period, 2002-2007, Anfield, Clubmoor, Everton, Speke-Garston[3] and Tuebrook-Stoneycroft were among the poorest wards in Liverpool on measures including household income and income deprivation affecting older people, both pointers to other deprivations. Despite improvement since 2003, significant areas in wards remain within the poorest 1-5 per cent in England (LCC, 2011): most (89 per cent) of Everton is within the poorest 1 per cent.High levels of poverty and deprivation persist in Liverpool, the once "proud second city of empire" (Belchem, 2006, p. 9). Table I illustrates the ranking of major cities in England on the Index of Multiple Deprivation (IMD) (LCC, 2011). Between 2007 and 2010, Liverpool remained the most deprived major city in England, in depth (concentration) and geographical extent of deprivation.Liverpool's overall weak IMD ranking suggests that low ranking areas within the city are considerably deprived (Table II). Socio-economic polarity within Liverpool is also evident. Among the study wards and Church, a comparator ward, Everton ranks first (worst) in Liverpool on most measures. Within other study wards, such as Tuebrook-Stoneycroft, there are small areas of relative affluence but deprivation and ill-health are more evident; both have been associated with poverty (Howse et al., 2011). Years of male life expectancy range from Anfield (72.9) to more affluent Church (83.8), with Liverpool at 78.8 years and England, 82.0 years. Worklessness, rates of benefit claim, low incomes and weak educational performance suggest current and potential future deprivations. Between 2002 and 2007, ACL commissioned studies in five deprived wards to inform policy-making. Liverpool PCT, part of the NHS, contributed informants. The work focussed on older people's needs, aspirations and barriers to active ageing.Participants and informants Most participants lived in the study wards and were aged fifty years or over. Additionally, 32 Black and Minority Ethnic (BME) older people in inner Liverpool discussed residential choice and racism. 38 key informants were selected for their experience of policy, practice and older people in the study areas. They included ACL and NHS staff; LCC officers and members; staff from housing associations, community and voluntary sectors including legal and financial advisory services.Methodology: data collection Triangulation and other cross-referencing developed with informant discussion preceding, parallel to, and following, different forms of contact with participants. Inclusion was important, so advertising in public places in and around the study areas preceded fieldwork. A mixed method approach to data collection (Tashakkori and Teddlie, 2003; Bryman, 2006) included fixed data requirements and opportunities for flexibility depending on circumstances (Cresswell, 2003).Questionnaires with closed and open-ended responses were used in 304 street interviews with samples of older people stratified by the most recent ward demographic data (ONS, 2002; LCC, 2004-2007). This research stage usefully captured the views of older people who might not have been encountered in social or health related settings (Emmel et al., 2007). Qualitative work followed in focus groups and individual discussion with a further 252 participants. A mixed method approach to data collection (Tashakkori and Teddlie, 2003; Bryman, 2006) had several advantages. Fixed data requirements allowed quantifiable data analyses based on samples of older people stratified by the most recent ward demographic data (ONS, 2002; LCC, 2004-2007). Qualitative work produced opportunities for flexibility depending on circumstances (Cresswell, 2003) nuanced responses.In each ward, 5-6 focus groups, typically of 6-7 older people, were held with participants' and gatekeepers' consent. To be non-intrusive, discussions "piggy-backed" on pre-arranged events. ACL arranged most contacts within sheltered accommodation; LJMU established groups in clubs, public houses, health centres, advice offices and community centres. All groups were open to older people, but some identified with particular racial or religious groups, a factor in the research strategy. The research focus and outcomes, consent forms and details of LJMU's secure data handling arrangements were fully explained to participants and feedback was provided if participants wished. Literacy and eye-sight could be problems, so the essential preparatory work could take time. Quantitative data was processed through SPSS. Qualitative information was recorded if groups gave permission and if they did not, researchers accompanying the rapporteurs took notes. Data were analysed using grounded theory (Glaser and Strauss, 1967). Key points were extracted from the text and using the constant comparative method grouped into thematic categories (Strauss and Corbin, 1998). A researcher training programme was part of the research strategy which was approved by LJMU's Ethics Committee. There are other factors, but lots of exclusion (from social and health activities) results from low incomes [...] though people may be too proud to admit it [...] income maximisation among older people is not only empowering, it can be a life saver. It means less anxiety and depression and more [...] fresh food, keeping warm, getting out [...] (Key informant).This observation was illustrated by comments such as: "keep-fit sessions cost a quid!" (widow aged 60); and "you've to spend PS25 now on a week's shop to get free delivery!" (elderly couple). The problem here was not cash flow, but long-standing forced economy. The following sections explore participants' understanding of "active ageing", patterns in their health and social activities and the implications of the findings for social policy.Active ageing: meanings "Active ageing" discussion arose naturally in focus groups since ACL had developed an Active Age Centre and Active Age health sessions with the PCT. "Active" was approved, but "ageing" and "age" met with resistance. Active ageing could be: "about the gym", but word association reflected more of the WHO's Active Age Framework:[...] getting around when you can, keeping up with news, walking, coming here (community centre), church, bingo, word puzzles, helping family, (planting) daffs - and not sitting around [...] feeling sorry for yourself - unless you're poorly.The comment also implies the role of physical and mental resilience in preserving activity in age. The 50s and 60s age groups were most likely to associate employment with "active" ageing and some would have liked to work: "It's bad enough being short of cash, but when you get to 60, you've no chance, ever, of earning more. It's awful". But there was ambivalence. Some work had been physically demanding and could no longer be attempted. Also, employment would have to be "worth it financially" and ideally not full-time "because of family".Street interviews produced the following patterns in social and health activity, and constraints in pursuing them (Table III).The results indicated generally constrained lives. Two-thirds of participants in street interviews were aged 50-64 years; others were aged 65 or older, so this is not an elderly sample. Yet only 58.6 per cent engaged in social activity outside home; 53.6 per cent in health activity. "Health activity" often meant walking from home to shops, relatives' homes and so on. There was rarely one constraint on participation in social and health activity.Ill-health and disability were the clearest barriers to social and health activities, with one in ten of participants housebound, or virtually so, without the "lifeline" of transport from ACL or family.Replies indicated lack of security, in going out, especially alone - in daytime as well as after dark, age and ageism. Care roles and lack of time, commonly in association, account for a fifth of replies. Perceived distance, from health or social provisions which participants found acceptable, also excluded. There was variation in activity by gender (women were generally the more socially active), disability, health status, social class and areas' geographies. The following sections present themes from qualitative discussions which contributed nuance to street interview responses.Ward differences in social and health activity North Liverpool street interviews indicated variation in health activity between 41 per cent in Anfield, an area with a history of manual labour and poor life expectancy, and 71 per cent in Tuebrook-Stoneycroft, with its pockets of relative affluence and "suitable" provisions in terms of activities and opening hours. Hilly Everton, worst in Liverpool on nearly every deprivation indicator, including poor health, recorded an initially surprising response 66 per cent participation in health activity. The most-cited health activity was walking, however, and as a 70-year old commented:[...] it's an aerobic bit of country - there's a lousy set of bus services - so if you can't afford a cab, you've to walk to do anything, even if it kills you!Social and health activity levels differed also between the two south Liverpool wards, Speke and Garston. Speke's social activity response (41 per cent) lagged considerably between adjacent Garston (76 per cent) which includes some middle class areas. Voluntary groups organised community activity in Speke, but local pubs could be "expensive" and "more for young ones" or "for men". Health activity in Speke (59 per cent) was boosted by Active Age sessions organised by ACL and PCT. Garston participants also made good use of the centre, getting lifts or using their own vehicles and engaging in a wider range of other activities, such as bowls and walking. Experienced community workers in Garston reflected upon widening differences between older people, based partly on age and on social class. Health activities such as yoga and relaxation were becoming more popular than the "old keep-fit". Bingo "in draughty halls" was becoming an "elderly" interest. "Consumer conscious", more affluent young/old people who were still earning, "may prefer EasyJet to a bus trip". Disengagement was a recurrent theme in Speke, as observed in a Joseph Rowntree Foundation study (JRF, 2010). Informants referred to difficulty in "getting anything going" and the incidence of depression. Participants generally avoided the "run-down centre - and gangs - did you ever see anything like it?" Informants and participants commented on Speke housing's close proximity to Liverpool John Lennon Airport, with "a lot of noise and pollution, which is not healthy". Estate refurbishment was "slow" and "rough types" had moved in. Speke was considered "miles from Liverpool - for hospital visits and good shops".Activity in volunteering and care Only 4 per cent of participants "volunteered" but this is an understatement. "Volunteering" included structured help set up by an external agency, such as ACL. More commonly, it was informal, neighbourly response to need, which could be spontaneous or longer-term, as in the case of an intrepid 80-year old widow whose home was a neighbourhood resource for help with homework and bureaucracy. Some participants would have liked a "little job" to provide "that bit extra" but many were actively involved in family care, for older and younger relatives. One asked if "Age Concern could provide childcare!".Commenting on the active, "keystone" role of many older people, an informant stated:Older people's social capital can be under-rated. To replace their informal labour [...] would cost Governments a fortune. But offer an older person a job and the family can be worse off; childcare has to be paid for [...] Get a young person a job and the older person can't work. Whole families can become poor - and what tips the balance is often childcare.Another informant looked at affordable care options and encouraged older people to work at least part-time:It's what quite a few would like, so it's good for morale [...] the impact of more income on poor families is huge, financially and it's a model of work (for younger family) where perhaps none exists [...] it's a potentially long-term lift from poverty.Ageing activity and personal biographies Most participants had retired from unskilled or semi-skilled work. Others were former professionals, highly skilled workers, or small business owners and generally these were the participants with the knowledge and confidence to discuss the research. It was apparent that better incomes during the working life and some occupational pension entitlement could enhance active ageing. Computers and newspapers might be afforded. A wider range of social and health activity could be managed. A cab could be taken occasionally to activities outside the home and to larger stores selling fresh food. A home in a "safer area, where there are decent families and fewer gangs" could be afforded.Transport, mobility and outdoor security "The bus pass is the best car I ever had" was a comment supported by several participants, but transport factors could exclude. Only 14.3 per cent of participants owned a vehicle, all were men and all owned their homes, possible indicators of gender and social class influencing mobility. Home location in relation to appropriate social and health opportunities, travel distances and routes which public transport routes which participants felt were affordable, physically manageable and safe, were critical factors in participation. Transport availability and security influenced a preference for daytime activities because: buses "finish practically at 6.0 pm, you can't wait ages at bus stops - you need a lift to be safe".Patterns of safety and insecurity emerged. Fear of attack by gangs even influenced choice of bus stop and route, with areas known locally for antisocial behaviour being avoided. Women expressed more vulnerability, but several men admitted being "wary" when certain clubs and public houses were open: "Now, you're no physical match for a younger fellow - or fellows". A focus group discussed security strategies including complex home exits: locking the front door, waiting, then checking the back of the house to confront potential housebreakers. A woman participant drew her indirect route to a community centre:[...] older people can be seen down the long straight streets here [...] I turn along x to my friend's house, turn near my doctor's, then hurry to the main road [...] I'm safe then.Fear of burglary could curtail activities outside the home, especially after dark.Some participants felt their contemporaries were over-cautious, but informants confirmed that fears were not groundless. Handbags had been snatched on pension day, so some older people clubbed up for a cab to return from a post office when Christmas pension bonuses were paid and CSOs assisted them. Crime rates per thousand population in the five wards, could range from 66.1 to 87.8 against a Liverpool average of 62.5 (LCC, 2007).Poor health and disability could also constrain outside activity. Some participants were housebound while others, of similar age and self-assessed state of health, went to social and learning events. Small income differences appeared important: the occasional cab to "get out and about" could be afforded; family and friends could give lifts. Three participants with mobility problems got lifts to learn computing, which could "help keep in touch", as one said matter-of-factly, "for when I really can't get out". Excel learning was to budget or help a grandson's homework; another participant was designing parish newsletters. Media and community events raised awareness of healthy diets but fresh foods were not readily available in deprived areas. Bus transport to adjacent areas could involve changing buses which was difficult, especially when carrying shopping.Social activities: gendered spaces[4] Social activity outside home was highly gendered. Men's socialising was primarily in pubs and clubs, including those connected with football and former workplaces. Women went somewhat less often to pubs and clubs, especially in certain areas, but church and hobbies (including bingo) were prevalent. Women appeared to have developed wider community links during childhood, child rearing and helping neighbours. Women also did most shopping, which could be the main "social" activity outside the home.Women focus group participants in a pleasant club referred to "rougher" pubs, where "ladies were not so welcome" [...] "especially if you're older". Local pubs could be well-defended, but not conspicuously more than in many cities. Pubs and clubs could be associated, however, with sport and major industries with predominantly male workforces, some now closed. "Male territory" was one woman's description. Several men referred to "getting out of the house", almost to a re-creation of work, with pub and club presenting opportunities to catch up with former work-mates "have a drink and doms - no bingo chatter!" Women generally felt easier about pubs with more mixed clientele, perhaps involving families.Whether in church or community settings, lunch clubs were popular with over half the participants. Older people are among the poorest and loneliest in these areas, and lunch clubs were usually subsidised by benefactors. Organisers confirmed that, for many older people, the two-course warm lunch could be the week's only full meal. Women's better life expectancy often meant solo attendance at events and lunch clubs were considered "respectable". Coming alone to a lunch club, with a formula in social interaction (lunch often followed by bingo), seemed preferable to clubs where interaction could be less predictable. Numbers of older women preferred not to go to evening events, except for something very special, and certainly not alone. Fear of crime was a deterrent, especially without a lift; also evening events could be "dressier", expensive, and attended primarily by couples.Social activities: other divided spaces Further exclusions related to religion and race. One informant commented: "some lovely people, poor, have little, but can be bitter about perceived difference - it's horrible". Religious discrimination was a factor in creating "no go"/"go warily" zones of social and health opportunity. Within the past three centuries, migrants left deprivation in Ireland, Scotland and Wales for further deprivation, particularly in north Liverpool. Sectarian graffiti is still evident and certain pubs and clubs suggest "Catholic" or "Protestant" traditions, gauged from names, insignia, colour schemes and football affiliations.During church-sponsored events, a participant could discreetly mention religious affiliation, for example: "I know (venue) [...] is Catholic/Protestant [...] I am Protestant/Catholic - but don't let on - I've come with (friend)". Asked what would happen if the rapporteur did "let on", the response was: "(other attendees at the event) wouldn't half give me the cold shoulder - and (friend) for bringing me - I'd feel awful for her". "Would the (organiser) behave the same way?" "No - she's younger - they don't care as much". Interestingly, a Salvation Army lunch centre seemed neutral territory where, as one Catholic 80-year old observed: "the Captain (does us) proud every week". Several key informants observed that, not exclusively among older people, sectarian allegiances are still apparent, culturally embedded, though not always accompanied by church-going, as in the past. Some transmission of these allegiances to the young was also in evidence.Racial discrimination was a further factor in exclusion, from participation and place of residence. The study wards are predominantly white British with a range of 4.5 per cent Black or Minority Ethnic (BME) residents in Anfield to 5.8 per cent in Tuebrook and Stoneycroft. BME participants contributed 5.7 per cent of the commentary. Some were professionals, some business owners, others were retired or unemployed. Their social activity appeared predominantly to be family-focused and some venues, more particularly in North Liverpool, were regarded as "no go" on grounds of racism and otherwise unsuitable for family. Businesses and residences may have been within a study area, but participation in social activity outside the home more typically took place in racially mixed areas. Focus groups narrated BME residential movement from deprived "white" Liverpool areas to more affluent suburbs or diverse inner Liverpool areas.Health activities: factors in participation and exclusion Health activity was associated with reasonable health and the availability of "suitable" venues which were affordable and offered personal and emotional security. Some women participants would be "going solo" and "feeling comfortable" was important. The most popular forms of health activity were walking and cycling which as one participant explained "come cheap". Two up-to-date health centres were described as "expensive" and "too Lycra". Participants could fear exposure to the youthful gaze and sessions for older people would have been appreciated: "you feel awkward"; "past it" (too "fat"; "old and wrinkly", "can't move [...] quickly"); the inhibitions of age. Affordability and security were also factors in choice of outdoor sites for health activity. Tuebrook's green spaces were considered accessible and easy to walk around, with "visible" space popular with older people and families. Tourist itineraries include parkland in north Liverpool with fine vistas of Liverpool, but older people shunned it because unobserved areas could invite crime. Satisfaction with "home" was important to well-being and active ageing. It could be the springboard to outside engagement, a place from which to contact the outside world through reading, telephones, computers in some cases, and somewhere to welcome family and friends. The home could be the repository of memory, an essential feature of personal security and identity, indeed as one participant commented, "my home is me".Home could signify loneliness, however, especially if mobility and health were difficult. It could be insecure within, if coping were problematic, and outside in troubled neighbourhoods. Across the wards, between 12 and 18 per cent of participants said they felt insecure in their homes, mainly through fear of burglary. Visits from family and friends were "cheaper than going out" but lack of maintenance, cleaning and adequate heating could constrain the reception of visitors. Home help services enabled many to remain in their homes, but there was fear of elder abuse and reluctance to allow help from strangers.Reaction to sheltered accommodation varied, depending mainly upon the previous home's characteristics, health, and whether a partner was alive. Loss of independence could sadden but for others, security and less "worry" compensated. The worries included burglary, "I can open my window now for fresh air"; organising budgets and home maintenance. Some older residents felt less lonely and happier. Others were ambivalent, liking the greater security, but disliking being observed, within others' structures of time, place and communal living. Some participants were currently homeless, primarily because of poverty and poor health. Homelessness was commonly preceded by intense difficulty, in relationships, mental and physical health and finance, although housing associations appeared generally more supportive than private landlords. "(In)active ageing" relates first to the contrast between narratives of older person as burden in relation to our observations of the pivotal role of many older people in family survival, working part-time, providing family care, or both. Some participants joked about their "inactivity", then gave accounts of selective disengagement, from post-employment "busy-ness" and "day-time television", to re-engagement in following interests: pigeons, local history and gardening were three examples. The title also recognises the barriers to active ageing, significantly built from lifetime inequity, in families and neighbourhoods.Lie et al. (2009) comment that older people's volunteering cannot substitute for funded, sustained and coherent care systems. In the UK, the Government's Big Society project (2010) has been criticised on a number of grounds: as rhetoric and cover for spending cuts (Corbett and Walker, 2012); contributory to the dismantling of the welfare state (Bone, 2011); and illustrative of short-term-ism (Ware, 2012). The "Big Society" is a wholly inadequate response to the needs of the deprived neighbourhoods we have studied. Here, the two major barriers to active ageing are long-term ill-health and disability underpinned by poverty. As Table I has indicated, Liverpool's deprivation is severe and persistent, so measures to effect improvement need to be well-supported and long-term. * The setting of a national minimum income for active and healthy living (O'Sullivan and Ashton, 2011).* Ensuring that incomes at least met this standard.* Emplementing the Marmot proposals (2010) for a "fair society" with reductions in health inequity from the beginning of life to the end.* Restoring the retirement pension - Retail Price Index link.* Establishing a National Care Service to address one of the major needs of older people and their carers.* Establishing a clear and effective focus within government for oversight and development of policy affecting older people.* Ensure that urban and rural areas are "friendly" to older people (ILC, 2011), which also means that they would be friendly to all.* Providing core funding for a range of opportunities in neighbourhoods for active participation by older people (Deeming, 2009).Lymbery (2012) points out that, especially in recession, an adequate governmental response to older people's needs is unlikely. We share that view. The conditions in the study areas are evidence of lack of social justice and the position is likely to deteriorate with further cuts in spending. Many informants were angered at the likely prospect of deterioration of already inadequate provisions in the wards, which is a fair response. In disseminating knowledge of Liverpool's situation, we hope to contribute to a growing body of knowledge and challenge about ageing in places characterised by these levels of poverty and deprivation. Opens in a new window.Table I IMD 2007 and 2010: ranking of Liverpool and seven major English citiesa Opens in a new window.Table II Liverpool ward study: deprivation in context Opens in a new window.Table III Levels of participation in social and health activities
|
- Ageing populations are national and global phenomena. These older residents are likely to be among the most disadvantaged nationally and in comparison with younger neighbours. The benefits of active ageing are attracting attention from policy makers globally, as it is increasingly recognised that age-friendly cities encourage active ageing. Resources to sustain active ageing are becoming scarce. Older people's health, social activity, needs, aspirations and the barriers to realising them are at the centre of this investigation. The purpose of this paper is to explore inclusion and exclusion within some of England's most deprived areas.
|
[SECTION: Method] I see they've been killing cats again [...] "Better bring mine in, then. Sighs, clear tea-cups after social, go home.This case study explores older people's experience of ageing in deprived neighbourhoods. Most cannot afford to move and may not wish to: "This is my home - it's where my heart is - it's been a lifetime".Prior to policy reviews and between 2002 and 2007, Age Concern Liverpool (ACL)[1] commissioned from Liverpool John Moores University (LJMU) consultations with 602 residents aged 50 years or older, in five of Liverpool's poorest wards. Other contributors included: Liverpool City Council (LCC) and Liverpool Primary Care Trust (PCT)[2], three housing associations and 15 voluntary sector bodies. Participation in health and social activities, opportunity and exclusion, needs and aspirations were discussed with the older people and key informants. Consultation reflected the active ageing policy of the World Health Organisation (WHO) and local and national initiatives to facilitate older people's autonomy, independence and participation in communities (WHO, 2002; JRF, 2004; Smith and Gay, 2005). The "general" national and supranational contexts of ageing will be linked to the "particular" (Zeilig, 2011, p. 8), "lived experiences" (Powell, 2006, p. 65) of older people living in deprivation in an advanced economy (Walker, 2009; Dorling, 2011; Zaidi, 2011). The study adds to a growing body of literature relating to deprivation and barriers to active ageing.The article starts with an introduction to supranational and UK active ageing policy. We then focus on the particularities of the study areas as sites of multiple deprivations. The account of study methods is followed by analysis of active ageing in the study localities, focussing on participation in social and health activities. We conclude with a summary of the policy implications of the research findings. Demographic change has stimulated reviews of concepts of ageing, with "active" ageing emerging as an important focus (WHO, 2002; Walker, 2010; DESA, 2011). Activity in older age may be limited by ageism, poverty in its many guises including ill-health (Howse et al., 2011) and policy and resource constraints which inadequately support older people's welfare (Dean, 2012; Lymbery, 2012). Disparities between poverty and wealth prompt examination of social justice and fairness (Harvey, 1973; Rawls, 2001; Barry, 2005; Dorling, 2011). A minority of older people are wealthy; others rely solely on the state retirement pension; others have occupational pensions and some savings, but these may be eroding relative to living costs (HC, 2009, 2012). Internationally and nationally, the elderly are likely to be among societies' poorest (Price, 2006; Walker, 2009; McKee, 2010; DWP, 2012; DESA, 2011).Phillipson (1998) explores changing ideology and social policy relating to ageing post-1945 with ageing's demands becoming individualised in a postmodern setting and generational interdependence being questioned. The complexities of work-retirement transitions, age discrimination and the changing roles of older people in work and society are meriting critical attention (Phillipson, 2004c, a; Macnicol, 2005). Coleman et al. (1993) illustrate negative calculations of age-related costs and changing elder support social networks, including family. Factors converge: global recession; neo-liberal policies of out-sourcing welfare to the private sector and voluntarism; and post-modern discourses of choice and consumerism (Walker, 2005; Coote, 2009). In different cultural settings, older people may find such socio-political change difficult (Moffatt et al., 2011; Miles, 2009).Concepts of ageing Global population change (Europa, 2009; WHO, 2008; ONS, 2012) challenges existing disciplinary concepts (Kohli, 1988; Walker and Naegele, 1999; Marcuse et al., 2009; Buffel et al., 2012; Plouffe and Kalache, 2010) and stimulates new study areas. Dissonance between traditional urban environments and the needs of vulnerable people, including elders (Antonietti and Garrett, 2012) have stimulated proposals for "age-friendly" urban environments (Phillipson, 2004b; WHO, 2007; Cinderby, 2010; ILC, 2011). Demographic change prompts re-examination of age and geography (Walker, 2010); population profiles, worker/dependent ratios (WHO, 2002; Kihara, 2009; ONS, 2012); and intergenerational relationships (Europa, 2009).Overviews of "ageing" (Bowling et al., 2005; Deeming, 2009; Stenner et al., 2011) illustrate conceptual strength and weaknesses. Boudiny (2012, p. 6) rejects "middle-aged perspectives". Societal constructions of age, including the imposition of limitations such as age-related retirement, are explored by Townsend (1989), Torres and Hammarstrom (2009) and Walker (2006a, b). Deeming (2009, p. 93) commends a conceptual and policy focus on "healthy life expectancy", while recognising the relationship between health and social class.Kalachea and Kickbusch (1997) contributed importantly to a focus on "active" ageing (WHO, 2002) as a positive experience combining "physical, social and mental well-being" with active involvement in "social, economic, cultural, spiritual and civic affairs". But "active" ageing is not without criticism (Bowling, 2008; Stenner et al., 2011; Estes et al., 2010). Concerns include the exclusionary potential of interpretations of "active" ageing relating to physical activity and continuing employment, activities specifically rejected in early conceptual formulations (WHO, 2002).Active ageing and supranational policies During deepening recession, active ageing policy directions, from the United Nations (UN), WHO, the Organisation for Economic Co-operation and Development (OECD) and the European Union (EU), differ most clearly with regard to employment. Both the EU (DGCOMM, 2012) and WHO (2011) reject ageing as a burden, commenting upon older people's potentially greater contributions to society if they are allowed to remain active, recognition of age's social construction. Policy and attitudinal changes are essential if older people are to continue to contribute, however (WHO, 2011). The Europe 2020 agenda focuses strongly on sustaining activity and health in order to benefit productivity and competitiveness, limit early retirement and reduce healthcare costs (CEC, 2011). OECD policy includes encouragement of older people's employment (OECD, 2004). Walker (2010, p. 586), however, considers that such policies could undermine older people's status outside the workforce, encourage ageism, elder dependency and social and economic distancing. Eurobarometer's exploration of solidarity between generations, including employment patterns (Europa, 2009), indicates national and generational differences. The mix of altruism and generational self-interest suggests policy-making complexity for the new demography.Ageing and UK policy There are contrasts and similarities between Labour policy on ageing (2002-2012) and Coalition policy (2010-2012) and between constituent UK countries. Work by the Centre for Policy on Ageing (CPA, 2012), Acts of Parliament, legislation in preparation, other official policy documents, and voluntary and campaigning bodies' reports inform this summary.Most UK social policy is decided at Westminster, and devolved powers to Wales, Scotland and Northern Ireland are constrained (Birrell, 2009). Nevertheless, divergencies exist, as in Scotland's approach to care costs (Scott and Mooney, 2009). A Welsh Assembly Government innovation in the UK was to appoint a Commissioner for Older People in Wales to focus on developing coherent age-related strategy (WAG, 2010). Throughout the UK, there are continuities in inequalities, past policy and public sector financial constraint, but devolution is expected to deliver change (Age Concern Northern Ireland, 2007). Further, divergence from Westminster may mean reduced emphasis on neo-liberal consumerism, possibly resulting in greater equity of outcomes (Moffatt et al., 2011; Mooney and Scott, 2005; Mooney and Wright, 2009; Scottish Parliament, 2012).UK policy on ageing refers, for example, to "Well-being" (DH, 2005), "Well" (DWP, 2010), "Healthy" (DH, 2010c), but specific reference to "active" ageing' is rare which, given the international focus on active ageing, appears surprising. An instance of reference to "active ageing" occurs within Labour's Opportunity Age (DWP, 2005), which proposes extending working lives, supporting "active ageing" in the community, extending choice and independence, and combating ageism.Health and adult social care feature strongly in Labour and Coalition policy between 2002 and 2012. Labour policies included proposals for health (DH, 2005); independence (HMG, 2008; DH, 2005); independence and choice, with individual care budgets (HMG, 2008). Potentially, these measures could have produced some equity in care and health, especially if accompanied by two measures: the implementation of the Marmot Review proposals (Marmot, 2010) for a "fair society" with reduction of socially influenced health inequity; and a national care and support system. Both were proposed by Labour (HMG, 2010). The Coalition's election in 2010, however, brought increased emphasis on "choice", "liberating" the NHS (DH, 2010a, b) through the Health and Social Care Act (HMG, 2012). Policy directions within the Act include privatisation, devolved powers and budgets to a range of providers, possibly in association with "Big Society" voluntary contributions (CLG, 2010a), all of which may be at odds with aims of cohesion in health and social care. Both are key areas in sustaining activity in later life. A major concern is the coincidence of recession with promotion of a neo-liberal, consumer-led agenda at the expense of society's most vulnerable, including older people (Walker, 2005).In a recent Treasury Committee, the National Association of Pension Funds and Saga, an organisation specialising in products and services for older people, criticised "loose monetary policy", arguing that "quantitative easing" (QE) was ineffective and particularly disadvantaged poor people, including some elders (HC, 2012, pp. 26-27). Representations in the Scottish Parliament (2012) made similar claims. Monetary policy, however, is likely to support "even further" QE (IFS, 2012). Moreover, pensions and benefits have been aligned with the Consumer Price Index (CPI), rather than the Retail Price Index, a policy described by the Royal Statistical Society (RSS) as "flawed" and unfair to poorer people since CPI fails to reflect their spending patterns and rising costs (Inman, 2012). Additionally, interest rates below inflation (Wilson, 2012) could deflate pension incomes for a lifetime (HC, 2012). AgeUK (2012a, b) lists further issues: treatment of age-related allowances; the position of supporting people (a programme directed to the most vulnerable) following cuts in local government budgets; delays in implementing the Equality Act (2012); and delays, also, in implementing the proposals for social care payment (Dilnot, 2011) which emerged during the labour government (HMG, 2009). Policy divergence has appeared within the Coalition. Most recently, it appears that further, major welfare cuts may be in preparation (Carrell and Watt, 2012), but the Liberal Democrats debate taxation of assets to support welfare (Watt, 2012). Overall, it is questionable how far current social policy directions, in England particularly, will support age equity, with active and healthy ageing. Most participants were primarily dependent on the state retirement pension. Additionally, ageing in poor neighbourhoods confers multiple disadvantages (Scharf et al., 2003) and the study areas are among the most deprived in England (CLG, 2010b; LCC, 2004-2011). During the study period, 2002-2007, Anfield, Clubmoor, Everton, Speke-Garston[3] and Tuebrook-Stoneycroft were among the poorest wards in Liverpool on measures including household income and income deprivation affecting older people, both pointers to other deprivations. Despite improvement since 2003, significant areas in wards remain within the poorest 1-5 per cent in England (LCC, 2011): most (89 per cent) of Everton is within the poorest 1 per cent.High levels of poverty and deprivation persist in Liverpool, the once "proud second city of empire" (Belchem, 2006, p. 9). Table I illustrates the ranking of major cities in England on the Index of Multiple Deprivation (IMD) (LCC, 2011). Between 2007 and 2010, Liverpool remained the most deprived major city in England, in depth (concentration) and geographical extent of deprivation.Liverpool's overall weak IMD ranking suggests that low ranking areas within the city are considerably deprived (Table II). Socio-economic polarity within Liverpool is also evident. Among the study wards and Church, a comparator ward, Everton ranks first (worst) in Liverpool on most measures. Within other study wards, such as Tuebrook-Stoneycroft, there are small areas of relative affluence but deprivation and ill-health are more evident; both have been associated with poverty (Howse et al., 2011). Years of male life expectancy range from Anfield (72.9) to more affluent Church (83.8), with Liverpool at 78.8 years and England, 82.0 years. Worklessness, rates of benefit claim, low incomes and weak educational performance suggest current and potential future deprivations. Between 2002 and 2007, ACL commissioned studies in five deprived wards to inform policy-making. Liverpool PCT, part of the NHS, contributed informants. The work focussed on older people's needs, aspirations and barriers to active ageing.Participants and informants Most participants lived in the study wards and were aged fifty years or over. Additionally, 32 Black and Minority Ethnic (BME) older people in inner Liverpool discussed residential choice and racism. 38 key informants were selected for their experience of policy, practice and older people in the study areas. They included ACL and NHS staff; LCC officers and members; staff from housing associations, community and voluntary sectors including legal and financial advisory services.Methodology: data collection Triangulation and other cross-referencing developed with informant discussion preceding, parallel to, and following, different forms of contact with participants. Inclusion was important, so advertising in public places in and around the study areas preceded fieldwork. A mixed method approach to data collection (Tashakkori and Teddlie, 2003; Bryman, 2006) included fixed data requirements and opportunities for flexibility depending on circumstances (Cresswell, 2003).Questionnaires with closed and open-ended responses were used in 304 street interviews with samples of older people stratified by the most recent ward demographic data (ONS, 2002; LCC, 2004-2007). This research stage usefully captured the views of older people who might not have been encountered in social or health related settings (Emmel et al., 2007). Qualitative work followed in focus groups and individual discussion with a further 252 participants. A mixed method approach to data collection (Tashakkori and Teddlie, 2003; Bryman, 2006) had several advantages. Fixed data requirements allowed quantifiable data analyses based on samples of older people stratified by the most recent ward demographic data (ONS, 2002; LCC, 2004-2007). Qualitative work produced opportunities for flexibility depending on circumstances (Cresswell, 2003) nuanced responses.In each ward, 5-6 focus groups, typically of 6-7 older people, were held with participants' and gatekeepers' consent. To be non-intrusive, discussions "piggy-backed" on pre-arranged events. ACL arranged most contacts within sheltered accommodation; LJMU established groups in clubs, public houses, health centres, advice offices and community centres. All groups were open to older people, but some identified with particular racial or religious groups, a factor in the research strategy. The research focus and outcomes, consent forms and details of LJMU's secure data handling arrangements were fully explained to participants and feedback was provided if participants wished. Literacy and eye-sight could be problems, so the essential preparatory work could take time. Quantitative data was processed through SPSS. Qualitative information was recorded if groups gave permission and if they did not, researchers accompanying the rapporteurs took notes. Data were analysed using grounded theory (Glaser and Strauss, 1967). Key points were extracted from the text and using the constant comparative method grouped into thematic categories (Strauss and Corbin, 1998). A researcher training programme was part of the research strategy which was approved by LJMU's Ethics Committee. There are other factors, but lots of exclusion (from social and health activities) results from low incomes [...] though people may be too proud to admit it [...] income maximisation among older people is not only empowering, it can be a life saver. It means less anxiety and depression and more [...] fresh food, keeping warm, getting out [...] (Key informant).This observation was illustrated by comments such as: "keep-fit sessions cost a quid!" (widow aged 60); and "you've to spend PS25 now on a week's shop to get free delivery!" (elderly couple). The problem here was not cash flow, but long-standing forced economy. The following sections explore participants' understanding of "active ageing", patterns in their health and social activities and the implications of the findings for social policy.Active ageing: meanings "Active ageing" discussion arose naturally in focus groups since ACL had developed an Active Age Centre and Active Age health sessions with the PCT. "Active" was approved, but "ageing" and "age" met with resistance. Active ageing could be: "about the gym", but word association reflected more of the WHO's Active Age Framework:[...] getting around when you can, keeping up with news, walking, coming here (community centre), church, bingo, word puzzles, helping family, (planting) daffs - and not sitting around [...] feeling sorry for yourself - unless you're poorly.The comment also implies the role of physical and mental resilience in preserving activity in age. The 50s and 60s age groups were most likely to associate employment with "active" ageing and some would have liked to work: "It's bad enough being short of cash, but when you get to 60, you've no chance, ever, of earning more. It's awful". But there was ambivalence. Some work had been physically demanding and could no longer be attempted. Also, employment would have to be "worth it financially" and ideally not full-time "because of family".Street interviews produced the following patterns in social and health activity, and constraints in pursuing them (Table III).The results indicated generally constrained lives. Two-thirds of participants in street interviews were aged 50-64 years; others were aged 65 or older, so this is not an elderly sample. Yet only 58.6 per cent engaged in social activity outside home; 53.6 per cent in health activity. "Health activity" often meant walking from home to shops, relatives' homes and so on. There was rarely one constraint on participation in social and health activity.Ill-health and disability were the clearest barriers to social and health activities, with one in ten of participants housebound, or virtually so, without the "lifeline" of transport from ACL or family.Replies indicated lack of security, in going out, especially alone - in daytime as well as after dark, age and ageism. Care roles and lack of time, commonly in association, account for a fifth of replies. Perceived distance, from health or social provisions which participants found acceptable, also excluded. There was variation in activity by gender (women were generally the more socially active), disability, health status, social class and areas' geographies. The following sections present themes from qualitative discussions which contributed nuance to street interview responses.Ward differences in social and health activity North Liverpool street interviews indicated variation in health activity between 41 per cent in Anfield, an area with a history of manual labour and poor life expectancy, and 71 per cent in Tuebrook-Stoneycroft, with its pockets of relative affluence and "suitable" provisions in terms of activities and opening hours. Hilly Everton, worst in Liverpool on nearly every deprivation indicator, including poor health, recorded an initially surprising response 66 per cent participation in health activity. The most-cited health activity was walking, however, and as a 70-year old commented:[...] it's an aerobic bit of country - there's a lousy set of bus services - so if you can't afford a cab, you've to walk to do anything, even if it kills you!Social and health activity levels differed also between the two south Liverpool wards, Speke and Garston. Speke's social activity response (41 per cent) lagged considerably between adjacent Garston (76 per cent) which includes some middle class areas. Voluntary groups organised community activity in Speke, but local pubs could be "expensive" and "more for young ones" or "for men". Health activity in Speke (59 per cent) was boosted by Active Age sessions organised by ACL and PCT. Garston participants also made good use of the centre, getting lifts or using their own vehicles and engaging in a wider range of other activities, such as bowls and walking. Experienced community workers in Garston reflected upon widening differences between older people, based partly on age and on social class. Health activities such as yoga and relaxation were becoming more popular than the "old keep-fit". Bingo "in draughty halls" was becoming an "elderly" interest. "Consumer conscious", more affluent young/old people who were still earning, "may prefer EasyJet to a bus trip". Disengagement was a recurrent theme in Speke, as observed in a Joseph Rowntree Foundation study (JRF, 2010). Informants referred to difficulty in "getting anything going" and the incidence of depression. Participants generally avoided the "run-down centre - and gangs - did you ever see anything like it?" Informants and participants commented on Speke housing's close proximity to Liverpool John Lennon Airport, with "a lot of noise and pollution, which is not healthy". Estate refurbishment was "slow" and "rough types" had moved in. Speke was considered "miles from Liverpool - for hospital visits and good shops".Activity in volunteering and care Only 4 per cent of participants "volunteered" but this is an understatement. "Volunteering" included structured help set up by an external agency, such as ACL. More commonly, it was informal, neighbourly response to need, which could be spontaneous or longer-term, as in the case of an intrepid 80-year old widow whose home was a neighbourhood resource for help with homework and bureaucracy. Some participants would have liked a "little job" to provide "that bit extra" but many were actively involved in family care, for older and younger relatives. One asked if "Age Concern could provide childcare!".Commenting on the active, "keystone" role of many older people, an informant stated:Older people's social capital can be under-rated. To replace their informal labour [...] would cost Governments a fortune. But offer an older person a job and the family can be worse off; childcare has to be paid for [...] Get a young person a job and the older person can't work. Whole families can become poor - and what tips the balance is often childcare.Another informant looked at affordable care options and encouraged older people to work at least part-time:It's what quite a few would like, so it's good for morale [...] the impact of more income on poor families is huge, financially and it's a model of work (for younger family) where perhaps none exists [...] it's a potentially long-term lift from poverty.Ageing activity and personal biographies Most participants had retired from unskilled or semi-skilled work. Others were former professionals, highly skilled workers, or small business owners and generally these were the participants with the knowledge and confidence to discuss the research. It was apparent that better incomes during the working life and some occupational pension entitlement could enhance active ageing. Computers and newspapers might be afforded. A wider range of social and health activity could be managed. A cab could be taken occasionally to activities outside the home and to larger stores selling fresh food. A home in a "safer area, where there are decent families and fewer gangs" could be afforded.Transport, mobility and outdoor security "The bus pass is the best car I ever had" was a comment supported by several participants, but transport factors could exclude. Only 14.3 per cent of participants owned a vehicle, all were men and all owned their homes, possible indicators of gender and social class influencing mobility. Home location in relation to appropriate social and health opportunities, travel distances and routes which public transport routes which participants felt were affordable, physically manageable and safe, were critical factors in participation. Transport availability and security influenced a preference for daytime activities because: buses "finish practically at 6.0 pm, you can't wait ages at bus stops - you need a lift to be safe".Patterns of safety and insecurity emerged. Fear of attack by gangs even influenced choice of bus stop and route, with areas known locally for antisocial behaviour being avoided. Women expressed more vulnerability, but several men admitted being "wary" when certain clubs and public houses were open: "Now, you're no physical match for a younger fellow - or fellows". A focus group discussed security strategies including complex home exits: locking the front door, waiting, then checking the back of the house to confront potential housebreakers. A woman participant drew her indirect route to a community centre:[...] older people can be seen down the long straight streets here [...] I turn along x to my friend's house, turn near my doctor's, then hurry to the main road [...] I'm safe then.Fear of burglary could curtail activities outside the home, especially after dark.Some participants felt their contemporaries were over-cautious, but informants confirmed that fears were not groundless. Handbags had been snatched on pension day, so some older people clubbed up for a cab to return from a post office when Christmas pension bonuses were paid and CSOs assisted them. Crime rates per thousand population in the five wards, could range from 66.1 to 87.8 against a Liverpool average of 62.5 (LCC, 2007).Poor health and disability could also constrain outside activity. Some participants were housebound while others, of similar age and self-assessed state of health, went to social and learning events. Small income differences appeared important: the occasional cab to "get out and about" could be afforded; family and friends could give lifts. Three participants with mobility problems got lifts to learn computing, which could "help keep in touch", as one said matter-of-factly, "for when I really can't get out". Excel learning was to budget or help a grandson's homework; another participant was designing parish newsletters. Media and community events raised awareness of healthy diets but fresh foods were not readily available in deprived areas. Bus transport to adjacent areas could involve changing buses which was difficult, especially when carrying shopping.Social activities: gendered spaces[4] Social activity outside home was highly gendered. Men's socialising was primarily in pubs and clubs, including those connected with football and former workplaces. Women went somewhat less often to pubs and clubs, especially in certain areas, but church and hobbies (including bingo) were prevalent. Women appeared to have developed wider community links during childhood, child rearing and helping neighbours. Women also did most shopping, which could be the main "social" activity outside the home.Women focus group participants in a pleasant club referred to "rougher" pubs, where "ladies were not so welcome" [...] "especially if you're older". Local pubs could be well-defended, but not conspicuously more than in many cities. Pubs and clubs could be associated, however, with sport and major industries with predominantly male workforces, some now closed. "Male territory" was one woman's description. Several men referred to "getting out of the house", almost to a re-creation of work, with pub and club presenting opportunities to catch up with former work-mates "have a drink and doms - no bingo chatter!" Women generally felt easier about pubs with more mixed clientele, perhaps involving families.Whether in church or community settings, lunch clubs were popular with over half the participants. Older people are among the poorest and loneliest in these areas, and lunch clubs were usually subsidised by benefactors. Organisers confirmed that, for many older people, the two-course warm lunch could be the week's only full meal. Women's better life expectancy often meant solo attendance at events and lunch clubs were considered "respectable". Coming alone to a lunch club, with a formula in social interaction (lunch often followed by bingo), seemed preferable to clubs where interaction could be less predictable. Numbers of older women preferred not to go to evening events, except for something very special, and certainly not alone. Fear of crime was a deterrent, especially without a lift; also evening events could be "dressier", expensive, and attended primarily by couples.Social activities: other divided spaces Further exclusions related to religion and race. One informant commented: "some lovely people, poor, have little, but can be bitter about perceived difference - it's horrible". Religious discrimination was a factor in creating "no go"/"go warily" zones of social and health opportunity. Within the past three centuries, migrants left deprivation in Ireland, Scotland and Wales for further deprivation, particularly in north Liverpool. Sectarian graffiti is still evident and certain pubs and clubs suggest "Catholic" or "Protestant" traditions, gauged from names, insignia, colour schemes and football affiliations.During church-sponsored events, a participant could discreetly mention religious affiliation, for example: "I know (venue) [...] is Catholic/Protestant [...] I am Protestant/Catholic - but don't let on - I've come with (friend)". Asked what would happen if the rapporteur did "let on", the response was: "(other attendees at the event) wouldn't half give me the cold shoulder - and (friend) for bringing me - I'd feel awful for her". "Would the (organiser) behave the same way?" "No - she's younger - they don't care as much". Interestingly, a Salvation Army lunch centre seemed neutral territory where, as one Catholic 80-year old observed: "the Captain (does us) proud every week". Several key informants observed that, not exclusively among older people, sectarian allegiances are still apparent, culturally embedded, though not always accompanied by church-going, as in the past. Some transmission of these allegiances to the young was also in evidence.Racial discrimination was a further factor in exclusion, from participation and place of residence. The study wards are predominantly white British with a range of 4.5 per cent Black or Minority Ethnic (BME) residents in Anfield to 5.8 per cent in Tuebrook and Stoneycroft. BME participants contributed 5.7 per cent of the commentary. Some were professionals, some business owners, others were retired or unemployed. Their social activity appeared predominantly to be family-focused and some venues, more particularly in North Liverpool, were regarded as "no go" on grounds of racism and otherwise unsuitable for family. Businesses and residences may have been within a study area, but participation in social activity outside the home more typically took place in racially mixed areas. Focus groups narrated BME residential movement from deprived "white" Liverpool areas to more affluent suburbs or diverse inner Liverpool areas.Health activities: factors in participation and exclusion Health activity was associated with reasonable health and the availability of "suitable" venues which were affordable and offered personal and emotional security. Some women participants would be "going solo" and "feeling comfortable" was important. The most popular forms of health activity were walking and cycling which as one participant explained "come cheap". Two up-to-date health centres were described as "expensive" and "too Lycra". Participants could fear exposure to the youthful gaze and sessions for older people would have been appreciated: "you feel awkward"; "past it" (too "fat"; "old and wrinkly", "can't move [...] quickly"); the inhibitions of age. Affordability and security were also factors in choice of outdoor sites for health activity. Tuebrook's green spaces were considered accessible and easy to walk around, with "visible" space popular with older people and families. Tourist itineraries include parkland in north Liverpool with fine vistas of Liverpool, but older people shunned it because unobserved areas could invite crime. Satisfaction with "home" was important to well-being and active ageing. It could be the springboard to outside engagement, a place from which to contact the outside world through reading, telephones, computers in some cases, and somewhere to welcome family and friends. The home could be the repository of memory, an essential feature of personal security and identity, indeed as one participant commented, "my home is me".Home could signify loneliness, however, especially if mobility and health were difficult. It could be insecure within, if coping were problematic, and outside in troubled neighbourhoods. Across the wards, between 12 and 18 per cent of participants said they felt insecure in their homes, mainly through fear of burglary. Visits from family and friends were "cheaper than going out" but lack of maintenance, cleaning and adequate heating could constrain the reception of visitors. Home help services enabled many to remain in their homes, but there was fear of elder abuse and reluctance to allow help from strangers.Reaction to sheltered accommodation varied, depending mainly upon the previous home's characteristics, health, and whether a partner was alive. Loss of independence could sadden but for others, security and less "worry" compensated. The worries included burglary, "I can open my window now for fresh air"; organising budgets and home maintenance. Some older residents felt less lonely and happier. Others were ambivalent, liking the greater security, but disliking being observed, within others' structures of time, place and communal living. Some participants were currently homeless, primarily because of poverty and poor health. Homelessness was commonly preceded by intense difficulty, in relationships, mental and physical health and finance, although housing associations appeared generally more supportive than private landlords. "(In)active ageing" relates first to the contrast between narratives of older person as burden in relation to our observations of the pivotal role of many older people in family survival, working part-time, providing family care, or both. Some participants joked about their "inactivity", then gave accounts of selective disengagement, from post-employment "busy-ness" and "day-time television", to re-engagement in following interests: pigeons, local history and gardening were three examples. The title also recognises the barriers to active ageing, significantly built from lifetime inequity, in families and neighbourhoods.Lie et al. (2009) comment that older people's volunteering cannot substitute for funded, sustained and coherent care systems. In the UK, the Government's Big Society project (2010) has been criticised on a number of grounds: as rhetoric and cover for spending cuts (Corbett and Walker, 2012); contributory to the dismantling of the welfare state (Bone, 2011); and illustrative of short-term-ism (Ware, 2012). The "Big Society" is a wholly inadequate response to the needs of the deprived neighbourhoods we have studied. Here, the two major barriers to active ageing are long-term ill-health and disability underpinned by poverty. As Table I has indicated, Liverpool's deprivation is severe and persistent, so measures to effect improvement need to be well-supported and long-term. * The setting of a national minimum income for active and healthy living (O'Sullivan and Ashton, 2011).* Ensuring that incomes at least met this standard.* Emplementing the Marmot proposals (2010) for a "fair society" with reductions in health inequity from the beginning of life to the end.* Restoring the retirement pension - Retail Price Index link.* Establishing a National Care Service to address one of the major needs of older people and their carers.* Establishing a clear and effective focus within government for oversight and development of policy affecting older people.* Ensure that urban and rural areas are "friendly" to older people (ILC, 2011), which also means that they would be friendly to all.* Providing core funding for a range of opportunities in neighbourhoods for active participation by older people (Deeming, 2009).Lymbery (2012) points out that, especially in recession, an adequate governmental response to older people's needs is unlikely. We share that view. The conditions in the study areas are evidence of lack of social justice and the position is likely to deteriorate with further cuts in spending. Many informants were angered at the likely prospect of deterioration of already inadequate provisions in the wards, which is a fair response. In disseminating knowledge of Liverpool's situation, we hope to contribute to a growing body of knowledge and challenge about ageing in places characterised by these levels of poverty and deprivation. Opens in a new window.Table I IMD 2007 and 2010: ranking of Liverpool and seven major English citiesa Opens in a new window.Table II Liverpool ward study: deprivation in context Opens in a new window.Table III Levels of participation in social and health activities
|
- Between 2002-2007, over 600 older Liverpool people, key informants and policy makers in five of Liverpool's poorest electoral wards were consulted via semi-structured questionnaires, key informant interviews and focus groups about their needs and aspirations.
|
[SECTION: Findings] I see they've been killing cats again [...] "Better bring mine in, then. Sighs, clear tea-cups after social, go home.This case study explores older people's experience of ageing in deprived neighbourhoods. Most cannot afford to move and may not wish to: "This is my home - it's where my heart is - it's been a lifetime".Prior to policy reviews and between 2002 and 2007, Age Concern Liverpool (ACL)[1] commissioned from Liverpool John Moores University (LJMU) consultations with 602 residents aged 50 years or older, in five of Liverpool's poorest wards. Other contributors included: Liverpool City Council (LCC) and Liverpool Primary Care Trust (PCT)[2], three housing associations and 15 voluntary sector bodies. Participation in health and social activities, opportunity and exclusion, needs and aspirations were discussed with the older people and key informants. Consultation reflected the active ageing policy of the World Health Organisation (WHO) and local and national initiatives to facilitate older people's autonomy, independence and participation in communities (WHO, 2002; JRF, 2004; Smith and Gay, 2005). The "general" national and supranational contexts of ageing will be linked to the "particular" (Zeilig, 2011, p. 8), "lived experiences" (Powell, 2006, p. 65) of older people living in deprivation in an advanced economy (Walker, 2009; Dorling, 2011; Zaidi, 2011). The study adds to a growing body of literature relating to deprivation and barriers to active ageing.The article starts with an introduction to supranational and UK active ageing policy. We then focus on the particularities of the study areas as sites of multiple deprivations. The account of study methods is followed by analysis of active ageing in the study localities, focussing on participation in social and health activities. We conclude with a summary of the policy implications of the research findings. Demographic change has stimulated reviews of concepts of ageing, with "active" ageing emerging as an important focus (WHO, 2002; Walker, 2010; DESA, 2011). Activity in older age may be limited by ageism, poverty in its many guises including ill-health (Howse et al., 2011) and policy and resource constraints which inadequately support older people's welfare (Dean, 2012; Lymbery, 2012). Disparities between poverty and wealth prompt examination of social justice and fairness (Harvey, 1973; Rawls, 2001; Barry, 2005; Dorling, 2011). A minority of older people are wealthy; others rely solely on the state retirement pension; others have occupational pensions and some savings, but these may be eroding relative to living costs (HC, 2009, 2012). Internationally and nationally, the elderly are likely to be among societies' poorest (Price, 2006; Walker, 2009; McKee, 2010; DWP, 2012; DESA, 2011).Phillipson (1998) explores changing ideology and social policy relating to ageing post-1945 with ageing's demands becoming individualised in a postmodern setting and generational interdependence being questioned. The complexities of work-retirement transitions, age discrimination and the changing roles of older people in work and society are meriting critical attention (Phillipson, 2004c, a; Macnicol, 2005). Coleman et al. (1993) illustrate negative calculations of age-related costs and changing elder support social networks, including family. Factors converge: global recession; neo-liberal policies of out-sourcing welfare to the private sector and voluntarism; and post-modern discourses of choice and consumerism (Walker, 2005; Coote, 2009). In different cultural settings, older people may find such socio-political change difficult (Moffatt et al., 2011; Miles, 2009).Concepts of ageing Global population change (Europa, 2009; WHO, 2008; ONS, 2012) challenges existing disciplinary concepts (Kohli, 1988; Walker and Naegele, 1999; Marcuse et al., 2009; Buffel et al., 2012; Plouffe and Kalache, 2010) and stimulates new study areas. Dissonance between traditional urban environments and the needs of vulnerable people, including elders (Antonietti and Garrett, 2012) have stimulated proposals for "age-friendly" urban environments (Phillipson, 2004b; WHO, 2007; Cinderby, 2010; ILC, 2011). Demographic change prompts re-examination of age and geography (Walker, 2010); population profiles, worker/dependent ratios (WHO, 2002; Kihara, 2009; ONS, 2012); and intergenerational relationships (Europa, 2009).Overviews of "ageing" (Bowling et al., 2005; Deeming, 2009; Stenner et al., 2011) illustrate conceptual strength and weaknesses. Boudiny (2012, p. 6) rejects "middle-aged perspectives". Societal constructions of age, including the imposition of limitations such as age-related retirement, are explored by Townsend (1989), Torres and Hammarstrom (2009) and Walker (2006a, b). Deeming (2009, p. 93) commends a conceptual and policy focus on "healthy life expectancy", while recognising the relationship between health and social class.Kalachea and Kickbusch (1997) contributed importantly to a focus on "active" ageing (WHO, 2002) as a positive experience combining "physical, social and mental well-being" with active involvement in "social, economic, cultural, spiritual and civic affairs". But "active" ageing is not without criticism (Bowling, 2008; Stenner et al., 2011; Estes et al., 2010). Concerns include the exclusionary potential of interpretations of "active" ageing relating to physical activity and continuing employment, activities specifically rejected in early conceptual formulations (WHO, 2002).Active ageing and supranational policies During deepening recession, active ageing policy directions, from the United Nations (UN), WHO, the Organisation for Economic Co-operation and Development (OECD) and the European Union (EU), differ most clearly with regard to employment. Both the EU (DGCOMM, 2012) and WHO (2011) reject ageing as a burden, commenting upon older people's potentially greater contributions to society if they are allowed to remain active, recognition of age's social construction. Policy and attitudinal changes are essential if older people are to continue to contribute, however (WHO, 2011). The Europe 2020 agenda focuses strongly on sustaining activity and health in order to benefit productivity and competitiveness, limit early retirement and reduce healthcare costs (CEC, 2011). OECD policy includes encouragement of older people's employment (OECD, 2004). Walker (2010, p. 586), however, considers that such policies could undermine older people's status outside the workforce, encourage ageism, elder dependency and social and economic distancing. Eurobarometer's exploration of solidarity between generations, including employment patterns (Europa, 2009), indicates national and generational differences. The mix of altruism and generational self-interest suggests policy-making complexity for the new demography.Ageing and UK policy There are contrasts and similarities between Labour policy on ageing (2002-2012) and Coalition policy (2010-2012) and between constituent UK countries. Work by the Centre for Policy on Ageing (CPA, 2012), Acts of Parliament, legislation in preparation, other official policy documents, and voluntary and campaigning bodies' reports inform this summary.Most UK social policy is decided at Westminster, and devolved powers to Wales, Scotland and Northern Ireland are constrained (Birrell, 2009). Nevertheless, divergencies exist, as in Scotland's approach to care costs (Scott and Mooney, 2009). A Welsh Assembly Government innovation in the UK was to appoint a Commissioner for Older People in Wales to focus on developing coherent age-related strategy (WAG, 2010). Throughout the UK, there are continuities in inequalities, past policy and public sector financial constraint, but devolution is expected to deliver change (Age Concern Northern Ireland, 2007). Further, divergence from Westminster may mean reduced emphasis on neo-liberal consumerism, possibly resulting in greater equity of outcomes (Moffatt et al., 2011; Mooney and Scott, 2005; Mooney and Wright, 2009; Scottish Parliament, 2012).UK policy on ageing refers, for example, to "Well-being" (DH, 2005), "Well" (DWP, 2010), "Healthy" (DH, 2010c), but specific reference to "active" ageing' is rare which, given the international focus on active ageing, appears surprising. An instance of reference to "active ageing" occurs within Labour's Opportunity Age (DWP, 2005), which proposes extending working lives, supporting "active ageing" in the community, extending choice and independence, and combating ageism.Health and adult social care feature strongly in Labour and Coalition policy between 2002 and 2012. Labour policies included proposals for health (DH, 2005); independence (HMG, 2008; DH, 2005); independence and choice, with individual care budgets (HMG, 2008). Potentially, these measures could have produced some equity in care and health, especially if accompanied by two measures: the implementation of the Marmot Review proposals (Marmot, 2010) for a "fair society" with reduction of socially influenced health inequity; and a national care and support system. Both were proposed by Labour (HMG, 2010). The Coalition's election in 2010, however, brought increased emphasis on "choice", "liberating" the NHS (DH, 2010a, b) through the Health and Social Care Act (HMG, 2012). Policy directions within the Act include privatisation, devolved powers and budgets to a range of providers, possibly in association with "Big Society" voluntary contributions (CLG, 2010a), all of which may be at odds with aims of cohesion in health and social care. Both are key areas in sustaining activity in later life. A major concern is the coincidence of recession with promotion of a neo-liberal, consumer-led agenda at the expense of society's most vulnerable, including older people (Walker, 2005).In a recent Treasury Committee, the National Association of Pension Funds and Saga, an organisation specialising in products and services for older people, criticised "loose monetary policy", arguing that "quantitative easing" (QE) was ineffective and particularly disadvantaged poor people, including some elders (HC, 2012, pp. 26-27). Representations in the Scottish Parliament (2012) made similar claims. Monetary policy, however, is likely to support "even further" QE (IFS, 2012). Moreover, pensions and benefits have been aligned with the Consumer Price Index (CPI), rather than the Retail Price Index, a policy described by the Royal Statistical Society (RSS) as "flawed" and unfair to poorer people since CPI fails to reflect their spending patterns and rising costs (Inman, 2012). Additionally, interest rates below inflation (Wilson, 2012) could deflate pension incomes for a lifetime (HC, 2012). AgeUK (2012a, b) lists further issues: treatment of age-related allowances; the position of supporting people (a programme directed to the most vulnerable) following cuts in local government budgets; delays in implementing the Equality Act (2012); and delays, also, in implementing the proposals for social care payment (Dilnot, 2011) which emerged during the labour government (HMG, 2009). Policy divergence has appeared within the Coalition. Most recently, it appears that further, major welfare cuts may be in preparation (Carrell and Watt, 2012), but the Liberal Democrats debate taxation of assets to support welfare (Watt, 2012). Overall, it is questionable how far current social policy directions, in England particularly, will support age equity, with active and healthy ageing. Most participants were primarily dependent on the state retirement pension. Additionally, ageing in poor neighbourhoods confers multiple disadvantages (Scharf et al., 2003) and the study areas are among the most deprived in England (CLG, 2010b; LCC, 2004-2011). During the study period, 2002-2007, Anfield, Clubmoor, Everton, Speke-Garston[3] and Tuebrook-Stoneycroft were among the poorest wards in Liverpool on measures including household income and income deprivation affecting older people, both pointers to other deprivations. Despite improvement since 2003, significant areas in wards remain within the poorest 1-5 per cent in England (LCC, 2011): most (89 per cent) of Everton is within the poorest 1 per cent.High levels of poverty and deprivation persist in Liverpool, the once "proud second city of empire" (Belchem, 2006, p. 9). Table I illustrates the ranking of major cities in England on the Index of Multiple Deprivation (IMD) (LCC, 2011). Between 2007 and 2010, Liverpool remained the most deprived major city in England, in depth (concentration) and geographical extent of deprivation.Liverpool's overall weak IMD ranking suggests that low ranking areas within the city are considerably deprived (Table II). Socio-economic polarity within Liverpool is also evident. Among the study wards and Church, a comparator ward, Everton ranks first (worst) in Liverpool on most measures. Within other study wards, such as Tuebrook-Stoneycroft, there are small areas of relative affluence but deprivation and ill-health are more evident; both have been associated with poverty (Howse et al., 2011). Years of male life expectancy range from Anfield (72.9) to more affluent Church (83.8), with Liverpool at 78.8 years and England, 82.0 years. Worklessness, rates of benefit claim, low incomes and weak educational performance suggest current and potential future deprivations. Between 2002 and 2007, ACL commissioned studies in five deprived wards to inform policy-making. Liverpool PCT, part of the NHS, contributed informants. The work focussed on older people's needs, aspirations and barriers to active ageing.Participants and informants Most participants lived in the study wards and were aged fifty years or over. Additionally, 32 Black and Minority Ethnic (BME) older people in inner Liverpool discussed residential choice and racism. 38 key informants were selected for their experience of policy, practice and older people in the study areas. They included ACL and NHS staff; LCC officers and members; staff from housing associations, community and voluntary sectors including legal and financial advisory services.Methodology: data collection Triangulation and other cross-referencing developed with informant discussion preceding, parallel to, and following, different forms of contact with participants. Inclusion was important, so advertising in public places in and around the study areas preceded fieldwork. A mixed method approach to data collection (Tashakkori and Teddlie, 2003; Bryman, 2006) included fixed data requirements and opportunities for flexibility depending on circumstances (Cresswell, 2003).Questionnaires with closed and open-ended responses were used in 304 street interviews with samples of older people stratified by the most recent ward demographic data (ONS, 2002; LCC, 2004-2007). This research stage usefully captured the views of older people who might not have been encountered in social or health related settings (Emmel et al., 2007). Qualitative work followed in focus groups and individual discussion with a further 252 participants. A mixed method approach to data collection (Tashakkori and Teddlie, 2003; Bryman, 2006) had several advantages. Fixed data requirements allowed quantifiable data analyses based on samples of older people stratified by the most recent ward demographic data (ONS, 2002; LCC, 2004-2007). Qualitative work produced opportunities for flexibility depending on circumstances (Cresswell, 2003) nuanced responses.In each ward, 5-6 focus groups, typically of 6-7 older people, were held with participants' and gatekeepers' consent. To be non-intrusive, discussions "piggy-backed" on pre-arranged events. ACL arranged most contacts within sheltered accommodation; LJMU established groups in clubs, public houses, health centres, advice offices and community centres. All groups were open to older people, but some identified with particular racial or religious groups, a factor in the research strategy. The research focus and outcomes, consent forms and details of LJMU's secure data handling arrangements were fully explained to participants and feedback was provided if participants wished. Literacy and eye-sight could be problems, so the essential preparatory work could take time. Quantitative data was processed through SPSS. Qualitative information was recorded if groups gave permission and if they did not, researchers accompanying the rapporteurs took notes. Data were analysed using grounded theory (Glaser and Strauss, 1967). Key points were extracted from the text and using the constant comparative method grouped into thematic categories (Strauss and Corbin, 1998). A researcher training programme was part of the research strategy which was approved by LJMU's Ethics Committee. There are other factors, but lots of exclusion (from social and health activities) results from low incomes [...] though people may be too proud to admit it [...] income maximisation among older people is not only empowering, it can be a life saver. It means less anxiety and depression and more [...] fresh food, keeping warm, getting out [...] (Key informant).This observation was illustrated by comments such as: "keep-fit sessions cost a quid!" (widow aged 60); and "you've to spend PS25 now on a week's shop to get free delivery!" (elderly couple). The problem here was not cash flow, but long-standing forced economy. The following sections explore participants' understanding of "active ageing", patterns in their health and social activities and the implications of the findings for social policy.Active ageing: meanings "Active ageing" discussion arose naturally in focus groups since ACL had developed an Active Age Centre and Active Age health sessions with the PCT. "Active" was approved, but "ageing" and "age" met with resistance. Active ageing could be: "about the gym", but word association reflected more of the WHO's Active Age Framework:[...] getting around when you can, keeping up with news, walking, coming here (community centre), church, bingo, word puzzles, helping family, (planting) daffs - and not sitting around [...] feeling sorry for yourself - unless you're poorly.The comment also implies the role of physical and mental resilience in preserving activity in age. The 50s and 60s age groups were most likely to associate employment with "active" ageing and some would have liked to work: "It's bad enough being short of cash, but when you get to 60, you've no chance, ever, of earning more. It's awful". But there was ambivalence. Some work had been physically demanding and could no longer be attempted. Also, employment would have to be "worth it financially" and ideally not full-time "because of family".Street interviews produced the following patterns in social and health activity, and constraints in pursuing them (Table III).The results indicated generally constrained lives. Two-thirds of participants in street interviews were aged 50-64 years; others were aged 65 or older, so this is not an elderly sample. Yet only 58.6 per cent engaged in social activity outside home; 53.6 per cent in health activity. "Health activity" often meant walking from home to shops, relatives' homes and so on. There was rarely one constraint on participation in social and health activity.Ill-health and disability were the clearest barriers to social and health activities, with one in ten of participants housebound, or virtually so, without the "lifeline" of transport from ACL or family.Replies indicated lack of security, in going out, especially alone - in daytime as well as after dark, age and ageism. Care roles and lack of time, commonly in association, account for a fifth of replies. Perceived distance, from health or social provisions which participants found acceptable, also excluded. There was variation in activity by gender (women were generally the more socially active), disability, health status, social class and areas' geographies. The following sections present themes from qualitative discussions which contributed nuance to street interview responses.Ward differences in social and health activity North Liverpool street interviews indicated variation in health activity between 41 per cent in Anfield, an area with a history of manual labour and poor life expectancy, and 71 per cent in Tuebrook-Stoneycroft, with its pockets of relative affluence and "suitable" provisions in terms of activities and opening hours. Hilly Everton, worst in Liverpool on nearly every deprivation indicator, including poor health, recorded an initially surprising response 66 per cent participation in health activity. The most-cited health activity was walking, however, and as a 70-year old commented:[...] it's an aerobic bit of country - there's a lousy set of bus services - so if you can't afford a cab, you've to walk to do anything, even if it kills you!Social and health activity levels differed also between the two south Liverpool wards, Speke and Garston. Speke's social activity response (41 per cent) lagged considerably between adjacent Garston (76 per cent) which includes some middle class areas. Voluntary groups organised community activity in Speke, but local pubs could be "expensive" and "more for young ones" or "for men". Health activity in Speke (59 per cent) was boosted by Active Age sessions organised by ACL and PCT. Garston participants also made good use of the centre, getting lifts or using their own vehicles and engaging in a wider range of other activities, such as bowls and walking. Experienced community workers in Garston reflected upon widening differences between older people, based partly on age and on social class. Health activities such as yoga and relaxation were becoming more popular than the "old keep-fit". Bingo "in draughty halls" was becoming an "elderly" interest. "Consumer conscious", more affluent young/old people who were still earning, "may prefer EasyJet to a bus trip". Disengagement was a recurrent theme in Speke, as observed in a Joseph Rowntree Foundation study (JRF, 2010). Informants referred to difficulty in "getting anything going" and the incidence of depression. Participants generally avoided the "run-down centre - and gangs - did you ever see anything like it?" Informants and participants commented on Speke housing's close proximity to Liverpool John Lennon Airport, with "a lot of noise and pollution, which is not healthy". Estate refurbishment was "slow" and "rough types" had moved in. Speke was considered "miles from Liverpool - for hospital visits and good shops".Activity in volunteering and care Only 4 per cent of participants "volunteered" but this is an understatement. "Volunteering" included structured help set up by an external agency, such as ACL. More commonly, it was informal, neighbourly response to need, which could be spontaneous or longer-term, as in the case of an intrepid 80-year old widow whose home was a neighbourhood resource for help with homework and bureaucracy. Some participants would have liked a "little job" to provide "that bit extra" but many were actively involved in family care, for older and younger relatives. One asked if "Age Concern could provide childcare!".Commenting on the active, "keystone" role of many older people, an informant stated:Older people's social capital can be under-rated. To replace their informal labour [...] would cost Governments a fortune. But offer an older person a job and the family can be worse off; childcare has to be paid for [...] Get a young person a job and the older person can't work. Whole families can become poor - and what tips the balance is often childcare.Another informant looked at affordable care options and encouraged older people to work at least part-time:It's what quite a few would like, so it's good for morale [...] the impact of more income on poor families is huge, financially and it's a model of work (for younger family) where perhaps none exists [...] it's a potentially long-term lift from poverty.Ageing activity and personal biographies Most participants had retired from unskilled or semi-skilled work. Others were former professionals, highly skilled workers, or small business owners and generally these were the participants with the knowledge and confidence to discuss the research. It was apparent that better incomes during the working life and some occupational pension entitlement could enhance active ageing. Computers and newspapers might be afforded. A wider range of social and health activity could be managed. A cab could be taken occasionally to activities outside the home and to larger stores selling fresh food. A home in a "safer area, where there are decent families and fewer gangs" could be afforded.Transport, mobility and outdoor security "The bus pass is the best car I ever had" was a comment supported by several participants, but transport factors could exclude. Only 14.3 per cent of participants owned a vehicle, all were men and all owned their homes, possible indicators of gender and social class influencing mobility. Home location in relation to appropriate social and health opportunities, travel distances and routes which public transport routes which participants felt were affordable, physically manageable and safe, were critical factors in participation. Transport availability and security influenced a preference for daytime activities because: buses "finish practically at 6.0 pm, you can't wait ages at bus stops - you need a lift to be safe".Patterns of safety and insecurity emerged. Fear of attack by gangs even influenced choice of bus stop and route, with areas known locally for antisocial behaviour being avoided. Women expressed more vulnerability, but several men admitted being "wary" when certain clubs and public houses were open: "Now, you're no physical match for a younger fellow - or fellows". A focus group discussed security strategies including complex home exits: locking the front door, waiting, then checking the back of the house to confront potential housebreakers. A woman participant drew her indirect route to a community centre:[...] older people can be seen down the long straight streets here [...] I turn along x to my friend's house, turn near my doctor's, then hurry to the main road [...] I'm safe then.Fear of burglary could curtail activities outside the home, especially after dark.Some participants felt their contemporaries were over-cautious, but informants confirmed that fears were not groundless. Handbags had been snatched on pension day, so some older people clubbed up for a cab to return from a post office when Christmas pension bonuses were paid and CSOs assisted them. Crime rates per thousand population in the five wards, could range from 66.1 to 87.8 against a Liverpool average of 62.5 (LCC, 2007).Poor health and disability could also constrain outside activity. Some participants were housebound while others, of similar age and self-assessed state of health, went to social and learning events. Small income differences appeared important: the occasional cab to "get out and about" could be afforded; family and friends could give lifts. Three participants with mobility problems got lifts to learn computing, which could "help keep in touch", as one said matter-of-factly, "for when I really can't get out". Excel learning was to budget or help a grandson's homework; another participant was designing parish newsletters. Media and community events raised awareness of healthy diets but fresh foods were not readily available in deprived areas. Bus transport to adjacent areas could involve changing buses which was difficult, especially when carrying shopping.Social activities: gendered spaces[4] Social activity outside home was highly gendered. Men's socialising was primarily in pubs and clubs, including those connected with football and former workplaces. Women went somewhat less often to pubs and clubs, especially in certain areas, but church and hobbies (including bingo) were prevalent. Women appeared to have developed wider community links during childhood, child rearing and helping neighbours. Women also did most shopping, which could be the main "social" activity outside the home.Women focus group participants in a pleasant club referred to "rougher" pubs, where "ladies were not so welcome" [...] "especially if you're older". Local pubs could be well-defended, but not conspicuously more than in many cities. Pubs and clubs could be associated, however, with sport and major industries with predominantly male workforces, some now closed. "Male territory" was one woman's description. Several men referred to "getting out of the house", almost to a re-creation of work, with pub and club presenting opportunities to catch up with former work-mates "have a drink and doms - no bingo chatter!" Women generally felt easier about pubs with more mixed clientele, perhaps involving families.Whether in church or community settings, lunch clubs were popular with over half the participants. Older people are among the poorest and loneliest in these areas, and lunch clubs were usually subsidised by benefactors. Organisers confirmed that, for many older people, the two-course warm lunch could be the week's only full meal. Women's better life expectancy often meant solo attendance at events and lunch clubs were considered "respectable". Coming alone to a lunch club, with a formula in social interaction (lunch often followed by bingo), seemed preferable to clubs where interaction could be less predictable. Numbers of older women preferred not to go to evening events, except for something very special, and certainly not alone. Fear of crime was a deterrent, especially without a lift; also evening events could be "dressier", expensive, and attended primarily by couples.Social activities: other divided spaces Further exclusions related to religion and race. One informant commented: "some lovely people, poor, have little, but can be bitter about perceived difference - it's horrible". Religious discrimination was a factor in creating "no go"/"go warily" zones of social and health opportunity. Within the past three centuries, migrants left deprivation in Ireland, Scotland and Wales for further deprivation, particularly in north Liverpool. Sectarian graffiti is still evident and certain pubs and clubs suggest "Catholic" or "Protestant" traditions, gauged from names, insignia, colour schemes and football affiliations.During church-sponsored events, a participant could discreetly mention religious affiliation, for example: "I know (venue) [...] is Catholic/Protestant [...] I am Protestant/Catholic - but don't let on - I've come with (friend)". Asked what would happen if the rapporteur did "let on", the response was: "(other attendees at the event) wouldn't half give me the cold shoulder - and (friend) for bringing me - I'd feel awful for her". "Would the (organiser) behave the same way?" "No - she's younger - they don't care as much". Interestingly, a Salvation Army lunch centre seemed neutral territory where, as one Catholic 80-year old observed: "the Captain (does us) proud every week". Several key informants observed that, not exclusively among older people, sectarian allegiances are still apparent, culturally embedded, though not always accompanied by church-going, as in the past. Some transmission of these allegiances to the young was also in evidence.Racial discrimination was a further factor in exclusion, from participation and place of residence. The study wards are predominantly white British with a range of 4.5 per cent Black or Minority Ethnic (BME) residents in Anfield to 5.8 per cent in Tuebrook and Stoneycroft. BME participants contributed 5.7 per cent of the commentary. Some were professionals, some business owners, others were retired or unemployed. Their social activity appeared predominantly to be family-focused and some venues, more particularly in North Liverpool, were regarded as "no go" on grounds of racism and otherwise unsuitable for family. Businesses and residences may have been within a study area, but participation in social activity outside the home more typically took place in racially mixed areas. Focus groups narrated BME residential movement from deprived "white" Liverpool areas to more affluent suburbs or diverse inner Liverpool areas.Health activities: factors in participation and exclusion Health activity was associated with reasonable health and the availability of "suitable" venues which were affordable and offered personal and emotional security. Some women participants would be "going solo" and "feeling comfortable" was important. The most popular forms of health activity were walking and cycling which as one participant explained "come cheap". Two up-to-date health centres were described as "expensive" and "too Lycra". Participants could fear exposure to the youthful gaze and sessions for older people would have been appreciated: "you feel awkward"; "past it" (too "fat"; "old and wrinkly", "can't move [...] quickly"); the inhibitions of age. Affordability and security were also factors in choice of outdoor sites for health activity. Tuebrook's green spaces were considered accessible and easy to walk around, with "visible" space popular with older people and families. Tourist itineraries include parkland in north Liverpool with fine vistas of Liverpool, but older people shunned it because unobserved areas could invite crime. Satisfaction with "home" was important to well-being and active ageing. It could be the springboard to outside engagement, a place from which to contact the outside world through reading, telephones, computers in some cases, and somewhere to welcome family and friends. The home could be the repository of memory, an essential feature of personal security and identity, indeed as one participant commented, "my home is me".Home could signify loneliness, however, especially if mobility and health were difficult. It could be insecure within, if coping were problematic, and outside in troubled neighbourhoods. Across the wards, between 12 and 18 per cent of participants said they felt insecure in their homes, mainly through fear of burglary. Visits from family and friends were "cheaper than going out" but lack of maintenance, cleaning and adequate heating could constrain the reception of visitors. Home help services enabled many to remain in their homes, but there was fear of elder abuse and reluctance to allow help from strangers.Reaction to sheltered accommodation varied, depending mainly upon the previous home's characteristics, health, and whether a partner was alive. Loss of independence could sadden but for others, security and less "worry" compensated. The worries included burglary, "I can open my window now for fresh air"; organising budgets and home maintenance. Some older residents felt less lonely and happier. Others were ambivalent, liking the greater security, but disliking being observed, within others' structures of time, place and communal living. Some participants were currently homeless, primarily because of poverty and poor health. Homelessness was commonly preceded by intense difficulty, in relationships, mental and physical health and finance, although housing associations appeared generally more supportive than private landlords. "(In)active ageing" relates first to the contrast between narratives of older person as burden in relation to our observations of the pivotal role of many older people in family survival, working part-time, providing family care, or both. Some participants joked about their "inactivity", then gave accounts of selective disengagement, from post-employment "busy-ness" and "day-time television", to re-engagement in following interests: pigeons, local history and gardening were three examples. The title also recognises the barriers to active ageing, significantly built from lifetime inequity, in families and neighbourhoods.Lie et al. (2009) comment that older people's volunteering cannot substitute for funded, sustained and coherent care systems. In the UK, the Government's Big Society project (2010) has been criticised on a number of grounds: as rhetoric and cover for spending cuts (Corbett and Walker, 2012); contributory to the dismantling of the welfare state (Bone, 2011); and illustrative of short-term-ism (Ware, 2012). The "Big Society" is a wholly inadequate response to the needs of the deprived neighbourhoods we have studied. Here, the two major barriers to active ageing are long-term ill-health and disability underpinned by poverty. As Table I has indicated, Liverpool's deprivation is severe and persistent, so measures to effect improvement need to be well-supported and long-term. * The setting of a national minimum income for active and healthy living (O'Sullivan and Ashton, 2011).* Ensuring that incomes at least met this standard.* Emplementing the Marmot proposals (2010) for a "fair society" with reductions in health inequity from the beginning of life to the end.* Restoring the retirement pension - Retail Price Index link.* Establishing a National Care Service to address one of the major needs of older people and their carers.* Establishing a clear and effective focus within government for oversight and development of policy affecting older people.* Ensure that urban and rural areas are "friendly" to older people (ILC, 2011), which also means that they would be friendly to all.* Providing core funding for a range of opportunities in neighbourhoods for active participation by older people (Deeming, 2009).Lymbery (2012) points out that, especially in recession, an adequate governmental response to older people's needs is unlikely. We share that view. The conditions in the study areas are evidence of lack of social justice and the position is likely to deteriorate with further cuts in spending. Many informants were angered at the likely prospect of deterioration of already inadequate provisions in the wards, which is a fair response. In disseminating knowledge of Liverpool's situation, we hope to contribute to a growing body of knowledge and challenge about ageing in places characterised by these levels of poverty and deprivation. Opens in a new window.Table I IMD 2007 and 2010: ranking of Liverpool and seven major English citiesa Opens in a new window.Table II Liverpool ward study: deprivation in context Opens in a new window.Table III Levels of participation in social and health activities
|
- Barriers to active ageing arose primarily from participants' poverty, ill-health and deprivation, poor neighbourhoods, ageism, and insecure, gendered, racialised and sectarian space.
|
[SECTION: Value] I see they've been killing cats again [...] "Better bring mine in, then. Sighs, clear tea-cups after social, go home.This case study explores older people's experience of ageing in deprived neighbourhoods. Most cannot afford to move and may not wish to: "This is my home - it's where my heart is - it's been a lifetime".Prior to policy reviews and between 2002 and 2007, Age Concern Liverpool (ACL)[1] commissioned from Liverpool John Moores University (LJMU) consultations with 602 residents aged 50 years or older, in five of Liverpool's poorest wards. Other contributors included: Liverpool City Council (LCC) and Liverpool Primary Care Trust (PCT)[2], three housing associations and 15 voluntary sector bodies. Participation in health and social activities, opportunity and exclusion, needs and aspirations were discussed with the older people and key informants. Consultation reflected the active ageing policy of the World Health Organisation (WHO) and local and national initiatives to facilitate older people's autonomy, independence and participation in communities (WHO, 2002; JRF, 2004; Smith and Gay, 2005). The "general" national and supranational contexts of ageing will be linked to the "particular" (Zeilig, 2011, p. 8), "lived experiences" (Powell, 2006, p. 65) of older people living in deprivation in an advanced economy (Walker, 2009; Dorling, 2011; Zaidi, 2011). The study adds to a growing body of literature relating to deprivation and barriers to active ageing.The article starts with an introduction to supranational and UK active ageing policy. We then focus on the particularities of the study areas as sites of multiple deprivations. The account of study methods is followed by analysis of active ageing in the study localities, focussing on participation in social and health activities. We conclude with a summary of the policy implications of the research findings. Demographic change has stimulated reviews of concepts of ageing, with "active" ageing emerging as an important focus (WHO, 2002; Walker, 2010; DESA, 2011). Activity in older age may be limited by ageism, poverty in its many guises including ill-health (Howse et al., 2011) and policy and resource constraints which inadequately support older people's welfare (Dean, 2012; Lymbery, 2012). Disparities between poverty and wealth prompt examination of social justice and fairness (Harvey, 1973; Rawls, 2001; Barry, 2005; Dorling, 2011). A minority of older people are wealthy; others rely solely on the state retirement pension; others have occupational pensions and some savings, but these may be eroding relative to living costs (HC, 2009, 2012). Internationally and nationally, the elderly are likely to be among societies' poorest (Price, 2006; Walker, 2009; McKee, 2010; DWP, 2012; DESA, 2011).Phillipson (1998) explores changing ideology and social policy relating to ageing post-1945 with ageing's demands becoming individualised in a postmodern setting and generational interdependence being questioned. The complexities of work-retirement transitions, age discrimination and the changing roles of older people in work and society are meriting critical attention (Phillipson, 2004c, a; Macnicol, 2005). Coleman et al. (1993) illustrate negative calculations of age-related costs and changing elder support social networks, including family. Factors converge: global recession; neo-liberal policies of out-sourcing welfare to the private sector and voluntarism; and post-modern discourses of choice and consumerism (Walker, 2005; Coote, 2009). In different cultural settings, older people may find such socio-political change difficult (Moffatt et al., 2011; Miles, 2009).Concepts of ageing Global population change (Europa, 2009; WHO, 2008; ONS, 2012) challenges existing disciplinary concepts (Kohli, 1988; Walker and Naegele, 1999; Marcuse et al., 2009; Buffel et al., 2012; Plouffe and Kalache, 2010) and stimulates new study areas. Dissonance between traditional urban environments and the needs of vulnerable people, including elders (Antonietti and Garrett, 2012) have stimulated proposals for "age-friendly" urban environments (Phillipson, 2004b; WHO, 2007; Cinderby, 2010; ILC, 2011). Demographic change prompts re-examination of age and geography (Walker, 2010); population profiles, worker/dependent ratios (WHO, 2002; Kihara, 2009; ONS, 2012); and intergenerational relationships (Europa, 2009).Overviews of "ageing" (Bowling et al., 2005; Deeming, 2009; Stenner et al., 2011) illustrate conceptual strength and weaknesses. Boudiny (2012, p. 6) rejects "middle-aged perspectives". Societal constructions of age, including the imposition of limitations such as age-related retirement, are explored by Townsend (1989), Torres and Hammarstrom (2009) and Walker (2006a, b). Deeming (2009, p. 93) commends a conceptual and policy focus on "healthy life expectancy", while recognising the relationship between health and social class.Kalachea and Kickbusch (1997) contributed importantly to a focus on "active" ageing (WHO, 2002) as a positive experience combining "physical, social and mental well-being" with active involvement in "social, economic, cultural, spiritual and civic affairs". But "active" ageing is not without criticism (Bowling, 2008; Stenner et al., 2011; Estes et al., 2010). Concerns include the exclusionary potential of interpretations of "active" ageing relating to physical activity and continuing employment, activities specifically rejected in early conceptual formulations (WHO, 2002).Active ageing and supranational policies During deepening recession, active ageing policy directions, from the United Nations (UN), WHO, the Organisation for Economic Co-operation and Development (OECD) and the European Union (EU), differ most clearly with regard to employment. Both the EU (DGCOMM, 2012) and WHO (2011) reject ageing as a burden, commenting upon older people's potentially greater contributions to society if they are allowed to remain active, recognition of age's social construction. Policy and attitudinal changes are essential if older people are to continue to contribute, however (WHO, 2011). The Europe 2020 agenda focuses strongly on sustaining activity and health in order to benefit productivity and competitiveness, limit early retirement and reduce healthcare costs (CEC, 2011). OECD policy includes encouragement of older people's employment (OECD, 2004). Walker (2010, p. 586), however, considers that such policies could undermine older people's status outside the workforce, encourage ageism, elder dependency and social and economic distancing. Eurobarometer's exploration of solidarity between generations, including employment patterns (Europa, 2009), indicates national and generational differences. The mix of altruism and generational self-interest suggests policy-making complexity for the new demography.Ageing and UK policy There are contrasts and similarities between Labour policy on ageing (2002-2012) and Coalition policy (2010-2012) and between constituent UK countries. Work by the Centre for Policy on Ageing (CPA, 2012), Acts of Parliament, legislation in preparation, other official policy documents, and voluntary and campaigning bodies' reports inform this summary.Most UK social policy is decided at Westminster, and devolved powers to Wales, Scotland and Northern Ireland are constrained (Birrell, 2009). Nevertheless, divergencies exist, as in Scotland's approach to care costs (Scott and Mooney, 2009). A Welsh Assembly Government innovation in the UK was to appoint a Commissioner for Older People in Wales to focus on developing coherent age-related strategy (WAG, 2010). Throughout the UK, there are continuities in inequalities, past policy and public sector financial constraint, but devolution is expected to deliver change (Age Concern Northern Ireland, 2007). Further, divergence from Westminster may mean reduced emphasis on neo-liberal consumerism, possibly resulting in greater equity of outcomes (Moffatt et al., 2011; Mooney and Scott, 2005; Mooney and Wright, 2009; Scottish Parliament, 2012).UK policy on ageing refers, for example, to "Well-being" (DH, 2005), "Well" (DWP, 2010), "Healthy" (DH, 2010c), but specific reference to "active" ageing' is rare which, given the international focus on active ageing, appears surprising. An instance of reference to "active ageing" occurs within Labour's Opportunity Age (DWP, 2005), which proposes extending working lives, supporting "active ageing" in the community, extending choice and independence, and combating ageism.Health and adult social care feature strongly in Labour and Coalition policy between 2002 and 2012. Labour policies included proposals for health (DH, 2005); independence (HMG, 2008; DH, 2005); independence and choice, with individual care budgets (HMG, 2008). Potentially, these measures could have produced some equity in care and health, especially if accompanied by two measures: the implementation of the Marmot Review proposals (Marmot, 2010) for a "fair society" with reduction of socially influenced health inequity; and a national care and support system. Both were proposed by Labour (HMG, 2010). The Coalition's election in 2010, however, brought increased emphasis on "choice", "liberating" the NHS (DH, 2010a, b) through the Health and Social Care Act (HMG, 2012). Policy directions within the Act include privatisation, devolved powers and budgets to a range of providers, possibly in association with "Big Society" voluntary contributions (CLG, 2010a), all of which may be at odds with aims of cohesion in health and social care. Both are key areas in sustaining activity in later life. A major concern is the coincidence of recession with promotion of a neo-liberal, consumer-led agenda at the expense of society's most vulnerable, including older people (Walker, 2005).In a recent Treasury Committee, the National Association of Pension Funds and Saga, an organisation specialising in products and services for older people, criticised "loose monetary policy", arguing that "quantitative easing" (QE) was ineffective and particularly disadvantaged poor people, including some elders (HC, 2012, pp. 26-27). Representations in the Scottish Parliament (2012) made similar claims. Monetary policy, however, is likely to support "even further" QE (IFS, 2012). Moreover, pensions and benefits have been aligned with the Consumer Price Index (CPI), rather than the Retail Price Index, a policy described by the Royal Statistical Society (RSS) as "flawed" and unfair to poorer people since CPI fails to reflect their spending patterns and rising costs (Inman, 2012). Additionally, interest rates below inflation (Wilson, 2012) could deflate pension incomes for a lifetime (HC, 2012). AgeUK (2012a, b) lists further issues: treatment of age-related allowances; the position of supporting people (a programme directed to the most vulnerable) following cuts in local government budgets; delays in implementing the Equality Act (2012); and delays, also, in implementing the proposals for social care payment (Dilnot, 2011) which emerged during the labour government (HMG, 2009). Policy divergence has appeared within the Coalition. Most recently, it appears that further, major welfare cuts may be in preparation (Carrell and Watt, 2012), but the Liberal Democrats debate taxation of assets to support welfare (Watt, 2012). Overall, it is questionable how far current social policy directions, in England particularly, will support age equity, with active and healthy ageing. Most participants were primarily dependent on the state retirement pension. Additionally, ageing in poor neighbourhoods confers multiple disadvantages (Scharf et al., 2003) and the study areas are among the most deprived in England (CLG, 2010b; LCC, 2004-2011). During the study period, 2002-2007, Anfield, Clubmoor, Everton, Speke-Garston[3] and Tuebrook-Stoneycroft were among the poorest wards in Liverpool on measures including household income and income deprivation affecting older people, both pointers to other deprivations. Despite improvement since 2003, significant areas in wards remain within the poorest 1-5 per cent in England (LCC, 2011): most (89 per cent) of Everton is within the poorest 1 per cent.High levels of poverty and deprivation persist in Liverpool, the once "proud second city of empire" (Belchem, 2006, p. 9). Table I illustrates the ranking of major cities in England on the Index of Multiple Deprivation (IMD) (LCC, 2011). Between 2007 and 2010, Liverpool remained the most deprived major city in England, in depth (concentration) and geographical extent of deprivation.Liverpool's overall weak IMD ranking suggests that low ranking areas within the city are considerably deprived (Table II). Socio-economic polarity within Liverpool is also evident. Among the study wards and Church, a comparator ward, Everton ranks first (worst) in Liverpool on most measures. Within other study wards, such as Tuebrook-Stoneycroft, there are small areas of relative affluence but deprivation and ill-health are more evident; both have been associated with poverty (Howse et al., 2011). Years of male life expectancy range from Anfield (72.9) to more affluent Church (83.8), with Liverpool at 78.8 years and England, 82.0 years. Worklessness, rates of benefit claim, low incomes and weak educational performance suggest current and potential future deprivations. Between 2002 and 2007, ACL commissioned studies in five deprived wards to inform policy-making. Liverpool PCT, part of the NHS, contributed informants. The work focussed on older people's needs, aspirations and barriers to active ageing.Participants and informants Most participants lived in the study wards and were aged fifty years or over. Additionally, 32 Black and Minority Ethnic (BME) older people in inner Liverpool discussed residential choice and racism. 38 key informants were selected for their experience of policy, practice and older people in the study areas. They included ACL and NHS staff; LCC officers and members; staff from housing associations, community and voluntary sectors including legal and financial advisory services.Methodology: data collection Triangulation and other cross-referencing developed with informant discussion preceding, parallel to, and following, different forms of contact with participants. Inclusion was important, so advertising in public places in and around the study areas preceded fieldwork. A mixed method approach to data collection (Tashakkori and Teddlie, 2003; Bryman, 2006) included fixed data requirements and opportunities for flexibility depending on circumstances (Cresswell, 2003).Questionnaires with closed and open-ended responses were used in 304 street interviews with samples of older people stratified by the most recent ward demographic data (ONS, 2002; LCC, 2004-2007). This research stage usefully captured the views of older people who might not have been encountered in social or health related settings (Emmel et al., 2007). Qualitative work followed in focus groups and individual discussion with a further 252 participants. A mixed method approach to data collection (Tashakkori and Teddlie, 2003; Bryman, 2006) had several advantages. Fixed data requirements allowed quantifiable data analyses based on samples of older people stratified by the most recent ward demographic data (ONS, 2002; LCC, 2004-2007). Qualitative work produced opportunities for flexibility depending on circumstances (Cresswell, 2003) nuanced responses.In each ward, 5-6 focus groups, typically of 6-7 older people, were held with participants' and gatekeepers' consent. To be non-intrusive, discussions "piggy-backed" on pre-arranged events. ACL arranged most contacts within sheltered accommodation; LJMU established groups in clubs, public houses, health centres, advice offices and community centres. All groups were open to older people, but some identified with particular racial or religious groups, a factor in the research strategy. The research focus and outcomes, consent forms and details of LJMU's secure data handling arrangements were fully explained to participants and feedback was provided if participants wished. Literacy and eye-sight could be problems, so the essential preparatory work could take time. Quantitative data was processed through SPSS. Qualitative information was recorded if groups gave permission and if they did not, researchers accompanying the rapporteurs took notes. Data were analysed using grounded theory (Glaser and Strauss, 1967). Key points were extracted from the text and using the constant comparative method grouped into thematic categories (Strauss and Corbin, 1998). A researcher training programme was part of the research strategy which was approved by LJMU's Ethics Committee. There are other factors, but lots of exclusion (from social and health activities) results from low incomes [...] though people may be too proud to admit it [...] income maximisation among older people is not only empowering, it can be a life saver. It means less anxiety and depression and more [...] fresh food, keeping warm, getting out [...] (Key informant).This observation was illustrated by comments such as: "keep-fit sessions cost a quid!" (widow aged 60); and "you've to spend PS25 now on a week's shop to get free delivery!" (elderly couple). The problem here was not cash flow, but long-standing forced economy. The following sections explore participants' understanding of "active ageing", patterns in their health and social activities and the implications of the findings for social policy.Active ageing: meanings "Active ageing" discussion arose naturally in focus groups since ACL had developed an Active Age Centre and Active Age health sessions with the PCT. "Active" was approved, but "ageing" and "age" met with resistance. Active ageing could be: "about the gym", but word association reflected more of the WHO's Active Age Framework:[...] getting around when you can, keeping up with news, walking, coming here (community centre), church, bingo, word puzzles, helping family, (planting) daffs - and not sitting around [...] feeling sorry for yourself - unless you're poorly.The comment also implies the role of physical and mental resilience in preserving activity in age. The 50s and 60s age groups were most likely to associate employment with "active" ageing and some would have liked to work: "It's bad enough being short of cash, but when you get to 60, you've no chance, ever, of earning more. It's awful". But there was ambivalence. Some work had been physically demanding and could no longer be attempted. Also, employment would have to be "worth it financially" and ideally not full-time "because of family".Street interviews produced the following patterns in social and health activity, and constraints in pursuing them (Table III).The results indicated generally constrained lives. Two-thirds of participants in street interviews were aged 50-64 years; others were aged 65 or older, so this is not an elderly sample. Yet only 58.6 per cent engaged in social activity outside home; 53.6 per cent in health activity. "Health activity" often meant walking from home to shops, relatives' homes and so on. There was rarely one constraint on participation in social and health activity.Ill-health and disability were the clearest barriers to social and health activities, with one in ten of participants housebound, or virtually so, without the "lifeline" of transport from ACL or family.Replies indicated lack of security, in going out, especially alone - in daytime as well as after dark, age and ageism. Care roles and lack of time, commonly in association, account for a fifth of replies. Perceived distance, from health or social provisions which participants found acceptable, also excluded. There was variation in activity by gender (women were generally the more socially active), disability, health status, social class and areas' geographies. The following sections present themes from qualitative discussions which contributed nuance to street interview responses.Ward differences in social and health activity North Liverpool street interviews indicated variation in health activity between 41 per cent in Anfield, an area with a history of manual labour and poor life expectancy, and 71 per cent in Tuebrook-Stoneycroft, with its pockets of relative affluence and "suitable" provisions in terms of activities and opening hours. Hilly Everton, worst in Liverpool on nearly every deprivation indicator, including poor health, recorded an initially surprising response 66 per cent participation in health activity. The most-cited health activity was walking, however, and as a 70-year old commented:[...] it's an aerobic bit of country - there's a lousy set of bus services - so if you can't afford a cab, you've to walk to do anything, even if it kills you!Social and health activity levels differed also between the two south Liverpool wards, Speke and Garston. Speke's social activity response (41 per cent) lagged considerably between adjacent Garston (76 per cent) which includes some middle class areas. Voluntary groups organised community activity in Speke, but local pubs could be "expensive" and "more for young ones" or "for men". Health activity in Speke (59 per cent) was boosted by Active Age sessions organised by ACL and PCT. Garston participants also made good use of the centre, getting lifts or using their own vehicles and engaging in a wider range of other activities, such as bowls and walking. Experienced community workers in Garston reflected upon widening differences between older people, based partly on age and on social class. Health activities such as yoga and relaxation were becoming more popular than the "old keep-fit". Bingo "in draughty halls" was becoming an "elderly" interest. "Consumer conscious", more affluent young/old people who were still earning, "may prefer EasyJet to a bus trip". Disengagement was a recurrent theme in Speke, as observed in a Joseph Rowntree Foundation study (JRF, 2010). Informants referred to difficulty in "getting anything going" and the incidence of depression. Participants generally avoided the "run-down centre - and gangs - did you ever see anything like it?" Informants and participants commented on Speke housing's close proximity to Liverpool John Lennon Airport, with "a lot of noise and pollution, which is not healthy". Estate refurbishment was "slow" and "rough types" had moved in. Speke was considered "miles from Liverpool - for hospital visits and good shops".Activity in volunteering and care Only 4 per cent of participants "volunteered" but this is an understatement. "Volunteering" included structured help set up by an external agency, such as ACL. More commonly, it was informal, neighbourly response to need, which could be spontaneous or longer-term, as in the case of an intrepid 80-year old widow whose home was a neighbourhood resource for help with homework and bureaucracy. Some participants would have liked a "little job" to provide "that bit extra" but many were actively involved in family care, for older and younger relatives. One asked if "Age Concern could provide childcare!".Commenting on the active, "keystone" role of many older people, an informant stated:Older people's social capital can be under-rated. To replace their informal labour [...] would cost Governments a fortune. But offer an older person a job and the family can be worse off; childcare has to be paid for [...] Get a young person a job and the older person can't work. Whole families can become poor - and what tips the balance is often childcare.Another informant looked at affordable care options and encouraged older people to work at least part-time:It's what quite a few would like, so it's good for morale [...] the impact of more income on poor families is huge, financially and it's a model of work (for younger family) where perhaps none exists [...] it's a potentially long-term lift from poverty.Ageing activity and personal biographies Most participants had retired from unskilled or semi-skilled work. Others were former professionals, highly skilled workers, or small business owners and generally these were the participants with the knowledge and confidence to discuss the research. It was apparent that better incomes during the working life and some occupational pension entitlement could enhance active ageing. Computers and newspapers might be afforded. A wider range of social and health activity could be managed. A cab could be taken occasionally to activities outside the home and to larger stores selling fresh food. A home in a "safer area, where there are decent families and fewer gangs" could be afforded.Transport, mobility and outdoor security "The bus pass is the best car I ever had" was a comment supported by several participants, but transport factors could exclude. Only 14.3 per cent of participants owned a vehicle, all were men and all owned their homes, possible indicators of gender and social class influencing mobility. Home location in relation to appropriate social and health opportunities, travel distances and routes which public transport routes which participants felt were affordable, physically manageable and safe, were critical factors in participation. Transport availability and security influenced a preference for daytime activities because: buses "finish practically at 6.0 pm, you can't wait ages at bus stops - you need a lift to be safe".Patterns of safety and insecurity emerged. Fear of attack by gangs even influenced choice of bus stop and route, with areas known locally for antisocial behaviour being avoided. Women expressed more vulnerability, but several men admitted being "wary" when certain clubs and public houses were open: "Now, you're no physical match for a younger fellow - or fellows". A focus group discussed security strategies including complex home exits: locking the front door, waiting, then checking the back of the house to confront potential housebreakers. A woman participant drew her indirect route to a community centre:[...] older people can be seen down the long straight streets here [...] I turn along x to my friend's house, turn near my doctor's, then hurry to the main road [...] I'm safe then.Fear of burglary could curtail activities outside the home, especially after dark.Some participants felt their contemporaries were over-cautious, but informants confirmed that fears were not groundless. Handbags had been snatched on pension day, so some older people clubbed up for a cab to return from a post office when Christmas pension bonuses were paid and CSOs assisted them. Crime rates per thousand population in the five wards, could range from 66.1 to 87.8 against a Liverpool average of 62.5 (LCC, 2007).Poor health and disability could also constrain outside activity. Some participants were housebound while others, of similar age and self-assessed state of health, went to social and learning events. Small income differences appeared important: the occasional cab to "get out and about" could be afforded; family and friends could give lifts. Three participants with mobility problems got lifts to learn computing, which could "help keep in touch", as one said matter-of-factly, "for when I really can't get out". Excel learning was to budget or help a grandson's homework; another participant was designing parish newsletters. Media and community events raised awareness of healthy diets but fresh foods were not readily available in deprived areas. Bus transport to adjacent areas could involve changing buses which was difficult, especially when carrying shopping.Social activities: gendered spaces[4] Social activity outside home was highly gendered. Men's socialising was primarily in pubs and clubs, including those connected with football and former workplaces. Women went somewhat less often to pubs and clubs, especially in certain areas, but church and hobbies (including bingo) were prevalent. Women appeared to have developed wider community links during childhood, child rearing and helping neighbours. Women also did most shopping, which could be the main "social" activity outside the home.Women focus group participants in a pleasant club referred to "rougher" pubs, where "ladies were not so welcome" [...] "especially if you're older". Local pubs could be well-defended, but not conspicuously more than in many cities. Pubs and clubs could be associated, however, with sport and major industries with predominantly male workforces, some now closed. "Male territory" was one woman's description. Several men referred to "getting out of the house", almost to a re-creation of work, with pub and club presenting opportunities to catch up with former work-mates "have a drink and doms - no bingo chatter!" Women generally felt easier about pubs with more mixed clientele, perhaps involving families.Whether in church or community settings, lunch clubs were popular with over half the participants. Older people are among the poorest and loneliest in these areas, and lunch clubs were usually subsidised by benefactors. Organisers confirmed that, for many older people, the two-course warm lunch could be the week's only full meal. Women's better life expectancy often meant solo attendance at events and lunch clubs were considered "respectable". Coming alone to a lunch club, with a formula in social interaction (lunch often followed by bingo), seemed preferable to clubs where interaction could be less predictable. Numbers of older women preferred not to go to evening events, except for something very special, and certainly not alone. Fear of crime was a deterrent, especially without a lift; also evening events could be "dressier", expensive, and attended primarily by couples.Social activities: other divided spaces Further exclusions related to religion and race. One informant commented: "some lovely people, poor, have little, but can be bitter about perceived difference - it's horrible". Religious discrimination was a factor in creating "no go"/"go warily" zones of social and health opportunity. Within the past three centuries, migrants left deprivation in Ireland, Scotland and Wales for further deprivation, particularly in north Liverpool. Sectarian graffiti is still evident and certain pubs and clubs suggest "Catholic" or "Protestant" traditions, gauged from names, insignia, colour schemes and football affiliations.During church-sponsored events, a participant could discreetly mention religious affiliation, for example: "I know (venue) [...] is Catholic/Protestant [...] I am Protestant/Catholic - but don't let on - I've come with (friend)". Asked what would happen if the rapporteur did "let on", the response was: "(other attendees at the event) wouldn't half give me the cold shoulder - and (friend) for bringing me - I'd feel awful for her". "Would the (organiser) behave the same way?" "No - she's younger - they don't care as much". Interestingly, a Salvation Army lunch centre seemed neutral territory where, as one Catholic 80-year old observed: "the Captain (does us) proud every week". Several key informants observed that, not exclusively among older people, sectarian allegiances are still apparent, culturally embedded, though not always accompanied by church-going, as in the past. Some transmission of these allegiances to the young was also in evidence.Racial discrimination was a further factor in exclusion, from participation and place of residence. The study wards are predominantly white British with a range of 4.5 per cent Black or Minority Ethnic (BME) residents in Anfield to 5.8 per cent in Tuebrook and Stoneycroft. BME participants contributed 5.7 per cent of the commentary. Some were professionals, some business owners, others were retired or unemployed. Their social activity appeared predominantly to be family-focused and some venues, more particularly in North Liverpool, were regarded as "no go" on grounds of racism and otherwise unsuitable for family. Businesses and residences may have been within a study area, but participation in social activity outside the home more typically took place in racially mixed areas. Focus groups narrated BME residential movement from deprived "white" Liverpool areas to more affluent suburbs or diverse inner Liverpool areas.Health activities: factors in participation and exclusion Health activity was associated with reasonable health and the availability of "suitable" venues which were affordable and offered personal and emotional security. Some women participants would be "going solo" and "feeling comfortable" was important. The most popular forms of health activity were walking and cycling which as one participant explained "come cheap". Two up-to-date health centres were described as "expensive" and "too Lycra". Participants could fear exposure to the youthful gaze and sessions for older people would have been appreciated: "you feel awkward"; "past it" (too "fat"; "old and wrinkly", "can't move [...] quickly"); the inhibitions of age. Affordability and security were also factors in choice of outdoor sites for health activity. Tuebrook's green spaces were considered accessible and easy to walk around, with "visible" space popular with older people and families. Tourist itineraries include parkland in north Liverpool with fine vistas of Liverpool, but older people shunned it because unobserved areas could invite crime. Satisfaction with "home" was important to well-being and active ageing. It could be the springboard to outside engagement, a place from which to contact the outside world through reading, telephones, computers in some cases, and somewhere to welcome family and friends. The home could be the repository of memory, an essential feature of personal security and identity, indeed as one participant commented, "my home is me".Home could signify loneliness, however, especially if mobility and health were difficult. It could be insecure within, if coping were problematic, and outside in troubled neighbourhoods. Across the wards, between 12 and 18 per cent of participants said they felt insecure in their homes, mainly through fear of burglary. Visits from family and friends were "cheaper than going out" but lack of maintenance, cleaning and adequate heating could constrain the reception of visitors. Home help services enabled many to remain in their homes, but there was fear of elder abuse and reluctance to allow help from strangers.Reaction to sheltered accommodation varied, depending mainly upon the previous home's characteristics, health, and whether a partner was alive. Loss of independence could sadden but for others, security and less "worry" compensated. The worries included burglary, "I can open my window now for fresh air"; organising budgets and home maintenance. Some older residents felt less lonely and happier. Others were ambivalent, liking the greater security, but disliking being observed, within others' structures of time, place and communal living. Some participants were currently homeless, primarily because of poverty and poor health. Homelessness was commonly preceded by intense difficulty, in relationships, mental and physical health and finance, although housing associations appeared generally more supportive than private landlords. "(In)active ageing" relates first to the contrast between narratives of older person as burden in relation to our observations of the pivotal role of many older people in family survival, working part-time, providing family care, or both. Some participants joked about their "inactivity", then gave accounts of selective disengagement, from post-employment "busy-ness" and "day-time television", to re-engagement in following interests: pigeons, local history and gardening were three examples. The title also recognises the barriers to active ageing, significantly built from lifetime inequity, in families and neighbourhoods.Lie et al. (2009) comment that older people's volunteering cannot substitute for funded, sustained and coherent care systems. In the UK, the Government's Big Society project (2010) has been criticised on a number of grounds: as rhetoric and cover for spending cuts (Corbett and Walker, 2012); contributory to the dismantling of the welfare state (Bone, 2011); and illustrative of short-term-ism (Ware, 2012). The "Big Society" is a wholly inadequate response to the needs of the deprived neighbourhoods we have studied. Here, the two major barriers to active ageing are long-term ill-health and disability underpinned by poverty. As Table I has indicated, Liverpool's deprivation is severe and persistent, so measures to effect improvement need to be well-supported and long-term. * The setting of a national minimum income for active and healthy living (O'Sullivan and Ashton, 2011).* Ensuring that incomes at least met this standard.* Emplementing the Marmot proposals (2010) for a "fair society" with reductions in health inequity from the beginning of life to the end.* Restoring the retirement pension - Retail Price Index link.* Establishing a National Care Service to address one of the major needs of older people and their carers.* Establishing a clear and effective focus within government for oversight and development of policy affecting older people.* Ensure that urban and rural areas are "friendly" to older people (ILC, 2011), which also means that they would be friendly to all.* Providing core funding for a range of opportunities in neighbourhoods for active participation by older people (Deeming, 2009).Lymbery (2012) points out that, especially in recession, an adequate governmental response to older people's needs is unlikely. We share that view. The conditions in the study areas are evidence of lack of social justice and the position is likely to deteriorate with further cuts in spending. Many informants were angered at the likely prospect of deterioration of already inadequate provisions in the wards, which is a fair response. In disseminating knowledge of Liverpool's situation, we hope to contribute to a growing body of knowledge and challenge about ageing in places characterised by these levels of poverty and deprivation. Opens in a new window.Table I IMD 2007 and 2010: ranking of Liverpool and seven major English citiesa Opens in a new window.Table II Liverpool ward study: deprivation in context Opens in a new window.Table III Levels of participation in social and health activities
|
- This in-depth investigation into active ageing consulted over 600 older people in some of Liverpool's poorest neighbourhoods. In disseminating knowledge of Liverpool's situation, the paper adds significantly to knowledge around the severe challenges to active ageing in localities characterised by multiple deprivation.
|
[SECTION: Purpose] Childhood obesity is a growing global health concern, with physical, emotional and social consequences frequently persisting into adulthood (Daniels et al., 2005; Must and Strauss, 1999; Doak et al., 2006; Reilly et al., 2003). Wang and Lobstein (2006) report that the obesity epidemic seems particularly prevalent among school age children. Accordingly, the World Health Organisation reports that an estimated 22 million children under five years old are currently obese, and that even developing countries are facing a similar problem. For example, in Thailand, the obesity prevalence rate for children aged five to 12 years old rose from 12.2 percent to 15-16 percent over the course of only two years (WHO, 2008). The rate of childhood obesity also seems to be on the increase, even in developed countries. For example, a WHO survey conducted in the USA revealed that obesity in children aged six to 11 years has more than doubled since the 1960's (WHO, 2008). In Australia, recent figures from the 2004 New South Wales Schools Physical Activity and Nutrition Survey (SPANS, Booth et al., 2006), indicate that 26 percent of boys and 24 percent of girls in the state of New South Wales aged between 5 and 16 years were overweight or obese. This is compared to the year 1985, when 11 percent of all young people aged 7 to 16 were overweight or obese (Booth et al., 2006).Childhood obesity has been shown to be associated with an increase in medical complications such as orthopaedic complications (Dietz, 1998), pulmonary consequences such as sleep apnea, asthma and exercise intolerance (Dietz, 1998; Ebbeling et al., 2002), and cardiovascular consequences, such as hypertension, dyslipedemia, chronic inflammation and blood clotting tendencies (Ebbeling et al., 2002). Children who are obese also face increased risks of non-alcoholic fatty liver disease (Daniels et al., 2005) and Type 2 diabetes (Daniels et al., 2005; Ebbeling et al., 2002).In addition to these physical consequences of obesity, children who are obese have also been shown to experience more psychosocial disturbances such as decreased social interaction, depression, impulse control problems and decreased perceived cognitive and athletic ability (Ells et al., 2006). Furthermore, Franklin et al. (2008) report a finding that obese boys and girls appeared to have lowered self concepts, with girls in particular experiencing significantly lower perceived social acceptance than their normal weight counterparts (Franklin et al., 2008). Accordingly, normal weight children's attitudes towards obese children or obese silhouettes have been found to be negative (Bell and Morgan, 2000) and unfavourable (Staffieri, 1967).Research has shown that accurate and developmentally sensitive understandings of chronic illness in children is associated with enhanced satisfaction, an improved emotional state, a better quality of life and, most importantly, better compliance with treatment (Veldtman et al., 2000). For children this understanding is contingent on a number of factors such as developmental levels and illness experience. It is a widely replicated and supported view that children's perceptions of illness vary in both content and depth, particularly as a child's understanding of their illness cognitions appear to be linked to their developmental level (Bibace and Walsh, 1980). Thus, attributions that children make about their own illness may be completely different from that of an adult or parent.For example, in the obesity/overweight literature, it has been found that adults understand a wide number of possible causes for obesity. These include junk food advertising during children's television viewing, genetic and endocrine factors, environmental factors, imbalance between energy taken into the body and energy expended, as well as increased calorie intake through diet and sedentary lifestyles (Saelens and Daniels, 2003). In an Australian study of lay persons' understanding of childhood obesity, Hardus et al. (2003) found that the public had similar sophisticated perceptions of the causes and preventions of obesity. Over half of the adults surveyed endorsed a view that media promotion of unhealthy foods and over consumption of fast food were the main contributors to childhood obesity.Despite this, it appears that the only aspects of children's views of obesity that have been investigated are perceived causes In a rare study examining childhood conceptions of the cause of obesity, Johnson et al. (1994) found that elementary school-aged children (in grades 1, 3 and 5) lacked much factual knowledge and held many misconceptions. For example, although children could identify "junk" foods as a cause, their definition of junk food included items like bread, pasta, potatoes, bananas and chicken. Many were confused about how fat, meat, salt and cholesterol led to obesity. In fact some thought that cholesterol was a type of food. Only 7 percent identified lack of exercise as a cause of obesity though many had no idea how exercise kept one from getting fat, or how a sedentary lifestyle could make a person fat. Although 4 percent mentioned calories as a cause of obesity, children did not know what calories were. Other causes of obesity identified were drinking too much milk or water and mother to unborn child transmission (Johnson et al., 1994). Thus, factual knowledge seemed to be limited, with children not fully comprehending the content of information, or how certain factors work to contribute to obesity.With regards to examining childhood conceptions of disease, an accepted, holistic and current framework is the common-sense model of illness representation (CSM; Leventhal et al., 1980). The CSM is primarily designed as an explanatory framework for the way people perceive and react to disease threat (Leventhal et al., 2003; Conner and Norman, 2005). It is hypothesised that these "lay views" of illness are formed along five dimensions, or interpretive schemas (Hagger and Orbell, 2003). The five dimensions of the CSM of illness representations are:1. Identity. Individual statements, beliefs, and labels for the illness. Leventhal et al. (2003) posit that this domain is critical as it joins feelings of vulnerability to general symptoms.2. Cause. Perceived causative agents of the disease, which may include genetics, lifestyles or infections.3. Timeline. The perceived duration of the illness.4. Consequences. The perceived impact that the illness has on overall lifestyle.5. Control/cure: Refers to perceived efficacy of treatment (Hagger and Orbell, 2003) as well as perceptions of how preventable, controllable or curable a condition is (Leventhal et al. 2003).Empirically, the CSM has been validated as being applicable to many illnesses, as it was developed primarily to apply to the health field (Leventhal et al., 2003). When applied to children, the CSM forms a good alternative to the Piagetian framework which is frequently used in the study of children's perceptions of illness (Eiser, 1989). First, the CSM addresses more domains of illness than simply causative factors, a criticism made of applications of the Piagetian model (Eiser, 1989). Secondly, the CSM allows for these five schema structures to be reshaped and develop in breadth as new information is gained from experience and other sources (Leventhal et al., 2003), a notion especially pertinent to children as they age and become more sophisticated in their cognitions.Empirically, previous studies have found that the CSM can be used to effectively describe children's perceptions of illnesses. For example, Goldman et al. (1991) reported that responses given by children aged four to six about the common cold tapped into the five domains articulated by the CSM. In addition, Paterson et al. (1999) also discovered that seven to 14 year olds' discussions on the common cold and asthma could also be conceptualised in terms of the illness representations model. The aim of the current study is to examine children's understandings of obesity. The current study aims to explore childhood perceptions of the identity, cause, timeline, consequences and cures/control of obesity. The current study recruited a total number of 33 children. Participants were as follows: 12 children aged seven to eight (36.4 percent), 13 were aged nine to ten (39.4 percent) and eight were 11-12 (24.2 percent). Age groups correspond to primary school grades 4, 5 and 6 respectively and are the group for which childhood obesity has shown the greatest increases over time (Behn and Ur, 2006; Wang and Lobstein, 2006). There were 20 boys in the sample. The final sample represented nine overweight/obese children (27.3 percent) and 24 normal weight children (62.7 percent). Children were recruited through the Catholic Schools system as well as a sample who were seeking appetite awareness training in Sydney, Australia. Children in this latter option were included in the study on the basis if their body mass indices fell within the childhood "overweight" or "obese" range defined by Cole et al. (2000). Children were interviewed using a semi-structured interview protocol which was designed to reflect the five dimensions of the CSM. Ethics approval for this study was granted by the University Human Ethics committee as well as the CEO of an Archdiocese of Catholic Schools in Sydney, Australia.The final sample is shown in graphical form in Figure 1.3.1 Materials The interview protocol was based on a proforma by Deborah Finney (Finney, 2003; personal communication) used in previous research on children's understanding of mental illness. The interview contained questions and tasks pertaining to the five areas of the CSM (Table I)3.2 Procedure After obtaining written permission from their parents to partake in the interview, children were interviewed in a quiet room at their school. Questioning followed the same order for each child, however, the interview was semi-structured as probing was needed for some children in order to elaborate their responses (for example, "What might happen to someone if they get 'sick'") and clarify some responses. Responses were audiotaped for purposes of clarity, and also to allow for the development of rapport with the child. Tapes were then transcribed by an independent agency.Direct questioning was performed to obtain responses pertaining to the various dimensions (e.g. how do you think some kids get to be obese?). The specific questions asked are shown in Table I. To ascertain features of labels, beliefs and statements of obesity, as relating to the domain of "identity", children were asked to draw a picture of someone they considered to be obese and a picture of themselves. They were then asked to state the differences between themselves and the "obese" picture to gain these beliefs and statements relating to obesity.3.3 Data analysis Qualitative content analysis was used to analyse the data obtained. Responses were coded according to the CSM dimensions - thus any section of text relating to any established dimension was included (Priest et al., 2002). To illustrate the type of responses which would have aptly fitted into a CSM dimensions, quotes from respondents are used. 4.1 Identity In the domain of "identity" of obesity, it was found that 6/9 overweight/obese children, and 9/24 of normal weight children identified a large stomach as a prominent feature of obesity. Large limbs and extremities were also a standout feature in 5/24 of the normal weight drawings and 3/9 overweight/obese drawings.Some further shared characteristics included:1. Different hair or lack of hair for 4/24 normal weight children and 1/9 of the overweight group, as shown below: Kevin[1] (normal weight): "Well, I've got different hair...I don't have my hair tied up in one of those things [hair net as in a sumo wrestler]"2. Perceptions of differences in clothes were present in 7/24 normal weight children's drawings and 1/9 overweight child's drawing. In this case, the "obese" drawing was often wearing no clothes or clothes seen as less fashionable, as seen below: Kevin (normal weight):"I have much better clothes on".Rachel (normal weight): "If I drew myself with a skirt on, I wouldn't have this [skin] pushed up and showing".One feature of the "obese" drawing found in 3/24 normal weight children and 1/9 overweight children was that the figure was drawn consuming junk food such as chocolate and lollies, for example: Matthew (normal weight): "He's eating chocolate"In addition, it was found that some normal weight children only drew their "obese" drawings as negative characters such as, the "Blueberry" from Charlie and the Chocolate Factory and Santa Claus.Examination of the obese drawings seemed to contain numerous negative features when compared to the child's self-drawing. This was generally found to be unanimous between both weight groups. Negative features of a physical nature found in the obese drawing (as opposed to the self drawing) included figures having no smile and figures wearing glasses.4.2 Cause With regards to causes of obesity, it was found that the majority of children were knowledgeable and correctly identified three major causes of obesity. These included:* The intake of junk food, cited by 15/24 of the normal weight group and 7/9 of the overweight children) Examples of such responses include:* Alan (overweight): "They just go and eat lots of McDonald's and stuff. Don't eat salad, healthy food".Chiara (normal weight): "Maybe they just don't get enough nice and healthy food".* Overeating, or eating too much food was also popularly endorsed as a cause of obesity with 7/24 normal weight and 2/9 overweight children making references such as these: Aaron (overweight): "Eat too much and snack all day ... need to break the habit of snacking 24/7".Joe (normal weight): "They eat way too much".* Lastly, decreases in exercise/activity levels were also commonly cited as a cause of obesity by 5/9 of the overweight/obese sample, and 13/24 of the normal weight sample. For example: Miranda (normal weight): "They might not exercise".Emma (overweight): "They can never like sport and they're not really sporty".It was of interest that only 3/9 of the overweight group and 6/24 of the normal weight group endorsed the engagement in sedentary activities as a cause of obesity, in the following ways: Aaron (overweight): "And also because they stay in front of the TV and computer all day and not exercise ... they just stay in the house all day really".Jared (normal weight): "Watching TV and not doing anything active".The analyses of responses also found that normal weight children solely identified a further two causes of obesity. These were:1. Attributions to parental behaviour, cited by 3/24 normal weight children, for example: Sally: "They're just let go by their parents ... their parents just say to them 'Oh don't worry about it, you can have whatever you want (of junk food) ... so that they don't know how to stop'".2. Familial transmission, as identified by 2/24 normal weight children in responses such as these: Carrie (normal weight): "Sometimes, their parents are large, so they're born with it".4.3 Timeline All children interviewed stated that duration of obesity was dependent on one's actions towards positive health behaviours. 4/9 overweight children and 18/24 normal weight children specified these positive health behaviours as eating healthily and engaging in exercise; for example:Brooke (normal weight): "If they don't start to exercise and eat healthy food, they will probably stay obese".Eric (overweight): "If they didn't get healthy soon, they'd stay obese".Of interest were some responses which reflected that duration of obesity was dependent on a cognitive change. In this case, 3/9 overweight / obese children, and 4/24 normal weight children interviewed made some reference to individual motivation in making a decision to lose weight, and the resisting of negative heath behaviours, as salient mediating features in the duration of obesity. For example:Alan (overweight): " ... depends if they get motivated to fitness or if they want to sit at home and have Macca's everyday".Robbie (normal weight): "If they don't try to stop being obese, they will (stay obese forever)".Children also perceived that an individual would remain obese unless they took part in exercise (3/9 overweight and 13/24 normal weight children), for example:Oliver (overweight): "I think they can exercise more and they can burn off all that overweight, like all that fat".Joe (normal weight): "Um, it depends if they started exercising more or something".4.4 Consequences In terms of immediate consequences, 3/9 of the overweight/obese children mentioned that one consequence of obesity was the person becoming more obese, for example:Brady (overweight): "um, they might get bigger".For the normal weight group, 4/24 children mentioned this as a consequence.Another popularly endorsed consequence was restricted mobility (3/9 for the overweight group and 11/24 of the normal weight group), for example:Aaron (overweight): "And also ... when you get obese, it's hard for you to, really hard for you to move I think".Vanessa (normal weight): "Sometimes, they can't move quickly".Further consequences to obesity as endorsed by the group can be viewed in Table II.In addition to the physical consequences of obesity, children were also asked about any possible social or emotional consequences linked to this condition. It was found that a common emotional consequence between both groups was the notion that obese children and adults would feel unhappy, or angry with themselves for remaining obese. 3/9 overweight and 4/24 normal weight children cited this consequence. For example:Emma (overweight): "They might be unhappy with themselves and get really stressed out".Rachel (normal weight): "Angry with themselves...like, how could I let this happen?".Also mentioned were that others would not like them as a person (3/9 overweight children and 6/24 normal weight children):Rachel (normal weight): "No one would want to be their friend".Paulo (overweight): "Um, well, for one, people might not like them".Half of normal weight children (12/24) endorsed the idea that obese people would get teased. This response was also endorsed by 3/9 overweight children for example:Charlotte (normal weight): "Because some people can tease them for being fat and being big".Emma (overweight): "Um, they might face teasing when they get older".Five children (four normal and one overweight) mentioned that obese children would be socially left out of activities, e.g.Samantha (normal weight): "They wouldn't get included in anything".Miranda (normal weight): "(People) might not want to play with him or her and then they'll have no friends".Lastly, five children spoke of sadness as an emotional consequence to obesity, four of these came from normal weight respondents, e.g.James (normal weight): "Probably make them really sad".Matthew (normal weight): "They might get sad because they get teased".Jason (overweight): "They might get depressed because no one likes them because they're big".4.5 Cure/control In the final dimension of Leventhal et al.'s (1980) common sense model of illness representation, the children's responses for cure or control of obesity were fairly homogeneous with 7/9 overweight/obese children and 23/24 normal weight children citing a need to exercise as a potent cure for obesity, e.g.Aaron (overweight): "The only way to get thin is to exercise really regularly and burn off the fat and calories and stuff".Brady (overweight): " ... exercise quite a bit and get to playing sports ..."Kevin (normal weight): "Go to the gym, like a couple of times a day for a week".The need to eat well, or decrease eating of junk food was also perceived as a cure for obesity by 6/9 overweight/obese children and 18/24 normal weight children, for example:Brady (overweight): " ... eat healthy fruit, food, veggies, not fried...deep fried chips".Emma (overweight): " ... eat some fruit and healthy things".Caroline (normal weight): " ... just eat your vegetables and fruit ... and you could stop eating so many calories".When questioned about the efficacy of such cures for obesity, 6/9 of the overweight and 23/24 of the normal weight children stated that these measures definitely worked if one implemented them. The normal weight children referred to their own weight as evidence that these methods worked, for example:Luke (normal weight): "Well that's what I do and I'm not fat".Caroline (normal weight): "Well, it works because my family rarely ever gets Macker's".This was interpreted as reflecting children's perceptions of the cures of obesity as being one's own responsibility as for the other categories. It was of further interest that exercise in particular was viewed as a powerful cure of obesity, whereas the causal account between obesity and a lack of exercise was not as strong. In the current study, children of primary school age were interviewed on their views of obesity as guided by the five dimensions of Leventhal et al.'s (1980) common sense model of illness representations. In general, it was found that children were quite knowledgeable in all areas.The most obvious identity feature of obesity depicted by the children was a large stomach. If the negative connotations of stereotyping are ignored, this is encouraging as children may be referring to a large waist circumference, which is associated with the most health complications in obesity (Wardle et al., 2008). However, the other identity features of obesity listed by the children were also primarily negative in nature. Such findings are consistent with previous research that have found negative societal attitudes of overweight people found in adults (Bell and Morgan, 2000; Pingitore et al., 1994; King et al., 2006; Hebl and Xu, 2001), and children (Hill and Silver, 1995; Staffieri, 1967).Children correctly identified consumption of junk food, overeating and a non-engagement in activity and exercise as prominent causes of obesity, reminiscent of findings by Eiser et al. (1983) whereby children were able to identify being "healthy" as exercising, having a balanced diet and being energetic in general.Almost half of all children interviewed did not mention sedentary behaviour as a cause of obesity. A similar finding was reported by Johnson et al. (1994). This finding has also been reflected in a recent study conducted by the NSW Schools Physical and Nutrition Survey (SPANS) (2004) (see Booth et al., 2006). The study reported that despite there being a recent increase in school-aged children fulfilling recommended exercise requirements over the year 2004-2005, there was also a corresponding increase in children undertaking sedentary activities such as television, videos/DVDs, and electronic and computer games (Booth et al., 2006). This generally poor understanding of sedentary behaviours is of particular interest as sedentary living is one of the major tenets in recent media campaigns by the NSW government to decrease the childhood obesity rate. Perhaps such interventions need to directly address how sedentary behaviours lead to obesity, and encourage children to identify their own patterns of sedentary behaviours.The majority of children in the current sample stated that the timeline or duration of obesity was reliant on people undertaking positive health behaviours. This could also be interpreted as reflective of an internal locus of control whereby the person is seen as being responsible for their own weight loss. As such, it may be that children believe that by either not engaging in these positive behaviours, or, alternatively having failed in weight loss despite these behaviours, that people may have "done the wrong thing". On the basis of their understanding of the causes of obesity, it appears that children seem to believe that the duration is short if people take appropriate action. This internal attribution to weight loss may mean that they fail to capture the multifactorial complexity of the issue.It was noted that normal weight children were more adept than their overweight counterparts at detecting the most severe consequences of obesity. First, it may be possible that the parents of overweight children protect them from such negative consequences, as it is understandable that parents would aim to minimise distress in their children. Second, it may also be that overweight children and their parents underestimate the true risks of obesity, as found by Jeffery et al. (2005).Lastly, the cures/control section of the interview yielded the finding that most children endorsed the idea that exercise was a potent cure for obesity. This finding may be seen as reflecting a general sense of an internal locus of control - in that it is seen as a person's own responsibility to carry out this behaviour to lose weight, and that exercise, being a vigorous, visible activity, is the most obvious behaviour to contribute to this. However, this is also limiting in that exercise is not the only component contributing to healthy lifestyles, and needs to be undertaken in conjunction with other positive health behaviours. Hence, the multiple components of obesity and overweight also seem to be underappreciated by children in the current study. Further interventions should make clear the multifactorial complexity of obesity to children, and further highlight the nature of sedentary behaviours in maintaining this condition.The current study was limited in that most of the sample came from the Catholic education system in Sydney's eastern suburbs, which are more affluent in terms of socioeconomic status and therefore impacts on the study's external validity. Nonetheless, recent figures from the Australian Bureau of Statistics (www.abs.gov.au) indicate that obesity and overweight actually tends to be more prevalent in low income households and living areas of greatest disadvantage. Future research in this area should therefore aim to recruit participants from areas of lower socio-economic status. In addition, a broader sample could also be obtained from public education settings. On a related note, the treatment-seeking sample in the current study also presents with limitations in the sense that results may be biased and not representative of the wider population. With this respect, wider sampling in a number of areas would be a further avenue to consider in future studies.In conclusion, this study has qualitatively examined children's views on facets of obesity as pertinent to the common sense model of illness representations. Results found that children do indeed have well developed perceptions of the identity, cause, timeline, consequences and control/cure of obesity but which nevertheless have some qualitatively different deviations from "expert" or "adult" understandings of these facets. Future obesity prevention interventions should take these childhood perceptions into account when considering improved adherence to regimes. Opens in a new window.Figure 1 Final sample of current study stratified in terms of age, gender and weight Opens in a new window.Table I CSM domain examined and examples of questions asked in each domain Opens in a new window.Table II Normal weight and overweight/obese responses to the consequence of obesity
|
- The aim of this paper is to qualitatively examine the ways in which primary school children, aged between 7 and 12, perceive various facets of obesity as defined by the common sense model of illness representation (CCM).
|
[SECTION: Method] Childhood obesity is a growing global health concern, with physical, emotional and social consequences frequently persisting into adulthood (Daniels et al., 2005; Must and Strauss, 1999; Doak et al., 2006; Reilly et al., 2003). Wang and Lobstein (2006) report that the obesity epidemic seems particularly prevalent among school age children. Accordingly, the World Health Organisation reports that an estimated 22 million children under five years old are currently obese, and that even developing countries are facing a similar problem. For example, in Thailand, the obesity prevalence rate for children aged five to 12 years old rose from 12.2 percent to 15-16 percent over the course of only two years (WHO, 2008). The rate of childhood obesity also seems to be on the increase, even in developed countries. For example, a WHO survey conducted in the USA revealed that obesity in children aged six to 11 years has more than doubled since the 1960's (WHO, 2008). In Australia, recent figures from the 2004 New South Wales Schools Physical Activity and Nutrition Survey (SPANS, Booth et al., 2006), indicate that 26 percent of boys and 24 percent of girls in the state of New South Wales aged between 5 and 16 years were overweight or obese. This is compared to the year 1985, when 11 percent of all young people aged 7 to 16 were overweight or obese (Booth et al., 2006).Childhood obesity has been shown to be associated with an increase in medical complications such as orthopaedic complications (Dietz, 1998), pulmonary consequences such as sleep apnea, asthma and exercise intolerance (Dietz, 1998; Ebbeling et al., 2002), and cardiovascular consequences, such as hypertension, dyslipedemia, chronic inflammation and blood clotting tendencies (Ebbeling et al., 2002). Children who are obese also face increased risks of non-alcoholic fatty liver disease (Daniels et al., 2005) and Type 2 diabetes (Daniels et al., 2005; Ebbeling et al., 2002).In addition to these physical consequences of obesity, children who are obese have also been shown to experience more psychosocial disturbances such as decreased social interaction, depression, impulse control problems and decreased perceived cognitive and athletic ability (Ells et al., 2006). Furthermore, Franklin et al. (2008) report a finding that obese boys and girls appeared to have lowered self concepts, with girls in particular experiencing significantly lower perceived social acceptance than their normal weight counterparts (Franklin et al., 2008). Accordingly, normal weight children's attitudes towards obese children or obese silhouettes have been found to be negative (Bell and Morgan, 2000) and unfavourable (Staffieri, 1967).Research has shown that accurate and developmentally sensitive understandings of chronic illness in children is associated with enhanced satisfaction, an improved emotional state, a better quality of life and, most importantly, better compliance with treatment (Veldtman et al., 2000). For children this understanding is contingent on a number of factors such as developmental levels and illness experience. It is a widely replicated and supported view that children's perceptions of illness vary in both content and depth, particularly as a child's understanding of their illness cognitions appear to be linked to their developmental level (Bibace and Walsh, 1980). Thus, attributions that children make about their own illness may be completely different from that of an adult or parent.For example, in the obesity/overweight literature, it has been found that adults understand a wide number of possible causes for obesity. These include junk food advertising during children's television viewing, genetic and endocrine factors, environmental factors, imbalance between energy taken into the body and energy expended, as well as increased calorie intake through diet and sedentary lifestyles (Saelens and Daniels, 2003). In an Australian study of lay persons' understanding of childhood obesity, Hardus et al. (2003) found that the public had similar sophisticated perceptions of the causes and preventions of obesity. Over half of the adults surveyed endorsed a view that media promotion of unhealthy foods and over consumption of fast food were the main contributors to childhood obesity.Despite this, it appears that the only aspects of children's views of obesity that have been investigated are perceived causes In a rare study examining childhood conceptions of the cause of obesity, Johnson et al. (1994) found that elementary school-aged children (in grades 1, 3 and 5) lacked much factual knowledge and held many misconceptions. For example, although children could identify "junk" foods as a cause, their definition of junk food included items like bread, pasta, potatoes, bananas and chicken. Many were confused about how fat, meat, salt and cholesterol led to obesity. In fact some thought that cholesterol was a type of food. Only 7 percent identified lack of exercise as a cause of obesity though many had no idea how exercise kept one from getting fat, or how a sedentary lifestyle could make a person fat. Although 4 percent mentioned calories as a cause of obesity, children did not know what calories were. Other causes of obesity identified were drinking too much milk or water and mother to unborn child transmission (Johnson et al., 1994). Thus, factual knowledge seemed to be limited, with children not fully comprehending the content of information, or how certain factors work to contribute to obesity.With regards to examining childhood conceptions of disease, an accepted, holistic and current framework is the common-sense model of illness representation (CSM; Leventhal et al., 1980). The CSM is primarily designed as an explanatory framework for the way people perceive and react to disease threat (Leventhal et al., 2003; Conner and Norman, 2005). It is hypothesised that these "lay views" of illness are formed along five dimensions, or interpretive schemas (Hagger and Orbell, 2003). The five dimensions of the CSM of illness representations are:1. Identity. Individual statements, beliefs, and labels for the illness. Leventhal et al. (2003) posit that this domain is critical as it joins feelings of vulnerability to general symptoms.2. Cause. Perceived causative agents of the disease, which may include genetics, lifestyles or infections.3. Timeline. The perceived duration of the illness.4. Consequences. The perceived impact that the illness has on overall lifestyle.5. Control/cure: Refers to perceived efficacy of treatment (Hagger and Orbell, 2003) as well as perceptions of how preventable, controllable or curable a condition is (Leventhal et al. 2003).Empirically, the CSM has been validated as being applicable to many illnesses, as it was developed primarily to apply to the health field (Leventhal et al., 2003). When applied to children, the CSM forms a good alternative to the Piagetian framework which is frequently used in the study of children's perceptions of illness (Eiser, 1989). First, the CSM addresses more domains of illness than simply causative factors, a criticism made of applications of the Piagetian model (Eiser, 1989). Secondly, the CSM allows for these five schema structures to be reshaped and develop in breadth as new information is gained from experience and other sources (Leventhal et al., 2003), a notion especially pertinent to children as they age and become more sophisticated in their cognitions.Empirically, previous studies have found that the CSM can be used to effectively describe children's perceptions of illnesses. For example, Goldman et al. (1991) reported that responses given by children aged four to six about the common cold tapped into the five domains articulated by the CSM. In addition, Paterson et al. (1999) also discovered that seven to 14 year olds' discussions on the common cold and asthma could also be conceptualised in terms of the illness representations model. The aim of the current study is to examine children's understandings of obesity. The current study aims to explore childhood perceptions of the identity, cause, timeline, consequences and cures/control of obesity. The current study recruited a total number of 33 children. Participants were as follows: 12 children aged seven to eight (36.4 percent), 13 were aged nine to ten (39.4 percent) and eight were 11-12 (24.2 percent). Age groups correspond to primary school grades 4, 5 and 6 respectively and are the group for which childhood obesity has shown the greatest increases over time (Behn and Ur, 2006; Wang and Lobstein, 2006). There were 20 boys in the sample. The final sample represented nine overweight/obese children (27.3 percent) and 24 normal weight children (62.7 percent). Children were recruited through the Catholic Schools system as well as a sample who were seeking appetite awareness training in Sydney, Australia. Children in this latter option were included in the study on the basis if their body mass indices fell within the childhood "overweight" or "obese" range defined by Cole et al. (2000). Children were interviewed using a semi-structured interview protocol which was designed to reflect the five dimensions of the CSM. Ethics approval for this study was granted by the University Human Ethics committee as well as the CEO of an Archdiocese of Catholic Schools in Sydney, Australia.The final sample is shown in graphical form in Figure 1.3.1 Materials The interview protocol was based on a proforma by Deborah Finney (Finney, 2003; personal communication) used in previous research on children's understanding of mental illness. The interview contained questions and tasks pertaining to the five areas of the CSM (Table I)3.2 Procedure After obtaining written permission from their parents to partake in the interview, children were interviewed in a quiet room at their school. Questioning followed the same order for each child, however, the interview was semi-structured as probing was needed for some children in order to elaborate their responses (for example, "What might happen to someone if they get 'sick'") and clarify some responses. Responses were audiotaped for purposes of clarity, and also to allow for the development of rapport with the child. Tapes were then transcribed by an independent agency.Direct questioning was performed to obtain responses pertaining to the various dimensions (e.g. how do you think some kids get to be obese?). The specific questions asked are shown in Table I. To ascertain features of labels, beliefs and statements of obesity, as relating to the domain of "identity", children were asked to draw a picture of someone they considered to be obese and a picture of themselves. They were then asked to state the differences between themselves and the "obese" picture to gain these beliefs and statements relating to obesity.3.3 Data analysis Qualitative content analysis was used to analyse the data obtained. Responses were coded according to the CSM dimensions - thus any section of text relating to any established dimension was included (Priest et al., 2002). To illustrate the type of responses which would have aptly fitted into a CSM dimensions, quotes from respondents are used. 4.1 Identity In the domain of "identity" of obesity, it was found that 6/9 overweight/obese children, and 9/24 of normal weight children identified a large stomach as a prominent feature of obesity. Large limbs and extremities were also a standout feature in 5/24 of the normal weight drawings and 3/9 overweight/obese drawings.Some further shared characteristics included:1. Different hair or lack of hair for 4/24 normal weight children and 1/9 of the overweight group, as shown below: Kevin[1] (normal weight): "Well, I've got different hair...I don't have my hair tied up in one of those things [hair net as in a sumo wrestler]"2. Perceptions of differences in clothes were present in 7/24 normal weight children's drawings and 1/9 overweight child's drawing. In this case, the "obese" drawing was often wearing no clothes or clothes seen as less fashionable, as seen below: Kevin (normal weight):"I have much better clothes on".Rachel (normal weight): "If I drew myself with a skirt on, I wouldn't have this [skin] pushed up and showing".One feature of the "obese" drawing found in 3/24 normal weight children and 1/9 overweight children was that the figure was drawn consuming junk food such as chocolate and lollies, for example: Matthew (normal weight): "He's eating chocolate"In addition, it was found that some normal weight children only drew their "obese" drawings as negative characters such as, the "Blueberry" from Charlie and the Chocolate Factory and Santa Claus.Examination of the obese drawings seemed to contain numerous negative features when compared to the child's self-drawing. This was generally found to be unanimous between both weight groups. Negative features of a physical nature found in the obese drawing (as opposed to the self drawing) included figures having no smile and figures wearing glasses.4.2 Cause With regards to causes of obesity, it was found that the majority of children were knowledgeable and correctly identified three major causes of obesity. These included:* The intake of junk food, cited by 15/24 of the normal weight group and 7/9 of the overweight children) Examples of such responses include:* Alan (overweight): "They just go and eat lots of McDonald's and stuff. Don't eat salad, healthy food".Chiara (normal weight): "Maybe they just don't get enough nice and healthy food".* Overeating, or eating too much food was also popularly endorsed as a cause of obesity with 7/24 normal weight and 2/9 overweight children making references such as these: Aaron (overweight): "Eat too much and snack all day ... need to break the habit of snacking 24/7".Joe (normal weight): "They eat way too much".* Lastly, decreases in exercise/activity levels were also commonly cited as a cause of obesity by 5/9 of the overweight/obese sample, and 13/24 of the normal weight sample. For example: Miranda (normal weight): "They might not exercise".Emma (overweight): "They can never like sport and they're not really sporty".It was of interest that only 3/9 of the overweight group and 6/24 of the normal weight group endorsed the engagement in sedentary activities as a cause of obesity, in the following ways: Aaron (overweight): "And also because they stay in front of the TV and computer all day and not exercise ... they just stay in the house all day really".Jared (normal weight): "Watching TV and not doing anything active".The analyses of responses also found that normal weight children solely identified a further two causes of obesity. These were:1. Attributions to parental behaviour, cited by 3/24 normal weight children, for example: Sally: "They're just let go by their parents ... their parents just say to them 'Oh don't worry about it, you can have whatever you want (of junk food) ... so that they don't know how to stop'".2. Familial transmission, as identified by 2/24 normal weight children in responses such as these: Carrie (normal weight): "Sometimes, their parents are large, so they're born with it".4.3 Timeline All children interviewed stated that duration of obesity was dependent on one's actions towards positive health behaviours. 4/9 overweight children and 18/24 normal weight children specified these positive health behaviours as eating healthily and engaging in exercise; for example:Brooke (normal weight): "If they don't start to exercise and eat healthy food, they will probably stay obese".Eric (overweight): "If they didn't get healthy soon, they'd stay obese".Of interest were some responses which reflected that duration of obesity was dependent on a cognitive change. In this case, 3/9 overweight / obese children, and 4/24 normal weight children interviewed made some reference to individual motivation in making a decision to lose weight, and the resisting of negative heath behaviours, as salient mediating features in the duration of obesity. For example:Alan (overweight): " ... depends if they get motivated to fitness or if they want to sit at home and have Macca's everyday".Robbie (normal weight): "If they don't try to stop being obese, they will (stay obese forever)".Children also perceived that an individual would remain obese unless they took part in exercise (3/9 overweight and 13/24 normal weight children), for example:Oliver (overweight): "I think they can exercise more and they can burn off all that overweight, like all that fat".Joe (normal weight): "Um, it depends if they started exercising more or something".4.4 Consequences In terms of immediate consequences, 3/9 of the overweight/obese children mentioned that one consequence of obesity was the person becoming more obese, for example:Brady (overweight): "um, they might get bigger".For the normal weight group, 4/24 children mentioned this as a consequence.Another popularly endorsed consequence was restricted mobility (3/9 for the overweight group and 11/24 of the normal weight group), for example:Aaron (overweight): "And also ... when you get obese, it's hard for you to, really hard for you to move I think".Vanessa (normal weight): "Sometimes, they can't move quickly".Further consequences to obesity as endorsed by the group can be viewed in Table II.In addition to the physical consequences of obesity, children were also asked about any possible social or emotional consequences linked to this condition. It was found that a common emotional consequence between both groups was the notion that obese children and adults would feel unhappy, or angry with themselves for remaining obese. 3/9 overweight and 4/24 normal weight children cited this consequence. For example:Emma (overweight): "They might be unhappy with themselves and get really stressed out".Rachel (normal weight): "Angry with themselves...like, how could I let this happen?".Also mentioned were that others would not like them as a person (3/9 overweight children and 6/24 normal weight children):Rachel (normal weight): "No one would want to be their friend".Paulo (overweight): "Um, well, for one, people might not like them".Half of normal weight children (12/24) endorsed the idea that obese people would get teased. This response was also endorsed by 3/9 overweight children for example:Charlotte (normal weight): "Because some people can tease them for being fat and being big".Emma (overweight): "Um, they might face teasing when they get older".Five children (four normal and one overweight) mentioned that obese children would be socially left out of activities, e.g.Samantha (normal weight): "They wouldn't get included in anything".Miranda (normal weight): "(People) might not want to play with him or her and then they'll have no friends".Lastly, five children spoke of sadness as an emotional consequence to obesity, four of these came from normal weight respondents, e.g.James (normal weight): "Probably make them really sad".Matthew (normal weight): "They might get sad because they get teased".Jason (overweight): "They might get depressed because no one likes them because they're big".4.5 Cure/control In the final dimension of Leventhal et al.'s (1980) common sense model of illness representation, the children's responses for cure or control of obesity were fairly homogeneous with 7/9 overweight/obese children and 23/24 normal weight children citing a need to exercise as a potent cure for obesity, e.g.Aaron (overweight): "The only way to get thin is to exercise really regularly and burn off the fat and calories and stuff".Brady (overweight): " ... exercise quite a bit and get to playing sports ..."Kevin (normal weight): "Go to the gym, like a couple of times a day for a week".The need to eat well, or decrease eating of junk food was also perceived as a cure for obesity by 6/9 overweight/obese children and 18/24 normal weight children, for example:Brady (overweight): " ... eat healthy fruit, food, veggies, not fried...deep fried chips".Emma (overweight): " ... eat some fruit and healthy things".Caroline (normal weight): " ... just eat your vegetables and fruit ... and you could stop eating so many calories".When questioned about the efficacy of such cures for obesity, 6/9 of the overweight and 23/24 of the normal weight children stated that these measures definitely worked if one implemented them. The normal weight children referred to their own weight as evidence that these methods worked, for example:Luke (normal weight): "Well that's what I do and I'm not fat".Caroline (normal weight): "Well, it works because my family rarely ever gets Macker's".This was interpreted as reflecting children's perceptions of the cures of obesity as being one's own responsibility as for the other categories. It was of further interest that exercise in particular was viewed as a powerful cure of obesity, whereas the causal account between obesity and a lack of exercise was not as strong. In the current study, children of primary school age were interviewed on their views of obesity as guided by the five dimensions of Leventhal et al.'s (1980) common sense model of illness representations. In general, it was found that children were quite knowledgeable in all areas.The most obvious identity feature of obesity depicted by the children was a large stomach. If the negative connotations of stereotyping are ignored, this is encouraging as children may be referring to a large waist circumference, which is associated with the most health complications in obesity (Wardle et al., 2008). However, the other identity features of obesity listed by the children were also primarily negative in nature. Such findings are consistent with previous research that have found negative societal attitudes of overweight people found in adults (Bell and Morgan, 2000; Pingitore et al., 1994; King et al., 2006; Hebl and Xu, 2001), and children (Hill and Silver, 1995; Staffieri, 1967).Children correctly identified consumption of junk food, overeating and a non-engagement in activity and exercise as prominent causes of obesity, reminiscent of findings by Eiser et al. (1983) whereby children were able to identify being "healthy" as exercising, having a balanced diet and being energetic in general.Almost half of all children interviewed did not mention sedentary behaviour as a cause of obesity. A similar finding was reported by Johnson et al. (1994). This finding has also been reflected in a recent study conducted by the NSW Schools Physical and Nutrition Survey (SPANS) (2004) (see Booth et al., 2006). The study reported that despite there being a recent increase in school-aged children fulfilling recommended exercise requirements over the year 2004-2005, there was also a corresponding increase in children undertaking sedentary activities such as television, videos/DVDs, and electronic and computer games (Booth et al., 2006). This generally poor understanding of sedentary behaviours is of particular interest as sedentary living is one of the major tenets in recent media campaigns by the NSW government to decrease the childhood obesity rate. Perhaps such interventions need to directly address how sedentary behaviours lead to obesity, and encourage children to identify their own patterns of sedentary behaviours.The majority of children in the current sample stated that the timeline or duration of obesity was reliant on people undertaking positive health behaviours. This could also be interpreted as reflective of an internal locus of control whereby the person is seen as being responsible for their own weight loss. As such, it may be that children believe that by either not engaging in these positive behaviours, or, alternatively having failed in weight loss despite these behaviours, that people may have "done the wrong thing". On the basis of their understanding of the causes of obesity, it appears that children seem to believe that the duration is short if people take appropriate action. This internal attribution to weight loss may mean that they fail to capture the multifactorial complexity of the issue.It was noted that normal weight children were more adept than their overweight counterparts at detecting the most severe consequences of obesity. First, it may be possible that the parents of overweight children protect them from such negative consequences, as it is understandable that parents would aim to minimise distress in their children. Second, it may also be that overweight children and their parents underestimate the true risks of obesity, as found by Jeffery et al. (2005).Lastly, the cures/control section of the interview yielded the finding that most children endorsed the idea that exercise was a potent cure for obesity. This finding may be seen as reflecting a general sense of an internal locus of control - in that it is seen as a person's own responsibility to carry out this behaviour to lose weight, and that exercise, being a vigorous, visible activity, is the most obvious behaviour to contribute to this. However, this is also limiting in that exercise is not the only component contributing to healthy lifestyles, and needs to be undertaken in conjunction with other positive health behaviours. Hence, the multiple components of obesity and overweight also seem to be underappreciated by children in the current study. Further interventions should make clear the multifactorial complexity of obesity to children, and further highlight the nature of sedentary behaviours in maintaining this condition.The current study was limited in that most of the sample came from the Catholic education system in Sydney's eastern suburbs, which are more affluent in terms of socioeconomic status and therefore impacts on the study's external validity. Nonetheless, recent figures from the Australian Bureau of Statistics (www.abs.gov.au) indicate that obesity and overweight actually tends to be more prevalent in low income households and living areas of greatest disadvantage. Future research in this area should therefore aim to recruit participants from areas of lower socio-economic status. In addition, a broader sample could also be obtained from public education settings. On a related note, the treatment-seeking sample in the current study also presents with limitations in the sense that results may be biased and not representative of the wider population. With this respect, wider sampling in a number of areas would be a further avenue to consider in future studies.In conclusion, this study has qualitatively examined children's views on facets of obesity as pertinent to the common sense model of illness representations. Results found that children do indeed have well developed perceptions of the identity, cause, timeline, consequences and control/cure of obesity but which nevertheless have some qualitatively different deviations from "expert" or "adult" understandings of these facets. Future obesity prevention interventions should take these childhood perceptions into account when considering improved adherence to regimes. Opens in a new window.Figure 1 Final sample of current study stratified in terms of age, gender and weight Opens in a new window.Table I CSM domain examined and examples of questions asked in each domain Opens in a new window.Table II Normal weight and overweight/obese responses to the consequence of obesity
|
- The study was qualitative in nature. Semi-structured interviews were conducted with 33 children on all dimensions of the CSM. Twenty four children were normal weight and nine were overweight. A drawing task formed the methodology for the "identity" section of the interview.
|
[SECTION: Findings] Childhood obesity is a growing global health concern, with physical, emotional and social consequences frequently persisting into adulthood (Daniels et al., 2005; Must and Strauss, 1999; Doak et al., 2006; Reilly et al., 2003). Wang and Lobstein (2006) report that the obesity epidemic seems particularly prevalent among school age children. Accordingly, the World Health Organisation reports that an estimated 22 million children under five years old are currently obese, and that even developing countries are facing a similar problem. For example, in Thailand, the obesity prevalence rate for children aged five to 12 years old rose from 12.2 percent to 15-16 percent over the course of only two years (WHO, 2008). The rate of childhood obesity also seems to be on the increase, even in developed countries. For example, a WHO survey conducted in the USA revealed that obesity in children aged six to 11 years has more than doubled since the 1960's (WHO, 2008). In Australia, recent figures from the 2004 New South Wales Schools Physical Activity and Nutrition Survey (SPANS, Booth et al., 2006), indicate that 26 percent of boys and 24 percent of girls in the state of New South Wales aged between 5 and 16 years were overweight or obese. This is compared to the year 1985, when 11 percent of all young people aged 7 to 16 were overweight or obese (Booth et al., 2006).Childhood obesity has been shown to be associated with an increase in medical complications such as orthopaedic complications (Dietz, 1998), pulmonary consequences such as sleep apnea, asthma and exercise intolerance (Dietz, 1998; Ebbeling et al., 2002), and cardiovascular consequences, such as hypertension, dyslipedemia, chronic inflammation and blood clotting tendencies (Ebbeling et al., 2002). Children who are obese also face increased risks of non-alcoholic fatty liver disease (Daniels et al., 2005) and Type 2 diabetes (Daniels et al., 2005; Ebbeling et al., 2002).In addition to these physical consequences of obesity, children who are obese have also been shown to experience more psychosocial disturbances such as decreased social interaction, depression, impulse control problems and decreased perceived cognitive and athletic ability (Ells et al., 2006). Furthermore, Franklin et al. (2008) report a finding that obese boys and girls appeared to have lowered self concepts, with girls in particular experiencing significantly lower perceived social acceptance than their normal weight counterparts (Franklin et al., 2008). Accordingly, normal weight children's attitudes towards obese children or obese silhouettes have been found to be negative (Bell and Morgan, 2000) and unfavourable (Staffieri, 1967).Research has shown that accurate and developmentally sensitive understandings of chronic illness in children is associated with enhanced satisfaction, an improved emotional state, a better quality of life and, most importantly, better compliance with treatment (Veldtman et al., 2000). For children this understanding is contingent on a number of factors such as developmental levels and illness experience. It is a widely replicated and supported view that children's perceptions of illness vary in both content and depth, particularly as a child's understanding of their illness cognitions appear to be linked to their developmental level (Bibace and Walsh, 1980). Thus, attributions that children make about their own illness may be completely different from that of an adult or parent.For example, in the obesity/overweight literature, it has been found that adults understand a wide number of possible causes for obesity. These include junk food advertising during children's television viewing, genetic and endocrine factors, environmental factors, imbalance between energy taken into the body and energy expended, as well as increased calorie intake through diet and sedentary lifestyles (Saelens and Daniels, 2003). In an Australian study of lay persons' understanding of childhood obesity, Hardus et al. (2003) found that the public had similar sophisticated perceptions of the causes and preventions of obesity. Over half of the adults surveyed endorsed a view that media promotion of unhealthy foods and over consumption of fast food were the main contributors to childhood obesity.Despite this, it appears that the only aspects of children's views of obesity that have been investigated are perceived causes In a rare study examining childhood conceptions of the cause of obesity, Johnson et al. (1994) found that elementary school-aged children (in grades 1, 3 and 5) lacked much factual knowledge and held many misconceptions. For example, although children could identify "junk" foods as a cause, their definition of junk food included items like bread, pasta, potatoes, bananas and chicken. Many were confused about how fat, meat, salt and cholesterol led to obesity. In fact some thought that cholesterol was a type of food. Only 7 percent identified lack of exercise as a cause of obesity though many had no idea how exercise kept one from getting fat, or how a sedentary lifestyle could make a person fat. Although 4 percent mentioned calories as a cause of obesity, children did not know what calories were. Other causes of obesity identified were drinking too much milk or water and mother to unborn child transmission (Johnson et al., 1994). Thus, factual knowledge seemed to be limited, with children not fully comprehending the content of information, or how certain factors work to contribute to obesity.With regards to examining childhood conceptions of disease, an accepted, holistic and current framework is the common-sense model of illness representation (CSM; Leventhal et al., 1980). The CSM is primarily designed as an explanatory framework for the way people perceive and react to disease threat (Leventhal et al., 2003; Conner and Norman, 2005). It is hypothesised that these "lay views" of illness are formed along five dimensions, or interpretive schemas (Hagger and Orbell, 2003). The five dimensions of the CSM of illness representations are:1. Identity. Individual statements, beliefs, and labels for the illness. Leventhal et al. (2003) posit that this domain is critical as it joins feelings of vulnerability to general symptoms.2. Cause. Perceived causative agents of the disease, which may include genetics, lifestyles or infections.3. Timeline. The perceived duration of the illness.4. Consequences. The perceived impact that the illness has on overall lifestyle.5. Control/cure: Refers to perceived efficacy of treatment (Hagger and Orbell, 2003) as well as perceptions of how preventable, controllable or curable a condition is (Leventhal et al. 2003).Empirically, the CSM has been validated as being applicable to many illnesses, as it was developed primarily to apply to the health field (Leventhal et al., 2003). When applied to children, the CSM forms a good alternative to the Piagetian framework which is frequently used in the study of children's perceptions of illness (Eiser, 1989). First, the CSM addresses more domains of illness than simply causative factors, a criticism made of applications of the Piagetian model (Eiser, 1989). Secondly, the CSM allows for these five schema structures to be reshaped and develop in breadth as new information is gained from experience and other sources (Leventhal et al., 2003), a notion especially pertinent to children as they age and become more sophisticated in their cognitions.Empirically, previous studies have found that the CSM can be used to effectively describe children's perceptions of illnesses. For example, Goldman et al. (1991) reported that responses given by children aged four to six about the common cold tapped into the five domains articulated by the CSM. In addition, Paterson et al. (1999) also discovered that seven to 14 year olds' discussions on the common cold and asthma could also be conceptualised in terms of the illness representations model. The aim of the current study is to examine children's understandings of obesity. The current study aims to explore childhood perceptions of the identity, cause, timeline, consequences and cures/control of obesity. The current study recruited a total number of 33 children. Participants were as follows: 12 children aged seven to eight (36.4 percent), 13 were aged nine to ten (39.4 percent) and eight were 11-12 (24.2 percent). Age groups correspond to primary school grades 4, 5 and 6 respectively and are the group for which childhood obesity has shown the greatest increases over time (Behn and Ur, 2006; Wang and Lobstein, 2006). There were 20 boys in the sample. The final sample represented nine overweight/obese children (27.3 percent) and 24 normal weight children (62.7 percent). Children were recruited through the Catholic Schools system as well as a sample who were seeking appetite awareness training in Sydney, Australia. Children in this latter option were included in the study on the basis if their body mass indices fell within the childhood "overweight" or "obese" range defined by Cole et al. (2000). Children were interviewed using a semi-structured interview protocol which was designed to reflect the five dimensions of the CSM. Ethics approval for this study was granted by the University Human Ethics committee as well as the CEO of an Archdiocese of Catholic Schools in Sydney, Australia.The final sample is shown in graphical form in Figure 1.3.1 Materials The interview protocol was based on a proforma by Deborah Finney (Finney, 2003; personal communication) used in previous research on children's understanding of mental illness. The interview contained questions and tasks pertaining to the five areas of the CSM (Table I)3.2 Procedure After obtaining written permission from their parents to partake in the interview, children were interviewed in a quiet room at their school. Questioning followed the same order for each child, however, the interview was semi-structured as probing was needed for some children in order to elaborate their responses (for example, "What might happen to someone if they get 'sick'") and clarify some responses. Responses were audiotaped for purposes of clarity, and also to allow for the development of rapport with the child. Tapes were then transcribed by an independent agency.Direct questioning was performed to obtain responses pertaining to the various dimensions (e.g. how do you think some kids get to be obese?). The specific questions asked are shown in Table I. To ascertain features of labels, beliefs and statements of obesity, as relating to the domain of "identity", children were asked to draw a picture of someone they considered to be obese and a picture of themselves. They were then asked to state the differences between themselves and the "obese" picture to gain these beliefs and statements relating to obesity.3.3 Data analysis Qualitative content analysis was used to analyse the data obtained. Responses were coded according to the CSM dimensions - thus any section of text relating to any established dimension was included (Priest et al., 2002). To illustrate the type of responses which would have aptly fitted into a CSM dimensions, quotes from respondents are used. 4.1 Identity In the domain of "identity" of obesity, it was found that 6/9 overweight/obese children, and 9/24 of normal weight children identified a large stomach as a prominent feature of obesity. Large limbs and extremities were also a standout feature in 5/24 of the normal weight drawings and 3/9 overweight/obese drawings.Some further shared characteristics included:1. Different hair or lack of hair for 4/24 normal weight children and 1/9 of the overweight group, as shown below: Kevin[1] (normal weight): "Well, I've got different hair...I don't have my hair tied up in one of those things [hair net as in a sumo wrestler]"2. Perceptions of differences in clothes were present in 7/24 normal weight children's drawings and 1/9 overweight child's drawing. In this case, the "obese" drawing was often wearing no clothes or clothes seen as less fashionable, as seen below: Kevin (normal weight):"I have much better clothes on".Rachel (normal weight): "If I drew myself with a skirt on, I wouldn't have this [skin] pushed up and showing".One feature of the "obese" drawing found in 3/24 normal weight children and 1/9 overweight children was that the figure was drawn consuming junk food such as chocolate and lollies, for example: Matthew (normal weight): "He's eating chocolate"In addition, it was found that some normal weight children only drew their "obese" drawings as negative characters such as, the "Blueberry" from Charlie and the Chocolate Factory and Santa Claus.Examination of the obese drawings seemed to contain numerous negative features when compared to the child's self-drawing. This was generally found to be unanimous between both weight groups. Negative features of a physical nature found in the obese drawing (as opposed to the self drawing) included figures having no smile and figures wearing glasses.4.2 Cause With regards to causes of obesity, it was found that the majority of children were knowledgeable and correctly identified three major causes of obesity. These included:* The intake of junk food, cited by 15/24 of the normal weight group and 7/9 of the overweight children) Examples of such responses include:* Alan (overweight): "They just go and eat lots of McDonald's and stuff. Don't eat salad, healthy food".Chiara (normal weight): "Maybe they just don't get enough nice and healthy food".* Overeating, or eating too much food was also popularly endorsed as a cause of obesity with 7/24 normal weight and 2/9 overweight children making references such as these: Aaron (overweight): "Eat too much and snack all day ... need to break the habit of snacking 24/7".Joe (normal weight): "They eat way too much".* Lastly, decreases in exercise/activity levels were also commonly cited as a cause of obesity by 5/9 of the overweight/obese sample, and 13/24 of the normal weight sample. For example: Miranda (normal weight): "They might not exercise".Emma (overweight): "They can never like sport and they're not really sporty".It was of interest that only 3/9 of the overweight group and 6/24 of the normal weight group endorsed the engagement in sedentary activities as a cause of obesity, in the following ways: Aaron (overweight): "And also because they stay in front of the TV and computer all day and not exercise ... they just stay in the house all day really".Jared (normal weight): "Watching TV and not doing anything active".The analyses of responses also found that normal weight children solely identified a further two causes of obesity. These were:1. Attributions to parental behaviour, cited by 3/24 normal weight children, for example: Sally: "They're just let go by their parents ... their parents just say to them 'Oh don't worry about it, you can have whatever you want (of junk food) ... so that they don't know how to stop'".2. Familial transmission, as identified by 2/24 normal weight children in responses such as these: Carrie (normal weight): "Sometimes, their parents are large, so they're born with it".4.3 Timeline All children interviewed stated that duration of obesity was dependent on one's actions towards positive health behaviours. 4/9 overweight children and 18/24 normal weight children specified these positive health behaviours as eating healthily and engaging in exercise; for example:Brooke (normal weight): "If they don't start to exercise and eat healthy food, they will probably stay obese".Eric (overweight): "If they didn't get healthy soon, they'd stay obese".Of interest were some responses which reflected that duration of obesity was dependent on a cognitive change. In this case, 3/9 overweight / obese children, and 4/24 normal weight children interviewed made some reference to individual motivation in making a decision to lose weight, and the resisting of negative heath behaviours, as salient mediating features in the duration of obesity. For example:Alan (overweight): " ... depends if they get motivated to fitness or if they want to sit at home and have Macca's everyday".Robbie (normal weight): "If they don't try to stop being obese, they will (stay obese forever)".Children also perceived that an individual would remain obese unless they took part in exercise (3/9 overweight and 13/24 normal weight children), for example:Oliver (overweight): "I think they can exercise more and they can burn off all that overweight, like all that fat".Joe (normal weight): "Um, it depends if they started exercising more or something".4.4 Consequences In terms of immediate consequences, 3/9 of the overweight/obese children mentioned that one consequence of obesity was the person becoming more obese, for example:Brady (overweight): "um, they might get bigger".For the normal weight group, 4/24 children mentioned this as a consequence.Another popularly endorsed consequence was restricted mobility (3/9 for the overweight group and 11/24 of the normal weight group), for example:Aaron (overweight): "And also ... when you get obese, it's hard for you to, really hard for you to move I think".Vanessa (normal weight): "Sometimes, they can't move quickly".Further consequences to obesity as endorsed by the group can be viewed in Table II.In addition to the physical consequences of obesity, children were also asked about any possible social or emotional consequences linked to this condition. It was found that a common emotional consequence between both groups was the notion that obese children and adults would feel unhappy, or angry with themselves for remaining obese. 3/9 overweight and 4/24 normal weight children cited this consequence. For example:Emma (overweight): "They might be unhappy with themselves and get really stressed out".Rachel (normal weight): "Angry with themselves...like, how could I let this happen?".Also mentioned were that others would not like them as a person (3/9 overweight children and 6/24 normal weight children):Rachel (normal weight): "No one would want to be their friend".Paulo (overweight): "Um, well, for one, people might not like them".Half of normal weight children (12/24) endorsed the idea that obese people would get teased. This response was also endorsed by 3/9 overweight children for example:Charlotte (normal weight): "Because some people can tease them for being fat and being big".Emma (overweight): "Um, they might face teasing when they get older".Five children (four normal and one overweight) mentioned that obese children would be socially left out of activities, e.g.Samantha (normal weight): "They wouldn't get included in anything".Miranda (normal weight): "(People) might not want to play with him or her and then they'll have no friends".Lastly, five children spoke of sadness as an emotional consequence to obesity, four of these came from normal weight respondents, e.g.James (normal weight): "Probably make them really sad".Matthew (normal weight): "They might get sad because they get teased".Jason (overweight): "They might get depressed because no one likes them because they're big".4.5 Cure/control In the final dimension of Leventhal et al.'s (1980) common sense model of illness representation, the children's responses for cure or control of obesity were fairly homogeneous with 7/9 overweight/obese children and 23/24 normal weight children citing a need to exercise as a potent cure for obesity, e.g.Aaron (overweight): "The only way to get thin is to exercise really regularly and burn off the fat and calories and stuff".Brady (overweight): " ... exercise quite a bit and get to playing sports ..."Kevin (normal weight): "Go to the gym, like a couple of times a day for a week".The need to eat well, or decrease eating of junk food was also perceived as a cure for obesity by 6/9 overweight/obese children and 18/24 normal weight children, for example:Brady (overweight): " ... eat healthy fruit, food, veggies, not fried...deep fried chips".Emma (overweight): " ... eat some fruit and healthy things".Caroline (normal weight): " ... just eat your vegetables and fruit ... and you could stop eating so many calories".When questioned about the efficacy of such cures for obesity, 6/9 of the overweight and 23/24 of the normal weight children stated that these measures definitely worked if one implemented them. The normal weight children referred to their own weight as evidence that these methods worked, for example:Luke (normal weight): "Well that's what I do and I'm not fat".Caroline (normal weight): "Well, it works because my family rarely ever gets Macker's".This was interpreted as reflecting children's perceptions of the cures of obesity as being one's own responsibility as for the other categories. It was of further interest that exercise in particular was viewed as a powerful cure of obesity, whereas the causal account between obesity and a lack of exercise was not as strong. In the current study, children of primary school age were interviewed on their views of obesity as guided by the five dimensions of Leventhal et al.'s (1980) common sense model of illness representations. In general, it was found that children were quite knowledgeable in all areas.The most obvious identity feature of obesity depicted by the children was a large stomach. If the negative connotations of stereotyping are ignored, this is encouraging as children may be referring to a large waist circumference, which is associated with the most health complications in obesity (Wardle et al., 2008). However, the other identity features of obesity listed by the children were also primarily negative in nature. Such findings are consistent with previous research that have found negative societal attitudes of overweight people found in adults (Bell and Morgan, 2000; Pingitore et al., 1994; King et al., 2006; Hebl and Xu, 2001), and children (Hill and Silver, 1995; Staffieri, 1967).Children correctly identified consumption of junk food, overeating and a non-engagement in activity and exercise as prominent causes of obesity, reminiscent of findings by Eiser et al. (1983) whereby children were able to identify being "healthy" as exercising, having a balanced diet and being energetic in general.Almost half of all children interviewed did not mention sedentary behaviour as a cause of obesity. A similar finding was reported by Johnson et al. (1994). This finding has also been reflected in a recent study conducted by the NSW Schools Physical and Nutrition Survey (SPANS) (2004) (see Booth et al., 2006). The study reported that despite there being a recent increase in school-aged children fulfilling recommended exercise requirements over the year 2004-2005, there was also a corresponding increase in children undertaking sedentary activities such as television, videos/DVDs, and electronic and computer games (Booth et al., 2006). This generally poor understanding of sedentary behaviours is of particular interest as sedentary living is one of the major tenets in recent media campaigns by the NSW government to decrease the childhood obesity rate. Perhaps such interventions need to directly address how sedentary behaviours lead to obesity, and encourage children to identify their own patterns of sedentary behaviours.The majority of children in the current sample stated that the timeline or duration of obesity was reliant on people undertaking positive health behaviours. This could also be interpreted as reflective of an internal locus of control whereby the person is seen as being responsible for their own weight loss. As such, it may be that children believe that by either not engaging in these positive behaviours, or, alternatively having failed in weight loss despite these behaviours, that people may have "done the wrong thing". On the basis of their understanding of the causes of obesity, it appears that children seem to believe that the duration is short if people take appropriate action. This internal attribution to weight loss may mean that they fail to capture the multifactorial complexity of the issue.It was noted that normal weight children were more adept than their overweight counterparts at detecting the most severe consequences of obesity. First, it may be possible that the parents of overweight children protect them from such negative consequences, as it is understandable that parents would aim to minimise distress in their children. Second, it may also be that overweight children and their parents underestimate the true risks of obesity, as found by Jeffery et al. (2005).Lastly, the cures/control section of the interview yielded the finding that most children endorsed the idea that exercise was a potent cure for obesity. This finding may be seen as reflecting a general sense of an internal locus of control - in that it is seen as a person's own responsibility to carry out this behaviour to lose weight, and that exercise, being a vigorous, visible activity, is the most obvious behaviour to contribute to this. However, this is also limiting in that exercise is not the only component contributing to healthy lifestyles, and needs to be undertaken in conjunction with other positive health behaviours. Hence, the multiple components of obesity and overweight also seem to be underappreciated by children in the current study. Further interventions should make clear the multifactorial complexity of obesity to children, and further highlight the nature of sedentary behaviours in maintaining this condition.The current study was limited in that most of the sample came from the Catholic education system in Sydney's eastern suburbs, which are more affluent in terms of socioeconomic status and therefore impacts on the study's external validity. Nonetheless, recent figures from the Australian Bureau of Statistics (www.abs.gov.au) indicate that obesity and overweight actually tends to be more prevalent in low income households and living areas of greatest disadvantage. Future research in this area should therefore aim to recruit participants from areas of lower socio-economic status. In addition, a broader sample could also be obtained from public education settings. On a related note, the treatment-seeking sample in the current study also presents with limitations in the sense that results may be biased and not representative of the wider population. With this respect, wider sampling in a number of areas would be a further avenue to consider in future studies.In conclusion, this study has qualitatively examined children's views on facets of obesity as pertinent to the common sense model of illness representations. Results found that children do indeed have well developed perceptions of the identity, cause, timeline, consequences and control/cure of obesity but which nevertheless have some qualitatively different deviations from "expert" or "adult" understandings of these facets. Future obesity prevention interventions should take these childhood perceptions into account when considering improved adherence to regimes. Opens in a new window.Figure 1 Final sample of current study stratified in terms of age, gender and weight Opens in a new window.Table I CSM domain examined and examples of questions asked in each domain Opens in a new window.Table II Normal weight and overweight/obese responses to the consequence of obesity
|
- Although children identified food intake as a main cause of obesity, almost half did not name sedentary behaviours as a cause of obesity. Duration (timeline) of obesity was regarded by most children as reliant on a person's undertaking of positive health behaviours. Normal weight children were found to list more severe consequences of obesity than the overweight group. It was found that experience contributed to the detailed knowledge of overweight children's perceptions of cures of obesity. Overweight children also spoke of personal incidents of barriers to cures.
|
[SECTION: Value] Childhood obesity is a growing global health concern, with physical, emotional and social consequences frequently persisting into adulthood (Daniels et al., 2005; Must and Strauss, 1999; Doak et al., 2006; Reilly et al., 2003). Wang and Lobstein (2006) report that the obesity epidemic seems particularly prevalent among school age children. Accordingly, the World Health Organisation reports that an estimated 22 million children under five years old are currently obese, and that even developing countries are facing a similar problem. For example, in Thailand, the obesity prevalence rate for children aged five to 12 years old rose from 12.2 percent to 15-16 percent over the course of only two years (WHO, 2008). The rate of childhood obesity also seems to be on the increase, even in developed countries. For example, a WHO survey conducted in the USA revealed that obesity in children aged six to 11 years has more than doubled since the 1960's (WHO, 2008). In Australia, recent figures from the 2004 New South Wales Schools Physical Activity and Nutrition Survey (SPANS, Booth et al., 2006), indicate that 26 percent of boys and 24 percent of girls in the state of New South Wales aged between 5 and 16 years were overweight or obese. This is compared to the year 1985, when 11 percent of all young people aged 7 to 16 were overweight or obese (Booth et al., 2006).Childhood obesity has been shown to be associated with an increase in medical complications such as orthopaedic complications (Dietz, 1998), pulmonary consequences such as sleep apnea, asthma and exercise intolerance (Dietz, 1998; Ebbeling et al., 2002), and cardiovascular consequences, such as hypertension, dyslipedemia, chronic inflammation and blood clotting tendencies (Ebbeling et al., 2002). Children who are obese also face increased risks of non-alcoholic fatty liver disease (Daniels et al., 2005) and Type 2 diabetes (Daniels et al., 2005; Ebbeling et al., 2002).In addition to these physical consequences of obesity, children who are obese have also been shown to experience more psychosocial disturbances such as decreased social interaction, depression, impulse control problems and decreased perceived cognitive and athletic ability (Ells et al., 2006). Furthermore, Franklin et al. (2008) report a finding that obese boys and girls appeared to have lowered self concepts, with girls in particular experiencing significantly lower perceived social acceptance than their normal weight counterparts (Franklin et al., 2008). Accordingly, normal weight children's attitudes towards obese children or obese silhouettes have been found to be negative (Bell and Morgan, 2000) and unfavourable (Staffieri, 1967).Research has shown that accurate and developmentally sensitive understandings of chronic illness in children is associated with enhanced satisfaction, an improved emotional state, a better quality of life and, most importantly, better compliance with treatment (Veldtman et al., 2000). For children this understanding is contingent on a number of factors such as developmental levels and illness experience. It is a widely replicated and supported view that children's perceptions of illness vary in both content and depth, particularly as a child's understanding of their illness cognitions appear to be linked to their developmental level (Bibace and Walsh, 1980). Thus, attributions that children make about their own illness may be completely different from that of an adult or parent.For example, in the obesity/overweight literature, it has been found that adults understand a wide number of possible causes for obesity. These include junk food advertising during children's television viewing, genetic and endocrine factors, environmental factors, imbalance between energy taken into the body and energy expended, as well as increased calorie intake through diet and sedentary lifestyles (Saelens and Daniels, 2003). In an Australian study of lay persons' understanding of childhood obesity, Hardus et al. (2003) found that the public had similar sophisticated perceptions of the causes and preventions of obesity. Over half of the adults surveyed endorsed a view that media promotion of unhealthy foods and over consumption of fast food were the main contributors to childhood obesity.Despite this, it appears that the only aspects of children's views of obesity that have been investigated are perceived causes In a rare study examining childhood conceptions of the cause of obesity, Johnson et al. (1994) found that elementary school-aged children (in grades 1, 3 and 5) lacked much factual knowledge and held many misconceptions. For example, although children could identify "junk" foods as a cause, their definition of junk food included items like bread, pasta, potatoes, bananas and chicken. Many were confused about how fat, meat, salt and cholesterol led to obesity. In fact some thought that cholesterol was a type of food. Only 7 percent identified lack of exercise as a cause of obesity though many had no idea how exercise kept one from getting fat, or how a sedentary lifestyle could make a person fat. Although 4 percent mentioned calories as a cause of obesity, children did not know what calories were. Other causes of obesity identified were drinking too much milk or water and mother to unborn child transmission (Johnson et al., 1994). Thus, factual knowledge seemed to be limited, with children not fully comprehending the content of information, or how certain factors work to contribute to obesity.With regards to examining childhood conceptions of disease, an accepted, holistic and current framework is the common-sense model of illness representation (CSM; Leventhal et al., 1980). The CSM is primarily designed as an explanatory framework for the way people perceive and react to disease threat (Leventhal et al., 2003; Conner and Norman, 2005). It is hypothesised that these "lay views" of illness are formed along five dimensions, or interpretive schemas (Hagger and Orbell, 2003). The five dimensions of the CSM of illness representations are:1. Identity. Individual statements, beliefs, and labels for the illness. Leventhal et al. (2003) posit that this domain is critical as it joins feelings of vulnerability to general symptoms.2. Cause. Perceived causative agents of the disease, which may include genetics, lifestyles or infections.3. Timeline. The perceived duration of the illness.4. Consequences. The perceived impact that the illness has on overall lifestyle.5. Control/cure: Refers to perceived efficacy of treatment (Hagger and Orbell, 2003) as well as perceptions of how preventable, controllable or curable a condition is (Leventhal et al. 2003).Empirically, the CSM has been validated as being applicable to many illnesses, as it was developed primarily to apply to the health field (Leventhal et al., 2003). When applied to children, the CSM forms a good alternative to the Piagetian framework which is frequently used in the study of children's perceptions of illness (Eiser, 1989). First, the CSM addresses more domains of illness than simply causative factors, a criticism made of applications of the Piagetian model (Eiser, 1989). Secondly, the CSM allows for these five schema structures to be reshaped and develop in breadth as new information is gained from experience and other sources (Leventhal et al., 2003), a notion especially pertinent to children as they age and become more sophisticated in their cognitions.Empirically, previous studies have found that the CSM can be used to effectively describe children's perceptions of illnesses. For example, Goldman et al. (1991) reported that responses given by children aged four to six about the common cold tapped into the five domains articulated by the CSM. In addition, Paterson et al. (1999) also discovered that seven to 14 year olds' discussions on the common cold and asthma could also be conceptualised in terms of the illness representations model. The aim of the current study is to examine children's understandings of obesity. The current study aims to explore childhood perceptions of the identity, cause, timeline, consequences and cures/control of obesity. The current study recruited a total number of 33 children. Participants were as follows: 12 children aged seven to eight (36.4 percent), 13 were aged nine to ten (39.4 percent) and eight were 11-12 (24.2 percent). Age groups correspond to primary school grades 4, 5 and 6 respectively and are the group for which childhood obesity has shown the greatest increases over time (Behn and Ur, 2006; Wang and Lobstein, 2006). There were 20 boys in the sample. The final sample represented nine overweight/obese children (27.3 percent) and 24 normal weight children (62.7 percent). Children were recruited through the Catholic Schools system as well as a sample who were seeking appetite awareness training in Sydney, Australia. Children in this latter option were included in the study on the basis if their body mass indices fell within the childhood "overweight" or "obese" range defined by Cole et al. (2000). Children were interviewed using a semi-structured interview protocol which was designed to reflect the five dimensions of the CSM. Ethics approval for this study was granted by the University Human Ethics committee as well as the CEO of an Archdiocese of Catholic Schools in Sydney, Australia.The final sample is shown in graphical form in Figure 1.3.1 Materials The interview protocol was based on a proforma by Deborah Finney (Finney, 2003; personal communication) used in previous research on children's understanding of mental illness. The interview contained questions and tasks pertaining to the five areas of the CSM (Table I)3.2 Procedure After obtaining written permission from their parents to partake in the interview, children were interviewed in a quiet room at their school. Questioning followed the same order for each child, however, the interview was semi-structured as probing was needed for some children in order to elaborate their responses (for example, "What might happen to someone if they get 'sick'") and clarify some responses. Responses were audiotaped for purposes of clarity, and also to allow for the development of rapport with the child. Tapes were then transcribed by an independent agency.Direct questioning was performed to obtain responses pertaining to the various dimensions (e.g. how do you think some kids get to be obese?). The specific questions asked are shown in Table I. To ascertain features of labels, beliefs and statements of obesity, as relating to the domain of "identity", children were asked to draw a picture of someone they considered to be obese and a picture of themselves. They were then asked to state the differences between themselves and the "obese" picture to gain these beliefs and statements relating to obesity.3.3 Data analysis Qualitative content analysis was used to analyse the data obtained. Responses were coded according to the CSM dimensions - thus any section of text relating to any established dimension was included (Priest et al., 2002). To illustrate the type of responses which would have aptly fitted into a CSM dimensions, quotes from respondents are used. 4.1 Identity In the domain of "identity" of obesity, it was found that 6/9 overweight/obese children, and 9/24 of normal weight children identified a large stomach as a prominent feature of obesity. Large limbs and extremities were also a standout feature in 5/24 of the normal weight drawings and 3/9 overweight/obese drawings.Some further shared characteristics included:1. Different hair or lack of hair for 4/24 normal weight children and 1/9 of the overweight group, as shown below: Kevin[1] (normal weight): "Well, I've got different hair...I don't have my hair tied up in one of those things [hair net as in a sumo wrestler]"2. Perceptions of differences in clothes were present in 7/24 normal weight children's drawings and 1/9 overweight child's drawing. In this case, the "obese" drawing was often wearing no clothes or clothes seen as less fashionable, as seen below: Kevin (normal weight):"I have much better clothes on".Rachel (normal weight): "If I drew myself with a skirt on, I wouldn't have this [skin] pushed up and showing".One feature of the "obese" drawing found in 3/24 normal weight children and 1/9 overweight children was that the figure was drawn consuming junk food such as chocolate and lollies, for example: Matthew (normal weight): "He's eating chocolate"In addition, it was found that some normal weight children only drew their "obese" drawings as negative characters such as, the "Blueberry" from Charlie and the Chocolate Factory and Santa Claus.Examination of the obese drawings seemed to contain numerous negative features when compared to the child's self-drawing. This was generally found to be unanimous between both weight groups. Negative features of a physical nature found in the obese drawing (as opposed to the self drawing) included figures having no smile and figures wearing glasses.4.2 Cause With regards to causes of obesity, it was found that the majority of children were knowledgeable and correctly identified three major causes of obesity. These included:* The intake of junk food, cited by 15/24 of the normal weight group and 7/9 of the overweight children) Examples of such responses include:* Alan (overweight): "They just go and eat lots of McDonald's and stuff. Don't eat salad, healthy food".Chiara (normal weight): "Maybe they just don't get enough nice and healthy food".* Overeating, or eating too much food was also popularly endorsed as a cause of obesity with 7/24 normal weight and 2/9 overweight children making references such as these: Aaron (overweight): "Eat too much and snack all day ... need to break the habit of snacking 24/7".Joe (normal weight): "They eat way too much".* Lastly, decreases in exercise/activity levels were also commonly cited as a cause of obesity by 5/9 of the overweight/obese sample, and 13/24 of the normal weight sample. For example: Miranda (normal weight): "They might not exercise".Emma (overweight): "They can never like sport and they're not really sporty".It was of interest that only 3/9 of the overweight group and 6/24 of the normal weight group endorsed the engagement in sedentary activities as a cause of obesity, in the following ways: Aaron (overweight): "And also because they stay in front of the TV and computer all day and not exercise ... they just stay in the house all day really".Jared (normal weight): "Watching TV and not doing anything active".The analyses of responses also found that normal weight children solely identified a further two causes of obesity. These were:1. Attributions to parental behaviour, cited by 3/24 normal weight children, for example: Sally: "They're just let go by their parents ... their parents just say to them 'Oh don't worry about it, you can have whatever you want (of junk food) ... so that they don't know how to stop'".2. Familial transmission, as identified by 2/24 normal weight children in responses such as these: Carrie (normal weight): "Sometimes, their parents are large, so they're born with it".4.3 Timeline All children interviewed stated that duration of obesity was dependent on one's actions towards positive health behaviours. 4/9 overweight children and 18/24 normal weight children specified these positive health behaviours as eating healthily and engaging in exercise; for example:Brooke (normal weight): "If they don't start to exercise and eat healthy food, they will probably stay obese".Eric (overweight): "If they didn't get healthy soon, they'd stay obese".Of interest were some responses which reflected that duration of obesity was dependent on a cognitive change. In this case, 3/9 overweight / obese children, and 4/24 normal weight children interviewed made some reference to individual motivation in making a decision to lose weight, and the resisting of negative heath behaviours, as salient mediating features in the duration of obesity. For example:Alan (overweight): " ... depends if they get motivated to fitness or if they want to sit at home and have Macca's everyday".Robbie (normal weight): "If they don't try to stop being obese, they will (stay obese forever)".Children also perceived that an individual would remain obese unless they took part in exercise (3/9 overweight and 13/24 normal weight children), for example:Oliver (overweight): "I think they can exercise more and they can burn off all that overweight, like all that fat".Joe (normal weight): "Um, it depends if they started exercising more or something".4.4 Consequences In terms of immediate consequences, 3/9 of the overweight/obese children mentioned that one consequence of obesity was the person becoming more obese, for example:Brady (overweight): "um, they might get bigger".For the normal weight group, 4/24 children mentioned this as a consequence.Another popularly endorsed consequence was restricted mobility (3/9 for the overweight group and 11/24 of the normal weight group), for example:Aaron (overweight): "And also ... when you get obese, it's hard for you to, really hard for you to move I think".Vanessa (normal weight): "Sometimes, they can't move quickly".Further consequences to obesity as endorsed by the group can be viewed in Table II.In addition to the physical consequences of obesity, children were also asked about any possible social or emotional consequences linked to this condition. It was found that a common emotional consequence between both groups was the notion that obese children and adults would feel unhappy, or angry with themselves for remaining obese. 3/9 overweight and 4/24 normal weight children cited this consequence. For example:Emma (overweight): "They might be unhappy with themselves and get really stressed out".Rachel (normal weight): "Angry with themselves...like, how could I let this happen?".Also mentioned were that others would not like them as a person (3/9 overweight children and 6/24 normal weight children):Rachel (normal weight): "No one would want to be their friend".Paulo (overweight): "Um, well, for one, people might not like them".Half of normal weight children (12/24) endorsed the idea that obese people would get teased. This response was also endorsed by 3/9 overweight children for example:Charlotte (normal weight): "Because some people can tease them for being fat and being big".Emma (overweight): "Um, they might face teasing when they get older".Five children (four normal and one overweight) mentioned that obese children would be socially left out of activities, e.g.Samantha (normal weight): "They wouldn't get included in anything".Miranda (normal weight): "(People) might not want to play with him or her and then they'll have no friends".Lastly, five children spoke of sadness as an emotional consequence to obesity, four of these came from normal weight respondents, e.g.James (normal weight): "Probably make them really sad".Matthew (normal weight): "They might get sad because they get teased".Jason (overweight): "They might get depressed because no one likes them because they're big".4.5 Cure/control In the final dimension of Leventhal et al.'s (1980) common sense model of illness representation, the children's responses for cure or control of obesity were fairly homogeneous with 7/9 overweight/obese children and 23/24 normal weight children citing a need to exercise as a potent cure for obesity, e.g.Aaron (overweight): "The only way to get thin is to exercise really regularly and burn off the fat and calories and stuff".Brady (overweight): " ... exercise quite a bit and get to playing sports ..."Kevin (normal weight): "Go to the gym, like a couple of times a day for a week".The need to eat well, or decrease eating of junk food was also perceived as a cure for obesity by 6/9 overweight/obese children and 18/24 normal weight children, for example:Brady (overweight): " ... eat healthy fruit, food, veggies, not fried...deep fried chips".Emma (overweight): " ... eat some fruit and healthy things".Caroline (normal weight): " ... just eat your vegetables and fruit ... and you could stop eating so many calories".When questioned about the efficacy of such cures for obesity, 6/9 of the overweight and 23/24 of the normal weight children stated that these measures definitely worked if one implemented them. The normal weight children referred to their own weight as evidence that these methods worked, for example:Luke (normal weight): "Well that's what I do and I'm not fat".Caroline (normal weight): "Well, it works because my family rarely ever gets Macker's".This was interpreted as reflecting children's perceptions of the cures of obesity as being one's own responsibility as for the other categories. It was of further interest that exercise in particular was viewed as a powerful cure of obesity, whereas the causal account between obesity and a lack of exercise was not as strong. In the current study, children of primary school age were interviewed on their views of obesity as guided by the five dimensions of Leventhal et al.'s (1980) common sense model of illness representations. In general, it was found that children were quite knowledgeable in all areas.The most obvious identity feature of obesity depicted by the children was a large stomach. If the negative connotations of stereotyping are ignored, this is encouraging as children may be referring to a large waist circumference, which is associated with the most health complications in obesity (Wardle et al., 2008). However, the other identity features of obesity listed by the children were also primarily negative in nature. Such findings are consistent with previous research that have found negative societal attitudes of overweight people found in adults (Bell and Morgan, 2000; Pingitore et al., 1994; King et al., 2006; Hebl and Xu, 2001), and children (Hill and Silver, 1995; Staffieri, 1967).Children correctly identified consumption of junk food, overeating and a non-engagement in activity and exercise as prominent causes of obesity, reminiscent of findings by Eiser et al. (1983) whereby children were able to identify being "healthy" as exercising, having a balanced diet and being energetic in general.Almost half of all children interviewed did not mention sedentary behaviour as a cause of obesity. A similar finding was reported by Johnson et al. (1994). This finding has also been reflected in a recent study conducted by the NSW Schools Physical and Nutrition Survey (SPANS) (2004) (see Booth et al., 2006). The study reported that despite there being a recent increase in school-aged children fulfilling recommended exercise requirements over the year 2004-2005, there was also a corresponding increase in children undertaking sedentary activities such as television, videos/DVDs, and electronic and computer games (Booth et al., 2006). This generally poor understanding of sedentary behaviours is of particular interest as sedentary living is one of the major tenets in recent media campaigns by the NSW government to decrease the childhood obesity rate. Perhaps such interventions need to directly address how sedentary behaviours lead to obesity, and encourage children to identify their own patterns of sedentary behaviours.The majority of children in the current sample stated that the timeline or duration of obesity was reliant on people undertaking positive health behaviours. This could also be interpreted as reflective of an internal locus of control whereby the person is seen as being responsible for their own weight loss. As such, it may be that children believe that by either not engaging in these positive behaviours, or, alternatively having failed in weight loss despite these behaviours, that people may have "done the wrong thing". On the basis of their understanding of the causes of obesity, it appears that children seem to believe that the duration is short if people take appropriate action. This internal attribution to weight loss may mean that they fail to capture the multifactorial complexity of the issue.It was noted that normal weight children were more adept than their overweight counterparts at detecting the most severe consequences of obesity. First, it may be possible that the parents of overweight children protect them from such negative consequences, as it is understandable that parents would aim to minimise distress in their children. Second, it may also be that overweight children and their parents underestimate the true risks of obesity, as found by Jeffery et al. (2005).Lastly, the cures/control section of the interview yielded the finding that most children endorsed the idea that exercise was a potent cure for obesity. This finding may be seen as reflecting a general sense of an internal locus of control - in that it is seen as a person's own responsibility to carry out this behaviour to lose weight, and that exercise, being a vigorous, visible activity, is the most obvious behaviour to contribute to this. However, this is also limiting in that exercise is not the only component contributing to healthy lifestyles, and needs to be undertaken in conjunction with other positive health behaviours. Hence, the multiple components of obesity and overweight also seem to be underappreciated by children in the current study. Further interventions should make clear the multifactorial complexity of obesity to children, and further highlight the nature of sedentary behaviours in maintaining this condition.The current study was limited in that most of the sample came from the Catholic education system in Sydney's eastern suburbs, which are more affluent in terms of socioeconomic status and therefore impacts on the study's external validity. Nonetheless, recent figures from the Australian Bureau of Statistics (www.abs.gov.au) indicate that obesity and overweight actually tends to be more prevalent in low income households and living areas of greatest disadvantage. Future research in this area should therefore aim to recruit participants from areas of lower socio-economic status. In addition, a broader sample could also be obtained from public education settings. On a related note, the treatment-seeking sample in the current study also presents with limitations in the sense that results may be biased and not representative of the wider population. With this respect, wider sampling in a number of areas would be a further avenue to consider in future studies.In conclusion, this study has qualitatively examined children's views on facets of obesity as pertinent to the common sense model of illness representations. Results found that children do indeed have well developed perceptions of the identity, cause, timeline, consequences and control/cure of obesity but which nevertheless have some qualitatively different deviations from "expert" or "adult" understandings of these facets. Future obesity prevention interventions should take these childhood perceptions into account when considering improved adherence to regimes. Opens in a new window.Figure 1 Final sample of current study stratified in terms of age, gender and weight Opens in a new window.Table I CSM domain examined and examples of questions asked in each domain Opens in a new window.Table II Normal weight and overweight/obese responses to the consequence of obesity
|
- To the authors' knowledge, no previous study has examined children's perceptions of obesity beyond perceived causes.
|
[SECTION: Purpose] Despite the ongoing increase of women in the top positions of hierarchy, they continue to be underrepresented in politics occupying 19.5 percent of seats worldwide, 22.8 percent in Europe, 22.6 percent in the Americas and 42.0 percent in Nordic countries (Inter-Parliamentary Union, 2017). In 2017, Italy ranks 44th out of 193 in the world classification compiled by the Inter-Parliamentary Union (2017), with a participation rate of Italian women to res public between 28 and 31 percent. Although the number of women elected in central institutions is increased compared to the past, in Italy women are still low represented in politics (Massa, 2013), and the number of women politicians decreases in the hierarchy of power (Fornengo and Guadagnini, 1999; Bonomi et al., 2013). Every president of the Italian Republic has been a man and women who have served in the institutional role of president of the Chamber of the Italian Republic from 1948 to date are 3 out of 14 (Italian Parliament, 2017). This position of women in politics, characterized by both a numerical increase compared to the past and a numerical decrease at the top of the power hierarchy, is well represented by the current government. The XVII Legislature of government, which started in 2013 and close to conclusion, is the legislature with the largest number of women in Parliament, in fact the women account for 32 percent of deputies and 30 percent of senators, but only 16 percent of women cover the roles of leader, president of the commission and the office of the presidency. Analyzing the composition of the Italian Parliament, it is clear that the disparity between men and women in politics is recognizable when roles are more prestigious. This condition is also confirmed at the regional level: women are present in the joint sessions in 29 percent of the councils and hold key positions (departments and offices with greater decision-making power) in 18 percent of the councils, while only 10 percent is president of the region and only 2 percent are leaders of the municipalities in the capital city (Openpolis, 2015). The Italian case is particularly interesting, as Italy is strongly characterized by conditions of gender inequality. In addition to the gender gap in political participation, as evidenced by the above data, in Italy the level of women's participation in the paid labor market is still very low (ISTAT, 2016): 54.4 percent compared with an OECD average of 62.6 percent (OECD, 2015). Although Italian women have a higher level of education than that of Italian men (ISTAT, 2016), gender segregation persists. Working women are concentrated in the tertiary sector and low-skilled professions (ISTAT, 2015b), with only 35 percent holding a managerial position. In this regard, Italy is ranked 82nd out of 142 according to the World Economic Forum's classification (WEF, 2014). In addition, in Italy the division of family work remains drastically imbalanced to women's disadvantage (Eurostat, 2011; ISTAT, 2015a; CNEL, 2014). Research shows that the time Italian mothers dedicate to family work is much higher than that dedicated by fathers, even in dual-earner families (Eurostat, 2011; ISTAT, 2015a). This strict division of family roles is the main obstacle to women reaching decision-making positions in politics, because Italy also continues to invest inadequately in services to support families (Panzeri and Viale, 2016). In Italy, maternity continues to have a significant impact on unemployment rates and the early exit of women from the labor market (CNEL, 2014). In Italy, therefore, the underrepresentation and low appreciation of women in society and politics together constitute a serious problem whose causes it is vital to investigate in order to intervene adequately (Lasio et al., 2013; Pacilli et al., 2012). Despite the substantial social and cultural transformations of men and women roles, women continue to be excluded or impeded in highly competitive and masculine hierarchical domains, such as politics (Lazar, 2005). Men continue to be the majority in power contexts that appear inaccessible and strewn with obstacles for women to grow in (Fox and Lawless, 2011; Paxton and Hughes, 2015). Although women have achieved full juridical parity and participate in the political life of the country, conditions of strong inequality still persist and limit their growth in contexts that are considered places of power (Davidson and Burke, 2016; Lazar, 2017). Several studies have investigated women in leadership (e.g. Fitzsimmons et al., 2014; Mavin and Grandy, 2016) and explored persistent systems of discrimination and inequality (e.g. Kelan, 2008; Lewis et al., 2017). The matter of women in politics continues to represent the context in which a gendered division of power has its highest expression and constitutes a macroscopic disadvantage in terms of "presence" and in the acquisition of roles (Childs, 2006). Many pieces of evidence attest the existence of strong constraints to full and complete participation women's policy. Politics is one area where male domination over women and resistance to change are fully expressed (Bourdieu, 1998). Politics tends to reproduce male models and rules presenting itself as an unwelcoming domain for women. Women then, although they can rely on the same opportunity as men, are blocked by systems of exclusion rooted in the culture of many societies (Inglehart and Norris, 2003; Krook, 2010) and very often exclude themselves. Recent research (Lasio and Serri, 2017; Lasio et al., 2018) has also showed how the resistance opposed in Italy to non heterosexual families reinforces normative standards on motherhood, thus contributing to maintain those asymmetries at the origin of women's invisibility as active citizens (Amancio and Oliveira, 2006). Women's underrepresentation in politics reproduces the imbalanced and unequal gender relations (Lovenduski, 2001; Santos et al., 2013). Since the 1990s, some scholars, mostly women from Anglo-Saxon and Northern European countries (Childs and Krook, 2006; Krook, 2010), have begun to investigate the relationship between women and politics, focusing especially on the numerical presence of women in the government and in the councils, on the trends and mechanisms of recruitment and selection of women by the parties (Kenny, 2007; Celis et al., 2015; Kenny and Verge, 2015). These studies found that women continue to be underrepresented in politics and despite there being a much greater number of women in elected positions compared to the past, this it is not a substantive representation and cannot be called substantive equality (Childs and Krook, 2006). The literature on this topic has addressed the potential factors that may explain this underrepresentation, and has highlighted the influence of patriarchal culture, the division of family roles, the lack of childcare, the resisting attitude of the parties and the fragmentation of women's movement (Lovenduski and Guadagnini, 2010; Bonomi et al., 2013). Others have analyzed the relationship between discourse, gender and political representation, focusing on political discourse and the construct of gender identity (Shaw, 2000; Walsh, 2001) and gender citizenship in politics (Poggio, 2004; Belle, 2012). This study is positioned in the field that investigates the relationship between women and politics using a discursive perspective. The research was conducted within a theoretical framework referred to as a discourse analysis (Potter and Wetherell, 1987), and to contribution from the critical feminist perspective (Lazar, 2007, 2017; Holmes and Marra, 2010), and proposes to understand the mechanisms that regulate the contemporary gender order in politics through the analysis of the discourses of a group of Italian politicians. The focus on discourses of politicians allows us to analyze the processes by which dominant symbolic gender order is reconstructed in politics and to show how the distribution of power between gender is acted out through practices of hegemonic masculinity which excludes women from the real seats of power. This study offers a reflection on gender inequalities in political participation, starting with a review of the literature and the results of research conducted on Italian politicians. The contribution of this paper is to shed light on continuing discourses concerning politics as a predominantly male context. We conclude by highlighting the study's theoretical and practical implications for enhancing equality in politics and further developing research in this area. The studies on women and politics continue to focus on women' numbers in politics and on sex differences, and with difficulty, they are able to embrace the concept of gender (Kenny, 2007) as a category that structures and makes sense of particular social practices. However, feminist studies have developed complex theories of gender, including both patriarchal and discursive conceptions of power. The power inequality between gender relies on the concept of patriarchy, a system of social structures and practices of domination founded in the subordination of women by men (Walby, 1990). Gender relations are power relations involving formal public structures as politics and private structures as the family. The distribution of power between gender, based on hegemonic patriarchal culture, which associates naturally the female to the house and the male to the labor and political activities, has determined an asymmetry in which the historical forced absence of women from prestigious positions, such as politics, would involve a vicious circle whereby the greater female participation in domestic work has become cause and consequence of their exclusion (Walby, 1990). Through the internalization of gendered norms produced routinely in discourses of everyday life, this gap of power between men and women has become invisible, misrecognized and recognized as legitimate and natural (Bourdieu, 1991), contributing to the consolidation of "hegemonic masculinity" that preserves, legitimizes and naturalizes men power and, consequently, women subordination (Connell, 1987, 2016; Connell and Messerschmidt, 2005). Discourse, therefore, plays a significant role in the reproduction of dominance, which is considered as an exercise of social power by elites, institutions or groups that have the effect of increasing social, political, class and gender inequality (van Dijk, 1993, 2011). Social power is enacted, reproduced or legitimized by the discourse of dominant groups or institutions (van Dijk, 1996). Relations of power and dominance (Bourdieu, 1991) can discursively challenge the interests at stake. Therefore, hegemonic power relations are discursively produced, perpetuated, negotiated and challenged. To share or refuse these contents means to contribute to the reproduction and maintenance of the social order and also to resist and transform that order (Lazar, 2005, 2017). Gender is socially constructed through discourses (Wodak, 1997), and the power relations associated to, and reproduced by, such discourses (Wodak, 1997). Gender expresses the relationships of power that constitute society: women and men, through their performances, can reify or on the contrary challenge the system of relations and gender privileges (Holmes and Marra, 2010). Feminist critical discourse perspective offers important contribution to theorize and analyze "the particularly insidious and oppressive nature of gender as an omni-relevant category in most social practices" (Lazar, 2007, p. 3). In particular, feminist critical discourse analysis aims to demystify the interrelationships of gender and power, in discourse and critiques the discourses that sustain a patriarchal social order, that systematically privilege men and disadvantage women (Holmes and Marra, 2010). The disclosing of the discursive strategies through which power structures and supports oppressive social relations is a form of "analytical resistance" that help to support the struggles of protest and social change (van Dijk, 1991). Feminist critical discourse contributes to social transformation through the focus on repressive norms and restrictive stereotypes reified and reinforced by routine discourses of people (Holmes and Marra, 2010). This study is based on 30 biographical interviews conducted with local Italian politicians, 15 of whom were men, 15 women. We chose to involve local politicians because figures on the number of women in local Italian institutions are lower than those at the national level (ISTAT, 2017). The average age of the politicians interviewed was 43 years (DS=12.5; range=24-64). The research was carried out in Italy and recruitment took place through the involvement of political parties. The face-to-face interviews were conducted by one of the authors, lasted between 60 and 90 min, and focused on two central topics: their experiences in politics and meaningful episodes in their political life, and the barriers that may hinder women's political participation. Respondents were asked to illustrate their political career starting from the reasons that led them to choose politics and to discuss the relationship between politics and gender, taking into consideration their own experiences and reflecting on the meanings attributed to the experiences. The study was carried out in accordance with the Declaration of Helsinki. Participants were informed of the aims, methods and other ethical aspects relevant to the study. After they had given informed consent, all interviews were digitally recorded and transcribed verbatim. Participants' confidentiality and anonymity have been protected through the use of pseudonyms. For the first step of the analysis, the authors independently examined the transcripts to identify the emerging main themes and recurrent repertoires. Subsequently, the researchers worked together on the discourse analysis (Billig, 1987; Potter and Wetherell, 1987), in order to come to an agreement on the identified interpretative repertoires and chosen illustrative key quotes drawn from the different interviews. Particular attention is given to discourse, as social action through which men and women can reify or challenge the system of power relations, gender imbalances and privileged places in which these mechanisms can be observed and unmasked (Holmes and Marra, 2010). Data analysis focused on the interpretative repertoires and ideological dilemmas. The repertoires are "clusters of terms organized around a central metaphor, often used with grammatical regularity" (Wiggins and Potter, 2008, p. 74) not defined by specific words or phrases, but by a textual constellation that refers to a specific theme (Potter and Wetherell, 1987). The ideological dilemmas (Billig et al., 1988) indicate the intrinsic contradictory nature of common sense and highlight the rhetorical origin of attitudes toward gender as they emerge in contexts where there is an argument. "[...] when contradictory positions overlap they provide a basis for awareness and reflexivity, just as they lead to problematization and change" (Fairclough, 1995, p. 82). Data analysis show how the dominant gender order is performed and reinforced through the discursive practices of the politicians interviewed. The participants tell their experience as politicians and produce discourses that clarify the norms that regulate the political field. These discourses are part of a wider social context in which choices, roles and careers are gendered. The politicians interviewed offer several arguments about women' low participation in politics, referring to different degrees of responsibility or states of helplessness. Results show the participants' tendency to either absolve or blame women for gender inequality in politics through different interpretive repertoires, which can be, in some cases, dilemmatics (Billig et al., 1988). In particular, the interpretative repertoire "Women's disinterest toward politics" tends to blame women for their low participation in politics, whilst the interpretative repertoire "Politics as masculine context" tends to absolve women and to attribute the responsibility of the gender gap in politics to the hegemonic social context and to the attitude of the parties. A third repertoire "Politics-family unbalance" identifies in the unequal distribution of family work the principal responsibility of gender inequalities in politics. The arguments referred to the latter repertoire concern the debate between nature and culture, fluctuating between the essentialist positions on the one hand, and progressive positions, on the other hand. Women's disinterest toward politics Though among the politicians interviewed the idea that men and women can perform the same roles in politics is common, they explain the women's low representation in politics referring to the lack of interest of women toward politics, the women voluntary abstention and their minor will to be involved (De Simone et al., 2017). The issue of women's lack of interest is built on beliefs that recall individual factors (Fornengo and Guadagnini, 1999) as "a choice for women" dictated by a lack of desire to engage. This choice would lead women to voluntarily keep away from the places of res public:Marco (39 years): Women can play the same role in politics, whatever role males can play; there is no difference. It is a choice for women who do not participate in politics. They are not interested. [...] because there is no woman that enforce herself! Marco does not display explicit conflicts that were unfavorable to full equality between men and women in politics, however affirming the claim of women's disinterest in politics, supports the counterarguments that identify in women the cause of the impossibility to put such equality into practice. The argumentation relying on women's responsibility for the gender gap in politics becomes even stronger when greater capacity and greater potential is attributed to women themselves:Giovanni (41 years): women voluntarily choose not to do politics [...] when they are brought into play, they have produced good results just as good or better than others. I think that women have all the possibilities and conditions for doing politics, and even in the best way. The notion of choice is another key element to the understanding of the gender gap in politics. The reference to women's disinterest in politics being a reflection of their own voluntary choice implies a lack of recognition of the social inequalities affecting women's access to politics. Reference to women having free choice moves attention away from the responsibility of the system to the responsibilities of the individual woman, indicating that women's choices are embedded within cultural, social and economic conditions that structure gender dynamics (Lewis et al., 2017). The quotations above show as the power, deeply ingrained, acts automatically and unconsciously, since is foregone and "buried" under consolidated ideologies (Lukes, 2005). Marco and Giovanni's words show how the gender order is preserved through the third dimension of power that prevents some issues from being expressed. Gender conflict is latent because the subjects holding positions of power, in this case two men politicians, do not explicitly express their interests. The ideologies influence people's wishes and thoughts, the "ideological power," the "real power," acts indirectly and unconsciously, influencing the social actors and ensures, in this way, their acquiescence, even when it goes against their self-interest (Lukes, 2005). Indeed Valeria (25 years), says:Women themselves are afraid of not being up. It depends very much on the lack of interest or fear to get involved because, in my opinion, we are the cause of the prejudice! I speak often of meritocracy, I hate rose quotas! There is no need for rose quotas. Valeria affirms that women have not the need for rose quotas. The rose quotas refer to those legal instruments aimed to protect gender equality within the representative bodies through different calculation mechanisms and allocation of seats. Valeria, with her statements, contributes to maintaining the status quo, by underplaying and silencing structural inequalities. Once again, it is not the politics to reiterate a gender order that does not allow women's participation, but women themselves are responsible for their own exclusion. For many respondents, including women, and as it is clear from Valeria's discourse, affirmative action promoted in favor of gender equity in political participation is perceived as offensive and not qualifying (Santos et al., 2016). Valeria denies gender inequalities, thereby oppositions are avoided, whereas asymmetric power relations between gender persist. The issue of voluntary women's abstention from politics justifies persistent power inequalities as the result of women's individual choices and underlies the non-recognition of gender social inequalities (Lewis, 2014). Marco, Giovanni and Valeria's extracts emphasize the individual responsibilities with the effect of concealing structural discrimination mechanisms and the inequality of power between men and women. Through those discourses seem to emerge a lack of awareness of the mechanisms of male dominance that push women, without their awareness, to adapt their hopes and their aspirations to opportunities, and consequently not to undertake actions that are not expected of them (Bourdieu, 1998). The order of things, in fact, is not a natural order, but a building of the world with which man pleases his thirst for domination, a vision so exclusive that the same women who are the victims have integrated in their own way of thinking by accepting their unconscious inferiority (Bourdieu, 1998). Valeria, as a woman, can be considered victims and author of gender relations of power acting in social practices upholding hegemonic culture. The hegemonic masculinity, as a pattern of practices that permit men's dominance over women (Connell and Messerschmidt, 2005), inspires discourses that aim to represent a reality of gender equality. Marco, Giovanni and Valeria's discourses lack an acknowledgment of this systematic imbalance referred to the rhetoric of sameness between gender and denial of inequalities. The hegemonic masculinity (Connell and Messerschmidt, 2005) inspires the discourses of equality that deny gender inequalities slowing social change (De Simone et al., 2015; De Simone and Scano, 2017). Politics as masculine context Interviewed politicians' discourses identify the cause of the women's low representation in politics in the sexism that characterizes the parties and politics in general, using the interpretive repertoire "Politics as masculine context." Women's access to some areas continues to be difficult and to represent a source of inequality especially in the "real" places of power (Lazar, 2005), as politics. Several interviewees emphasize the persistence of the gender gap in politics highlighting that despite the number of women in politics has increased compared to the past, high positions continue to be held predominantly by men reiterating a gender order that relegates women to subordinate positions (Connell, 1987, 2016):Cristina (53 years): Certain roads were blocked to women. There is a social structure that does not allow access to women in politics. Since it is been a long time, and now it seems that everything is established a norm. Women, despite the battles, block themselves because this is the social structure which they know and consequently everything remains unchanged. Cristina's words recount as the politics represents a system of power relations aimed at gaining control of resources and opportunities of majority group over the minority group (van Dijk, 2001). According to respondents also the attitude of the parties that have opposed resistance and have offered few opportunities for women helps to explain the low representation of women in politics (Guadagnini, 2005; Yirmibesoglu, 2008; Lovenduski and Guadagnini, 2010; Belle, 2012; De Simone et al., 2017). About this Sara, 60 years, says:I have encountered many difficulties in my political career. The political parties are governed by men, who support other men. Women do not vote for women because there are no women on the list and if they are there, they are already put in the last places, and they will never be elected. The parties are a male lobby. Sara and other respondents attribute the barriers encountered by women in politics to the discriminatory practices carried out by the parties (Lovenduski and Guadagnini, 2010). Sara describes the difficulties encountered in the party and the sense of exclusion generated by belonging to a minority. Sara's words perfectly convey the idea of the dynamics that develop in power institutions, always male dominance, helping to define the segregating phenomena of women. The lower representation of women in political institutions is explained by reference to the selection and recruitment methods in force in the various political systems and implemented by the various parties (Fornengo and Guadagnini, 1999; Belle, 2012; De Simone et al., 2017). The construction of male-dominated lists is fostered by party logics that build blocked lists where women are at the end of the list, without any chance of affirmation, according to the political gatekeeping phenomenon (Bain and Cummings, 2000). In this way, politics tends to exclude women by proposing predominantly male patterns and discriminatory practices (Annesley and Gains, 2010). Carlo, 45 years, explains the mechanisms of political parties that discourage women participation (Fox and Lawless, 2011) and contribute to the consolidation of hegemonic masculinity (Connell and Messerschmidt, 2005):In political parties, women are called by the secretaries, who are mostly men, in the election period. But after the election, when women did not take votes, they are no longer considered and they feel cheated. So, they say: do not go back there again." The existence of the glass ceiling (Morrison et al., 1992) in politics represents an invisible barrier that prevents women participation in politics and, on the contrary, privileges men. The prevalence of votes of men candidates becomes an obvious, logical, natural phenomenon, as told also by Giacomo, 38 years:There is a greater tendency by most people to vote for men, the system is structured in a sexist way. First, since women are less, there are ten women and ninety men; clearly, it is easier for a man to achieve, than for a woman. Second, a man wants to support himself, so he wants to keep that power and a woman who arrives so high would compromise the rules. Giacomo exemplifies as the vote reifies the dominant gender order according to specific logic of power. His discourse operates for the naturalization of a gender order socially accepted and shared, which mystifies the diversity in the distribution of power and inequality between women and men (Lazar, 2005, 2017). Politics-family unbalance The underrepresentation of women in politics is linked to family-work life balance (De Simone et al., 2018) as women's responsibility (Lovenduski and Guadagnini, 2010; De Simone et al., 2017). Some interviewees assert that women would not be interested in doing politics because they organize their lives around the family. For example, Giovanni, 38 years, say:I also understand that for a woman who decides to have a family and children it is more complicated because family work falls more on the woman than on the man. Women are interested in organising their lives around the family. Women should not be pretentious and have unrealistic demands toward politics, they must organize themselves with family and adapt to the rules of politics. Giovanni's extract describes a rigid division of gender roles and a strong unbalance in the distribution of family work as a disadvantage for women because of the limitation of time and resources for professional and political affirmation (Lazar, 2005; Lasio et al., 2017) and as a cause of the gender gap in politics (Guadagnini, 2005). This imbalance reflects the unequal distribution of power in gender relations (Inglehart and Norris, 2003; Lazar, 2017) that finds expression also in politics:Kayla (63 years): So, there is a problem when a woman with family decides to go into politics. Because we are good at giving ourselves the responsibility of all that is not all right. We feel guilty at work because we are not at home, and we feel guilty at home because we are not at work. It is evident by Kayla's discourse, the logic of the hegemonic power is internalized and shared by women themselves (Bourdieu, 1991), involved in that relationship of dominance-submissiveness dictated by patriarchal power, which places men and women in specific roles, public and private. The dichotomy between private domestic sphere, traditionally considered feminine, and political-public sphere, historically a men prerogative, is proposed again in the interviewees' discourses. The family-political life balance as women's responsibility is explained by the respondents referring both to "natural" both to "cultural" factors:Elisa (47 years): Many women do not present themselves as candidates because they would not be able to balance family and politics, unlike men, clearly, probably by nature, by the trend. It is normal to be like this, but there is the difference. There is a different way of approaching the family, but that is DNA.Carlo (45 years): Women have a natural limitation because they can become mothers. Being a mother influences life. A woman has different needs than a man; she is more careful toward the family and, therefore, feels more strongly the need to go back home - she wants to prepare the soup for the baby. She is worried, but these worries are a limitation in the practice of an administrator. Because you have to be free and distractions are not allowed. All this penalizes the woman. The focus is on the alleged inability of women to reduce in their favor domestic and family responsibilities. Although there is an awareness of family burdens on women, that affects their opportunities for growth, the organization of the system seems to be conceived as an essential and unchangeable rule. Respondents that adhere to this interpretation tend to minimize and naturalize (Lazar, 2005) inequalities and women disadvantage resulting from strong adherence to a dominant gender order, which places women and men, respectively, inside and outside the home, mystifying diversity in the distribution of power. However, as clearly expressed by Valeria, the explanation for family-political life balance as women's responsibility not only comes in terms of "natural" factors, but an important role is also played by "cultural" factors that reiterate the status quo, and are accepted and considerate as appropriate by women as well:Valeria (25 years): There is still the mentality of the woman that stays at home, cleans, takes care of the children and cannot do it late at night. The low participation of women in politics may also be caused by this. This depends on culture; for a man is different, it is inevitable! A man goes to a working dinner, but I know that my children are at home, and my "mother's heart" makes me come back home. The social, political, and cultural context offers a dominant gender order (Connell, 1987, 2016), proposing a "natural" division between gender and roles played by men and women. Women's domestic overload is cause and consequence of the exclusion of women from prestigious activities like politics, a vicious circle whereby the hegemonic patriarchal culture, associating the women to the home and the men to the work, lays the foundation to determine an asymmetry between gender (Walby, 1990). The characteristics of the cultural context are another factor that plays a key role in explaining the gender gap in politics (Fornengo and Guadagnini, 1999). In particular, Giacomo's reflection underlines a dichotomous division of roles (production - men and reproductive - women), as result of a hegemonic patriarchal culture. The politics and its "times" are described not as a natural and obvious dimension, but as subject to renegotiations and changes. Giacomo, 38 years, supports the role of "cultural" factors:Now it is happening that since the woman has to stay home and take care of the kids, the man goes to do politics, because the timing in politics is done for men. Our society has this idea about politics. We must change it [...] we must change the culture. We must educate children. We must create a new political establishment and the conditions so that things will be different in the future, otherwise, it is normal that in politics you will repeat what is the everyday life. This quotation identifies the social and cultural dimensions that influence common sense featured in an essentialist nuance and refers to the patriarchal culture that categorizes men and women according to hierarchical relationships of dominance and submission. This distribution of roles as told by Giacomo is the product of practices and power relations in a social context that conveys social expectations that reproduce existing power unbalances between the gender. Giacomo expresses the need to change the patriarchal hegemonic culture that influences a gender order also in the political life. Some respondents think that adequate reconciliation services would help to solve the problem of underrepresentation of women in Italian politics (Lovenduski and Guadagnini, 2010):Ambra (61 years): Here are no services, or at least not enough. A good plan of services that allow women to enter the labour market, just a systematic project that needs to go in the direction of eliminating disparities. The balancing of family life and political career is a woman prerogative because even currently women more than men are engaged both in the public and private spheres. According to our respondents, the lack of adequate support from government policies and inequities in the distribution of family work are among the causes of women's choice to not devote to institutional policy. The political and family systems are still structured for the men advantage and for the oppressive presence of a gender order (Connell, 1987, 2016) that still put the exercise of social power in the hands of elite groups or specific men institutions (Lazar, 2005, 2007, 2017). This paper explores the topic of the gender gap in politics through a discourse analysis of a group of Italian politicians and shows the patterns whereby a dominant gender order is constructed and reproduced. Discourse analysis discloses some interpretive repertoires used by men and women to confirm and reinforce the hegemonic gender order. Data analysis confirms that the Italian context is strongly characterized by a dominant gender order that assumes that most men are involved in the public sphere and in paid employment, while women are more likely to operate in the private sphere and family work. This gender order defines the boundaries of symbolic practices and materials of masculinity and femininity, positioning women and men in different positions of power. While the presence of women in politics breaks the dominant gender order (Connell, 1987, 2016), the other gender practices in politics tend to rebuild that order, repositioning women in subordinate roles. The participants tend to either absolve or blame women for gender inequality in politics through different interpretive repertoires: ascribe to women the responsibility of their low representation in politics relying on women's disinterest toward politics; absolve women attributing the responsibility of the gender gap in politics to the reproduction by politics of masculine rules, rhythms and symbols that do not correspond to the characteristics and needs of women. The participants' discourses also highlight how work-family imbalanced distribution influences women's participation in politics attributing this both to natural factors and cultural factors. These discourses highlight an ideological dilemma (Billig et al., 1988) with respect to the causes that would lead to the clear separation between private and public spheres and, consequently, the different roles assumed between women and men. The findings of this paper allow us to understand the processes by which dominant symbolic gender orders are reconstructed and show how power relations are permeated by practices of hegemonic masculinity which reinforce this order. Discourses on low representation of women in politics contributes to the consolidation of hegemonic masculinity that conserves and legitimizes men power and women subordination (Connell, 1987; Connell and Messerschmidt, 2005). Hegemonic patriarchal culture continues to act undisturbed and, through everyday discourses about natural gender binarism, legitimizes and becomes the direct expression of the power relationship between men and women (Walby, 1990). Behind the rhetoric of the growing democratization of modern organizations, gender inequalities have become more powerful because they act through subtle discursive mechanisms, much more difficult to unmask (Fairclough, 1992). Rhetoric of equality, framed into a broader context of male hegemony, guides their sense-making into denying gender inequalities in politics. Those discourses represent a subtle, pervasive, insidious and seemingly innocuous form of power in modern societies (Lazar, 2005). In fact, processes of naturalization of gender inequality (Lazar, 2005) are proposed also in those discursive, apparently egalitarian, contexts that absolve women. Women in dilemma are the main witnesses (unaware) that gender differences are built in discourses and acted out in daily practices (Lazar, 2017). For many women interviewed, it was very difficult to recognize their position of inequality because it is produced by a latent ideological and hegemonic power (Lukes, 2005). This has the effect of driving choices and waivers and systematically and unknowingly supports men as a privileged social subject in the exercise of political and social power. This study has some limitations that may be overcome with further research. First, the small sample size and voluntary participation of politicians means that our results are not generalizable. Thus, this research project should be replicated using an expanded sample and comparing local and national politicians. A second limitation is the exclusive reliance on qualitative data, which in future should be supplemented with quantitative data. Despite these limitations, the findings highlight the importance of exploring issues relating to the gender gap in politics and offer some suggestions for implementing actions to promote gender equality in this area. The rebalancing of gender could be promoted by quotas guaranteeing women a percentage of posts in the political system. Pink quotas, along with actions to promote awareness of gender discrimination, could be a response to women's disinterest in politics. The presence of women as candidates or in political office also assumes a symbolic function and conveys the unequivocal message that politics "is not for men only." Previous studies have shown that the presence of women in institutional political roles increases other women's interest in politics (Burns et al., 2001). The virtuous effect of quotas in countries that have adopted them is widely documented in the literature (Dahlerup and Freidenvall, 2010). Also, work-life balance policies (such as monetary transfers, and children and care services) aimed at alleviating the burden of women's familial responsibilities could increase women's participation in politics by promoting gender equality inside and outside the family (Gornick and Meyers, 2003; Lewis, 2009). Indeed, some studies have shown a relationship between different welfare regimes and the way in which men and women build their life paths (Gornick and Meyers, 2003; Geist, 2005). To promote social change and challenge the dominant gender order it is necessary not only to take into account psychosocial and cultural factors, but also to leverage institutional factors. The research allows disclosure of the social practices based on gender and the patriarchal ideology that contributes to the construction and preservation of women political disadvantage (Holmes and Marra, 2010). The feminist critique of political practices and social relations based on gender aims ultimately at inspiring a social change. The status quo is challenged in favor of a feminist humanist vision of a just society in which gender does not predetermine or mediate relations with others or our sense of who we are and what we can become. Discourse analysis, which reveals how power works to support oppressive social structures and relations, is in itself a form of "analytical resistance" (van Dijk, 1991) and contributes to conducting power struggles and social change (Lazar, 2007). The study shows how discourses can represent an instrument of resistance to change and reinforcement of hegemonic gender order and offers some suggestions for future research. Future studies could use discourse analysis and the contribution of the critical feminist perspective as instruments to raise awareness to understand the mechanisms underlying gender inequalities and the gender gap also in other domains of power in which women are still excluded.
|
Despite the ongoing increase of women in the top positions, they are still underrepresented in politics. The studies that primarily focus on women's underrepresentation in politics neglect the role of gender as a category that structures and makes sense of social practices. The purpose of this paper is to investigate the mechanisms that regulate the contemporary gender order in politics through discourse analysis and the contribution of the critical feminist perspective.
|
[SECTION: Method] Despite the ongoing increase of women in the top positions of hierarchy, they continue to be underrepresented in politics occupying 19.5 percent of seats worldwide, 22.8 percent in Europe, 22.6 percent in the Americas and 42.0 percent in Nordic countries (Inter-Parliamentary Union, 2017). In 2017, Italy ranks 44th out of 193 in the world classification compiled by the Inter-Parliamentary Union (2017), with a participation rate of Italian women to res public between 28 and 31 percent. Although the number of women elected in central institutions is increased compared to the past, in Italy women are still low represented in politics (Massa, 2013), and the number of women politicians decreases in the hierarchy of power (Fornengo and Guadagnini, 1999; Bonomi et al., 2013). Every president of the Italian Republic has been a man and women who have served in the institutional role of president of the Chamber of the Italian Republic from 1948 to date are 3 out of 14 (Italian Parliament, 2017). This position of women in politics, characterized by both a numerical increase compared to the past and a numerical decrease at the top of the power hierarchy, is well represented by the current government. The XVII Legislature of government, which started in 2013 and close to conclusion, is the legislature with the largest number of women in Parliament, in fact the women account for 32 percent of deputies and 30 percent of senators, but only 16 percent of women cover the roles of leader, president of the commission and the office of the presidency. Analyzing the composition of the Italian Parliament, it is clear that the disparity between men and women in politics is recognizable when roles are more prestigious. This condition is also confirmed at the regional level: women are present in the joint sessions in 29 percent of the councils and hold key positions (departments and offices with greater decision-making power) in 18 percent of the councils, while only 10 percent is president of the region and only 2 percent are leaders of the municipalities in the capital city (Openpolis, 2015). The Italian case is particularly interesting, as Italy is strongly characterized by conditions of gender inequality. In addition to the gender gap in political participation, as evidenced by the above data, in Italy the level of women's participation in the paid labor market is still very low (ISTAT, 2016): 54.4 percent compared with an OECD average of 62.6 percent (OECD, 2015). Although Italian women have a higher level of education than that of Italian men (ISTAT, 2016), gender segregation persists. Working women are concentrated in the tertiary sector and low-skilled professions (ISTAT, 2015b), with only 35 percent holding a managerial position. In this regard, Italy is ranked 82nd out of 142 according to the World Economic Forum's classification (WEF, 2014). In addition, in Italy the division of family work remains drastically imbalanced to women's disadvantage (Eurostat, 2011; ISTAT, 2015a; CNEL, 2014). Research shows that the time Italian mothers dedicate to family work is much higher than that dedicated by fathers, even in dual-earner families (Eurostat, 2011; ISTAT, 2015a). This strict division of family roles is the main obstacle to women reaching decision-making positions in politics, because Italy also continues to invest inadequately in services to support families (Panzeri and Viale, 2016). In Italy, maternity continues to have a significant impact on unemployment rates and the early exit of women from the labor market (CNEL, 2014). In Italy, therefore, the underrepresentation and low appreciation of women in society and politics together constitute a serious problem whose causes it is vital to investigate in order to intervene adequately (Lasio et al., 2013; Pacilli et al., 2012). Despite the substantial social and cultural transformations of men and women roles, women continue to be excluded or impeded in highly competitive and masculine hierarchical domains, such as politics (Lazar, 2005). Men continue to be the majority in power contexts that appear inaccessible and strewn with obstacles for women to grow in (Fox and Lawless, 2011; Paxton and Hughes, 2015). Although women have achieved full juridical parity and participate in the political life of the country, conditions of strong inequality still persist and limit their growth in contexts that are considered places of power (Davidson and Burke, 2016; Lazar, 2017). Several studies have investigated women in leadership (e.g. Fitzsimmons et al., 2014; Mavin and Grandy, 2016) and explored persistent systems of discrimination and inequality (e.g. Kelan, 2008; Lewis et al., 2017). The matter of women in politics continues to represent the context in which a gendered division of power has its highest expression and constitutes a macroscopic disadvantage in terms of "presence" and in the acquisition of roles (Childs, 2006). Many pieces of evidence attest the existence of strong constraints to full and complete participation women's policy. Politics is one area where male domination over women and resistance to change are fully expressed (Bourdieu, 1998). Politics tends to reproduce male models and rules presenting itself as an unwelcoming domain for women. Women then, although they can rely on the same opportunity as men, are blocked by systems of exclusion rooted in the culture of many societies (Inglehart and Norris, 2003; Krook, 2010) and very often exclude themselves. Recent research (Lasio and Serri, 2017; Lasio et al., 2018) has also showed how the resistance opposed in Italy to non heterosexual families reinforces normative standards on motherhood, thus contributing to maintain those asymmetries at the origin of women's invisibility as active citizens (Amancio and Oliveira, 2006). Women's underrepresentation in politics reproduces the imbalanced and unequal gender relations (Lovenduski, 2001; Santos et al., 2013). Since the 1990s, some scholars, mostly women from Anglo-Saxon and Northern European countries (Childs and Krook, 2006; Krook, 2010), have begun to investigate the relationship between women and politics, focusing especially on the numerical presence of women in the government and in the councils, on the trends and mechanisms of recruitment and selection of women by the parties (Kenny, 2007; Celis et al., 2015; Kenny and Verge, 2015). These studies found that women continue to be underrepresented in politics and despite there being a much greater number of women in elected positions compared to the past, this it is not a substantive representation and cannot be called substantive equality (Childs and Krook, 2006). The literature on this topic has addressed the potential factors that may explain this underrepresentation, and has highlighted the influence of patriarchal culture, the division of family roles, the lack of childcare, the resisting attitude of the parties and the fragmentation of women's movement (Lovenduski and Guadagnini, 2010; Bonomi et al., 2013). Others have analyzed the relationship between discourse, gender and political representation, focusing on political discourse and the construct of gender identity (Shaw, 2000; Walsh, 2001) and gender citizenship in politics (Poggio, 2004; Belle, 2012). This study is positioned in the field that investigates the relationship between women and politics using a discursive perspective. The research was conducted within a theoretical framework referred to as a discourse analysis (Potter and Wetherell, 1987), and to contribution from the critical feminist perspective (Lazar, 2007, 2017; Holmes and Marra, 2010), and proposes to understand the mechanisms that regulate the contemporary gender order in politics through the analysis of the discourses of a group of Italian politicians. The focus on discourses of politicians allows us to analyze the processes by which dominant symbolic gender order is reconstructed in politics and to show how the distribution of power between gender is acted out through practices of hegemonic masculinity which excludes women from the real seats of power. This study offers a reflection on gender inequalities in political participation, starting with a review of the literature and the results of research conducted on Italian politicians. The contribution of this paper is to shed light on continuing discourses concerning politics as a predominantly male context. We conclude by highlighting the study's theoretical and practical implications for enhancing equality in politics and further developing research in this area. The studies on women and politics continue to focus on women' numbers in politics and on sex differences, and with difficulty, they are able to embrace the concept of gender (Kenny, 2007) as a category that structures and makes sense of particular social practices. However, feminist studies have developed complex theories of gender, including both patriarchal and discursive conceptions of power. The power inequality between gender relies on the concept of patriarchy, a system of social structures and practices of domination founded in the subordination of women by men (Walby, 1990). Gender relations are power relations involving formal public structures as politics and private structures as the family. The distribution of power between gender, based on hegemonic patriarchal culture, which associates naturally the female to the house and the male to the labor and political activities, has determined an asymmetry in which the historical forced absence of women from prestigious positions, such as politics, would involve a vicious circle whereby the greater female participation in domestic work has become cause and consequence of their exclusion (Walby, 1990). Through the internalization of gendered norms produced routinely in discourses of everyday life, this gap of power between men and women has become invisible, misrecognized and recognized as legitimate and natural (Bourdieu, 1991), contributing to the consolidation of "hegemonic masculinity" that preserves, legitimizes and naturalizes men power and, consequently, women subordination (Connell, 1987, 2016; Connell and Messerschmidt, 2005). Discourse, therefore, plays a significant role in the reproduction of dominance, which is considered as an exercise of social power by elites, institutions or groups that have the effect of increasing social, political, class and gender inequality (van Dijk, 1993, 2011). Social power is enacted, reproduced or legitimized by the discourse of dominant groups or institutions (van Dijk, 1996). Relations of power and dominance (Bourdieu, 1991) can discursively challenge the interests at stake. Therefore, hegemonic power relations are discursively produced, perpetuated, negotiated and challenged. To share or refuse these contents means to contribute to the reproduction and maintenance of the social order and also to resist and transform that order (Lazar, 2005, 2017). Gender is socially constructed through discourses (Wodak, 1997), and the power relations associated to, and reproduced by, such discourses (Wodak, 1997). Gender expresses the relationships of power that constitute society: women and men, through their performances, can reify or on the contrary challenge the system of relations and gender privileges (Holmes and Marra, 2010). Feminist critical discourse perspective offers important contribution to theorize and analyze "the particularly insidious and oppressive nature of gender as an omni-relevant category in most social practices" (Lazar, 2007, p. 3). In particular, feminist critical discourse analysis aims to demystify the interrelationships of gender and power, in discourse and critiques the discourses that sustain a patriarchal social order, that systematically privilege men and disadvantage women (Holmes and Marra, 2010). The disclosing of the discursive strategies through which power structures and supports oppressive social relations is a form of "analytical resistance" that help to support the struggles of protest and social change (van Dijk, 1991). Feminist critical discourse contributes to social transformation through the focus on repressive norms and restrictive stereotypes reified and reinforced by routine discourses of people (Holmes and Marra, 2010). This study is based on 30 biographical interviews conducted with local Italian politicians, 15 of whom were men, 15 women. We chose to involve local politicians because figures on the number of women in local Italian institutions are lower than those at the national level (ISTAT, 2017). The average age of the politicians interviewed was 43 years (DS=12.5; range=24-64). The research was carried out in Italy and recruitment took place through the involvement of political parties. The face-to-face interviews were conducted by one of the authors, lasted between 60 and 90 min, and focused on two central topics: their experiences in politics and meaningful episodes in their political life, and the barriers that may hinder women's political participation. Respondents were asked to illustrate their political career starting from the reasons that led them to choose politics and to discuss the relationship between politics and gender, taking into consideration their own experiences and reflecting on the meanings attributed to the experiences. The study was carried out in accordance with the Declaration of Helsinki. Participants were informed of the aims, methods and other ethical aspects relevant to the study. After they had given informed consent, all interviews were digitally recorded and transcribed verbatim. Participants' confidentiality and anonymity have been protected through the use of pseudonyms. For the first step of the analysis, the authors independently examined the transcripts to identify the emerging main themes and recurrent repertoires. Subsequently, the researchers worked together on the discourse analysis (Billig, 1987; Potter and Wetherell, 1987), in order to come to an agreement on the identified interpretative repertoires and chosen illustrative key quotes drawn from the different interviews. Particular attention is given to discourse, as social action through which men and women can reify or challenge the system of power relations, gender imbalances and privileged places in which these mechanisms can be observed and unmasked (Holmes and Marra, 2010). Data analysis focused on the interpretative repertoires and ideological dilemmas. The repertoires are "clusters of terms organized around a central metaphor, often used with grammatical regularity" (Wiggins and Potter, 2008, p. 74) not defined by specific words or phrases, but by a textual constellation that refers to a specific theme (Potter and Wetherell, 1987). The ideological dilemmas (Billig et al., 1988) indicate the intrinsic contradictory nature of common sense and highlight the rhetorical origin of attitudes toward gender as they emerge in contexts where there is an argument. "[...] when contradictory positions overlap they provide a basis for awareness and reflexivity, just as they lead to problematization and change" (Fairclough, 1995, p. 82). Data analysis show how the dominant gender order is performed and reinforced through the discursive practices of the politicians interviewed. The participants tell their experience as politicians and produce discourses that clarify the norms that regulate the political field. These discourses are part of a wider social context in which choices, roles and careers are gendered. The politicians interviewed offer several arguments about women' low participation in politics, referring to different degrees of responsibility or states of helplessness. Results show the participants' tendency to either absolve or blame women for gender inequality in politics through different interpretive repertoires, which can be, in some cases, dilemmatics (Billig et al., 1988). In particular, the interpretative repertoire "Women's disinterest toward politics" tends to blame women for their low participation in politics, whilst the interpretative repertoire "Politics as masculine context" tends to absolve women and to attribute the responsibility of the gender gap in politics to the hegemonic social context and to the attitude of the parties. A third repertoire "Politics-family unbalance" identifies in the unequal distribution of family work the principal responsibility of gender inequalities in politics. The arguments referred to the latter repertoire concern the debate between nature and culture, fluctuating between the essentialist positions on the one hand, and progressive positions, on the other hand. Women's disinterest toward politics Though among the politicians interviewed the idea that men and women can perform the same roles in politics is common, they explain the women's low representation in politics referring to the lack of interest of women toward politics, the women voluntary abstention and their minor will to be involved (De Simone et al., 2017). The issue of women's lack of interest is built on beliefs that recall individual factors (Fornengo and Guadagnini, 1999) as "a choice for women" dictated by a lack of desire to engage. This choice would lead women to voluntarily keep away from the places of res public:Marco (39 years): Women can play the same role in politics, whatever role males can play; there is no difference. It is a choice for women who do not participate in politics. They are not interested. [...] because there is no woman that enforce herself! Marco does not display explicit conflicts that were unfavorable to full equality between men and women in politics, however affirming the claim of women's disinterest in politics, supports the counterarguments that identify in women the cause of the impossibility to put such equality into practice. The argumentation relying on women's responsibility for the gender gap in politics becomes even stronger when greater capacity and greater potential is attributed to women themselves:Giovanni (41 years): women voluntarily choose not to do politics [...] when they are brought into play, they have produced good results just as good or better than others. I think that women have all the possibilities and conditions for doing politics, and even in the best way. The notion of choice is another key element to the understanding of the gender gap in politics. The reference to women's disinterest in politics being a reflection of their own voluntary choice implies a lack of recognition of the social inequalities affecting women's access to politics. Reference to women having free choice moves attention away from the responsibility of the system to the responsibilities of the individual woman, indicating that women's choices are embedded within cultural, social and economic conditions that structure gender dynamics (Lewis et al., 2017). The quotations above show as the power, deeply ingrained, acts automatically and unconsciously, since is foregone and "buried" under consolidated ideologies (Lukes, 2005). Marco and Giovanni's words show how the gender order is preserved through the third dimension of power that prevents some issues from being expressed. Gender conflict is latent because the subjects holding positions of power, in this case two men politicians, do not explicitly express their interests. The ideologies influence people's wishes and thoughts, the "ideological power," the "real power," acts indirectly and unconsciously, influencing the social actors and ensures, in this way, their acquiescence, even when it goes against their self-interest (Lukes, 2005). Indeed Valeria (25 years), says:Women themselves are afraid of not being up. It depends very much on the lack of interest or fear to get involved because, in my opinion, we are the cause of the prejudice! I speak often of meritocracy, I hate rose quotas! There is no need for rose quotas. Valeria affirms that women have not the need for rose quotas. The rose quotas refer to those legal instruments aimed to protect gender equality within the representative bodies through different calculation mechanisms and allocation of seats. Valeria, with her statements, contributes to maintaining the status quo, by underplaying and silencing structural inequalities. Once again, it is not the politics to reiterate a gender order that does not allow women's participation, but women themselves are responsible for their own exclusion. For many respondents, including women, and as it is clear from Valeria's discourse, affirmative action promoted in favor of gender equity in political participation is perceived as offensive and not qualifying (Santos et al., 2016). Valeria denies gender inequalities, thereby oppositions are avoided, whereas asymmetric power relations between gender persist. The issue of voluntary women's abstention from politics justifies persistent power inequalities as the result of women's individual choices and underlies the non-recognition of gender social inequalities (Lewis, 2014). Marco, Giovanni and Valeria's extracts emphasize the individual responsibilities with the effect of concealing structural discrimination mechanisms and the inequality of power between men and women. Through those discourses seem to emerge a lack of awareness of the mechanisms of male dominance that push women, without their awareness, to adapt their hopes and their aspirations to opportunities, and consequently not to undertake actions that are not expected of them (Bourdieu, 1998). The order of things, in fact, is not a natural order, but a building of the world with which man pleases his thirst for domination, a vision so exclusive that the same women who are the victims have integrated in their own way of thinking by accepting their unconscious inferiority (Bourdieu, 1998). Valeria, as a woman, can be considered victims and author of gender relations of power acting in social practices upholding hegemonic culture. The hegemonic masculinity, as a pattern of practices that permit men's dominance over women (Connell and Messerschmidt, 2005), inspires discourses that aim to represent a reality of gender equality. Marco, Giovanni and Valeria's discourses lack an acknowledgment of this systematic imbalance referred to the rhetoric of sameness between gender and denial of inequalities. The hegemonic masculinity (Connell and Messerschmidt, 2005) inspires the discourses of equality that deny gender inequalities slowing social change (De Simone et al., 2015; De Simone and Scano, 2017). Politics as masculine context Interviewed politicians' discourses identify the cause of the women's low representation in politics in the sexism that characterizes the parties and politics in general, using the interpretive repertoire "Politics as masculine context." Women's access to some areas continues to be difficult and to represent a source of inequality especially in the "real" places of power (Lazar, 2005), as politics. Several interviewees emphasize the persistence of the gender gap in politics highlighting that despite the number of women in politics has increased compared to the past, high positions continue to be held predominantly by men reiterating a gender order that relegates women to subordinate positions (Connell, 1987, 2016):Cristina (53 years): Certain roads were blocked to women. There is a social structure that does not allow access to women in politics. Since it is been a long time, and now it seems that everything is established a norm. Women, despite the battles, block themselves because this is the social structure which they know and consequently everything remains unchanged. Cristina's words recount as the politics represents a system of power relations aimed at gaining control of resources and opportunities of majority group over the minority group (van Dijk, 2001). According to respondents also the attitude of the parties that have opposed resistance and have offered few opportunities for women helps to explain the low representation of women in politics (Guadagnini, 2005; Yirmibesoglu, 2008; Lovenduski and Guadagnini, 2010; Belle, 2012; De Simone et al., 2017). About this Sara, 60 years, says:I have encountered many difficulties in my political career. The political parties are governed by men, who support other men. Women do not vote for women because there are no women on the list and if they are there, they are already put in the last places, and they will never be elected. The parties are a male lobby. Sara and other respondents attribute the barriers encountered by women in politics to the discriminatory practices carried out by the parties (Lovenduski and Guadagnini, 2010). Sara describes the difficulties encountered in the party and the sense of exclusion generated by belonging to a minority. Sara's words perfectly convey the idea of the dynamics that develop in power institutions, always male dominance, helping to define the segregating phenomena of women. The lower representation of women in political institutions is explained by reference to the selection and recruitment methods in force in the various political systems and implemented by the various parties (Fornengo and Guadagnini, 1999; Belle, 2012; De Simone et al., 2017). The construction of male-dominated lists is fostered by party logics that build blocked lists where women are at the end of the list, without any chance of affirmation, according to the political gatekeeping phenomenon (Bain and Cummings, 2000). In this way, politics tends to exclude women by proposing predominantly male patterns and discriminatory practices (Annesley and Gains, 2010). Carlo, 45 years, explains the mechanisms of political parties that discourage women participation (Fox and Lawless, 2011) and contribute to the consolidation of hegemonic masculinity (Connell and Messerschmidt, 2005):In political parties, women are called by the secretaries, who are mostly men, in the election period. But after the election, when women did not take votes, they are no longer considered and they feel cheated. So, they say: do not go back there again." The existence of the glass ceiling (Morrison et al., 1992) in politics represents an invisible barrier that prevents women participation in politics and, on the contrary, privileges men. The prevalence of votes of men candidates becomes an obvious, logical, natural phenomenon, as told also by Giacomo, 38 years:There is a greater tendency by most people to vote for men, the system is structured in a sexist way. First, since women are less, there are ten women and ninety men; clearly, it is easier for a man to achieve, than for a woman. Second, a man wants to support himself, so he wants to keep that power and a woman who arrives so high would compromise the rules. Giacomo exemplifies as the vote reifies the dominant gender order according to specific logic of power. His discourse operates for the naturalization of a gender order socially accepted and shared, which mystifies the diversity in the distribution of power and inequality between women and men (Lazar, 2005, 2017). Politics-family unbalance The underrepresentation of women in politics is linked to family-work life balance (De Simone et al., 2018) as women's responsibility (Lovenduski and Guadagnini, 2010; De Simone et al., 2017). Some interviewees assert that women would not be interested in doing politics because they organize their lives around the family. For example, Giovanni, 38 years, say:I also understand that for a woman who decides to have a family and children it is more complicated because family work falls more on the woman than on the man. Women are interested in organising their lives around the family. Women should not be pretentious and have unrealistic demands toward politics, they must organize themselves with family and adapt to the rules of politics. Giovanni's extract describes a rigid division of gender roles and a strong unbalance in the distribution of family work as a disadvantage for women because of the limitation of time and resources for professional and political affirmation (Lazar, 2005; Lasio et al., 2017) and as a cause of the gender gap in politics (Guadagnini, 2005). This imbalance reflects the unequal distribution of power in gender relations (Inglehart and Norris, 2003; Lazar, 2017) that finds expression also in politics:Kayla (63 years): So, there is a problem when a woman with family decides to go into politics. Because we are good at giving ourselves the responsibility of all that is not all right. We feel guilty at work because we are not at home, and we feel guilty at home because we are not at work. It is evident by Kayla's discourse, the logic of the hegemonic power is internalized and shared by women themselves (Bourdieu, 1991), involved in that relationship of dominance-submissiveness dictated by patriarchal power, which places men and women in specific roles, public and private. The dichotomy between private domestic sphere, traditionally considered feminine, and political-public sphere, historically a men prerogative, is proposed again in the interviewees' discourses. The family-political life balance as women's responsibility is explained by the respondents referring both to "natural" both to "cultural" factors:Elisa (47 years): Many women do not present themselves as candidates because they would not be able to balance family and politics, unlike men, clearly, probably by nature, by the trend. It is normal to be like this, but there is the difference. There is a different way of approaching the family, but that is DNA.Carlo (45 years): Women have a natural limitation because they can become mothers. Being a mother influences life. A woman has different needs than a man; she is more careful toward the family and, therefore, feels more strongly the need to go back home - she wants to prepare the soup for the baby. She is worried, but these worries are a limitation in the practice of an administrator. Because you have to be free and distractions are not allowed. All this penalizes the woman. The focus is on the alleged inability of women to reduce in their favor domestic and family responsibilities. Although there is an awareness of family burdens on women, that affects their opportunities for growth, the organization of the system seems to be conceived as an essential and unchangeable rule. Respondents that adhere to this interpretation tend to minimize and naturalize (Lazar, 2005) inequalities and women disadvantage resulting from strong adherence to a dominant gender order, which places women and men, respectively, inside and outside the home, mystifying diversity in the distribution of power. However, as clearly expressed by Valeria, the explanation for family-political life balance as women's responsibility not only comes in terms of "natural" factors, but an important role is also played by "cultural" factors that reiterate the status quo, and are accepted and considerate as appropriate by women as well:Valeria (25 years): There is still the mentality of the woman that stays at home, cleans, takes care of the children and cannot do it late at night. The low participation of women in politics may also be caused by this. This depends on culture; for a man is different, it is inevitable! A man goes to a working dinner, but I know that my children are at home, and my "mother's heart" makes me come back home. The social, political, and cultural context offers a dominant gender order (Connell, 1987, 2016), proposing a "natural" division between gender and roles played by men and women. Women's domestic overload is cause and consequence of the exclusion of women from prestigious activities like politics, a vicious circle whereby the hegemonic patriarchal culture, associating the women to the home and the men to the work, lays the foundation to determine an asymmetry between gender (Walby, 1990). The characteristics of the cultural context are another factor that plays a key role in explaining the gender gap in politics (Fornengo and Guadagnini, 1999). In particular, Giacomo's reflection underlines a dichotomous division of roles (production - men and reproductive - women), as result of a hegemonic patriarchal culture. The politics and its "times" are described not as a natural and obvious dimension, but as subject to renegotiations and changes. Giacomo, 38 years, supports the role of "cultural" factors:Now it is happening that since the woman has to stay home and take care of the kids, the man goes to do politics, because the timing in politics is done for men. Our society has this idea about politics. We must change it [...] we must change the culture. We must educate children. We must create a new political establishment and the conditions so that things will be different in the future, otherwise, it is normal that in politics you will repeat what is the everyday life. This quotation identifies the social and cultural dimensions that influence common sense featured in an essentialist nuance and refers to the patriarchal culture that categorizes men and women according to hierarchical relationships of dominance and submission. This distribution of roles as told by Giacomo is the product of practices and power relations in a social context that conveys social expectations that reproduce existing power unbalances between the gender. Giacomo expresses the need to change the patriarchal hegemonic culture that influences a gender order also in the political life. Some respondents think that adequate reconciliation services would help to solve the problem of underrepresentation of women in Italian politics (Lovenduski and Guadagnini, 2010):Ambra (61 years): Here are no services, or at least not enough. A good plan of services that allow women to enter the labour market, just a systematic project that needs to go in the direction of eliminating disparities. The balancing of family life and political career is a woman prerogative because even currently women more than men are engaged both in the public and private spheres. According to our respondents, the lack of adequate support from government policies and inequities in the distribution of family work are among the causes of women's choice to not devote to institutional policy. The political and family systems are still structured for the men advantage and for the oppressive presence of a gender order (Connell, 1987, 2016) that still put the exercise of social power in the hands of elite groups or specific men institutions (Lazar, 2005, 2007, 2017). This paper explores the topic of the gender gap in politics through a discourse analysis of a group of Italian politicians and shows the patterns whereby a dominant gender order is constructed and reproduced. Discourse analysis discloses some interpretive repertoires used by men and women to confirm and reinforce the hegemonic gender order. Data analysis confirms that the Italian context is strongly characterized by a dominant gender order that assumes that most men are involved in the public sphere and in paid employment, while women are more likely to operate in the private sphere and family work. This gender order defines the boundaries of symbolic practices and materials of masculinity and femininity, positioning women and men in different positions of power. While the presence of women in politics breaks the dominant gender order (Connell, 1987, 2016), the other gender practices in politics tend to rebuild that order, repositioning women in subordinate roles. The participants tend to either absolve or blame women for gender inequality in politics through different interpretive repertoires: ascribe to women the responsibility of their low representation in politics relying on women's disinterest toward politics; absolve women attributing the responsibility of the gender gap in politics to the reproduction by politics of masculine rules, rhythms and symbols that do not correspond to the characteristics and needs of women. The participants' discourses also highlight how work-family imbalanced distribution influences women's participation in politics attributing this both to natural factors and cultural factors. These discourses highlight an ideological dilemma (Billig et al., 1988) with respect to the causes that would lead to the clear separation between private and public spheres and, consequently, the different roles assumed between women and men. The findings of this paper allow us to understand the processes by which dominant symbolic gender orders are reconstructed and show how power relations are permeated by practices of hegemonic masculinity which reinforce this order. Discourses on low representation of women in politics contributes to the consolidation of hegemonic masculinity that conserves and legitimizes men power and women subordination (Connell, 1987; Connell and Messerschmidt, 2005). Hegemonic patriarchal culture continues to act undisturbed and, through everyday discourses about natural gender binarism, legitimizes and becomes the direct expression of the power relationship between men and women (Walby, 1990). Behind the rhetoric of the growing democratization of modern organizations, gender inequalities have become more powerful because they act through subtle discursive mechanisms, much more difficult to unmask (Fairclough, 1992). Rhetoric of equality, framed into a broader context of male hegemony, guides their sense-making into denying gender inequalities in politics. Those discourses represent a subtle, pervasive, insidious and seemingly innocuous form of power in modern societies (Lazar, 2005). In fact, processes of naturalization of gender inequality (Lazar, 2005) are proposed also in those discursive, apparently egalitarian, contexts that absolve women. Women in dilemma are the main witnesses (unaware) that gender differences are built in discourses and acted out in daily practices (Lazar, 2017). For many women interviewed, it was very difficult to recognize their position of inequality because it is produced by a latent ideological and hegemonic power (Lukes, 2005). This has the effect of driving choices and waivers and systematically and unknowingly supports men as a privileged social subject in the exercise of political and social power. This study has some limitations that may be overcome with further research. First, the small sample size and voluntary participation of politicians means that our results are not generalizable. Thus, this research project should be replicated using an expanded sample and comparing local and national politicians. A second limitation is the exclusive reliance on qualitative data, which in future should be supplemented with quantitative data. Despite these limitations, the findings highlight the importance of exploring issues relating to the gender gap in politics and offer some suggestions for implementing actions to promote gender equality in this area. The rebalancing of gender could be promoted by quotas guaranteeing women a percentage of posts in the political system. Pink quotas, along with actions to promote awareness of gender discrimination, could be a response to women's disinterest in politics. The presence of women as candidates or in political office also assumes a symbolic function and conveys the unequivocal message that politics "is not for men only." Previous studies have shown that the presence of women in institutional political roles increases other women's interest in politics (Burns et al., 2001). The virtuous effect of quotas in countries that have adopted them is widely documented in the literature (Dahlerup and Freidenvall, 2010). Also, work-life balance policies (such as monetary transfers, and children and care services) aimed at alleviating the burden of women's familial responsibilities could increase women's participation in politics by promoting gender equality inside and outside the family (Gornick and Meyers, 2003; Lewis, 2009). Indeed, some studies have shown a relationship between different welfare regimes and the way in which men and women build their life paths (Gornick and Meyers, 2003; Geist, 2005). To promote social change and challenge the dominant gender order it is necessary not only to take into account psychosocial and cultural factors, but also to leverage institutional factors. The research allows disclosure of the social practices based on gender and the patriarchal ideology that contributes to the construction and preservation of women political disadvantage (Holmes and Marra, 2010). The feminist critique of political practices and social relations based on gender aims ultimately at inspiring a social change. The status quo is challenged in favor of a feminist humanist vision of a just society in which gender does not predetermine or mediate relations with others or our sense of who we are and what we can become. Discourse analysis, which reveals how power works to support oppressive social structures and relations, is in itself a form of "analytical resistance" (van Dijk, 1991) and contributes to conducting power struggles and social change (Lazar, 2007). The study shows how discourses can represent an instrument of resistance to change and reinforcement of hegemonic gender order and offers some suggestions for future research. Future studies could use discourse analysis and the contribution of the critical feminist perspective as instruments to raise awareness to understand the mechanisms underlying gender inequalities and the gender gap also in other domains of power in which women are still excluded.
|
The study is based on 30 biographical interviews with Italian politicians and focuses on the account of their political experiences and on the meanings attributed to these.
|
[SECTION: Findings] Despite the ongoing increase of women in the top positions of hierarchy, they continue to be underrepresented in politics occupying 19.5 percent of seats worldwide, 22.8 percent in Europe, 22.6 percent in the Americas and 42.0 percent in Nordic countries (Inter-Parliamentary Union, 2017). In 2017, Italy ranks 44th out of 193 in the world classification compiled by the Inter-Parliamentary Union (2017), with a participation rate of Italian women to res public between 28 and 31 percent. Although the number of women elected in central institutions is increased compared to the past, in Italy women are still low represented in politics (Massa, 2013), and the number of women politicians decreases in the hierarchy of power (Fornengo and Guadagnini, 1999; Bonomi et al., 2013). Every president of the Italian Republic has been a man and women who have served in the institutional role of president of the Chamber of the Italian Republic from 1948 to date are 3 out of 14 (Italian Parliament, 2017). This position of women in politics, characterized by both a numerical increase compared to the past and a numerical decrease at the top of the power hierarchy, is well represented by the current government. The XVII Legislature of government, which started in 2013 and close to conclusion, is the legislature with the largest number of women in Parliament, in fact the women account for 32 percent of deputies and 30 percent of senators, but only 16 percent of women cover the roles of leader, president of the commission and the office of the presidency. Analyzing the composition of the Italian Parliament, it is clear that the disparity between men and women in politics is recognizable when roles are more prestigious. This condition is also confirmed at the regional level: women are present in the joint sessions in 29 percent of the councils and hold key positions (departments and offices with greater decision-making power) in 18 percent of the councils, while only 10 percent is president of the region and only 2 percent are leaders of the municipalities in the capital city (Openpolis, 2015). The Italian case is particularly interesting, as Italy is strongly characterized by conditions of gender inequality. In addition to the gender gap in political participation, as evidenced by the above data, in Italy the level of women's participation in the paid labor market is still very low (ISTAT, 2016): 54.4 percent compared with an OECD average of 62.6 percent (OECD, 2015). Although Italian women have a higher level of education than that of Italian men (ISTAT, 2016), gender segregation persists. Working women are concentrated in the tertiary sector and low-skilled professions (ISTAT, 2015b), with only 35 percent holding a managerial position. In this regard, Italy is ranked 82nd out of 142 according to the World Economic Forum's classification (WEF, 2014). In addition, in Italy the division of family work remains drastically imbalanced to women's disadvantage (Eurostat, 2011; ISTAT, 2015a; CNEL, 2014). Research shows that the time Italian mothers dedicate to family work is much higher than that dedicated by fathers, even in dual-earner families (Eurostat, 2011; ISTAT, 2015a). This strict division of family roles is the main obstacle to women reaching decision-making positions in politics, because Italy also continues to invest inadequately in services to support families (Panzeri and Viale, 2016). In Italy, maternity continues to have a significant impact on unemployment rates and the early exit of women from the labor market (CNEL, 2014). In Italy, therefore, the underrepresentation and low appreciation of women in society and politics together constitute a serious problem whose causes it is vital to investigate in order to intervene adequately (Lasio et al., 2013; Pacilli et al., 2012). Despite the substantial social and cultural transformations of men and women roles, women continue to be excluded or impeded in highly competitive and masculine hierarchical domains, such as politics (Lazar, 2005). Men continue to be the majority in power contexts that appear inaccessible and strewn with obstacles for women to grow in (Fox and Lawless, 2011; Paxton and Hughes, 2015). Although women have achieved full juridical parity and participate in the political life of the country, conditions of strong inequality still persist and limit their growth in contexts that are considered places of power (Davidson and Burke, 2016; Lazar, 2017). Several studies have investigated women in leadership (e.g. Fitzsimmons et al., 2014; Mavin and Grandy, 2016) and explored persistent systems of discrimination and inequality (e.g. Kelan, 2008; Lewis et al., 2017). The matter of women in politics continues to represent the context in which a gendered division of power has its highest expression and constitutes a macroscopic disadvantage in terms of "presence" and in the acquisition of roles (Childs, 2006). Many pieces of evidence attest the existence of strong constraints to full and complete participation women's policy. Politics is one area where male domination over women and resistance to change are fully expressed (Bourdieu, 1998). Politics tends to reproduce male models and rules presenting itself as an unwelcoming domain for women. Women then, although they can rely on the same opportunity as men, are blocked by systems of exclusion rooted in the culture of many societies (Inglehart and Norris, 2003; Krook, 2010) and very often exclude themselves. Recent research (Lasio and Serri, 2017; Lasio et al., 2018) has also showed how the resistance opposed in Italy to non heterosexual families reinforces normative standards on motherhood, thus contributing to maintain those asymmetries at the origin of women's invisibility as active citizens (Amancio and Oliveira, 2006). Women's underrepresentation in politics reproduces the imbalanced and unequal gender relations (Lovenduski, 2001; Santos et al., 2013). Since the 1990s, some scholars, mostly women from Anglo-Saxon and Northern European countries (Childs and Krook, 2006; Krook, 2010), have begun to investigate the relationship between women and politics, focusing especially on the numerical presence of women in the government and in the councils, on the trends and mechanisms of recruitment and selection of women by the parties (Kenny, 2007; Celis et al., 2015; Kenny and Verge, 2015). These studies found that women continue to be underrepresented in politics and despite there being a much greater number of women in elected positions compared to the past, this it is not a substantive representation and cannot be called substantive equality (Childs and Krook, 2006). The literature on this topic has addressed the potential factors that may explain this underrepresentation, and has highlighted the influence of patriarchal culture, the division of family roles, the lack of childcare, the resisting attitude of the parties and the fragmentation of women's movement (Lovenduski and Guadagnini, 2010; Bonomi et al., 2013). Others have analyzed the relationship between discourse, gender and political representation, focusing on political discourse and the construct of gender identity (Shaw, 2000; Walsh, 2001) and gender citizenship in politics (Poggio, 2004; Belle, 2012). This study is positioned in the field that investigates the relationship between women and politics using a discursive perspective. The research was conducted within a theoretical framework referred to as a discourse analysis (Potter and Wetherell, 1987), and to contribution from the critical feminist perspective (Lazar, 2007, 2017; Holmes and Marra, 2010), and proposes to understand the mechanisms that regulate the contemporary gender order in politics through the analysis of the discourses of a group of Italian politicians. The focus on discourses of politicians allows us to analyze the processes by which dominant symbolic gender order is reconstructed in politics and to show how the distribution of power between gender is acted out through practices of hegemonic masculinity which excludes women from the real seats of power. This study offers a reflection on gender inequalities in political participation, starting with a review of the literature and the results of research conducted on Italian politicians. The contribution of this paper is to shed light on continuing discourses concerning politics as a predominantly male context. We conclude by highlighting the study's theoretical and practical implications for enhancing equality in politics and further developing research in this area. The studies on women and politics continue to focus on women' numbers in politics and on sex differences, and with difficulty, they are able to embrace the concept of gender (Kenny, 2007) as a category that structures and makes sense of particular social practices. However, feminist studies have developed complex theories of gender, including both patriarchal and discursive conceptions of power. The power inequality between gender relies on the concept of patriarchy, a system of social structures and practices of domination founded in the subordination of women by men (Walby, 1990). Gender relations are power relations involving formal public structures as politics and private structures as the family. The distribution of power between gender, based on hegemonic patriarchal culture, which associates naturally the female to the house and the male to the labor and political activities, has determined an asymmetry in which the historical forced absence of women from prestigious positions, such as politics, would involve a vicious circle whereby the greater female participation in domestic work has become cause and consequence of their exclusion (Walby, 1990). Through the internalization of gendered norms produced routinely in discourses of everyday life, this gap of power between men and women has become invisible, misrecognized and recognized as legitimate and natural (Bourdieu, 1991), contributing to the consolidation of "hegemonic masculinity" that preserves, legitimizes and naturalizes men power and, consequently, women subordination (Connell, 1987, 2016; Connell and Messerschmidt, 2005). Discourse, therefore, plays a significant role in the reproduction of dominance, which is considered as an exercise of social power by elites, institutions or groups that have the effect of increasing social, political, class and gender inequality (van Dijk, 1993, 2011). Social power is enacted, reproduced or legitimized by the discourse of dominant groups or institutions (van Dijk, 1996). Relations of power and dominance (Bourdieu, 1991) can discursively challenge the interests at stake. Therefore, hegemonic power relations are discursively produced, perpetuated, negotiated and challenged. To share or refuse these contents means to contribute to the reproduction and maintenance of the social order and also to resist and transform that order (Lazar, 2005, 2017). Gender is socially constructed through discourses (Wodak, 1997), and the power relations associated to, and reproduced by, such discourses (Wodak, 1997). Gender expresses the relationships of power that constitute society: women and men, through their performances, can reify or on the contrary challenge the system of relations and gender privileges (Holmes and Marra, 2010). Feminist critical discourse perspective offers important contribution to theorize and analyze "the particularly insidious and oppressive nature of gender as an omni-relevant category in most social practices" (Lazar, 2007, p. 3). In particular, feminist critical discourse analysis aims to demystify the interrelationships of gender and power, in discourse and critiques the discourses that sustain a patriarchal social order, that systematically privilege men and disadvantage women (Holmes and Marra, 2010). The disclosing of the discursive strategies through which power structures and supports oppressive social relations is a form of "analytical resistance" that help to support the struggles of protest and social change (van Dijk, 1991). Feminist critical discourse contributes to social transformation through the focus on repressive norms and restrictive stereotypes reified and reinforced by routine discourses of people (Holmes and Marra, 2010). This study is based on 30 biographical interviews conducted with local Italian politicians, 15 of whom were men, 15 women. We chose to involve local politicians because figures on the number of women in local Italian institutions are lower than those at the national level (ISTAT, 2017). The average age of the politicians interviewed was 43 years (DS=12.5; range=24-64). The research was carried out in Italy and recruitment took place through the involvement of political parties. The face-to-face interviews were conducted by one of the authors, lasted between 60 and 90 min, and focused on two central topics: their experiences in politics and meaningful episodes in their political life, and the barriers that may hinder women's political participation. Respondents were asked to illustrate their political career starting from the reasons that led them to choose politics and to discuss the relationship between politics and gender, taking into consideration their own experiences and reflecting on the meanings attributed to the experiences. The study was carried out in accordance with the Declaration of Helsinki. Participants were informed of the aims, methods and other ethical aspects relevant to the study. After they had given informed consent, all interviews were digitally recorded and transcribed verbatim. Participants' confidentiality and anonymity have been protected through the use of pseudonyms. For the first step of the analysis, the authors independently examined the transcripts to identify the emerging main themes and recurrent repertoires. Subsequently, the researchers worked together on the discourse analysis (Billig, 1987; Potter and Wetherell, 1987), in order to come to an agreement on the identified interpretative repertoires and chosen illustrative key quotes drawn from the different interviews. Particular attention is given to discourse, as social action through which men and women can reify or challenge the system of power relations, gender imbalances and privileged places in which these mechanisms can be observed and unmasked (Holmes and Marra, 2010). Data analysis focused on the interpretative repertoires and ideological dilemmas. The repertoires are "clusters of terms organized around a central metaphor, often used with grammatical regularity" (Wiggins and Potter, 2008, p. 74) not defined by specific words or phrases, but by a textual constellation that refers to a specific theme (Potter and Wetherell, 1987). The ideological dilemmas (Billig et al., 1988) indicate the intrinsic contradictory nature of common sense and highlight the rhetorical origin of attitudes toward gender as they emerge in contexts where there is an argument. "[...] when contradictory positions overlap they provide a basis for awareness and reflexivity, just as they lead to problematization and change" (Fairclough, 1995, p. 82). Data analysis show how the dominant gender order is performed and reinforced through the discursive practices of the politicians interviewed. The participants tell their experience as politicians and produce discourses that clarify the norms that regulate the political field. These discourses are part of a wider social context in which choices, roles and careers are gendered. The politicians interviewed offer several arguments about women' low participation in politics, referring to different degrees of responsibility or states of helplessness. Results show the participants' tendency to either absolve or blame women for gender inequality in politics through different interpretive repertoires, which can be, in some cases, dilemmatics (Billig et al., 1988). In particular, the interpretative repertoire "Women's disinterest toward politics" tends to blame women for their low participation in politics, whilst the interpretative repertoire "Politics as masculine context" tends to absolve women and to attribute the responsibility of the gender gap in politics to the hegemonic social context and to the attitude of the parties. A third repertoire "Politics-family unbalance" identifies in the unequal distribution of family work the principal responsibility of gender inequalities in politics. The arguments referred to the latter repertoire concern the debate between nature and culture, fluctuating between the essentialist positions on the one hand, and progressive positions, on the other hand. Women's disinterest toward politics Though among the politicians interviewed the idea that men and women can perform the same roles in politics is common, they explain the women's low representation in politics referring to the lack of interest of women toward politics, the women voluntary abstention and their minor will to be involved (De Simone et al., 2017). The issue of women's lack of interest is built on beliefs that recall individual factors (Fornengo and Guadagnini, 1999) as "a choice for women" dictated by a lack of desire to engage. This choice would lead women to voluntarily keep away from the places of res public:Marco (39 years): Women can play the same role in politics, whatever role males can play; there is no difference. It is a choice for women who do not participate in politics. They are not interested. [...] because there is no woman that enforce herself! Marco does not display explicit conflicts that were unfavorable to full equality between men and women in politics, however affirming the claim of women's disinterest in politics, supports the counterarguments that identify in women the cause of the impossibility to put such equality into practice. The argumentation relying on women's responsibility for the gender gap in politics becomes even stronger when greater capacity and greater potential is attributed to women themselves:Giovanni (41 years): women voluntarily choose not to do politics [...] when they are brought into play, they have produced good results just as good or better than others. I think that women have all the possibilities and conditions for doing politics, and even in the best way. The notion of choice is another key element to the understanding of the gender gap in politics. The reference to women's disinterest in politics being a reflection of their own voluntary choice implies a lack of recognition of the social inequalities affecting women's access to politics. Reference to women having free choice moves attention away from the responsibility of the system to the responsibilities of the individual woman, indicating that women's choices are embedded within cultural, social and economic conditions that structure gender dynamics (Lewis et al., 2017). The quotations above show as the power, deeply ingrained, acts automatically and unconsciously, since is foregone and "buried" under consolidated ideologies (Lukes, 2005). Marco and Giovanni's words show how the gender order is preserved through the third dimension of power that prevents some issues from being expressed. Gender conflict is latent because the subjects holding positions of power, in this case two men politicians, do not explicitly express their interests. The ideologies influence people's wishes and thoughts, the "ideological power," the "real power," acts indirectly and unconsciously, influencing the social actors and ensures, in this way, their acquiescence, even when it goes against their self-interest (Lukes, 2005). Indeed Valeria (25 years), says:Women themselves are afraid of not being up. It depends very much on the lack of interest or fear to get involved because, in my opinion, we are the cause of the prejudice! I speak often of meritocracy, I hate rose quotas! There is no need for rose quotas. Valeria affirms that women have not the need for rose quotas. The rose quotas refer to those legal instruments aimed to protect gender equality within the representative bodies through different calculation mechanisms and allocation of seats. Valeria, with her statements, contributes to maintaining the status quo, by underplaying and silencing structural inequalities. Once again, it is not the politics to reiterate a gender order that does not allow women's participation, but women themselves are responsible for their own exclusion. For many respondents, including women, and as it is clear from Valeria's discourse, affirmative action promoted in favor of gender equity in political participation is perceived as offensive and not qualifying (Santos et al., 2016). Valeria denies gender inequalities, thereby oppositions are avoided, whereas asymmetric power relations between gender persist. The issue of voluntary women's abstention from politics justifies persistent power inequalities as the result of women's individual choices and underlies the non-recognition of gender social inequalities (Lewis, 2014). Marco, Giovanni and Valeria's extracts emphasize the individual responsibilities with the effect of concealing structural discrimination mechanisms and the inequality of power between men and women. Through those discourses seem to emerge a lack of awareness of the mechanisms of male dominance that push women, without their awareness, to adapt their hopes and their aspirations to opportunities, and consequently not to undertake actions that are not expected of them (Bourdieu, 1998). The order of things, in fact, is not a natural order, but a building of the world with which man pleases his thirst for domination, a vision so exclusive that the same women who are the victims have integrated in their own way of thinking by accepting their unconscious inferiority (Bourdieu, 1998). Valeria, as a woman, can be considered victims and author of gender relations of power acting in social practices upholding hegemonic culture. The hegemonic masculinity, as a pattern of practices that permit men's dominance over women (Connell and Messerschmidt, 2005), inspires discourses that aim to represent a reality of gender equality. Marco, Giovanni and Valeria's discourses lack an acknowledgment of this systematic imbalance referred to the rhetoric of sameness between gender and denial of inequalities. The hegemonic masculinity (Connell and Messerschmidt, 2005) inspires the discourses of equality that deny gender inequalities slowing social change (De Simone et al., 2015; De Simone and Scano, 2017). Politics as masculine context Interviewed politicians' discourses identify the cause of the women's low representation in politics in the sexism that characterizes the parties and politics in general, using the interpretive repertoire "Politics as masculine context." Women's access to some areas continues to be difficult and to represent a source of inequality especially in the "real" places of power (Lazar, 2005), as politics. Several interviewees emphasize the persistence of the gender gap in politics highlighting that despite the number of women in politics has increased compared to the past, high positions continue to be held predominantly by men reiterating a gender order that relegates women to subordinate positions (Connell, 1987, 2016):Cristina (53 years): Certain roads were blocked to women. There is a social structure that does not allow access to women in politics. Since it is been a long time, and now it seems that everything is established a norm. Women, despite the battles, block themselves because this is the social structure which they know and consequently everything remains unchanged. Cristina's words recount as the politics represents a system of power relations aimed at gaining control of resources and opportunities of majority group over the minority group (van Dijk, 2001). According to respondents also the attitude of the parties that have opposed resistance and have offered few opportunities for women helps to explain the low representation of women in politics (Guadagnini, 2005; Yirmibesoglu, 2008; Lovenduski and Guadagnini, 2010; Belle, 2012; De Simone et al., 2017). About this Sara, 60 years, says:I have encountered many difficulties in my political career. The political parties are governed by men, who support other men. Women do not vote for women because there are no women on the list and if they are there, they are already put in the last places, and they will never be elected. The parties are a male lobby. Sara and other respondents attribute the barriers encountered by women in politics to the discriminatory practices carried out by the parties (Lovenduski and Guadagnini, 2010). Sara describes the difficulties encountered in the party and the sense of exclusion generated by belonging to a minority. Sara's words perfectly convey the idea of the dynamics that develop in power institutions, always male dominance, helping to define the segregating phenomena of women. The lower representation of women in political institutions is explained by reference to the selection and recruitment methods in force in the various political systems and implemented by the various parties (Fornengo and Guadagnini, 1999; Belle, 2012; De Simone et al., 2017). The construction of male-dominated lists is fostered by party logics that build blocked lists where women are at the end of the list, without any chance of affirmation, according to the political gatekeeping phenomenon (Bain and Cummings, 2000). In this way, politics tends to exclude women by proposing predominantly male patterns and discriminatory practices (Annesley and Gains, 2010). Carlo, 45 years, explains the mechanisms of political parties that discourage women participation (Fox and Lawless, 2011) and contribute to the consolidation of hegemonic masculinity (Connell and Messerschmidt, 2005):In political parties, women are called by the secretaries, who are mostly men, in the election period. But after the election, when women did not take votes, they are no longer considered and they feel cheated. So, they say: do not go back there again." The existence of the glass ceiling (Morrison et al., 1992) in politics represents an invisible barrier that prevents women participation in politics and, on the contrary, privileges men. The prevalence of votes of men candidates becomes an obvious, logical, natural phenomenon, as told also by Giacomo, 38 years:There is a greater tendency by most people to vote for men, the system is structured in a sexist way. First, since women are less, there are ten women and ninety men; clearly, it is easier for a man to achieve, than for a woman. Second, a man wants to support himself, so he wants to keep that power and a woman who arrives so high would compromise the rules. Giacomo exemplifies as the vote reifies the dominant gender order according to specific logic of power. His discourse operates for the naturalization of a gender order socially accepted and shared, which mystifies the diversity in the distribution of power and inequality between women and men (Lazar, 2005, 2017). Politics-family unbalance The underrepresentation of women in politics is linked to family-work life balance (De Simone et al., 2018) as women's responsibility (Lovenduski and Guadagnini, 2010; De Simone et al., 2017). Some interviewees assert that women would not be interested in doing politics because they organize their lives around the family. For example, Giovanni, 38 years, say:I also understand that for a woman who decides to have a family and children it is more complicated because family work falls more on the woman than on the man. Women are interested in organising their lives around the family. Women should not be pretentious and have unrealistic demands toward politics, they must organize themselves with family and adapt to the rules of politics. Giovanni's extract describes a rigid division of gender roles and a strong unbalance in the distribution of family work as a disadvantage for women because of the limitation of time and resources for professional and political affirmation (Lazar, 2005; Lasio et al., 2017) and as a cause of the gender gap in politics (Guadagnini, 2005). This imbalance reflects the unequal distribution of power in gender relations (Inglehart and Norris, 2003; Lazar, 2017) that finds expression also in politics:Kayla (63 years): So, there is a problem when a woman with family decides to go into politics. Because we are good at giving ourselves the responsibility of all that is not all right. We feel guilty at work because we are not at home, and we feel guilty at home because we are not at work. It is evident by Kayla's discourse, the logic of the hegemonic power is internalized and shared by women themselves (Bourdieu, 1991), involved in that relationship of dominance-submissiveness dictated by patriarchal power, which places men and women in specific roles, public and private. The dichotomy between private domestic sphere, traditionally considered feminine, and political-public sphere, historically a men prerogative, is proposed again in the interviewees' discourses. The family-political life balance as women's responsibility is explained by the respondents referring both to "natural" both to "cultural" factors:Elisa (47 years): Many women do not present themselves as candidates because they would not be able to balance family and politics, unlike men, clearly, probably by nature, by the trend. It is normal to be like this, but there is the difference. There is a different way of approaching the family, but that is DNA.Carlo (45 years): Women have a natural limitation because they can become mothers. Being a mother influences life. A woman has different needs than a man; she is more careful toward the family and, therefore, feels more strongly the need to go back home - she wants to prepare the soup for the baby. She is worried, but these worries are a limitation in the practice of an administrator. Because you have to be free and distractions are not allowed. All this penalizes the woman. The focus is on the alleged inability of women to reduce in their favor domestic and family responsibilities. Although there is an awareness of family burdens on women, that affects their opportunities for growth, the organization of the system seems to be conceived as an essential and unchangeable rule. Respondents that adhere to this interpretation tend to minimize and naturalize (Lazar, 2005) inequalities and women disadvantage resulting from strong adherence to a dominant gender order, which places women and men, respectively, inside and outside the home, mystifying diversity in the distribution of power. However, as clearly expressed by Valeria, the explanation for family-political life balance as women's responsibility not only comes in terms of "natural" factors, but an important role is also played by "cultural" factors that reiterate the status quo, and are accepted and considerate as appropriate by women as well:Valeria (25 years): There is still the mentality of the woman that stays at home, cleans, takes care of the children and cannot do it late at night. The low participation of women in politics may also be caused by this. This depends on culture; for a man is different, it is inevitable! A man goes to a working dinner, but I know that my children are at home, and my "mother's heart" makes me come back home. The social, political, and cultural context offers a dominant gender order (Connell, 1987, 2016), proposing a "natural" division between gender and roles played by men and women. Women's domestic overload is cause and consequence of the exclusion of women from prestigious activities like politics, a vicious circle whereby the hegemonic patriarchal culture, associating the women to the home and the men to the work, lays the foundation to determine an asymmetry between gender (Walby, 1990). The characteristics of the cultural context are another factor that plays a key role in explaining the gender gap in politics (Fornengo and Guadagnini, 1999). In particular, Giacomo's reflection underlines a dichotomous division of roles (production - men and reproductive - women), as result of a hegemonic patriarchal culture. The politics and its "times" are described not as a natural and obvious dimension, but as subject to renegotiations and changes. Giacomo, 38 years, supports the role of "cultural" factors:Now it is happening that since the woman has to stay home and take care of the kids, the man goes to do politics, because the timing in politics is done for men. Our society has this idea about politics. We must change it [...] we must change the culture. We must educate children. We must create a new political establishment and the conditions so that things will be different in the future, otherwise, it is normal that in politics you will repeat what is the everyday life. This quotation identifies the social and cultural dimensions that influence common sense featured in an essentialist nuance and refers to the patriarchal culture that categorizes men and women according to hierarchical relationships of dominance and submission. This distribution of roles as told by Giacomo is the product of practices and power relations in a social context that conveys social expectations that reproduce existing power unbalances between the gender. Giacomo expresses the need to change the patriarchal hegemonic culture that influences a gender order also in the political life. Some respondents think that adequate reconciliation services would help to solve the problem of underrepresentation of women in Italian politics (Lovenduski and Guadagnini, 2010):Ambra (61 years): Here are no services, or at least not enough. A good plan of services that allow women to enter the labour market, just a systematic project that needs to go in the direction of eliminating disparities. The balancing of family life and political career is a woman prerogative because even currently women more than men are engaged both in the public and private spheres. According to our respondents, the lack of adequate support from government policies and inequities in the distribution of family work are among the causes of women's choice to not devote to institutional policy. The political and family systems are still structured for the men advantage and for the oppressive presence of a gender order (Connell, 1987, 2016) that still put the exercise of social power in the hands of elite groups or specific men institutions (Lazar, 2005, 2007, 2017). This paper explores the topic of the gender gap in politics through a discourse analysis of a group of Italian politicians and shows the patterns whereby a dominant gender order is constructed and reproduced. Discourse analysis discloses some interpretive repertoires used by men and women to confirm and reinforce the hegemonic gender order. Data analysis confirms that the Italian context is strongly characterized by a dominant gender order that assumes that most men are involved in the public sphere and in paid employment, while women are more likely to operate in the private sphere and family work. This gender order defines the boundaries of symbolic practices and materials of masculinity and femininity, positioning women and men in different positions of power. While the presence of women in politics breaks the dominant gender order (Connell, 1987, 2016), the other gender practices in politics tend to rebuild that order, repositioning women in subordinate roles. The participants tend to either absolve or blame women for gender inequality in politics through different interpretive repertoires: ascribe to women the responsibility of their low representation in politics relying on women's disinterest toward politics; absolve women attributing the responsibility of the gender gap in politics to the reproduction by politics of masculine rules, rhythms and symbols that do not correspond to the characteristics and needs of women. The participants' discourses also highlight how work-family imbalanced distribution influences women's participation in politics attributing this both to natural factors and cultural factors. These discourses highlight an ideological dilemma (Billig et al., 1988) with respect to the causes that would lead to the clear separation between private and public spheres and, consequently, the different roles assumed between women and men. The findings of this paper allow us to understand the processes by which dominant symbolic gender orders are reconstructed and show how power relations are permeated by practices of hegemonic masculinity which reinforce this order. Discourses on low representation of women in politics contributes to the consolidation of hegemonic masculinity that conserves and legitimizes men power and women subordination (Connell, 1987; Connell and Messerschmidt, 2005). Hegemonic patriarchal culture continues to act undisturbed and, through everyday discourses about natural gender binarism, legitimizes and becomes the direct expression of the power relationship between men and women (Walby, 1990). Behind the rhetoric of the growing democratization of modern organizations, gender inequalities have become more powerful because they act through subtle discursive mechanisms, much more difficult to unmask (Fairclough, 1992). Rhetoric of equality, framed into a broader context of male hegemony, guides their sense-making into denying gender inequalities in politics. Those discourses represent a subtle, pervasive, insidious and seemingly innocuous form of power in modern societies (Lazar, 2005). In fact, processes of naturalization of gender inequality (Lazar, 2005) are proposed also in those discursive, apparently egalitarian, contexts that absolve women. Women in dilemma are the main witnesses (unaware) that gender differences are built in discourses and acted out in daily practices (Lazar, 2017). For many women interviewed, it was very difficult to recognize their position of inequality because it is produced by a latent ideological and hegemonic power (Lukes, 2005). This has the effect of driving choices and waivers and systematically and unknowingly supports men as a privileged social subject in the exercise of political and social power. This study has some limitations that may be overcome with further research. First, the small sample size and voluntary participation of politicians means that our results are not generalizable. Thus, this research project should be replicated using an expanded sample and comparing local and national politicians. A second limitation is the exclusive reliance on qualitative data, which in future should be supplemented with quantitative data. Despite these limitations, the findings highlight the importance of exploring issues relating to the gender gap in politics and offer some suggestions for implementing actions to promote gender equality in this area. The rebalancing of gender could be promoted by quotas guaranteeing women a percentage of posts in the political system. Pink quotas, along with actions to promote awareness of gender discrimination, could be a response to women's disinterest in politics. The presence of women as candidates or in political office also assumes a symbolic function and conveys the unequivocal message that politics "is not for men only." Previous studies have shown that the presence of women in institutional political roles increases other women's interest in politics (Burns et al., 2001). The virtuous effect of quotas in countries that have adopted them is widely documented in the literature (Dahlerup and Freidenvall, 2010). Also, work-life balance policies (such as monetary transfers, and children and care services) aimed at alleviating the burden of women's familial responsibilities could increase women's participation in politics by promoting gender equality inside and outside the family (Gornick and Meyers, 2003; Lewis, 2009). Indeed, some studies have shown a relationship between different welfare regimes and the way in which men and women build their life paths (Gornick and Meyers, 2003; Geist, 2005). To promote social change and challenge the dominant gender order it is necessary not only to take into account psychosocial and cultural factors, but also to leverage institutional factors. The research allows disclosure of the social practices based on gender and the patriarchal ideology that contributes to the construction and preservation of women political disadvantage (Holmes and Marra, 2010). The feminist critique of political practices and social relations based on gender aims ultimately at inspiring a social change. The status quo is challenged in favor of a feminist humanist vision of a just society in which gender does not predetermine or mediate relations with others or our sense of who we are and what we can become. Discourse analysis, which reveals how power works to support oppressive social structures and relations, is in itself a form of "analytical resistance" (van Dijk, 1991) and contributes to conducting power struggles and social change (Lazar, 2007). The study shows how discourses can represent an instrument of resistance to change and reinforcement of hegemonic gender order and offers some suggestions for future research. Future studies could use discourse analysis and the contribution of the critical feminist perspective as instruments to raise awareness to understand the mechanisms underlying gender inequalities and the gender gap also in other domains of power in which women are still excluded.
|
The results of this paper underline the tendency to either absolve or blame women for gender inequality in politics through different interpretative repertoires: "Women's disinterest toward politics," "Politics as masculine context" and "Politics-family unbalance." The analysis allowed to unravel the way in which the discursive practices create and reproduce the hegemonic gender order in politics.
|
[SECTION: Value] Despite the ongoing increase of women in the top positions of hierarchy, they continue to be underrepresented in politics occupying 19.5 percent of seats worldwide, 22.8 percent in Europe, 22.6 percent in the Americas and 42.0 percent in Nordic countries (Inter-Parliamentary Union, 2017). In 2017, Italy ranks 44th out of 193 in the world classification compiled by the Inter-Parliamentary Union (2017), with a participation rate of Italian women to res public between 28 and 31 percent. Although the number of women elected in central institutions is increased compared to the past, in Italy women are still low represented in politics (Massa, 2013), and the number of women politicians decreases in the hierarchy of power (Fornengo and Guadagnini, 1999; Bonomi et al., 2013). Every president of the Italian Republic has been a man and women who have served in the institutional role of president of the Chamber of the Italian Republic from 1948 to date are 3 out of 14 (Italian Parliament, 2017). This position of women in politics, characterized by both a numerical increase compared to the past and a numerical decrease at the top of the power hierarchy, is well represented by the current government. The XVII Legislature of government, which started in 2013 and close to conclusion, is the legislature with the largest number of women in Parliament, in fact the women account for 32 percent of deputies and 30 percent of senators, but only 16 percent of women cover the roles of leader, president of the commission and the office of the presidency. Analyzing the composition of the Italian Parliament, it is clear that the disparity between men and women in politics is recognizable when roles are more prestigious. This condition is also confirmed at the regional level: women are present in the joint sessions in 29 percent of the councils and hold key positions (departments and offices with greater decision-making power) in 18 percent of the councils, while only 10 percent is president of the region and only 2 percent are leaders of the municipalities in the capital city (Openpolis, 2015). The Italian case is particularly interesting, as Italy is strongly characterized by conditions of gender inequality. In addition to the gender gap in political participation, as evidenced by the above data, in Italy the level of women's participation in the paid labor market is still very low (ISTAT, 2016): 54.4 percent compared with an OECD average of 62.6 percent (OECD, 2015). Although Italian women have a higher level of education than that of Italian men (ISTAT, 2016), gender segregation persists. Working women are concentrated in the tertiary sector and low-skilled professions (ISTAT, 2015b), with only 35 percent holding a managerial position. In this regard, Italy is ranked 82nd out of 142 according to the World Economic Forum's classification (WEF, 2014). In addition, in Italy the division of family work remains drastically imbalanced to women's disadvantage (Eurostat, 2011; ISTAT, 2015a; CNEL, 2014). Research shows that the time Italian mothers dedicate to family work is much higher than that dedicated by fathers, even in dual-earner families (Eurostat, 2011; ISTAT, 2015a). This strict division of family roles is the main obstacle to women reaching decision-making positions in politics, because Italy also continues to invest inadequately in services to support families (Panzeri and Viale, 2016). In Italy, maternity continues to have a significant impact on unemployment rates and the early exit of women from the labor market (CNEL, 2014). In Italy, therefore, the underrepresentation and low appreciation of women in society and politics together constitute a serious problem whose causes it is vital to investigate in order to intervene adequately (Lasio et al., 2013; Pacilli et al., 2012). Despite the substantial social and cultural transformations of men and women roles, women continue to be excluded or impeded in highly competitive and masculine hierarchical domains, such as politics (Lazar, 2005). Men continue to be the majority in power contexts that appear inaccessible and strewn with obstacles for women to grow in (Fox and Lawless, 2011; Paxton and Hughes, 2015). Although women have achieved full juridical parity and participate in the political life of the country, conditions of strong inequality still persist and limit their growth in contexts that are considered places of power (Davidson and Burke, 2016; Lazar, 2017). Several studies have investigated women in leadership (e.g. Fitzsimmons et al., 2014; Mavin and Grandy, 2016) and explored persistent systems of discrimination and inequality (e.g. Kelan, 2008; Lewis et al., 2017). The matter of women in politics continues to represent the context in which a gendered division of power has its highest expression and constitutes a macroscopic disadvantage in terms of "presence" and in the acquisition of roles (Childs, 2006). Many pieces of evidence attest the existence of strong constraints to full and complete participation women's policy. Politics is one area where male domination over women and resistance to change are fully expressed (Bourdieu, 1998). Politics tends to reproduce male models and rules presenting itself as an unwelcoming domain for women. Women then, although they can rely on the same opportunity as men, are blocked by systems of exclusion rooted in the culture of many societies (Inglehart and Norris, 2003; Krook, 2010) and very often exclude themselves. Recent research (Lasio and Serri, 2017; Lasio et al., 2018) has also showed how the resistance opposed in Italy to non heterosexual families reinforces normative standards on motherhood, thus contributing to maintain those asymmetries at the origin of women's invisibility as active citizens (Amancio and Oliveira, 2006). Women's underrepresentation in politics reproduces the imbalanced and unequal gender relations (Lovenduski, 2001; Santos et al., 2013). Since the 1990s, some scholars, mostly women from Anglo-Saxon and Northern European countries (Childs and Krook, 2006; Krook, 2010), have begun to investigate the relationship between women and politics, focusing especially on the numerical presence of women in the government and in the councils, on the trends and mechanisms of recruitment and selection of women by the parties (Kenny, 2007; Celis et al., 2015; Kenny and Verge, 2015). These studies found that women continue to be underrepresented in politics and despite there being a much greater number of women in elected positions compared to the past, this it is not a substantive representation and cannot be called substantive equality (Childs and Krook, 2006). The literature on this topic has addressed the potential factors that may explain this underrepresentation, and has highlighted the influence of patriarchal culture, the division of family roles, the lack of childcare, the resisting attitude of the parties and the fragmentation of women's movement (Lovenduski and Guadagnini, 2010; Bonomi et al., 2013). Others have analyzed the relationship between discourse, gender and political representation, focusing on political discourse and the construct of gender identity (Shaw, 2000; Walsh, 2001) and gender citizenship in politics (Poggio, 2004; Belle, 2012). This study is positioned in the field that investigates the relationship between women and politics using a discursive perspective. The research was conducted within a theoretical framework referred to as a discourse analysis (Potter and Wetherell, 1987), and to contribution from the critical feminist perspective (Lazar, 2007, 2017; Holmes and Marra, 2010), and proposes to understand the mechanisms that regulate the contemporary gender order in politics through the analysis of the discourses of a group of Italian politicians. The focus on discourses of politicians allows us to analyze the processes by which dominant symbolic gender order is reconstructed in politics and to show how the distribution of power between gender is acted out through practices of hegemonic masculinity which excludes women from the real seats of power. This study offers a reflection on gender inequalities in political participation, starting with a review of the literature and the results of research conducted on Italian politicians. The contribution of this paper is to shed light on continuing discourses concerning politics as a predominantly male context. We conclude by highlighting the study's theoretical and practical implications for enhancing equality in politics and further developing research in this area. The studies on women and politics continue to focus on women' numbers in politics and on sex differences, and with difficulty, they are able to embrace the concept of gender (Kenny, 2007) as a category that structures and makes sense of particular social practices. However, feminist studies have developed complex theories of gender, including both patriarchal and discursive conceptions of power. The power inequality between gender relies on the concept of patriarchy, a system of social structures and practices of domination founded in the subordination of women by men (Walby, 1990). Gender relations are power relations involving formal public structures as politics and private structures as the family. The distribution of power between gender, based on hegemonic patriarchal culture, which associates naturally the female to the house and the male to the labor and political activities, has determined an asymmetry in which the historical forced absence of women from prestigious positions, such as politics, would involve a vicious circle whereby the greater female participation in domestic work has become cause and consequence of their exclusion (Walby, 1990). Through the internalization of gendered norms produced routinely in discourses of everyday life, this gap of power between men and women has become invisible, misrecognized and recognized as legitimate and natural (Bourdieu, 1991), contributing to the consolidation of "hegemonic masculinity" that preserves, legitimizes and naturalizes men power and, consequently, women subordination (Connell, 1987, 2016; Connell and Messerschmidt, 2005). Discourse, therefore, plays a significant role in the reproduction of dominance, which is considered as an exercise of social power by elites, institutions or groups that have the effect of increasing social, political, class and gender inequality (van Dijk, 1993, 2011). Social power is enacted, reproduced or legitimized by the discourse of dominant groups or institutions (van Dijk, 1996). Relations of power and dominance (Bourdieu, 1991) can discursively challenge the interests at stake. Therefore, hegemonic power relations are discursively produced, perpetuated, negotiated and challenged. To share or refuse these contents means to contribute to the reproduction and maintenance of the social order and also to resist and transform that order (Lazar, 2005, 2017). Gender is socially constructed through discourses (Wodak, 1997), and the power relations associated to, and reproduced by, such discourses (Wodak, 1997). Gender expresses the relationships of power that constitute society: women and men, through their performances, can reify or on the contrary challenge the system of relations and gender privileges (Holmes and Marra, 2010). Feminist critical discourse perspective offers important contribution to theorize and analyze "the particularly insidious and oppressive nature of gender as an omni-relevant category in most social practices" (Lazar, 2007, p. 3). In particular, feminist critical discourse analysis aims to demystify the interrelationships of gender and power, in discourse and critiques the discourses that sustain a patriarchal social order, that systematically privilege men and disadvantage women (Holmes and Marra, 2010). The disclosing of the discursive strategies through which power structures and supports oppressive social relations is a form of "analytical resistance" that help to support the struggles of protest and social change (van Dijk, 1991). Feminist critical discourse contributes to social transformation through the focus on repressive norms and restrictive stereotypes reified and reinforced by routine discourses of people (Holmes and Marra, 2010). This study is based on 30 biographical interviews conducted with local Italian politicians, 15 of whom were men, 15 women. We chose to involve local politicians because figures on the number of women in local Italian institutions are lower than those at the national level (ISTAT, 2017). The average age of the politicians interviewed was 43 years (DS=12.5; range=24-64). The research was carried out in Italy and recruitment took place through the involvement of political parties. The face-to-face interviews were conducted by one of the authors, lasted between 60 and 90 min, and focused on two central topics: their experiences in politics and meaningful episodes in their political life, and the barriers that may hinder women's political participation. Respondents were asked to illustrate their political career starting from the reasons that led them to choose politics and to discuss the relationship between politics and gender, taking into consideration their own experiences and reflecting on the meanings attributed to the experiences. The study was carried out in accordance with the Declaration of Helsinki. Participants were informed of the aims, methods and other ethical aspects relevant to the study. After they had given informed consent, all interviews were digitally recorded and transcribed verbatim. Participants' confidentiality and anonymity have been protected through the use of pseudonyms. For the first step of the analysis, the authors independently examined the transcripts to identify the emerging main themes and recurrent repertoires. Subsequently, the researchers worked together on the discourse analysis (Billig, 1987; Potter and Wetherell, 1987), in order to come to an agreement on the identified interpretative repertoires and chosen illustrative key quotes drawn from the different interviews. Particular attention is given to discourse, as social action through which men and women can reify or challenge the system of power relations, gender imbalances and privileged places in which these mechanisms can be observed and unmasked (Holmes and Marra, 2010). Data analysis focused on the interpretative repertoires and ideological dilemmas. The repertoires are "clusters of terms organized around a central metaphor, often used with grammatical regularity" (Wiggins and Potter, 2008, p. 74) not defined by specific words or phrases, but by a textual constellation that refers to a specific theme (Potter and Wetherell, 1987). The ideological dilemmas (Billig et al., 1988) indicate the intrinsic contradictory nature of common sense and highlight the rhetorical origin of attitudes toward gender as they emerge in contexts where there is an argument. "[...] when contradictory positions overlap they provide a basis for awareness and reflexivity, just as they lead to problematization and change" (Fairclough, 1995, p. 82). Data analysis show how the dominant gender order is performed and reinforced through the discursive practices of the politicians interviewed. The participants tell their experience as politicians and produce discourses that clarify the norms that regulate the political field. These discourses are part of a wider social context in which choices, roles and careers are gendered. The politicians interviewed offer several arguments about women' low participation in politics, referring to different degrees of responsibility or states of helplessness. Results show the participants' tendency to either absolve or blame women for gender inequality in politics through different interpretive repertoires, which can be, in some cases, dilemmatics (Billig et al., 1988). In particular, the interpretative repertoire "Women's disinterest toward politics" tends to blame women for their low participation in politics, whilst the interpretative repertoire "Politics as masculine context" tends to absolve women and to attribute the responsibility of the gender gap in politics to the hegemonic social context and to the attitude of the parties. A third repertoire "Politics-family unbalance" identifies in the unequal distribution of family work the principal responsibility of gender inequalities in politics. The arguments referred to the latter repertoire concern the debate between nature and culture, fluctuating between the essentialist positions on the one hand, and progressive positions, on the other hand. Women's disinterest toward politics Though among the politicians interviewed the idea that men and women can perform the same roles in politics is common, they explain the women's low representation in politics referring to the lack of interest of women toward politics, the women voluntary abstention and their minor will to be involved (De Simone et al., 2017). The issue of women's lack of interest is built on beliefs that recall individual factors (Fornengo and Guadagnini, 1999) as "a choice for women" dictated by a lack of desire to engage. This choice would lead women to voluntarily keep away from the places of res public:Marco (39 years): Women can play the same role in politics, whatever role males can play; there is no difference. It is a choice for women who do not participate in politics. They are not interested. [...] because there is no woman that enforce herself! Marco does not display explicit conflicts that were unfavorable to full equality between men and women in politics, however affirming the claim of women's disinterest in politics, supports the counterarguments that identify in women the cause of the impossibility to put such equality into practice. The argumentation relying on women's responsibility for the gender gap in politics becomes even stronger when greater capacity and greater potential is attributed to women themselves:Giovanni (41 years): women voluntarily choose not to do politics [...] when they are brought into play, they have produced good results just as good or better than others. I think that women have all the possibilities and conditions for doing politics, and even in the best way. The notion of choice is another key element to the understanding of the gender gap in politics. The reference to women's disinterest in politics being a reflection of their own voluntary choice implies a lack of recognition of the social inequalities affecting women's access to politics. Reference to women having free choice moves attention away from the responsibility of the system to the responsibilities of the individual woman, indicating that women's choices are embedded within cultural, social and economic conditions that structure gender dynamics (Lewis et al., 2017). The quotations above show as the power, deeply ingrained, acts automatically and unconsciously, since is foregone and "buried" under consolidated ideologies (Lukes, 2005). Marco and Giovanni's words show how the gender order is preserved through the third dimension of power that prevents some issues from being expressed. Gender conflict is latent because the subjects holding positions of power, in this case two men politicians, do not explicitly express their interests. The ideologies influence people's wishes and thoughts, the "ideological power," the "real power," acts indirectly and unconsciously, influencing the social actors and ensures, in this way, their acquiescence, even when it goes against their self-interest (Lukes, 2005). Indeed Valeria (25 years), says:Women themselves are afraid of not being up. It depends very much on the lack of interest or fear to get involved because, in my opinion, we are the cause of the prejudice! I speak often of meritocracy, I hate rose quotas! There is no need for rose quotas. Valeria affirms that women have not the need for rose quotas. The rose quotas refer to those legal instruments aimed to protect gender equality within the representative bodies through different calculation mechanisms and allocation of seats. Valeria, with her statements, contributes to maintaining the status quo, by underplaying and silencing structural inequalities. Once again, it is not the politics to reiterate a gender order that does not allow women's participation, but women themselves are responsible for their own exclusion. For many respondents, including women, and as it is clear from Valeria's discourse, affirmative action promoted in favor of gender equity in political participation is perceived as offensive and not qualifying (Santos et al., 2016). Valeria denies gender inequalities, thereby oppositions are avoided, whereas asymmetric power relations between gender persist. The issue of voluntary women's abstention from politics justifies persistent power inequalities as the result of women's individual choices and underlies the non-recognition of gender social inequalities (Lewis, 2014). Marco, Giovanni and Valeria's extracts emphasize the individual responsibilities with the effect of concealing structural discrimination mechanisms and the inequality of power between men and women. Through those discourses seem to emerge a lack of awareness of the mechanisms of male dominance that push women, without their awareness, to adapt their hopes and their aspirations to opportunities, and consequently not to undertake actions that are not expected of them (Bourdieu, 1998). The order of things, in fact, is not a natural order, but a building of the world with which man pleases his thirst for domination, a vision so exclusive that the same women who are the victims have integrated in their own way of thinking by accepting their unconscious inferiority (Bourdieu, 1998). Valeria, as a woman, can be considered victims and author of gender relations of power acting in social practices upholding hegemonic culture. The hegemonic masculinity, as a pattern of practices that permit men's dominance over women (Connell and Messerschmidt, 2005), inspires discourses that aim to represent a reality of gender equality. Marco, Giovanni and Valeria's discourses lack an acknowledgment of this systematic imbalance referred to the rhetoric of sameness between gender and denial of inequalities. The hegemonic masculinity (Connell and Messerschmidt, 2005) inspires the discourses of equality that deny gender inequalities slowing social change (De Simone et al., 2015; De Simone and Scano, 2017). Politics as masculine context Interviewed politicians' discourses identify the cause of the women's low representation in politics in the sexism that characterizes the parties and politics in general, using the interpretive repertoire "Politics as masculine context." Women's access to some areas continues to be difficult and to represent a source of inequality especially in the "real" places of power (Lazar, 2005), as politics. Several interviewees emphasize the persistence of the gender gap in politics highlighting that despite the number of women in politics has increased compared to the past, high positions continue to be held predominantly by men reiterating a gender order that relegates women to subordinate positions (Connell, 1987, 2016):Cristina (53 years): Certain roads were blocked to women. There is a social structure that does not allow access to women in politics. Since it is been a long time, and now it seems that everything is established a norm. Women, despite the battles, block themselves because this is the social structure which they know and consequently everything remains unchanged. Cristina's words recount as the politics represents a system of power relations aimed at gaining control of resources and opportunities of majority group over the minority group (van Dijk, 2001). According to respondents also the attitude of the parties that have opposed resistance and have offered few opportunities for women helps to explain the low representation of women in politics (Guadagnini, 2005; Yirmibesoglu, 2008; Lovenduski and Guadagnini, 2010; Belle, 2012; De Simone et al., 2017). About this Sara, 60 years, says:I have encountered many difficulties in my political career. The political parties are governed by men, who support other men. Women do not vote for women because there are no women on the list and if they are there, they are already put in the last places, and they will never be elected. The parties are a male lobby. Sara and other respondents attribute the barriers encountered by women in politics to the discriminatory practices carried out by the parties (Lovenduski and Guadagnini, 2010). Sara describes the difficulties encountered in the party and the sense of exclusion generated by belonging to a minority. Sara's words perfectly convey the idea of the dynamics that develop in power institutions, always male dominance, helping to define the segregating phenomena of women. The lower representation of women in political institutions is explained by reference to the selection and recruitment methods in force in the various political systems and implemented by the various parties (Fornengo and Guadagnini, 1999; Belle, 2012; De Simone et al., 2017). The construction of male-dominated lists is fostered by party logics that build blocked lists where women are at the end of the list, without any chance of affirmation, according to the political gatekeeping phenomenon (Bain and Cummings, 2000). In this way, politics tends to exclude women by proposing predominantly male patterns and discriminatory practices (Annesley and Gains, 2010). Carlo, 45 years, explains the mechanisms of political parties that discourage women participation (Fox and Lawless, 2011) and contribute to the consolidation of hegemonic masculinity (Connell and Messerschmidt, 2005):In political parties, women are called by the secretaries, who are mostly men, in the election period. But after the election, when women did not take votes, they are no longer considered and they feel cheated. So, they say: do not go back there again." The existence of the glass ceiling (Morrison et al., 1992) in politics represents an invisible barrier that prevents women participation in politics and, on the contrary, privileges men. The prevalence of votes of men candidates becomes an obvious, logical, natural phenomenon, as told also by Giacomo, 38 years:There is a greater tendency by most people to vote for men, the system is structured in a sexist way. First, since women are less, there are ten women and ninety men; clearly, it is easier for a man to achieve, than for a woman. Second, a man wants to support himself, so he wants to keep that power and a woman who arrives so high would compromise the rules. Giacomo exemplifies as the vote reifies the dominant gender order according to specific logic of power. His discourse operates for the naturalization of a gender order socially accepted and shared, which mystifies the diversity in the distribution of power and inequality between women and men (Lazar, 2005, 2017). Politics-family unbalance The underrepresentation of women in politics is linked to family-work life balance (De Simone et al., 2018) as women's responsibility (Lovenduski and Guadagnini, 2010; De Simone et al., 2017). Some interviewees assert that women would not be interested in doing politics because they organize their lives around the family. For example, Giovanni, 38 years, say:I also understand that for a woman who decides to have a family and children it is more complicated because family work falls more on the woman than on the man. Women are interested in organising their lives around the family. Women should not be pretentious and have unrealistic demands toward politics, they must organize themselves with family and adapt to the rules of politics. Giovanni's extract describes a rigid division of gender roles and a strong unbalance in the distribution of family work as a disadvantage for women because of the limitation of time and resources for professional and political affirmation (Lazar, 2005; Lasio et al., 2017) and as a cause of the gender gap in politics (Guadagnini, 2005). This imbalance reflects the unequal distribution of power in gender relations (Inglehart and Norris, 2003; Lazar, 2017) that finds expression also in politics:Kayla (63 years): So, there is a problem when a woman with family decides to go into politics. Because we are good at giving ourselves the responsibility of all that is not all right. We feel guilty at work because we are not at home, and we feel guilty at home because we are not at work. It is evident by Kayla's discourse, the logic of the hegemonic power is internalized and shared by women themselves (Bourdieu, 1991), involved in that relationship of dominance-submissiveness dictated by patriarchal power, which places men and women in specific roles, public and private. The dichotomy between private domestic sphere, traditionally considered feminine, and political-public sphere, historically a men prerogative, is proposed again in the interviewees' discourses. The family-political life balance as women's responsibility is explained by the respondents referring both to "natural" both to "cultural" factors:Elisa (47 years): Many women do not present themselves as candidates because they would not be able to balance family and politics, unlike men, clearly, probably by nature, by the trend. It is normal to be like this, but there is the difference. There is a different way of approaching the family, but that is DNA.Carlo (45 years): Women have a natural limitation because they can become mothers. Being a mother influences life. A woman has different needs than a man; she is more careful toward the family and, therefore, feels more strongly the need to go back home - she wants to prepare the soup for the baby. She is worried, but these worries are a limitation in the practice of an administrator. Because you have to be free and distractions are not allowed. All this penalizes the woman. The focus is on the alleged inability of women to reduce in their favor domestic and family responsibilities. Although there is an awareness of family burdens on women, that affects their opportunities for growth, the organization of the system seems to be conceived as an essential and unchangeable rule. Respondents that adhere to this interpretation tend to minimize and naturalize (Lazar, 2005) inequalities and women disadvantage resulting from strong adherence to a dominant gender order, which places women and men, respectively, inside and outside the home, mystifying diversity in the distribution of power. However, as clearly expressed by Valeria, the explanation for family-political life balance as women's responsibility not only comes in terms of "natural" factors, but an important role is also played by "cultural" factors that reiterate the status quo, and are accepted and considerate as appropriate by women as well:Valeria (25 years): There is still the mentality of the woman that stays at home, cleans, takes care of the children and cannot do it late at night. The low participation of women in politics may also be caused by this. This depends on culture; for a man is different, it is inevitable! A man goes to a working dinner, but I know that my children are at home, and my "mother's heart" makes me come back home. The social, political, and cultural context offers a dominant gender order (Connell, 1987, 2016), proposing a "natural" division between gender and roles played by men and women. Women's domestic overload is cause and consequence of the exclusion of women from prestigious activities like politics, a vicious circle whereby the hegemonic patriarchal culture, associating the women to the home and the men to the work, lays the foundation to determine an asymmetry between gender (Walby, 1990). The characteristics of the cultural context are another factor that plays a key role in explaining the gender gap in politics (Fornengo and Guadagnini, 1999). In particular, Giacomo's reflection underlines a dichotomous division of roles (production - men and reproductive - women), as result of a hegemonic patriarchal culture. The politics and its "times" are described not as a natural and obvious dimension, but as subject to renegotiations and changes. Giacomo, 38 years, supports the role of "cultural" factors:Now it is happening that since the woman has to stay home and take care of the kids, the man goes to do politics, because the timing in politics is done for men. Our society has this idea about politics. We must change it [...] we must change the culture. We must educate children. We must create a new political establishment and the conditions so that things will be different in the future, otherwise, it is normal that in politics you will repeat what is the everyday life. This quotation identifies the social and cultural dimensions that influence common sense featured in an essentialist nuance and refers to the patriarchal culture that categorizes men and women according to hierarchical relationships of dominance and submission. This distribution of roles as told by Giacomo is the product of practices and power relations in a social context that conveys social expectations that reproduce existing power unbalances between the gender. Giacomo expresses the need to change the patriarchal hegemonic culture that influences a gender order also in the political life. Some respondents think that adequate reconciliation services would help to solve the problem of underrepresentation of women in Italian politics (Lovenduski and Guadagnini, 2010):Ambra (61 years): Here are no services, or at least not enough. A good plan of services that allow women to enter the labour market, just a systematic project that needs to go in the direction of eliminating disparities. The balancing of family life and political career is a woman prerogative because even currently women more than men are engaged both in the public and private spheres. According to our respondents, the lack of adequate support from government policies and inequities in the distribution of family work are among the causes of women's choice to not devote to institutional policy. The political and family systems are still structured for the men advantage and for the oppressive presence of a gender order (Connell, 1987, 2016) that still put the exercise of social power in the hands of elite groups or specific men institutions (Lazar, 2005, 2007, 2017). This paper explores the topic of the gender gap in politics through a discourse analysis of a group of Italian politicians and shows the patterns whereby a dominant gender order is constructed and reproduced. Discourse analysis discloses some interpretive repertoires used by men and women to confirm and reinforce the hegemonic gender order. Data analysis confirms that the Italian context is strongly characterized by a dominant gender order that assumes that most men are involved in the public sphere and in paid employment, while women are more likely to operate in the private sphere and family work. This gender order defines the boundaries of symbolic practices and materials of masculinity and femininity, positioning women and men in different positions of power. While the presence of women in politics breaks the dominant gender order (Connell, 1987, 2016), the other gender practices in politics tend to rebuild that order, repositioning women in subordinate roles. The participants tend to either absolve or blame women for gender inequality in politics through different interpretive repertoires: ascribe to women the responsibility of their low representation in politics relying on women's disinterest toward politics; absolve women attributing the responsibility of the gender gap in politics to the reproduction by politics of masculine rules, rhythms and symbols that do not correspond to the characteristics and needs of women. The participants' discourses also highlight how work-family imbalanced distribution influences women's participation in politics attributing this both to natural factors and cultural factors. These discourses highlight an ideological dilemma (Billig et al., 1988) with respect to the causes that would lead to the clear separation between private and public spheres and, consequently, the different roles assumed between women and men. The findings of this paper allow us to understand the processes by which dominant symbolic gender orders are reconstructed and show how power relations are permeated by practices of hegemonic masculinity which reinforce this order. Discourses on low representation of women in politics contributes to the consolidation of hegemonic masculinity that conserves and legitimizes men power and women subordination (Connell, 1987; Connell and Messerschmidt, 2005). Hegemonic patriarchal culture continues to act undisturbed and, through everyday discourses about natural gender binarism, legitimizes and becomes the direct expression of the power relationship between men and women (Walby, 1990). Behind the rhetoric of the growing democratization of modern organizations, gender inequalities have become more powerful because they act through subtle discursive mechanisms, much more difficult to unmask (Fairclough, 1992). Rhetoric of equality, framed into a broader context of male hegemony, guides their sense-making into denying gender inequalities in politics. Those discourses represent a subtle, pervasive, insidious and seemingly innocuous form of power in modern societies (Lazar, 2005). In fact, processes of naturalization of gender inequality (Lazar, 2005) are proposed also in those discursive, apparently egalitarian, contexts that absolve women. Women in dilemma are the main witnesses (unaware) that gender differences are built in discourses and acted out in daily practices (Lazar, 2017). For many women interviewed, it was very difficult to recognize their position of inequality because it is produced by a latent ideological and hegemonic power (Lukes, 2005). This has the effect of driving choices and waivers and systematically and unknowingly supports men as a privileged social subject in the exercise of political and social power. This study has some limitations that may be overcome with further research. First, the small sample size and voluntary participation of politicians means that our results are not generalizable. Thus, this research project should be replicated using an expanded sample and comparing local and national politicians. A second limitation is the exclusive reliance on qualitative data, which in future should be supplemented with quantitative data. Despite these limitations, the findings highlight the importance of exploring issues relating to the gender gap in politics and offer some suggestions for implementing actions to promote gender equality in this area. The rebalancing of gender could be promoted by quotas guaranteeing women a percentage of posts in the political system. Pink quotas, along with actions to promote awareness of gender discrimination, could be a response to women's disinterest in politics. The presence of women as candidates or in political office also assumes a symbolic function and conveys the unequivocal message that politics "is not for men only." Previous studies have shown that the presence of women in institutional political roles increases other women's interest in politics (Burns et al., 2001). The virtuous effect of quotas in countries that have adopted them is widely documented in the literature (Dahlerup and Freidenvall, 2010). Also, work-life balance policies (such as monetary transfers, and children and care services) aimed at alleviating the burden of women's familial responsibilities could increase women's participation in politics by promoting gender equality inside and outside the family (Gornick and Meyers, 2003; Lewis, 2009). Indeed, some studies have shown a relationship between different welfare regimes and the way in which men and women build their life paths (Gornick and Meyers, 2003; Geist, 2005). To promote social change and challenge the dominant gender order it is necessary not only to take into account psychosocial and cultural factors, but also to leverage institutional factors. The research allows disclosure of the social practices based on gender and the patriarchal ideology that contributes to the construction and preservation of women political disadvantage (Holmes and Marra, 2010). The feminist critique of political practices and social relations based on gender aims ultimately at inspiring a social change. The status quo is challenged in favor of a feminist humanist vision of a just society in which gender does not predetermine or mediate relations with others or our sense of who we are and what we can become. Discourse analysis, which reveals how power works to support oppressive social structures and relations, is in itself a form of "analytical resistance" (van Dijk, 1991) and contributes to conducting power struggles and social change (Lazar, 2007). The study shows how discourses can represent an instrument of resistance to change and reinforcement of hegemonic gender order and offers some suggestions for future research. Future studies could use discourse analysis and the contribution of the critical feminist perspective as instruments to raise awareness to understand the mechanisms underlying gender inequalities and the gender gap also in other domains of power in which women are still excluded.
|
The paper contributes to an understanding of women's underrepresentation in politics and offers causes for reflection on a phenomenon that has profound implications for our society.
|
[SECTION: Purpose] As the popularity of strategic alliances is increasing and such alliances are becoming an integral component in business development, attention in the research community has moved towards an exploration of their role in firm performance and innovation. Formal and informal interactions with external actors have long being argued to play a fundamental role in firm innovation (Von Hippel, 1988; Frankort et al., 2012; Rice et al., 2012; Colombo et al., 2011; Demirkan, 2018). However, it is accepted that alliances carry coordination costs and risks of misappropriation, and these frequently diminish chances of success or preclude full acquisition of anticipated benefits (e.g. De Man and Duysters, 2005; Gkypali et al., 2017; Faems et al., 2010). A substantial body of literature finds a positive relationship between the extent of alliances and firm innovation performance, whilst other work identifies diminishing and even negative returns (Hoang and Rothaermel, 2005; Sampson, 2005; Laursen and Salter, 2006; Rothaermel and Deeds, 2006). As a result, a growing strand in the literature has examined the factors enabling firms to generate and capture value from alliances. One such factor is alliance experience accumulation: as firms develop greater experience in managing alliances they become better at coordinating cross-organisational tasks and knowledge flows (e.g. Sampson, 2005). Another factor that can improve performance in alliances is developing formal and codified processes and routines for alliance management (e.g. the establishment of dedicated alliance functions), which are argued to capture, or to be reflective of, firm-specific alliance management capabilities (Kale et al., 2002; Kale and Singh, 2009, 2007; Heimeriks and Duysters, 2007; Sampson, 2005; Schreiner et al., 2009; Di Guardo and Harrigan, 2016; Shukla and Mital, 2018). Whilst these contributions offer useful insights, much the greater part of current theorising has been constructed via the use of cross-sectional data. Longitudinal explorations are scarce, thus we lack a nuanced understanding of the type of changes that occur within firms over time with respect to enhanced innovation potential from engaging in alliances. The contribution of this paper is that, on the basis of a longitudinal approach, it examines how frequent engagement in alliances and expanding alliance portfolios (a compilation of contemporaneous alliances) impact on the likelihood of firm innovation; this offers a longitudinal approach that is lacking in the current literature (Draulans et al., 2003; Rothaermel and Deeds, 2006) with few notable exceptions (Di Guardo and Harrigan, 2016; Belderbos et al., 2015). A dynamic approach is appealing as it captures the impact on innovation of firm learning with respect to management of alliances. Such learning is evident in the use of past alliance experience to support conflict avoidance or management, and the effective coordination of knowledge sharing and communication with partners. Improved management can affect value capture in alliances especially as firms manage expanding alliance portfolios due to both potential of cross-partner asymmetries and knowledge redundancies, which are common in the bio-pharmaceuticals sector (Diestre and Rajagopalan, 2012; Caner and Tyler, 2013). The paper uses longitudinal econometric panel data analysis to explore its hypotheses. Secondary to this, we deploy a case analysis to capture the nuanced changes due to learning, shedding more light on the role of frequent engagement in alliances in compelling firms to direct their attention and resources in developing practices and routines for alliance management. This unpacks the antecedents to alliance capabilities and travels beyond current understandings in relation to experience accumulation. As highlighted in a recent review (Wang and Rajagopalan, 2015), research on the antecedents to alliance and alliance portfolio capabilities is rather thin. This paper contributes to exploring this link further. Our arguments are explored in the context of a unique, bespoke and specially assembled data set, comprising the population of firms in the UK bio-pharmaceutical sector (110 firms "alive" in 2003) and active during the period 1991-2001. The UK bio-pharmaceutical sector presents an ideal setting for investigation. The UK accounts for a significant share of the worldwide bio-pharmaceutical sector (Mckinsey, 2014) and enjoys a revealed technological advantage (OECD, 2017) and comparative advantage (OECD, 2008) in this high-tech sector. Moreover, the UK is a prime example of the Liberal Market Economy variant detailed in the Varieties of Capitalism literature (Hall and Soskice, 2001; Hancke, 2009), where the socio-economic institutional configuration is associated with international competitiveness of industries that are characterised by radical innovations (such as the pharma and biotech industries). The period 1991 and 2001 is perceived to be particularly apposite for our analysis as it exhibits several critical features: a first instance and general upsurge in alliances (not only in the UK bio-pharmaceutical sector, but also in other sectors and territories, Kang and Sakai, 2000); the emergence of alliances as an integral component in firm R&D strategies in the bio-pharmaceutical industry (Hagedoorn, 2002; Demirkan, 2018; Powell et al., 2005) but also more generally (Rice et al., 2012); and the presence of significant variation - visible in terms of identifiable peaks and troughs in alliance activity - especially towards the end of the period (Hagedoorn, 2002). Moreover, this marks a paradigmatic era of new technology evolution, with new entrants and established firms developing capabilities in biotechnologies through alliance activity (Hopkins et al., 2007): thus it provides a historical backdrop that aids understanding of development of similar technologies, for example, nanotechnologies (Rothaermel and Thursby, 2007). The remainder of this paper is structured as follows. Section 2 sets-out the theoretical background to the study and reviews the literature on alliance experience and alliance/alliance portfolio capabilities. Sections 3 and 4 develop our hypotheses. Section 5 discusses our sample, data sources and methodological considerations and details the variables for our longitudinal econometric analysis. Section 6 is dedicated to estimation and results, and Section 7 provides results from post-estimation robustness checks. Section 8 discusses our findings together with the case study analysis and Section 9 highlights the implications of the study for management theory and practice. To explain heterogeneity with respect to firm's abilities to benefit from alliances, the alliance literature resonates strongly with knowledge and capability-based theories of the firm (e.g. Kogut and Zander, 1992; Helfat and Peteraf, 2003). Here, we adopt a perspective that is informed by both evolutionary theory of the firm (Nelson and Winter, 1982) and dynamic approaches to the resource-based view (RBV) (Helfat et al., 2007; Helfat and Peteraf, 2003). These approaches take a dynamic view of organisational development, emphasising the role of experience and knowledge accumulation in supporting improved management and coordination of organisational tasks and activities. Given their dynamic and longitudinal orientation, they are closely aligned with our own analytical approach, informing our exploration of the roles of alliance experience accumulation and frequent engagement in alliances in organisational learning and enhanced innovation. Evolutionary theory argues that by repetitively engaging in organisational tasks, or by gaining experience, organisations benefit from efficiency improvements as a result of "learning-by-doing". As firms follow different paths in accumulating and making sense of experience, they become heterogeneous in their abilities to carry out similar organisational activities. Experience assists firms in assessing which processes and practices are likely to perform better or to confer improved results, by judging on similar situations in the past. Organisations gradually create routines to efficiently coordinate organisational activities and develop capabilities to achieve outcomes important for competitive advantage (Dosi et al., 2000; Winter, 2000, 2003; Schilke and Goerzen, 2010; Nelson and Winter, 1982). Experience accumulation is an antecedent to developing capabilities; however, there can be a deliberate and conscious process of organisational learning as firms make investments in time, resources, training and knowledge to further improve their capabilities (Zollo and Winter, 2002), or to improve them through reflecting on newly acquired experience (Helfat and Peteraf, 2003). Dynamic approaches to RBV are illuminating on the process of capability development by suggesting that there is an evolutionary path to their creation whereby capabilities emerge and improve through several developmental stages (Helfat and Peteraf, 2003). First, there is a "capability founding" stage wherein organisations realise the need to develop a capability to meet desirable objectives and to orchestrate joint organisational action towards meeting these. Second, a "capability development" stage: here improvements to capabilities arise from a range of factors including further experience accumulation, investments in capital, organisational processes and learning mechanisms. Experience accumulation is crucial to both "founding" and "development" stages of a capability development. A third stage, "capability maturity" is also identified: in this phase capabilities may be replicated, retrenched, renewed, redeployed or re-combined with other capabilities. Even though alliances differ compared to other organisational tasks, in that firms collaborate with different partners in agreements with varying content, purpose and governance structure, gaining experience in managing alliances can lead to superior coordination, value creation and capture in alliances (Kale and Singh, 2007; Heimeriks and Duysters, 2007). Alliance experience is found to enhance the performance of alliances between recurrent partners, as firms use their experience and routines developed during their first interactions in managing subsequent collaborations within the dyad (Zollo et al., 2002). Moreover, firms mobilise experience and lessons gained across alliances with different partners (Anand and Khanna, 2000), with partner-specific alliance experience making a greater contribution to performance compared to general (i.e. non-partner specific) experience (Gulati et al., 2009). Alliance experience (a firm's cumulative number of alliances) is an antecedent to both superior coordination of alliance tasks (alliance capabilities), and to the efficient management of synergies and redundancies across alliances in alliance portfolios (alliance portfolio capabilities) (for a review see Wang and Rajagopalan, 2015). Kale and Singh (2007) were the first to detail how firms proactively learn from their alliance experience by externalising, codifying, diffusing and internalising individual and group-level alliance know-how. A stream within the alliance literature explores the link between cumulative alliance experience, the development of dedicated alliance management functions and other processes and practices connected with upgrading alliance management routines (Kale et al., 2002; Heimeriks and Duysters, 2007; Kale and Singh, 2007). Alliance functions reflect alliance capabilities and can lead to superior outcomes from alliances. Alliance experience indirectly affects firm performance via such capabilities, indicating that dedicated alliance functions embody and absorb the impact of alliance experience, suggesting that firms leverage all the learning gained from experience in their alliance management functions (Kale et al., 2002). Other research suggests that alliance experience directly influences both financial (Kale et al., 2002; Anand and Khanna, 2000) and non-financial outcomes such as innovation performance (Sampson, 2005), and that this is in addition to any impact conferred by dedicated alliance functions, suggesting a complementary relationship between the two (Heimeriks and Duysters, 2007). Indeed, systematic alliance practices are improved through training, mentoring programmes and external consulting. Such learning processes aim at introducing new alliance management practices and/or at improving existing ones in response to dissatisfaction with current performance (Zollo and Winter, 2002). The following section revisits existing but still contested hypotheses on the expected impact of alliance experience on firm innovation performance. We also develop our hypothesis on the impact of alliance frequency, which from a longitudinal perspective can be better positioned to capture nuancing and conditioning factors that reflect the impact of organisational learning from alliance experience on innovation. Firms with greater experience can draw from a greater pool of situations about what has worked in practice, when making decisions and inferences with respect to the performance of organisational practices (Levitt and March, 1988; Argote et al., 1990). Alliance experience (cumulative number of alliances) improves firms' abilities: to manage and coordinate alliances, to improve coordination of inter-organisational relationships and joint tasks, to form efficient arrangements for knowledge sharing, to deal effectively with unforeseen contingencies and to identify ways to overcome and resolve inter-partner conflict (Anand and Khanna, 2000; Sampson, 2005; Belderbos et al., 2015; Rothaermel and Deeds, 2006). Due to the link between alliance experience and organisational learning, experience is seen as a fundamental antecedent to both alliance and alliance portfolio capabilities (Wang and Rajagopalan, 2015; Kale and Singh, 2009; Shukla and Mital, 2018). Literature suggests that firms may not be in a position to benefit from learning from experience and superior coordination of alliances, when facing power asymmetries and resource dependence in alliances. Conflict is more frequent in such alliances which affects value creation and capture, especially for the weak partner as they are at a comparative disadvantage (Diestre and Rajagopalan, 2012). Power asymmetries are likely in the bio-pharmaceutical sector, as large pharmaceutical firms may be collaborating with small-dedicated biotech firms and due to their longer commitment to alliances and historic investments in downstream capabilities may be at a comparative advantage in deriving value from alliances (Caner and Tyler, 2013). We argue that by gaining greater alliance experience, firms gradually establish more fruitful conditions for effective knowledge sharing with their partners, and generate a basis for circumventing or resolving alliance conflicts. Because conflicts are more likely to occur in the bio-pharmaceuticals sector we expect that the higher the levels of experience the more likely the firms are to capture value from alliances and enhance their likelihood of innovation: H1. Alliance experience accumulation positively affects a firm's likelihood of innovation. A body of work suggests that efficiency improvements from experience accumulation make a gradually diminishing contribution to firm performance. There are two likely explanations for this. First, as firms form more alliances and accumulate greater experience, the contribution of any additional alliance to accumulated learning becomes increasingly reduced. This has been identified irrespective of type of partner and across upstream, downstream and horizontal alliances (Rothaermel and Deeds, 2006). Firms may experience diminishing returns when managing an increasing number of alliances due to the increased costs associated with identifying suitable partners, establishing and maintaining contracts, difficulties in absorbing knowledge from diverse sources, conflicts due to knowledge asymmetries and increased possibilities of knowledge redundancies (Diestre and Rajagopalan, 2012; Rothaermel and Deeds, 2006; White and Lui, 2005). A second likely explanation is that the value of lessons learned from cumulative experience depreciates over time as knowledge becomes increasingly obsolete (e.g. Argote et al., 1990). Past knowledge may depreciate as task requirements change over time. Further, the value of past lessons may dissipate gradually (or suddenly) as a result of employee turnover or a failure to store intelligence in organisational routines, or in a way that permits its re-application (Levitt and March, 1988; Dutton and Thomas, 1984). Unless firms make appropriate investments in retaining and leveraging lessons learned from their experience, its effects will increasingly dissipate. Research in the alliance literature has demonstrated a positive and linear relationship between alliance experience and firm-level indicators of financial (Anand and Khanna, 2000) and innovation performance (Shan et al., 1994). Only a slender body of work has explored diminishing returns to alliance experience. Hoang and Rothaermel (2005) find that, whilst alliance experience of small firms positively impacts the performance of their joint projects with large firms, the alliance experience of large firms makes no significant impact on their joint alliance project performance. Benefits stemming from alliance experience differ between small and large firms because they are positioned at different levels along the diminishing/concave alliance experience curve. In related research, Sampson (2005), commenting on patenting performance in the telecom equipment industry, identifies similarly diminishing returns to cumulative alliance experience, which she attributes to a diminishing contribution of distant experience, suggesting that distant alliance experience plays less important role than recent experience in efficient management of alliance portfolios (Shukla and Mital, 2018). As a consequence, we derive the following hypothesis: H2. The effect of cumulative alliance experience on likelihood of innovation exhibits diminishing marginal returns: as alliance experience accumulates, its contribution to the likelihood of firm innovation decreases. Deciphering, coding and measuring capabilities is notoriously difficult (Godfrey and Hill, 1995). As a result, the alliance literature has, in the main, relied on identifying alliance management practices (e.g. alliance functions) as a way of documenting alliance capabilities (Kale et al., 2002; Kale and Singh, 2007, 2009). An exception is found in the work of Rothaermel and Deeds (2006). They explore an inverted U-shaped relationship between cumulative alliance experience and new product development. They argue that, as the inflection point of the inverted U-shaped curve corresponds to the level of experience beyond which firms start experiencing inefficiencies in alliance management, it can reflect the level of their alliance capability. The notion that dedicated alliance management functions reflect alliance capabilities has being challenged in recent literature. First, anecdotal evidence suggests that firms establish dedicated alliance functions only after engaging in alliances over a protracted period, and after initiating a substantial number of such agreements (Kale and Singh, 2009). As a result such functions cannot reflect alliance capabilities, particularly when capabilities are at the initial stages of their development. Research indicates that firms - especially those motivated by perceived market signalling requirements - may establish dedicated alliance functions as a result of mimitisism or isomorphism (Heimeriks, 2010)[1]. Indeed, isolated mechanisms for institutionalising learning (e.g. dedicated alliance functions) from cumulative alliance experience may prove to be less effective compared to mechanisms designed to integrate and embed alliance knowledge throughout the organisation (Heimeriks et al., 2015). Despite some tentative and tangential work, explorations with respect to the underlying factors that lead firms to become aware of the need to leverage their alliance experience in developing a capability, and to invest efforts and resources in processes for deliberate learning from alliances are sparse in extant literature (see Wang and Rajagopalan, 2015). We contend that the first instances of frequent engagement in alliances can form such underlying factors and we turn to evolutionary theory and dynamic RBV to assist in elaborating our argument. Specifically, we suggest that frequent engagement in alliances may compel firms to found an alliance capability development lifecycle, as they both protract the period of engagement in alliances, and engage more substantially in this activity (Helfat and Peteraf, 2003). By engaging in alliances frequently, firms expand, augment and renew their stock of cumulative alliance experience. Frequent engagement in alliances induces firms to invest in developing higher order routines and practises to initiate purposeful and directed organisational learning. Developing processes for alliance management and practices for their further improvement requires managerial commitment, commitment that is highly demanding in terms of time and employee resources (Winter, 2000, 2003; Dosi et al., 2000). As a result, firms need to justify the allocation of resources to building dedicated routines for alliance management: frequent task reoccurrence can trigger and motivate firms to deploy resources in such a direction. It can also justify such investments to external stakeholders, building internal support and commitment to the purpose. Sparse empirical literature explores the role of frequent engagement in alliances in developing alliance capabilities (see Wang and Rajagopalan, 2015 for a review). There is a particular dearth of longitudinal studies in this field (Draulans et al., 2003; Rothaermel and Deeds, 2006). Research on CIS longitudinal Spanish data indicates that persistent and not sporadic collaboration exhibits a systematic positive relationship with firm innovativeness, which runs parallel to our argument (Belderbos et al., 2015). Moreover, firms that renew and expand their collaborations and alliances over time become more embedded in a network and experience higher growth over time (Powell et al., 2005). At the cross-sectional level, there is only limited research that links larger alliance portfolios to the development of alliance capability (as this is reflected in alliance management practices). Larger alliance portfolios are associated with the use of enhanced alliance management practices and processes, with the latter being linked to improved firm performance (Heimeriks et al., 2009). Our longitudinal approach, capturing the tendencies of firms to engage in alliances more frequently over time, complements such accounts which are based on the role of stock measures of alliance experience in firm performance (Kale et al., 2002; Heimeriks and Duysters, 2007; Kale and Singh, 2007). Therefore we propose that firms that engage in alliances frequently are more likely to innovate compared to those firms that engage in alliances irregularly or sporadically, or that retain a static engagement and a stagnating alliance portfolio over time: H3. Frequent engagement in alliances and the ensuing expanding alliance portfolios positively affect firm likelihood of firm innovation over time. 5.1 Sample We test our hypotheses on a panel data set observing the population of firms in the UK bio-pharma sector (110 firms in 2003) during the period 1991-2001. The empirical exploration in this paper relies fundamentally on econometric analysis. However, secondarily the paper employs a case analysis to shed further light on - and nuance - some of its arguments. To identify our sample, we used two editions of the UK biotechnology directory (Coombs and Alston, 2000, 2002). The directory lists all firms in the sector that undertake research using biotechnologies. It provides a more comprehensive account of the firms in the UK bio-pharma sector than that available via the use of SIC classifications, as biotechnologies are used by firms from a range of industries in the life sciences as broadly described (Walsh, 2004). To form our working sample, we used the "Who Owns Who" directory to identify all firms that were active in 2003. Out of that list, we identified 110 firms that publish accounts in FAME[2]. For the sample of 110 firms, we collected data on alliances from ReCap.com and BioScan, the two most popular sources of alliance data in empirical research on this sector (e.g. Deeds and Hill, 1996; Schilling, 2009). The databases report alliances established in the bio-pharmaceutical sector for innovation purposes, including R&D alliances, research alliances and alliances for technology licensing and product development[3]. Innovation is crucial to firm survival and prosperity in the bio-pharmaceutical sector and patents are perceived as an appropriate proxy for innovations due to the science-based nature of the sector (e.g. Pavitt, 1984). Moreover, patents are an appropriate indicator to capture firm-level returns from innovation alliances (Demirkan, 2018; Shan et al., 1994; Sampson, 2005). We collected data on patents granted to the 110 firms in our sample between 1991 and 2001 by the UKPTO through the publicly available database Esp@cenet. Via deployment of Boolean searches, it was possible to assemble a comprehensive list by matching both the name and the address of firms in our sample with those of the patent assignees (Arora et al., 2011). For firms in our sample that belong to MNEs with non-UK-based HQs, we examined information on inventor location, and allocated any patents assigned to the MNE HQ with UK-based inventors to the UK-based subsidiary in our sample. Since our research focusses on UK-based firms, the UKPTO is perceived to be an appropriate source for the collection of patent data. Filing patent applications at national patent offices is less costly (in terms of application and renewal fees) and less time consuming when compared with international patent applications. As international applications also involve costs of unfamiliar IP regulations and legal frameworks, many firms - at least those with constrained resources - are more likely to patent at national patent offices than internationally (Archambault, 2002). Finally, information on firm accounts was obtained from FAME, and information on investments in R&D was gathered from Thomson's Analytics and the UK DTI's R&D Scoreboard. 5.2 Methods Patents take positive integer values, making count dependent variable models, such as the Poisson and Negative Binomial, appropriate (Greene, 2003). Instead of count dependent variable models, we employ discrete dependent variable models as they permit the capture of factors that alter firm abilities enabling them to perform over a threshold level: successful filing of a patent (Long, 1997). This is better aligned to the core aim of this work, which is to explore whether frequent alliance engagement is linked to organisational learning in alliance management (and alliance capabilities), reflected in firms exhibiting over threshold performance (Hoang and Rothaermel, 2005). Due to the longitudinal dimension of panels, we can explore factors that affect firms' changing abilities to innovate over time (fixed effects specification), as well as factors that can explain cross-firm variation in innovation abilities (random effects specification). The former, dynamic consideration of panels is of particular interest in this paper, as explained above. The fixed effects specification has a further appeal as its estimates are not based on the assumption of strict exogeneity of the independent variables (Greene, 2003; Baltagi, 1995), which is desirable, as decisions to form alliances may not be independent of past innovation performance (Colombo and Garrone, 1996; Gkypali et al., 2017). To explore further the nuanced factors underlying the development of alliance capabilities, we report on the findings of an in-depth case analysis that explores the link between frequency of engagement in alliances, and the development of practices and processes to manage alliances, an indication of alliance capabilities (Schreiner et al., 2009; Kale and Singh, 2007); this is a secondary analysis providing an illuminating complement. The case focusses on a UK bio-pharmaceutical firm that intensified its engagement in alliances particularly after the early 1990s when it started to invest in biotechnology R&D. Interviews were conducted in 2005 as the firm was forming a dedicated alliance function and entailed detailed reflection on the factors that underpin the creation of processes established to improve efficiency in alliance management. The processes identified were matched with those identified in existing literature on alliance capabilities (Kale et al., 2002; Kale and Singh, 2007, 2009; Heimeriks and Duysters, 2007). The case is based on interviews with directors of alliances and intellectual property, with a long working experience with the case firm and focussed on alliances that involve cooperation in research and that have led to patents. All materials and associated analyses were reviewed and approved by the interviewees. 5.3 Variables Patents (dependent variable) To test our hypotheses we create a dummy variable that is equal to 1 for each firm that is granted a patent at year t (and equal to zero otherwise). We use a patent dummy variable as we aim at capturing the impact of alliance experience in enabling firms to perform over a threshold, i.e. to innovate. To track the firm's innovation activities over time, we use patent filing dates. Filing dates have certain advantages over publication dates: first, they better reflect the originating time of innovations; and, second, they are not influenced by regulatory changes or fluctuations in resource availability in patent offices over time, as these can delay publication dates (e.g. Jaffe, 1986). Patents are widely used to capture firm returns from innovation alliances (Sampson, 2005, 2007; Shan et al., 1994; Demirkan, 2018). Alliance experience Research has explored the role of alliance experience in improving financial returns of firms from alliances (Kale et al., 2002), alliance project success (Hoang and Rothaermel, 2005), returns to firm innovation performance (Sampson, 2005, 2007; Belderbos et al., 2015; Di Guardo and Harrigan, 2016) and number of products under development (Deeds and Hill, 1996; Rothaermel and Deeds, 2006). Studies have operationalised alliance experience as a stock variable reflected either by the cumulative number of alliances formed throughout a firm's history (Kale and Singh, 2007; Hoang and Rothaermel, 2005; Kale et al., 2002; Rothaermel and Deeds, 2006), or by the cumulative number of years of experience in engaging in alliances (Rothaermel and Deeds, 2006). Only a handful of these studies consider diminishing (Sampson, 2005) or non-linear (Rothaermel and Deeds, 2006) relationships between alliance experience and firm returns from alliances and most studies base their estimates on cross-sectional data (an exception is Kale et al., 2002). Following existing literature, we operationalise alliance experience at each year t as the cumulative number of alliances throughout a firm's history. To aid interpretation of results, and following recent contributions (Lavie et al., 2011) we use one year lagged values of the above variable to more accurately capture the effect of cumulative alliance experience. Our variable is left-censored to 1991, the start year for our sampling. This is experienced elsewhere in alliance literature (e.g. Kale and Singh, 2007). Also, because in our study 87 per cent of total alliances from 1979 and 2001 fall within the 1991-2001 period, left censoring is not introducing considerable bias. As is customary in the literature (Sampson, 2005), we use the natural logarithmic transformation of cumulative alliance experience to test for diminishing returns (H2). As a robustness check, we test for a concave relationship (inverted U-shaped) by using both the linear and quadratic values of the cumulative number of alliances: again, this approach follows relevant literature (e.g. Deeds and Hill, 1996; Rothaermel and Deeds, 2006)[4]. Frequent engagement in alliances and expanding alliance portfolios To explore the effects of frequent engagement in alliances on firm abilities to innovate, we measure the total number of on-going alliances that a firm manages simultaneously at any point in time. As alliances last for several years, this requires information on alliance duration which is not available via the alliance data sources consulted in this study (RECap.com and BioScan). As lack of information on alliance duration has also been encountered in other work, we follow established tradition and rely on estimates of the average duration of alliances, i.e., five years (Kogut, 1988; Di Guardo and Harrigan, 2016). Therefore, we operationalise frequent engagement in alliances by calculating the number of alliances a firm manages at any given point in time by considering that alliances last on average for five years. Increases in this variable reflect the tendency of firms to engage more frequently in alliances over time, and on average, to initiate more alliances than they terminate, managing potentially an expanding portfolio of on-going alliances. As the use of a blanket measure for alliance duration might be considered simplistic, we cross-validated this information with alliance managers in the sector, and it was considered an acceptable approximation for the duration of the majority of alliances. We also performed robustness checks by using different estimates for alliance duration as detailed in our analysis. Control variables For parsimonious empirical models, we use the most appropriate control variables that have been employed in the literature on alliances and innovation. First, we account for the effects of well-established indicators of firm innovation, such as investments in R&D and firm size (Griliches, 1990) as they can influence firm abilities to innovate. We used a measure of R&D intensity that captures the amount of R&D expenditure per employee. Firm size is captured by annual turnover and we include a natural logarithm to overcome the possible problem of skewed distributions. Second, and following relevant research (Zollo et al., 2002), we account for differences in the content of alliances, as it influences innovation opportunities in alliances and affects firm performance (e.g. Anand and Khanna, 2000). Therefore, we include a dummy that takes the value of one when the focal firm establishes an alliance that involves R&D, and the value of zero for each year that a focal firm establishes alliances that do not involve R&D. Table I presents descriptive statistics and bivariate correlations. The panel is unbalanced, as a result of either missing information or the inclusion of less-mature dedicated biotechnology firms (i.e. a third of firms in the sample were established after 1995). The unbalanced data set includes firms observed for a maximum period of ten years and others for as little as four years. On average, firms are observed for approximately seven years (the range is from 6.6 to 7.4 years across the three estimated models) and form 8.28 alliances between 1991 and 2001. We estimated both the fixed and random effects specifications (FE and RE, respectively) of the Logit model for panel data for all three models (Greene, 2003; Baltagi, 1995)[5]. Results are obtained in STATA(c) 11. The r parameter is significant across all three models, indicating significant cross-firm heterogeneity. As Hausman's (1978) specification test has limited discerning power in discrete dependent variable models, we assess the relative merits of the two specifications on a conceptual basis and for reasons discussed in the methods section we present Logit FE estimates. Moreover, the residuals of the FE specification are corrected for potential heteroscedasticity (White, 1980; Baltagi, 1995). Based on model kh2 statistics in Table II, all three models are significant at 5 per cent levels and above. This suggests that our variables have a significant interpretative power with respect to the likelihood of firm innovation. We find support for all hypotheses. In Model 1, the cumulative number of alliances positively affects firm's likelihood of innovation over time, but levels of significance are low (10 per cent). Our results are consistent with studies showing a positive relationship between cumulative alliance experience and firm innovation, but at low significance levels (Hagedoorn and Schakenraad, 1994; Shan et al., 1994). The FE estimates of Model 2 provide support for diminishing returns to firm innovation from cumulative alliance experience, with the corresponding variable being positive and highly significant. The results of Model 2 are consistent with findings testing similar hypotheses in cross-sectional settings in the US bio-pharmaceutical and the telecom equipment industries (Hoang and Rothaermel, 2005; Sampson, 2005). Our results are robust to alternative operationalisations, as reflected by exploring concave relationships (inverted U-shaped), via use of both linear and quadratic values of cumulative alliance experience, with estimates (both FE and RE) confirming non-linear diminishing returns to likelihood of firm innovation. To unpack the extent of diminishing returns we estimate the inflection point based on estimates of the curvilinear relationship; this is equal to 127.515 points of cumulative experience, which corresponds to firms managing 18 alliances annually (average length of time per firm in the panel is seven years). Our results imply that at such a level of annual alliance activity, firms will accrue no additional benefits from gaining more experience in managing alliances. It should be noted that in our sample the average number of new alliances formed every year is 8.28, which suggests that the majority of firms in our sample are below the optimal level of alliance engagement. The concluding section discusses the implications of this. Model 3 provides support for our third hypothesis. The coefficient of the core independent variable is positive and significant (it marginally fails to pass the highest 1 per cent level of significance). Model 3 suggests that firms that establish alliances frequently and that continuously manage an expanding portfolio can enjoy improved likelihood of innovation. This provides fresh insights as, to date, there has been a scarcity of longitudinal studies in the literature (see Draulans et al., 2003; Rothaermel and Deeds, 2006). As a test of robustness, we estimated Model 3 by using four and six years as the average duration for alliances. The results remained robust to different operationalisations. Among the control variables, firm size is positively and significantly associated with firm's likelihood of innovation in Models 1 and 3. Investments in R&D per number of employees are negatively associated with likelihood of firm innovation but only in Model 2. Alliance content is insignificant across all three models. Finally, all three model estimates are robust to different measures of the control variables. For example, all models lead to the same outcomes in terms of significance when controls are collapsed to dichotomous variables indicating above and below average firm size and intensity of investments in R&D per number of employees. A further robustness check is undertaken to examine whether our results are sensitive to the statistical assumptions of the distribution followed by the error term in the Logit and Probit specifications. The Probit model only provides RE estimates for panel data (STATA(c) 11) and these further confirm our results. Moreover, we estimate our model specifications with the Negative Binomial model to explore differences between our dependent variable and a count variable capturing innovation performance with the number of patents. Both FE and RE estimates confirm our hypotheses. The results suggest that there are diminishing returns to cumulative alliance experience. This can signal that ageing experience contributes less to current outcomes and that recent experience may have a higher contribution (Shukla and Mital, 2018; Sampson, 2005). Following Sampson (2005), we explore the contributions of recent and past alliance experience. We develop a set of variables capturing alliance experience between 1 and 6 years prior to our year of observation. So, for example, in 2001 alliance experience of one year corresponds to the number of alliances initiated in 2000 and alliance experience of two years corresponds to those initiated in 1999 and so on. None of these variables appear to be significant, with the exception of alliance experience of four years prior to the year of observation, which appears with a negative and significant sign. Therefore, the diminishing returns to cumulative alliance experience identified in our paper cannot be attributed to decreasing contributions of distant experience. Our results most likely reflect that firms, by just forming more alliances and learning from experience how to improve alliance management and coordination, cannot experience improved efficiency ad infinitum. With regard to H3 we undertake further estimations to isolate the impact of frequent engagement in alliances from that of portfolio size. We break down the variable testing H3 into two components. First, capturing changes in the frequency of forming alliances over time, we construct a variable reflecting a decreasing, static and increasing trend in alliance portfolios (the variable takes the values of: -1, 0 and 1, respectively, and refers to comparisons between the current and the previous year). Second, we include a variable to capture the number of new alliances formed within every year (one year lagged values are used in the models). Our results show that the trend variable has a negative sign while the annual figures of newly formed alliances appear with a positive and significant sign (both for the Logit and the Negative Binomial models and the FE and RE specifications). It is worth noting that when the values of the trend variable are lagged by one year, their impact becomes insignificant (i.e. they retain a negative sign). This means that only the contemporaneous impact of alliance frequency on innovation is negative, implying that when firms expand their alliance portfolio there is an immediate adverse effect. This may result from immediate strains on their resources and cognitive abilities, which however, cease to have an adverse effect within a year's time. The results remain the same when the trend variable is substituted with a dummy variable capturing only increases in the frequency of forming alliances between current values and those of the previous year. Finally, we isolate partner-specific alliance experience from general (i.e. non-partner specific) experience to further explore the importance of alliance management learning within partner dyads and in alliance portfolios. None of these variables has a significant effect, suggesting that the separation lacks interpretative power. The paper contributes to the literature on the role of alliances in innovation and alliance capabilities by using a longitudinal analysis which enables capture of the impact of organisational learning in managing and coordinating alliances on innovation. As it is difficult to capture learning, the paper uses a longitudinal approach to capture dynamic changes within firms and it observes learning through its impact on outcome variables such as innovation. The paper adds to a slender body of work exploring antecedents to firm-level alliance capabilities and their impact on enhancing firm's abilities to innovate (for a review see Wang and Rajagopalan, 2015). First, we find that the impact of learning from gaining greater alliance experience in coordinating alliances and portfolios on firm innovation is positive but of weak significance. This is an important finding, as alliances in the bio-pharmaceutical sector can be characterised by power asymmetries and unbalanced resource dependence between partners, which are likely to create conflicts affecting the creation and appropriation of alliance value (Diestre and Rajagopalan, 2012; Caner and Tyler, 2013). The period of our study forms the first window of exploitation of biotechnologies via alliances which potentially creates disincentives to collaborating firms to behave opportunistically, as joining forces during this close to pre-competitive stage could be more beneficial than not (Powell et al., 2005). Second, our analysis proceeds by exploring the factors that affect learning in managing and coordinating alliances in this sector. The first arm to this exploration sheds light on the role of cumulative-historic experience in enhancing innovation performance, while the second arm focusses on the role of such experience as an antecedent to the creation of capabilities to manage interactions with alliance partners and to create synergies and minimise redundancies in alliance portfolios. In that endeavour, we first confirm diminishing returns to the benefits that can be accrued by gaining alliance experience for more effective management of alliances and capturing innovation returns. This supports a thin body of research which suggests that learning from gaining experience cannot confer benefits ad infinitum (e.g. Rothaermel and Deeds, 2006; Sampson, 2005). Our in-depth exploration shows that this cannot be attributed to diminishing contributions of ageing experience as identified in other work (Sampson, 2005; Shukla and Mital, 2018). It could be attributed to failure of firms to capture learning in a systematic way, or to costs associated with identifying more partners, establishing more agreements and coordinating increased diversity (White and Lui, 2005; Rothaermel and Deeds, 2006). Our last set of findings show that engaging in alliances more frequently over time, with firms managing an expanding portfolio, by forming more alliances than they terminate, enhances firms' likelihood of innovation. Our longitudinal approach is consistent with other work showing that persistence in collaboration has a positive impact on firm innovation as opposed to sporadic collaboration (Belderbos et al., 2015). Longitudinal explorations are currently almost absent in alliance literature, and their potential contribution has been particularly welcomed (Draulans et al., 2003; Rothaermel and Deeds, 2006), specifically as they might shed further light on the type of changes that can occur within firms, in particular, with respect to the dynamics of learning on alliance management and how it enhances innovation potential. Contributions within evolutionary theory (Winter, 2000, 2003) argue that frequent engagement in organisational tasks is linked to capability development, while the dynamic resource-based view (Helfat and Peteraf, 2003) argues that capabilities develop through several stages: experience accumulation compels firms to recognise the need to establish a capability, and this initiates a process of systematic learning and alliance management. Our results clearly pertain to this literature and specifically on how alliance experience forms an antecedent to alliance and alliance portfolio capabilities (Wang and Rajagopalan, 2015). To explore, in a situated and more granular fashion, whether frequent engagement in alliances and a dynamically expanding alliance portfolio contribute to the development of alliance capabilities, we conducted a detailed case analysis as a complement to our econometric analysis. The case analysis explores the link between an increasing engagement in alliances and the development of novel or changed practices and processes for alliance management, i.e., those factors that normally sit within the "black-box" of alliance capability studies (Anand and Khanna, 2000; Heimeriks and Duysters, 2007; Kale and Singh, 2007). The case analysis clearly reveals such a relationship. First, it shows that alliance management practices and routines are established after alliances become an integral part of firm R&D and innovation strategy and firms intensify and commit to alliance activity. As the Alliance Director interviewed for our study stated: "[...] we have standardised approaches for contractual compliance [with respect to alliances] [...] it's inconceivable to me that one wouldn't have, and yet five years ago we didn't [...] You've got one deal and you do it however you do it which largely depends on who you employ and who it's [the alliance] with [...] I am not sure how many deals you need to have before you get yourself a standardised process but it's probably substantially more than ten or twenty [...][7]". Second, several of the practices identified in the extant literature that are used to institutionalise and diffuse alliance management learning (namely, employment of alliance professionals, use of alliance metrics, creation of databases to solicit potential alliance partners and establishment of processes to monitor alliance performance) are deployed by the case firm (Heimeriks and Duysters, 2007; Kale and Singh, 2007). Ten years after intensifying alliance engagement, it developed a dedicated alliance management function to coordinate alliance activity and engage in monitoring and improving existing alliance management practices. The case analysis illustrates the links between frequent engagement in alliances and the ensuing expanding the alliance portfolio, with the motivation to invest resources in developing firm-specific capabilities for alliance management. It also reflects the time required to recognise that alliance management can be improved via the development of specific and specialist processes for the governance of cross-partner interaction (a finding that resonates strongly with existing literature, Kale and Singh, 2009). In addition, the case reflects the requirement for firms to continuously review and develop their capabilities in order to confront new challenges and intensified competition. Indeed, the case analysis, combined with the findings of our econometric exploration, suggest that renewing alliance experience and committing to forming more alliances over time are crucial in enabling firms to create and capture value in alliances. Our research contributes to the literature on the antecedents of alliance and alliance portfolio capabilities and on the conditions that shape superior firm outcomes from alliances, such as innovation. It suggests a need to delve into the antecedents to alliance capabilities, a thin body of research and to identify nascent factors that provide potential foundations for alliance capability development (Helfat and Peteraf, 2003), shifting attention away from the role played by alliance management practices (such as dedicated alliance functions) that have dominated existing research. This is particularly important as firms may establish such practices during the advanced stages of the alliance capability development process (Kale and Singh, 2009), and as such they may not appropriately reflect the foundational stages of alliance capability development. Here, we echo calls to delve deeper in understanding alliance management capabilities and the antecedents to alliance portfolio capabilities (Wang and Rajagopalan, 2015), especially in the context of alliances involving a higher learning potential (Heimeriks, 2010). Moreover, recent research shows that codification of alliance learning and systematic approaches to alliance management contribute to efficient partner selection and alliance termination, but may restrict flexibility and adaptability which are important for efficient management during the course of the alliance (Heimeriks et al., 2015; Wang and Rajagopalan, 2015). Our research has several implications for managerial practice and policy. First, we find that there are diminishing returns to benefits in terms of enhanced innovation accruing from gaining greater alliance experience. Specifically firms reach a plateau at 18 alliances in annual terms, suggesting that on average this is the optimal level of alliance engagement for enhancing firm likelihood of innovation. This has implications for managing alliance portfolio size for innovation efficiency (e.g. Hoffmann, 2007; Demirkan, 2018). On average firms in our sample are below the optimal level of alliance engagement (forming eight alliances annually). This suggests that innovation likelihood can be marginally enhanced by expanding alliance engagement. This is in line with findings of other research (Rothaermel and Deeds, 2006; Sampson, 2005; Hewitt-Dundas and Roper, 2018) which in the context of micro-businesses, for example, shows that sub-optimal engagement can be attributed to reluctance of firms to cooperate due to fear of trustworthiness and lack of knowledge on benefits of collaboration and partner capabilities (Hewitt-Dundas and Roper, 2018). This emphasises our findings with respect to the need to facilitate a culture of systematic management of alliances. It also supports research suggesting the need for policy interventions, for example in the form of intermediaries, which can support and encourage firms to establish collaborations and manage openness (Howells, 2006). This is not only due to the benefits of collaboration to firm innovation but also due to the positive externalities from openness, suggesting broader economic and societal benefits (Roper et al., 2013). At the firm level, our results suggest that it is important for firms to replenish their alliance portfolios and to gradually form more alliances than they terminate over time if they wish to enhance their innovation likelihood. In the bio-pharmaceuticals sector where there is high rate of alliance formation, it is important for firm competitiveness and future growth that they renew and expand their networks (Powell et al., 2005). This is linked to firms becoming more aware of the need to make alliance management learning more systematic. Firms should use such systematic practices in a discretionary manner and should also consider altering and upgrading such practices to respond to the challenges and demands of new alliances. Although our results are specific in terms of sector and period, they derive from the examination of a paradigmatic period in the development of new technologies, and address the implications of changing innovation processes in an established industry. As such, they provide useful insights that assist understanding of trajectories and dynamics in industries with similarities to biotech, for example, nanotechnology (e.g. Rothaermel and Thursby, 2007). Our aspiration is that our research will provide some re-focussing and clarity in an established field of study. We acknowledge that analysis of further sectors and periods will nuance our understanding of both the role of gaining experience in improving firm alliance management and in developing alliance capabilities, and hope that the ideas and methods explored above will assist in this on-going quest.
|
The purpose of this paper is to explore the role of alliance experience in firm innovation; it argues that, while cumulative alliance experience has a marginally diminishing contribution to likelihood of firm innovation over time, frequent engagement in alliances and an expanding alliance portfolio inhabit an enhancing role. This reveals new dimensions to the role of alliance experience as an antecedent to firm learning in managing alliances and to the development of alliance capabilities.
|
[SECTION: Method] As the popularity of strategic alliances is increasing and such alliances are becoming an integral component in business development, attention in the research community has moved towards an exploration of their role in firm performance and innovation. Formal and informal interactions with external actors have long being argued to play a fundamental role in firm innovation (Von Hippel, 1988; Frankort et al., 2012; Rice et al., 2012; Colombo et al., 2011; Demirkan, 2018). However, it is accepted that alliances carry coordination costs and risks of misappropriation, and these frequently diminish chances of success or preclude full acquisition of anticipated benefits (e.g. De Man and Duysters, 2005; Gkypali et al., 2017; Faems et al., 2010). A substantial body of literature finds a positive relationship between the extent of alliances and firm innovation performance, whilst other work identifies diminishing and even negative returns (Hoang and Rothaermel, 2005; Sampson, 2005; Laursen and Salter, 2006; Rothaermel and Deeds, 2006). As a result, a growing strand in the literature has examined the factors enabling firms to generate and capture value from alliances. One such factor is alliance experience accumulation: as firms develop greater experience in managing alliances they become better at coordinating cross-organisational tasks and knowledge flows (e.g. Sampson, 2005). Another factor that can improve performance in alliances is developing formal and codified processes and routines for alliance management (e.g. the establishment of dedicated alliance functions), which are argued to capture, or to be reflective of, firm-specific alliance management capabilities (Kale et al., 2002; Kale and Singh, 2009, 2007; Heimeriks and Duysters, 2007; Sampson, 2005; Schreiner et al., 2009; Di Guardo and Harrigan, 2016; Shukla and Mital, 2018). Whilst these contributions offer useful insights, much the greater part of current theorising has been constructed via the use of cross-sectional data. Longitudinal explorations are scarce, thus we lack a nuanced understanding of the type of changes that occur within firms over time with respect to enhanced innovation potential from engaging in alliances. The contribution of this paper is that, on the basis of a longitudinal approach, it examines how frequent engagement in alliances and expanding alliance portfolios (a compilation of contemporaneous alliances) impact on the likelihood of firm innovation; this offers a longitudinal approach that is lacking in the current literature (Draulans et al., 2003; Rothaermel and Deeds, 2006) with few notable exceptions (Di Guardo and Harrigan, 2016; Belderbos et al., 2015). A dynamic approach is appealing as it captures the impact on innovation of firm learning with respect to management of alliances. Such learning is evident in the use of past alliance experience to support conflict avoidance or management, and the effective coordination of knowledge sharing and communication with partners. Improved management can affect value capture in alliances especially as firms manage expanding alliance portfolios due to both potential of cross-partner asymmetries and knowledge redundancies, which are common in the bio-pharmaceuticals sector (Diestre and Rajagopalan, 2012; Caner and Tyler, 2013). The paper uses longitudinal econometric panel data analysis to explore its hypotheses. Secondary to this, we deploy a case analysis to capture the nuanced changes due to learning, shedding more light on the role of frequent engagement in alliances in compelling firms to direct their attention and resources in developing practices and routines for alliance management. This unpacks the antecedents to alliance capabilities and travels beyond current understandings in relation to experience accumulation. As highlighted in a recent review (Wang and Rajagopalan, 2015), research on the antecedents to alliance and alliance portfolio capabilities is rather thin. This paper contributes to exploring this link further. Our arguments are explored in the context of a unique, bespoke and specially assembled data set, comprising the population of firms in the UK bio-pharmaceutical sector (110 firms "alive" in 2003) and active during the period 1991-2001. The UK bio-pharmaceutical sector presents an ideal setting for investigation. The UK accounts for a significant share of the worldwide bio-pharmaceutical sector (Mckinsey, 2014) and enjoys a revealed technological advantage (OECD, 2017) and comparative advantage (OECD, 2008) in this high-tech sector. Moreover, the UK is a prime example of the Liberal Market Economy variant detailed in the Varieties of Capitalism literature (Hall and Soskice, 2001; Hancke, 2009), where the socio-economic institutional configuration is associated with international competitiveness of industries that are characterised by radical innovations (such as the pharma and biotech industries). The period 1991 and 2001 is perceived to be particularly apposite for our analysis as it exhibits several critical features: a first instance and general upsurge in alliances (not only in the UK bio-pharmaceutical sector, but also in other sectors and territories, Kang and Sakai, 2000); the emergence of alliances as an integral component in firm R&D strategies in the bio-pharmaceutical industry (Hagedoorn, 2002; Demirkan, 2018; Powell et al., 2005) but also more generally (Rice et al., 2012); and the presence of significant variation - visible in terms of identifiable peaks and troughs in alliance activity - especially towards the end of the period (Hagedoorn, 2002). Moreover, this marks a paradigmatic era of new technology evolution, with new entrants and established firms developing capabilities in biotechnologies through alliance activity (Hopkins et al., 2007): thus it provides a historical backdrop that aids understanding of development of similar technologies, for example, nanotechnologies (Rothaermel and Thursby, 2007). The remainder of this paper is structured as follows. Section 2 sets-out the theoretical background to the study and reviews the literature on alliance experience and alliance/alliance portfolio capabilities. Sections 3 and 4 develop our hypotheses. Section 5 discusses our sample, data sources and methodological considerations and details the variables for our longitudinal econometric analysis. Section 6 is dedicated to estimation and results, and Section 7 provides results from post-estimation robustness checks. Section 8 discusses our findings together with the case study analysis and Section 9 highlights the implications of the study for management theory and practice. To explain heterogeneity with respect to firm's abilities to benefit from alliances, the alliance literature resonates strongly with knowledge and capability-based theories of the firm (e.g. Kogut and Zander, 1992; Helfat and Peteraf, 2003). Here, we adopt a perspective that is informed by both evolutionary theory of the firm (Nelson and Winter, 1982) and dynamic approaches to the resource-based view (RBV) (Helfat et al., 2007; Helfat and Peteraf, 2003). These approaches take a dynamic view of organisational development, emphasising the role of experience and knowledge accumulation in supporting improved management and coordination of organisational tasks and activities. Given their dynamic and longitudinal orientation, they are closely aligned with our own analytical approach, informing our exploration of the roles of alliance experience accumulation and frequent engagement in alliances in organisational learning and enhanced innovation. Evolutionary theory argues that by repetitively engaging in organisational tasks, or by gaining experience, organisations benefit from efficiency improvements as a result of "learning-by-doing". As firms follow different paths in accumulating and making sense of experience, they become heterogeneous in their abilities to carry out similar organisational activities. Experience assists firms in assessing which processes and practices are likely to perform better or to confer improved results, by judging on similar situations in the past. Organisations gradually create routines to efficiently coordinate organisational activities and develop capabilities to achieve outcomes important for competitive advantage (Dosi et al., 2000; Winter, 2000, 2003; Schilke and Goerzen, 2010; Nelson and Winter, 1982). Experience accumulation is an antecedent to developing capabilities; however, there can be a deliberate and conscious process of organisational learning as firms make investments in time, resources, training and knowledge to further improve their capabilities (Zollo and Winter, 2002), or to improve them through reflecting on newly acquired experience (Helfat and Peteraf, 2003). Dynamic approaches to RBV are illuminating on the process of capability development by suggesting that there is an evolutionary path to their creation whereby capabilities emerge and improve through several developmental stages (Helfat and Peteraf, 2003). First, there is a "capability founding" stage wherein organisations realise the need to develop a capability to meet desirable objectives and to orchestrate joint organisational action towards meeting these. Second, a "capability development" stage: here improvements to capabilities arise from a range of factors including further experience accumulation, investments in capital, organisational processes and learning mechanisms. Experience accumulation is crucial to both "founding" and "development" stages of a capability development. A third stage, "capability maturity" is also identified: in this phase capabilities may be replicated, retrenched, renewed, redeployed or re-combined with other capabilities. Even though alliances differ compared to other organisational tasks, in that firms collaborate with different partners in agreements with varying content, purpose and governance structure, gaining experience in managing alliances can lead to superior coordination, value creation and capture in alliances (Kale and Singh, 2007; Heimeriks and Duysters, 2007). Alliance experience is found to enhance the performance of alliances between recurrent partners, as firms use their experience and routines developed during their first interactions in managing subsequent collaborations within the dyad (Zollo et al., 2002). Moreover, firms mobilise experience and lessons gained across alliances with different partners (Anand and Khanna, 2000), with partner-specific alliance experience making a greater contribution to performance compared to general (i.e. non-partner specific) experience (Gulati et al., 2009). Alliance experience (a firm's cumulative number of alliances) is an antecedent to both superior coordination of alliance tasks (alliance capabilities), and to the efficient management of synergies and redundancies across alliances in alliance portfolios (alliance portfolio capabilities) (for a review see Wang and Rajagopalan, 2015). Kale and Singh (2007) were the first to detail how firms proactively learn from their alliance experience by externalising, codifying, diffusing and internalising individual and group-level alliance know-how. A stream within the alliance literature explores the link between cumulative alliance experience, the development of dedicated alliance management functions and other processes and practices connected with upgrading alliance management routines (Kale et al., 2002; Heimeriks and Duysters, 2007; Kale and Singh, 2007). Alliance functions reflect alliance capabilities and can lead to superior outcomes from alliances. Alliance experience indirectly affects firm performance via such capabilities, indicating that dedicated alliance functions embody and absorb the impact of alliance experience, suggesting that firms leverage all the learning gained from experience in their alliance management functions (Kale et al., 2002). Other research suggests that alliance experience directly influences both financial (Kale et al., 2002; Anand and Khanna, 2000) and non-financial outcomes such as innovation performance (Sampson, 2005), and that this is in addition to any impact conferred by dedicated alliance functions, suggesting a complementary relationship between the two (Heimeriks and Duysters, 2007). Indeed, systematic alliance practices are improved through training, mentoring programmes and external consulting. Such learning processes aim at introducing new alliance management practices and/or at improving existing ones in response to dissatisfaction with current performance (Zollo and Winter, 2002). The following section revisits existing but still contested hypotheses on the expected impact of alliance experience on firm innovation performance. We also develop our hypothesis on the impact of alliance frequency, which from a longitudinal perspective can be better positioned to capture nuancing and conditioning factors that reflect the impact of organisational learning from alliance experience on innovation. Firms with greater experience can draw from a greater pool of situations about what has worked in practice, when making decisions and inferences with respect to the performance of organisational practices (Levitt and March, 1988; Argote et al., 1990). Alliance experience (cumulative number of alliances) improves firms' abilities: to manage and coordinate alliances, to improve coordination of inter-organisational relationships and joint tasks, to form efficient arrangements for knowledge sharing, to deal effectively with unforeseen contingencies and to identify ways to overcome and resolve inter-partner conflict (Anand and Khanna, 2000; Sampson, 2005; Belderbos et al., 2015; Rothaermel and Deeds, 2006). Due to the link between alliance experience and organisational learning, experience is seen as a fundamental antecedent to both alliance and alliance portfolio capabilities (Wang and Rajagopalan, 2015; Kale and Singh, 2009; Shukla and Mital, 2018). Literature suggests that firms may not be in a position to benefit from learning from experience and superior coordination of alliances, when facing power asymmetries and resource dependence in alliances. Conflict is more frequent in such alliances which affects value creation and capture, especially for the weak partner as they are at a comparative disadvantage (Diestre and Rajagopalan, 2012). Power asymmetries are likely in the bio-pharmaceutical sector, as large pharmaceutical firms may be collaborating with small-dedicated biotech firms and due to their longer commitment to alliances and historic investments in downstream capabilities may be at a comparative advantage in deriving value from alliances (Caner and Tyler, 2013). We argue that by gaining greater alliance experience, firms gradually establish more fruitful conditions for effective knowledge sharing with their partners, and generate a basis for circumventing or resolving alliance conflicts. Because conflicts are more likely to occur in the bio-pharmaceuticals sector we expect that the higher the levels of experience the more likely the firms are to capture value from alliances and enhance their likelihood of innovation: H1. Alliance experience accumulation positively affects a firm's likelihood of innovation. A body of work suggests that efficiency improvements from experience accumulation make a gradually diminishing contribution to firm performance. There are two likely explanations for this. First, as firms form more alliances and accumulate greater experience, the contribution of any additional alliance to accumulated learning becomes increasingly reduced. This has been identified irrespective of type of partner and across upstream, downstream and horizontal alliances (Rothaermel and Deeds, 2006). Firms may experience diminishing returns when managing an increasing number of alliances due to the increased costs associated with identifying suitable partners, establishing and maintaining contracts, difficulties in absorbing knowledge from diverse sources, conflicts due to knowledge asymmetries and increased possibilities of knowledge redundancies (Diestre and Rajagopalan, 2012; Rothaermel and Deeds, 2006; White and Lui, 2005). A second likely explanation is that the value of lessons learned from cumulative experience depreciates over time as knowledge becomes increasingly obsolete (e.g. Argote et al., 1990). Past knowledge may depreciate as task requirements change over time. Further, the value of past lessons may dissipate gradually (or suddenly) as a result of employee turnover or a failure to store intelligence in organisational routines, or in a way that permits its re-application (Levitt and March, 1988; Dutton and Thomas, 1984). Unless firms make appropriate investments in retaining and leveraging lessons learned from their experience, its effects will increasingly dissipate. Research in the alliance literature has demonstrated a positive and linear relationship between alliance experience and firm-level indicators of financial (Anand and Khanna, 2000) and innovation performance (Shan et al., 1994). Only a slender body of work has explored diminishing returns to alliance experience. Hoang and Rothaermel (2005) find that, whilst alliance experience of small firms positively impacts the performance of their joint projects with large firms, the alliance experience of large firms makes no significant impact on their joint alliance project performance. Benefits stemming from alliance experience differ between small and large firms because they are positioned at different levels along the diminishing/concave alliance experience curve. In related research, Sampson (2005), commenting on patenting performance in the telecom equipment industry, identifies similarly diminishing returns to cumulative alliance experience, which she attributes to a diminishing contribution of distant experience, suggesting that distant alliance experience plays less important role than recent experience in efficient management of alliance portfolios (Shukla and Mital, 2018). As a consequence, we derive the following hypothesis: H2. The effect of cumulative alliance experience on likelihood of innovation exhibits diminishing marginal returns: as alliance experience accumulates, its contribution to the likelihood of firm innovation decreases. Deciphering, coding and measuring capabilities is notoriously difficult (Godfrey and Hill, 1995). As a result, the alliance literature has, in the main, relied on identifying alliance management practices (e.g. alliance functions) as a way of documenting alliance capabilities (Kale et al., 2002; Kale and Singh, 2007, 2009). An exception is found in the work of Rothaermel and Deeds (2006). They explore an inverted U-shaped relationship between cumulative alliance experience and new product development. They argue that, as the inflection point of the inverted U-shaped curve corresponds to the level of experience beyond which firms start experiencing inefficiencies in alliance management, it can reflect the level of their alliance capability. The notion that dedicated alliance management functions reflect alliance capabilities has being challenged in recent literature. First, anecdotal evidence suggests that firms establish dedicated alliance functions only after engaging in alliances over a protracted period, and after initiating a substantial number of such agreements (Kale and Singh, 2009). As a result such functions cannot reflect alliance capabilities, particularly when capabilities are at the initial stages of their development. Research indicates that firms - especially those motivated by perceived market signalling requirements - may establish dedicated alliance functions as a result of mimitisism or isomorphism (Heimeriks, 2010)[1]. Indeed, isolated mechanisms for institutionalising learning (e.g. dedicated alliance functions) from cumulative alliance experience may prove to be less effective compared to mechanisms designed to integrate and embed alliance knowledge throughout the organisation (Heimeriks et al., 2015). Despite some tentative and tangential work, explorations with respect to the underlying factors that lead firms to become aware of the need to leverage their alliance experience in developing a capability, and to invest efforts and resources in processes for deliberate learning from alliances are sparse in extant literature (see Wang and Rajagopalan, 2015). We contend that the first instances of frequent engagement in alliances can form such underlying factors and we turn to evolutionary theory and dynamic RBV to assist in elaborating our argument. Specifically, we suggest that frequent engagement in alliances may compel firms to found an alliance capability development lifecycle, as they both protract the period of engagement in alliances, and engage more substantially in this activity (Helfat and Peteraf, 2003). By engaging in alliances frequently, firms expand, augment and renew their stock of cumulative alliance experience. Frequent engagement in alliances induces firms to invest in developing higher order routines and practises to initiate purposeful and directed organisational learning. Developing processes for alliance management and practices for their further improvement requires managerial commitment, commitment that is highly demanding in terms of time and employee resources (Winter, 2000, 2003; Dosi et al., 2000). As a result, firms need to justify the allocation of resources to building dedicated routines for alliance management: frequent task reoccurrence can trigger and motivate firms to deploy resources in such a direction. It can also justify such investments to external stakeholders, building internal support and commitment to the purpose. Sparse empirical literature explores the role of frequent engagement in alliances in developing alliance capabilities (see Wang and Rajagopalan, 2015 for a review). There is a particular dearth of longitudinal studies in this field (Draulans et al., 2003; Rothaermel and Deeds, 2006). Research on CIS longitudinal Spanish data indicates that persistent and not sporadic collaboration exhibits a systematic positive relationship with firm innovativeness, which runs parallel to our argument (Belderbos et al., 2015). Moreover, firms that renew and expand their collaborations and alliances over time become more embedded in a network and experience higher growth over time (Powell et al., 2005). At the cross-sectional level, there is only limited research that links larger alliance portfolios to the development of alliance capability (as this is reflected in alliance management practices). Larger alliance portfolios are associated with the use of enhanced alliance management practices and processes, with the latter being linked to improved firm performance (Heimeriks et al., 2009). Our longitudinal approach, capturing the tendencies of firms to engage in alliances more frequently over time, complements such accounts which are based on the role of stock measures of alliance experience in firm performance (Kale et al., 2002; Heimeriks and Duysters, 2007; Kale and Singh, 2007). Therefore we propose that firms that engage in alliances frequently are more likely to innovate compared to those firms that engage in alliances irregularly or sporadically, or that retain a static engagement and a stagnating alliance portfolio over time: H3. Frequent engagement in alliances and the ensuing expanding alliance portfolios positively affect firm likelihood of firm innovation over time. 5.1 Sample We test our hypotheses on a panel data set observing the population of firms in the UK bio-pharma sector (110 firms in 2003) during the period 1991-2001. The empirical exploration in this paper relies fundamentally on econometric analysis. However, secondarily the paper employs a case analysis to shed further light on - and nuance - some of its arguments. To identify our sample, we used two editions of the UK biotechnology directory (Coombs and Alston, 2000, 2002). The directory lists all firms in the sector that undertake research using biotechnologies. It provides a more comprehensive account of the firms in the UK bio-pharma sector than that available via the use of SIC classifications, as biotechnologies are used by firms from a range of industries in the life sciences as broadly described (Walsh, 2004). To form our working sample, we used the "Who Owns Who" directory to identify all firms that were active in 2003. Out of that list, we identified 110 firms that publish accounts in FAME[2]. For the sample of 110 firms, we collected data on alliances from ReCap.com and BioScan, the two most popular sources of alliance data in empirical research on this sector (e.g. Deeds and Hill, 1996; Schilling, 2009). The databases report alliances established in the bio-pharmaceutical sector for innovation purposes, including R&D alliances, research alliances and alliances for technology licensing and product development[3]. Innovation is crucial to firm survival and prosperity in the bio-pharmaceutical sector and patents are perceived as an appropriate proxy for innovations due to the science-based nature of the sector (e.g. Pavitt, 1984). Moreover, patents are an appropriate indicator to capture firm-level returns from innovation alliances (Demirkan, 2018; Shan et al., 1994; Sampson, 2005). We collected data on patents granted to the 110 firms in our sample between 1991 and 2001 by the UKPTO through the publicly available database Esp@cenet. Via deployment of Boolean searches, it was possible to assemble a comprehensive list by matching both the name and the address of firms in our sample with those of the patent assignees (Arora et al., 2011). For firms in our sample that belong to MNEs with non-UK-based HQs, we examined information on inventor location, and allocated any patents assigned to the MNE HQ with UK-based inventors to the UK-based subsidiary in our sample. Since our research focusses on UK-based firms, the UKPTO is perceived to be an appropriate source for the collection of patent data. Filing patent applications at national patent offices is less costly (in terms of application and renewal fees) and less time consuming when compared with international patent applications. As international applications also involve costs of unfamiliar IP regulations and legal frameworks, many firms - at least those with constrained resources - are more likely to patent at national patent offices than internationally (Archambault, 2002). Finally, information on firm accounts was obtained from FAME, and information on investments in R&D was gathered from Thomson's Analytics and the UK DTI's R&D Scoreboard. 5.2 Methods Patents take positive integer values, making count dependent variable models, such as the Poisson and Negative Binomial, appropriate (Greene, 2003). Instead of count dependent variable models, we employ discrete dependent variable models as they permit the capture of factors that alter firm abilities enabling them to perform over a threshold level: successful filing of a patent (Long, 1997). This is better aligned to the core aim of this work, which is to explore whether frequent alliance engagement is linked to organisational learning in alliance management (and alliance capabilities), reflected in firms exhibiting over threshold performance (Hoang and Rothaermel, 2005). Due to the longitudinal dimension of panels, we can explore factors that affect firms' changing abilities to innovate over time (fixed effects specification), as well as factors that can explain cross-firm variation in innovation abilities (random effects specification). The former, dynamic consideration of panels is of particular interest in this paper, as explained above. The fixed effects specification has a further appeal as its estimates are not based on the assumption of strict exogeneity of the independent variables (Greene, 2003; Baltagi, 1995), which is desirable, as decisions to form alliances may not be independent of past innovation performance (Colombo and Garrone, 1996; Gkypali et al., 2017). To explore further the nuanced factors underlying the development of alliance capabilities, we report on the findings of an in-depth case analysis that explores the link between frequency of engagement in alliances, and the development of practices and processes to manage alliances, an indication of alliance capabilities (Schreiner et al., 2009; Kale and Singh, 2007); this is a secondary analysis providing an illuminating complement. The case focusses on a UK bio-pharmaceutical firm that intensified its engagement in alliances particularly after the early 1990s when it started to invest in biotechnology R&D. Interviews were conducted in 2005 as the firm was forming a dedicated alliance function and entailed detailed reflection on the factors that underpin the creation of processes established to improve efficiency in alliance management. The processes identified were matched with those identified in existing literature on alliance capabilities (Kale et al., 2002; Kale and Singh, 2007, 2009; Heimeriks and Duysters, 2007). The case is based on interviews with directors of alliances and intellectual property, with a long working experience with the case firm and focussed on alliances that involve cooperation in research and that have led to patents. All materials and associated analyses were reviewed and approved by the interviewees. 5.3 Variables Patents (dependent variable) To test our hypotheses we create a dummy variable that is equal to 1 for each firm that is granted a patent at year t (and equal to zero otherwise). We use a patent dummy variable as we aim at capturing the impact of alliance experience in enabling firms to perform over a threshold, i.e. to innovate. To track the firm's innovation activities over time, we use patent filing dates. Filing dates have certain advantages over publication dates: first, they better reflect the originating time of innovations; and, second, they are not influenced by regulatory changes or fluctuations in resource availability in patent offices over time, as these can delay publication dates (e.g. Jaffe, 1986). Patents are widely used to capture firm returns from innovation alliances (Sampson, 2005, 2007; Shan et al., 1994; Demirkan, 2018). Alliance experience Research has explored the role of alliance experience in improving financial returns of firms from alliances (Kale et al., 2002), alliance project success (Hoang and Rothaermel, 2005), returns to firm innovation performance (Sampson, 2005, 2007; Belderbos et al., 2015; Di Guardo and Harrigan, 2016) and number of products under development (Deeds and Hill, 1996; Rothaermel and Deeds, 2006). Studies have operationalised alliance experience as a stock variable reflected either by the cumulative number of alliances formed throughout a firm's history (Kale and Singh, 2007; Hoang and Rothaermel, 2005; Kale et al., 2002; Rothaermel and Deeds, 2006), or by the cumulative number of years of experience in engaging in alliances (Rothaermel and Deeds, 2006). Only a handful of these studies consider diminishing (Sampson, 2005) or non-linear (Rothaermel and Deeds, 2006) relationships between alliance experience and firm returns from alliances and most studies base their estimates on cross-sectional data (an exception is Kale et al., 2002). Following existing literature, we operationalise alliance experience at each year t as the cumulative number of alliances throughout a firm's history. To aid interpretation of results, and following recent contributions (Lavie et al., 2011) we use one year lagged values of the above variable to more accurately capture the effect of cumulative alliance experience. Our variable is left-censored to 1991, the start year for our sampling. This is experienced elsewhere in alliance literature (e.g. Kale and Singh, 2007). Also, because in our study 87 per cent of total alliances from 1979 and 2001 fall within the 1991-2001 period, left censoring is not introducing considerable bias. As is customary in the literature (Sampson, 2005), we use the natural logarithmic transformation of cumulative alliance experience to test for diminishing returns (H2). As a robustness check, we test for a concave relationship (inverted U-shaped) by using both the linear and quadratic values of the cumulative number of alliances: again, this approach follows relevant literature (e.g. Deeds and Hill, 1996; Rothaermel and Deeds, 2006)[4]. Frequent engagement in alliances and expanding alliance portfolios To explore the effects of frequent engagement in alliances on firm abilities to innovate, we measure the total number of on-going alliances that a firm manages simultaneously at any point in time. As alliances last for several years, this requires information on alliance duration which is not available via the alliance data sources consulted in this study (RECap.com and BioScan). As lack of information on alliance duration has also been encountered in other work, we follow established tradition and rely on estimates of the average duration of alliances, i.e., five years (Kogut, 1988; Di Guardo and Harrigan, 2016). Therefore, we operationalise frequent engagement in alliances by calculating the number of alliances a firm manages at any given point in time by considering that alliances last on average for five years. Increases in this variable reflect the tendency of firms to engage more frequently in alliances over time, and on average, to initiate more alliances than they terminate, managing potentially an expanding portfolio of on-going alliances. As the use of a blanket measure for alliance duration might be considered simplistic, we cross-validated this information with alliance managers in the sector, and it was considered an acceptable approximation for the duration of the majority of alliances. We also performed robustness checks by using different estimates for alliance duration as detailed in our analysis. Control variables For parsimonious empirical models, we use the most appropriate control variables that have been employed in the literature on alliances and innovation. First, we account for the effects of well-established indicators of firm innovation, such as investments in R&D and firm size (Griliches, 1990) as they can influence firm abilities to innovate. We used a measure of R&D intensity that captures the amount of R&D expenditure per employee. Firm size is captured by annual turnover and we include a natural logarithm to overcome the possible problem of skewed distributions. Second, and following relevant research (Zollo et al., 2002), we account for differences in the content of alliances, as it influences innovation opportunities in alliances and affects firm performance (e.g. Anand and Khanna, 2000). Therefore, we include a dummy that takes the value of one when the focal firm establishes an alliance that involves R&D, and the value of zero for each year that a focal firm establishes alliances that do not involve R&D. Table I presents descriptive statistics and bivariate correlations. The panel is unbalanced, as a result of either missing information or the inclusion of less-mature dedicated biotechnology firms (i.e. a third of firms in the sample were established after 1995). The unbalanced data set includes firms observed for a maximum period of ten years and others for as little as four years. On average, firms are observed for approximately seven years (the range is from 6.6 to 7.4 years across the three estimated models) and form 8.28 alliances between 1991 and 2001. We estimated both the fixed and random effects specifications (FE and RE, respectively) of the Logit model for panel data for all three models (Greene, 2003; Baltagi, 1995)[5]. Results are obtained in STATA(c) 11. The r parameter is significant across all three models, indicating significant cross-firm heterogeneity. As Hausman's (1978) specification test has limited discerning power in discrete dependent variable models, we assess the relative merits of the two specifications on a conceptual basis and for reasons discussed in the methods section we present Logit FE estimates. Moreover, the residuals of the FE specification are corrected for potential heteroscedasticity (White, 1980; Baltagi, 1995). Based on model kh2 statistics in Table II, all three models are significant at 5 per cent levels and above. This suggests that our variables have a significant interpretative power with respect to the likelihood of firm innovation. We find support for all hypotheses. In Model 1, the cumulative number of alliances positively affects firm's likelihood of innovation over time, but levels of significance are low (10 per cent). Our results are consistent with studies showing a positive relationship between cumulative alliance experience and firm innovation, but at low significance levels (Hagedoorn and Schakenraad, 1994; Shan et al., 1994). The FE estimates of Model 2 provide support for diminishing returns to firm innovation from cumulative alliance experience, with the corresponding variable being positive and highly significant. The results of Model 2 are consistent with findings testing similar hypotheses in cross-sectional settings in the US bio-pharmaceutical and the telecom equipment industries (Hoang and Rothaermel, 2005; Sampson, 2005). Our results are robust to alternative operationalisations, as reflected by exploring concave relationships (inverted U-shaped), via use of both linear and quadratic values of cumulative alliance experience, with estimates (both FE and RE) confirming non-linear diminishing returns to likelihood of firm innovation. To unpack the extent of diminishing returns we estimate the inflection point based on estimates of the curvilinear relationship; this is equal to 127.515 points of cumulative experience, which corresponds to firms managing 18 alliances annually (average length of time per firm in the panel is seven years). Our results imply that at such a level of annual alliance activity, firms will accrue no additional benefits from gaining more experience in managing alliances. It should be noted that in our sample the average number of new alliances formed every year is 8.28, which suggests that the majority of firms in our sample are below the optimal level of alliance engagement. The concluding section discusses the implications of this. Model 3 provides support for our third hypothesis. The coefficient of the core independent variable is positive and significant (it marginally fails to pass the highest 1 per cent level of significance). Model 3 suggests that firms that establish alliances frequently and that continuously manage an expanding portfolio can enjoy improved likelihood of innovation. This provides fresh insights as, to date, there has been a scarcity of longitudinal studies in the literature (see Draulans et al., 2003; Rothaermel and Deeds, 2006). As a test of robustness, we estimated Model 3 by using four and six years as the average duration for alliances. The results remained robust to different operationalisations. Among the control variables, firm size is positively and significantly associated with firm's likelihood of innovation in Models 1 and 3. Investments in R&D per number of employees are negatively associated with likelihood of firm innovation but only in Model 2. Alliance content is insignificant across all three models. Finally, all three model estimates are robust to different measures of the control variables. For example, all models lead to the same outcomes in terms of significance when controls are collapsed to dichotomous variables indicating above and below average firm size and intensity of investments in R&D per number of employees. A further robustness check is undertaken to examine whether our results are sensitive to the statistical assumptions of the distribution followed by the error term in the Logit and Probit specifications. The Probit model only provides RE estimates for panel data (STATA(c) 11) and these further confirm our results. Moreover, we estimate our model specifications with the Negative Binomial model to explore differences between our dependent variable and a count variable capturing innovation performance with the number of patents. Both FE and RE estimates confirm our hypotheses. The results suggest that there are diminishing returns to cumulative alliance experience. This can signal that ageing experience contributes less to current outcomes and that recent experience may have a higher contribution (Shukla and Mital, 2018; Sampson, 2005). Following Sampson (2005), we explore the contributions of recent and past alliance experience. We develop a set of variables capturing alliance experience between 1 and 6 years prior to our year of observation. So, for example, in 2001 alliance experience of one year corresponds to the number of alliances initiated in 2000 and alliance experience of two years corresponds to those initiated in 1999 and so on. None of these variables appear to be significant, with the exception of alliance experience of four years prior to the year of observation, which appears with a negative and significant sign. Therefore, the diminishing returns to cumulative alliance experience identified in our paper cannot be attributed to decreasing contributions of distant experience. Our results most likely reflect that firms, by just forming more alliances and learning from experience how to improve alliance management and coordination, cannot experience improved efficiency ad infinitum. With regard to H3 we undertake further estimations to isolate the impact of frequent engagement in alliances from that of portfolio size. We break down the variable testing H3 into two components. First, capturing changes in the frequency of forming alliances over time, we construct a variable reflecting a decreasing, static and increasing trend in alliance portfolios (the variable takes the values of: -1, 0 and 1, respectively, and refers to comparisons between the current and the previous year). Second, we include a variable to capture the number of new alliances formed within every year (one year lagged values are used in the models). Our results show that the trend variable has a negative sign while the annual figures of newly formed alliances appear with a positive and significant sign (both for the Logit and the Negative Binomial models and the FE and RE specifications). It is worth noting that when the values of the trend variable are lagged by one year, their impact becomes insignificant (i.e. they retain a negative sign). This means that only the contemporaneous impact of alliance frequency on innovation is negative, implying that when firms expand their alliance portfolio there is an immediate adverse effect. This may result from immediate strains on their resources and cognitive abilities, which however, cease to have an adverse effect within a year's time. The results remain the same when the trend variable is substituted with a dummy variable capturing only increases in the frequency of forming alliances between current values and those of the previous year. Finally, we isolate partner-specific alliance experience from general (i.e. non-partner specific) experience to further explore the importance of alliance management learning within partner dyads and in alliance portfolios. None of these variables has a significant effect, suggesting that the separation lacks interpretative power. The paper contributes to the literature on the role of alliances in innovation and alliance capabilities by using a longitudinal analysis which enables capture of the impact of organisational learning in managing and coordinating alliances on innovation. As it is difficult to capture learning, the paper uses a longitudinal approach to capture dynamic changes within firms and it observes learning through its impact on outcome variables such as innovation. The paper adds to a slender body of work exploring antecedents to firm-level alliance capabilities and their impact on enhancing firm's abilities to innovate (for a review see Wang and Rajagopalan, 2015). First, we find that the impact of learning from gaining greater alliance experience in coordinating alliances and portfolios on firm innovation is positive but of weak significance. This is an important finding, as alliances in the bio-pharmaceutical sector can be characterised by power asymmetries and unbalanced resource dependence between partners, which are likely to create conflicts affecting the creation and appropriation of alliance value (Diestre and Rajagopalan, 2012; Caner and Tyler, 2013). The period of our study forms the first window of exploitation of biotechnologies via alliances which potentially creates disincentives to collaborating firms to behave opportunistically, as joining forces during this close to pre-competitive stage could be more beneficial than not (Powell et al., 2005). Second, our analysis proceeds by exploring the factors that affect learning in managing and coordinating alliances in this sector. The first arm to this exploration sheds light on the role of cumulative-historic experience in enhancing innovation performance, while the second arm focusses on the role of such experience as an antecedent to the creation of capabilities to manage interactions with alliance partners and to create synergies and minimise redundancies in alliance portfolios. In that endeavour, we first confirm diminishing returns to the benefits that can be accrued by gaining alliance experience for more effective management of alliances and capturing innovation returns. This supports a thin body of research which suggests that learning from gaining experience cannot confer benefits ad infinitum (e.g. Rothaermel and Deeds, 2006; Sampson, 2005). Our in-depth exploration shows that this cannot be attributed to diminishing contributions of ageing experience as identified in other work (Sampson, 2005; Shukla and Mital, 2018). It could be attributed to failure of firms to capture learning in a systematic way, or to costs associated with identifying more partners, establishing more agreements and coordinating increased diversity (White and Lui, 2005; Rothaermel and Deeds, 2006). Our last set of findings show that engaging in alliances more frequently over time, with firms managing an expanding portfolio, by forming more alliances than they terminate, enhances firms' likelihood of innovation. Our longitudinal approach is consistent with other work showing that persistence in collaboration has a positive impact on firm innovation as opposed to sporadic collaboration (Belderbos et al., 2015). Longitudinal explorations are currently almost absent in alliance literature, and their potential contribution has been particularly welcomed (Draulans et al., 2003; Rothaermel and Deeds, 2006), specifically as they might shed further light on the type of changes that can occur within firms, in particular, with respect to the dynamics of learning on alliance management and how it enhances innovation potential. Contributions within evolutionary theory (Winter, 2000, 2003) argue that frequent engagement in organisational tasks is linked to capability development, while the dynamic resource-based view (Helfat and Peteraf, 2003) argues that capabilities develop through several stages: experience accumulation compels firms to recognise the need to establish a capability, and this initiates a process of systematic learning and alliance management. Our results clearly pertain to this literature and specifically on how alliance experience forms an antecedent to alliance and alliance portfolio capabilities (Wang and Rajagopalan, 2015). To explore, in a situated and more granular fashion, whether frequent engagement in alliances and a dynamically expanding alliance portfolio contribute to the development of alliance capabilities, we conducted a detailed case analysis as a complement to our econometric analysis. The case analysis explores the link between an increasing engagement in alliances and the development of novel or changed practices and processes for alliance management, i.e., those factors that normally sit within the "black-box" of alliance capability studies (Anand and Khanna, 2000; Heimeriks and Duysters, 2007; Kale and Singh, 2007). The case analysis clearly reveals such a relationship. First, it shows that alliance management practices and routines are established after alliances become an integral part of firm R&D and innovation strategy and firms intensify and commit to alliance activity. As the Alliance Director interviewed for our study stated: "[...] we have standardised approaches for contractual compliance [with respect to alliances] [...] it's inconceivable to me that one wouldn't have, and yet five years ago we didn't [...] You've got one deal and you do it however you do it which largely depends on who you employ and who it's [the alliance] with [...] I am not sure how many deals you need to have before you get yourself a standardised process but it's probably substantially more than ten or twenty [...][7]". Second, several of the practices identified in the extant literature that are used to institutionalise and diffuse alliance management learning (namely, employment of alliance professionals, use of alliance metrics, creation of databases to solicit potential alliance partners and establishment of processes to monitor alliance performance) are deployed by the case firm (Heimeriks and Duysters, 2007; Kale and Singh, 2007). Ten years after intensifying alliance engagement, it developed a dedicated alliance management function to coordinate alliance activity and engage in monitoring and improving existing alliance management practices. The case analysis illustrates the links between frequent engagement in alliances and the ensuing expanding the alliance portfolio, with the motivation to invest resources in developing firm-specific capabilities for alliance management. It also reflects the time required to recognise that alliance management can be improved via the development of specific and specialist processes for the governance of cross-partner interaction (a finding that resonates strongly with existing literature, Kale and Singh, 2009). In addition, the case reflects the requirement for firms to continuously review and develop their capabilities in order to confront new challenges and intensified competition. Indeed, the case analysis, combined with the findings of our econometric exploration, suggest that renewing alliance experience and committing to forming more alliances over time are crucial in enabling firms to create and capture value in alliances. Our research contributes to the literature on the antecedents of alliance and alliance portfolio capabilities and on the conditions that shape superior firm outcomes from alliances, such as innovation. It suggests a need to delve into the antecedents to alliance capabilities, a thin body of research and to identify nascent factors that provide potential foundations for alliance capability development (Helfat and Peteraf, 2003), shifting attention away from the role played by alliance management practices (such as dedicated alliance functions) that have dominated existing research. This is particularly important as firms may establish such practices during the advanced stages of the alliance capability development process (Kale and Singh, 2009), and as such they may not appropriately reflect the foundational stages of alliance capability development. Here, we echo calls to delve deeper in understanding alliance management capabilities and the antecedents to alliance portfolio capabilities (Wang and Rajagopalan, 2015), especially in the context of alliances involving a higher learning potential (Heimeriks, 2010). Moreover, recent research shows that codification of alliance learning and systematic approaches to alliance management contribute to efficient partner selection and alliance termination, but may restrict flexibility and adaptability which are important for efficient management during the course of the alliance (Heimeriks et al., 2015; Wang and Rajagopalan, 2015). Our research has several implications for managerial practice and policy. First, we find that there are diminishing returns to benefits in terms of enhanced innovation accruing from gaining greater alliance experience. Specifically firms reach a plateau at 18 alliances in annual terms, suggesting that on average this is the optimal level of alliance engagement for enhancing firm likelihood of innovation. This has implications for managing alliance portfolio size for innovation efficiency (e.g. Hoffmann, 2007; Demirkan, 2018). On average firms in our sample are below the optimal level of alliance engagement (forming eight alliances annually). This suggests that innovation likelihood can be marginally enhanced by expanding alliance engagement. This is in line with findings of other research (Rothaermel and Deeds, 2006; Sampson, 2005; Hewitt-Dundas and Roper, 2018) which in the context of micro-businesses, for example, shows that sub-optimal engagement can be attributed to reluctance of firms to cooperate due to fear of trustworthiness and lack of knowledge on benefits of collaboration and partner capabilities (Hewitt-Dundas and Roper, 2018). This emphasises our findings with respect to the need to facilitate a culture of systematic management of alliances. It also supports research suggesting the need for policy interventions, for example in the form of intermediaries, which can support and encourage firms to establish collaborations and manage openness (Howells, 2006). This is not only due to the benefits of collaboration to firm innovation but also due to the positive externalities from openness, suggesting broader economic and societal benefits (Roper et al., 2013). At the firm level, our results suggest that it is important for firms to replenish their alliance portfolios and to gradually form more alliances than they terminate over time if they wish to enhance their innovation likelihood. In the bio-pharmaceuticals sector where there is high rate of alliance formation, it is important for firm competitiveness and future growth that they renew and expand their networks (Powell et al., 2005). This is linked to firms becoming more aware of the need to make alliance management learning more systematic. Firms should use such systematic practices in a discretionary manner and should also consider altering and upgrading such practices to respond to the challenges and demands of new alliances. Although our results are specific in terms of sector and period, they derive from the examination of a paradigmatic period in the development of new technologies, and address the implications of changing innovation processes in an established industry. As such, they provide useful insights that assist understanding of trajectories and dynamics in industries with similarities to biotech, for example, nanotechnology (e.g. Rothaermel and Thursby, 2007). Our aspiration is that our research will provide some re-focussing and clarity in an established field of study. We acknowledge that analysis of further sectors and periods will nuance our understanding of both the role of gaining experience in improving firm alliance management and in developing alliance capabilities, and hope that the ideas and methods explored above will assist in this on-going quest.
|
The paper estimates a range of models identifying the relationship between alliance experience and firm innovation. The panel data sample captures the full range of firms active in the UK bio-pharmaceuticals sector during the early stages of its development observing them from 1991 to 2001. An exploratory case study analysis is employed to shed light on the nuanced factors linking frequent engagement in alliances to the development of practices for efficient alliance management.
|
[SECTION: Findings] As the popularity of strategic alliances is increasing and such alliances are becoming an integral component in business development, attention in the research community has moved towards an exploration of their role in firm performance and innovation. Formal and informal interactions with external actors have long being argued to play a fundamental role in firm innovation (Von Hippel, 1988; Frankort et al., 2012; Rice et al., 2012; Colombo et al., 2011; Demirkan, 2018). However, it is accepted that alliances carry coordination costs and risks of misappropriation, and these frequently diminish chances of success or preclude full acquisition of anticipated benefits (e.g. De Man and Duysters, 2005; Gkypali et al., 2017; Faems et al., 2010). A substantial body of literature finds a positive relationship between the extent of alliances and firm innovation performance, whilst other work identifies diminishing and even negative returns (Hoang and Rothaermel, 2005; Sampson, 2005; Laursen and Salter, 2006; Rothaermel and Deeds, 2006). As a result, a growing strand in the literature has examined the factors enabling firms to generate and capture value from alliances. One such factor is alliance experience accumulation: as firms develop greater experience in managing alliances they become better at coordinating cross-organisational tasks and knowledge flows (e.g. Sampson, 2005). Another factor that can improve performance in alliances is developing formal and codified processes and routines for alliance management (e.g. the establishment of dedicated alliance functions), which are argued to capture, or to be reflective of, firm-specific alliance management capabilities (Kale et al., 2002; Kale and Singh, 2009, 2007; Heimeriks and Duysters, 2007; Sampson, 2005; Schreiner et al., 2009; Di Guardo and Harrigan, 2016; Shukla and Mital, 2018). Whilst these contributions offer useful insights, much the greater part of current theorising has been constructed via the use of cross-sectional data. Longitudinal explorations are scarce, thus we lack a nuanced understanding of the type of changes that occur within firms over time with respect to enhanced innovation potential from engaging in alliances. The contribution of this paper is that, on the basis of a longitudinal approach, it examines how frequent engagement in alliances and expanding alliance portfolios (a compilation of contemporaneous alliances) impact on the likelihood of firm innovation; this offers a longitudinal approach that is lacking in the current literature (Draulans et al., 2003; Rothaermel and Deeds, 2006) with few notable exceptions (Di Guardo and Harrigan, 2016; Belderbos et al., 2015). A dynamic approach is appealing as it captures the impact on innovation of firm learning with respect to management of alliances. Such learning is evident in the use of past alliance experience to support conflict avoidance or management, and the effective coordination of knowledge sharing and communication with partners. Improved management can affect value capture in alliances especially as firms manage expanding alliance portfolios due to both potential of cross-partner asymmetries and knowledge redundancies, which are common in the bio-pharmaceuticals sector (Diestre and Rajagopalan, 2012; Caner and Tyler, 2013). The paper uses longitudinal econometric panel data analysis to explore its hypotheses. Secondary to this, we deploy a case analysis to capture the nuanced changes due to learning, shedding more light on the role of frequent engagement in alliances in compelling firms to direct their attention and resources in developing practices and routines for alliance management. This unpacks the antecedents to alliance capabilities and travels beyond current understandings in relation to experience accumulation. As highlighted in a recent review (Wang and Rajagopalan, 2015), research on the antecedents to alliance and alliance portfolio capabilities is rather thin. This paper contributes to exploring this link further. Our arguments are explored in the context of a unique, bespoke and specially assembled data set, comprising the population of firms in the UK bio-pharmaceutical sector (110 firms "alive" in 2003) and active during the period 1991-2001. The UK bio-pharmaceutical sector presents an ideal setting for investigation. The UK accounts for a significant share of the worldwide bio-pharmaceutical sector (Mckinsey, 2014) and enjoys a revealed technological advantage (OECD, 2017) and comparative advantage (OECD, 2008) in this high-tech sector. Moreover, the UK is a prime example of the Liberal Market Economy variant detailed in the Varieties of Capitalism literature (Hall and Soskice, 2001; Hancke, 2009), where the socio-economic institutional configuration is associated with international competitiveness of industries that are characterised by radical innovations (such as the pharma and biotech industries). The period 1991 and 2001 is perceived to be particularly apposite for our analysis as it exhibits several critical features: a first instance and general upsurge in alliances (not only in the UK bio-pharmaceutical sector, but also in other sectors and territories, Kang and Sakai, 2000); the emergence of alliances as an integral component in firm R&D strategies in the bio-pharmaceutical industry (Hagedoorn, 2002; Demirkan, 2018; Powell et al., 2005) but also more generally (Rice et al., 2012); and the presence of significant variation - visible in terms of identifiable peaks and troughs in alliance activity - especially towards the end of the period (Hagedoorn, 2002). Moreover, this marks a paradigmatic era of new technology evolution, with new entrants and established firms developing capabilities in biotechnologies through alliance activity (Hopkins et al., 2007): thus it provides a historical backdrop that aids understanding of development of similar technologies, for example, nanotechnologies (Rothaermel and Thursby, 2007). The remainder of this paper is structured as follows. Section 2 sets-out the theoretical background to the study and reviews the literature on alliance experience and alliance/alliance portfolio capabilities. Sections 3 and 4 develop our hypotheses. Section 5 discusses our sample, data sources and methodological considerations and details the variables for our longitudinal econometric analysis. Section 6 is dedicated to estimation and results, and Section 7 provides results from post-estimation robustness checks. Section 8 discusses our findings together with the case study analysis and Section 9 highlights the implications of the study for management theory and practice. To explain heterogeneity with respect to firm's abilities to benefit from alliances, the alliance literature resonates strongly with knowledge and capability-based theories of the firm (e.g. Kogut and Zander, 1992; Helfat and Peteraf, 2003). Here, we adopt a perspective that is informed by both evolutionary theory of the firm (Nelson and Winter, 1982) and dynamic approaches to the resource-based view (RBV) (Helfat et al., 2007; Helfat and Peteraf, 2003). These approaches take a dynamic view of organisational development, emphasising the role of experience and knowledge accumulation in supporting improved management and coordination of organisational tasks and activities. Given their dynamic and longitudinal orientation, they are closely aligned with our own analytical approach, informing our exploration of the roles of alliance experience accumulation and frequent engagement in alliances in organisational learning and enhanced innovation. Evolutionary theory argues that by repetitively engaging in organisational tasks, or by gaining experience, organisations benefit from efficiency improvements as a result of "learning-by-doing". As firms follow different paths in accumulating and making sense of experience, they become heterogeneous in their abilities to carry out similar organisational activities. Experience assists firms in assessing which processes and practices are likely to perform better or to confer improved results, by judging on similar situations in the past. Organisations gradually create routines to efficiently coordinate organisational activities and develop capabilities to achieve outcomes important for competitive advantage (Dosi et al., 2000; Winter, 2000, 2003; Schilke and Goerzen, 2010; Nelson and Winter, 1982). Experience accumulation is an antecedent to developing capabilities; however, there can be a deliberate and conscious process of organisational learning as firms make investments in time, resources, training and knowledge to further improve their capabilities (Zollo and Winter, 2002), or to improve them through reflecting on newly acquired experience (Helfat and Peteraf, 2003). Dynamic approaches to RBV are illuminating on the process of capability development by suggesting that there is an evolutionary path to their creation whereby capabilities emerge and improve through several developmental stages (Helfat and Peteraf, 2003). First, there is a "capability founding" stage wherein organisations realise the need to develop a capability to meet desirable objectives and to orchestrate joint organisational action towards meeting these. Second, a "capability development" stage: here improvements to capabilities arise from a range of factors including further experience accumulation, investments in capital, organisational processes and learning mechanisms. Experience accumulation is crucial to both "founding" and "development" stages of a capability development. A third stage, "capability maturity" is also identified: in this phase capabilities may be replicated, retrenched, renewed, redeployed or re-combined with other capabilities. Even though alliances differ compared to other organisational tasks, in that firms collaborate with different partners in agreements with varying content, purpose and governance structure, gaining experience in managing alliances can lead to superior coordination, value creation and capture in alliances (Kale and Singh, 2007; Heimeriks and Duysters, 2007). Alliance experience is found to enhance the performance of alliances between recurrent partners, as firms use their experience and routines developed during their first interactions in managing subsequent collaborations within the dyad (Zollo et al., 2002). Moreover, firms mobilise experience and lessons gained across alliances with different partners (Anand and Khanna, 2000), with partner-specific alliance experience making a greater contribution to performance compared to general (i.e. non-partner specific) experience (Gulati et al., 2009). Alliance experience (a firm's cumulative number of alliances) is an antecedent to both superior coordination of alliance tasks (alliance capabilities), and to the efficient management of synergies and redundancies across alliances in alliance portfolios (alliance portfolio capabilities) (for a review see Wang and Rajagopalan, 2015). Kale and Singh (2007) were the first to detail how firms proactively learn from their alliance experience by externalising, codifying, diffusing and internalising individual and group-level alliance know-how. A stream within the alliance literature explores the link between cumulative alliance experience, the development of dedicated alliance management functions and other processes and practices connected with upgrading alliance management routines (Kale et al., 2002; Heimeriks and Duysters, 2007; Kale and Singh, 2007). Alliance functions reflect alliance capabilities and can lead to superior outcomes from alliances. Alliance experience indirectly affects firm performance via such capabilities, indicating that dedicated alliance functions embody and absorb the impact of alliance experience, suggesting that firms leverage all the learning gained from experience in their alliance management functions (Kale et al., 2002). Other research suggests that alliance experience directly influences both financial (Kale et al., 2002; Anand and Khanna, 2000) and non-financial outcomes such as innovation performance (Sampson, 2005), and that this is in addition to any impact conferred by dedicated alliance functions, suggesting a complementary relationship between the two (Heimeriks and Duysters, 2007). Indeed, systematic alliance practices are improved through training, mentoring programmes and external consulting. Such learning processes aim at introducing new alliance management practices and/or at improving existing ones in response to dissatisfaction with current performance (Zollo and Winter, 2002). The following section revisits existing but still contested hypotheses on the expected impact of alliance experience on firm innovation performance. We also develop our hypothesis on the impact of alliance frequency, which from a longitudinal perspective can be better positioned to capture nuancing and conditioning factors that reflect the impact of organisational learning from alliance experience on innovation. Firms with greater experience can draw from a greater pool of situations about what has worked in practice, when making decisions and inferences with respect to the performance of organisational practices (Levitt and March, 1988; Argote et al., 1990). Alliance experience (cumulative number of alliances) improves firms' abilities: to manage and coordinate alliances, to improve coordination of inter-organisational relationships and joint tasks, to form efficient arrangements for knowledge sharing, to deal effectively with unforeseen contingencies and to identify ways to overcome and resolve inter-partner conflict (Anand and Khanna, 2000; Sampson, 2005; Belderbos et al., 2015; Rothaermel and Deeds, 2006). Due to the link between alliance experience and organisational learning, experience is seen as a fundamental antecedent to both alliance and alliance portfolio capabilities (Wang and Rajagopalan, 2015; Kale and Singh, 2009; Shukla and Mital, 2018). Literature suggests that firms may not be in a position to benefit from learning from experience and superior coordination of alliances, when facing power asymmetries and resource dependence in alliances. Conflict is more frequent in such alliances which affects value creation and capture, especially for the weak partner as they are at a comparative disadvantage (Diestre and Rajagopalan, 2012). Power asymmetries are likely in the bio-pharmaceutical sector, as large pharmaceutical firms may be collaborating with small-dedicated biotech firms and due to their longer commitment to alliances and historic investments in downstream capabilities may be at a comparative advantage in deriving value from alliances (Caner and Tyler, 2013). We argue that by gaining greater alliance experience, firms gradually establish more fruitful conditions for effective knowledge sharing with their partners, and generate a basis for circumventing or resolving alliance conflicts. Because conflicts are more likely to occur in the bio-pharmaceuticals sector we expect that the higher the levels of experience the more likely the firms are to capture value from alliances and enhance their likelihood of innovation: H1. Alliance experience accumulation positively affects a firm's likelihood of innovation. A body of work suggests that efficiency improvements from experience accumulation make a gradually diminishing contribution to firm performance. There are two likely explanations for this. First, as firms form more alliances and accumulate greater experience, the contribution of any additional alliance to accumulated learning becomes increasingly reduced. This has been identified irrespective of type of partner and across upstream, downstream and horizontal alliances (Rothaermel and Deeds, 2006). Firms may experience diminishing returns when managing an increasing number of alliances due to the increased costs associated with identifying suitable partners, establishing and maintaining contracts, difficulties in absorbing knowledge from diverse sources, conflicts due to knowledge asymmetries and increased possibilities of knowledge redundancies (Diestre and Rajagopalan, 2012; Rothaermel and Deeds, 2006; White and Lui, 2005). A second likely explanation is that the value of lessons learned from cumulative experience depreciates over time as knowledge becomes increasingly obsolete (e.g. Argote et al., 1990). Past knowledge may depreciate as task requirements change over time. Further, the value of past lessons may dissipate gradually (or suddenly) as a result of employee turnover or a failure to store intelligence in organisational routines, or in a way that permits its re-application (Levitt and March, 1988; Dutton and Thomas, 1984). Unless firms make appropriate investments in retaining and leveraging lessons learned from their experience, its effects will increasingly dissipate. Research in the alliance literature has demonstrated a positive and linear relationship between alliance experience and firm-level indicators of financial (Anand and Khanna, 2000) and innovation performance (Shan et al., 1994). Only a slender body of work has explored diminishing returns to alliance experience. Hoang and Rothaermel (2005) find that, whilst alliance experience of small firms positively impacts the performance of their joint projects with large firms, the alliance experience of large firms makes no significant impact on their joint alliance project performance. Benefits stemming from alliance experience differ between small and large firms because they are positioned at different levels along the diminishing/concave alliance experience curve. In related research, Sampson (2005), commenting on patenting performance in the telecom equipment industry, identifies similarly diminishing returns to cumulative alliance experience, which she attributes to a diminishing contribution of distant experience, suggesting that distant alliance experience plays less important role than recent experience in efficient management of alliance portfolios (Shukla and Mital, 2018). As a consequence, we derive the following hypothesis: H2. The effect of cumulative alliance experience on likelihood of innovation exhibits diminishing marginal returns: as alliance experience accumulates, its contribution to the likelihood of firm innovation decreases. Deciphering, coding and measuring capabilities is notoriously difficult (Godfrey and Hill, 1995). As a result, the alliance literature has, in the main, relied on identifying alliance management practices (e.g. alliance functions) as a way of documenting alliance capabilities (Kale et al., 2002; Kale and Singh, 2007, 2009). An exception is found in the work of Rothaermel and Deeds (2006). They explore an inverted U-shaped relationship between cumulative alliance experience and new product development. They argue that, as the inflection point of the inverted U-shaped curve corresponds to the level of experience beyond which firms start experiencing inefficiencies in alliance management, it can reflect the level of their alliance capability. The notion that dedicated alliance management functions reflect alliance capabilities has being challenged in recent literature. First, anecdotal evidence suggests that firms establish dedicated alliance functions only after engaging in alliances over a protracted period, and after initiating a substantial number of such agreements (Kale and Singh, 2009). As a result such functions cannot reflect alliance capabilities, particularly when capabilities are at the initial stages of their development. Research indicates that firms - especially those motivated by perceived market signalling requirements - may establish dedicated alliance functions as a result of mimitisism or isomorphism (Heimeriks, 2010)[1]. Indeed, isolated mechanisms for institutionalising learning (e.g. dedicated alliance functions) from cumulative alliance experience may prove to be less effective compared to mechanisms designed to integrate and embed alliance knowledge throughout the organisation (Heimeriks et al., 2015). Despite some tentative and tangential work, explorations with respect to the underlying factors that lead firms to become aware of the need to leverage their alliance experience in developing a capability, and to invest efforts and resources in processes for deliberate learning from alliances are sparse in extant literature (see Wang and Rajagopalan, 2015). We contend that the first instances of frequent engagement in alliances can form such underlying factors and we turn to evolutionary theory and dynamic RBV to assist in elaborating our argument. Specifically, we suggest that frequent engagement in alliances may compel firms to found an alliance capability development lifecycle, as they both protract the period of engagement in alliances, and engage more substantially in this activity (Helfat and Peteraf, 2003). By engaging in alliances frequently, firms expand, augment and renew their stock of cumulative alliance experience. Frequent engagement in alliances induces firms to invest in developing higher order routines and practises to initiate purposeful and directed organisational learning. Developing processes for alliance management and practices for their further improvement requires managerial commitment, commitment that is highly demanding in terms of time and employee resources (Winter, 2000, 2003; Dosi et al., 2000). As a result, firms need to justify the allocation of resources to building dedicated routines for alliance management: frequent task reoccurrence can trigger and motivate firms to deploy resources in such a direction. It can also justify such investments to external stakeholders, building internal support and commitment to the purpose. Sparse empirical literature explores the role of frequent engagement in alliances in developing alliance capabilities (see Wang and Rajagopalan, 2015 for a review). There is a particular dearth of longitudinal studies in this field (Draulans et al., 2003; Rothaermel and Deeds, 2006). Research on CIS longitudinal Spanish data indicates that persistent and not sporadic collaboration exhibits a systematic positive relationship with firm innovativeness, which runs parallel to our argument (Belderbos et al., 2015). Moreover, firms that renew and expand their collaborations and alliances over time become more embedded in a network and experience higher growth over time (Powell et al., 2005). At the cross-sectional level, there is only limited research that links larger alliance portfolios to the development of alliance capability (as this is reflected in alliance management practices). Larger alliance portfolios are associated with the use of enhanced alliance management practices and processes, with the latter being linked to improved firm performance (Heimeriks et al., 2009). Our longitudinal approach, capturing the tendencies of firms to engage in alliances more frequently over time, complements such accounts which are based on the role of stock measures of alliance experience in firm performance (Kale et al., 2002; Heimeriks and Duysters, 2007; Kale and Singh, 2007). Therefore we propose that firms that engage in alliances frequently are more likely to innovate compared to those firms that engage in alliances irregularly or sporadically, or that retain a static engagement and a stagnating alliance portfolio over time: H3. Frequent engagement in alliances and the ensuing expanding alliance portfolios positively affect firm likelihood of firm innovation over time. 5.1 Sample We test our hypotheses on a panel data set observing the population of firms in the UK bio-pharma sector (110 firms in 2003) during the period 1991-2001. The empirical exploration in this paper relies fundamentally on econometric analysis. However, secondarily the paper employs a case analysis to shed further light on - and nuance - some of its arguments. To identify our sample, we used two editions of the UK biotechnology directory (Coombs and Alston, 2000, 2002). The directory lists all firms in the sector that undertake research using biotechnologies. It provides a more comprehensive account of the firms in the UK bio-pharma sector than that available via the use of SIC classifications, as biotechnologies are used by firms from a range of industries in the life sciences as broadly described (Walsh, 2004). To form our working sample, we used the "Who Owns Who" directory to identify all firms that were active in 2003. Out of that list, we identified 110 firms that publish accounts in FAME[2]. For the sample of 110 firms, we collected data on alliances from ReCap.com and BioScan, the two most popular sources of alliance data in empirical research on this sector (e.g. Deeds and Hill, 1996; Schilling, 2009). The databases report alliances established in the bio-pharmaceutical sector for innovation purposes, including R&D alliances, research alliances and alliances for technology licensing and product development[3]. Innovation is crucial to firm survival and prosperity in the bio-pharmaceutical sector and patents are perceived as an appropriate proxy for innovations due to the science-based nature of the sector (e.g. Pavitt, 1984). Moreover, patents are an appropriate indicator to capture firm-level returns from innovation alliances (Demirkan, 2018; Shan et al., 1994; Sampson, 2005). We collected data on patents granted to the 110 firms in our sample between 1991 and 2001 by the UKPTO through the publicly available database Esp@cenet. Via deployment of Boolean searches, it was possible to assemble a comprehensive list by matching both the name and the address of firms in our sample with those of the patent assignees (Arora et al., 2011). For firms in our sample that belong to MNEs with non-UK-based HQs, we examined information on inventor location, and allocated any patents assigned to the MNE HQ with UK-based inventors to the UK-based subsidiary in our sample. Since our research focusses on UK-based firms, the UKPTO is perceived to be an appropriate source for the collection of patent data. Filing patent applications at national patent offices is less costly (in terms of application and renewal fees) and less time consuming when compared with international patent applications. As international applications also involve costs of unfamiliar IP regulations and legal frameworks, many firms - at least those with constrained resources - are more likely to patent at national patent offices than internationally (Archambault, 2002). Finally, information on firm accounts was obtained from FAME, and information on investments in R&D was gathered from Thomson's Analytics and the UK DTI's R&D Scoreboard. 5.2 Methods Patents take positive integer values, making count dependent variable models, such as the Poisson and Negative Binomial, appropriate (Greene, 2003). Instead of count dependent variable models, we employ discrete dependent variable models as they permit the capture of factors that alter firm abilities enabling them to perform over a threshold level: successful filing of a patent (Long, 1997). This is better aligned to the core aim of this work, which is to explore whether frequent alliance engagement is linked to organisational learning in alliance management (and alliance capabilities), reflected in firms exhibiting over threshold performance (Hoang and Rothaermel, 2005). Due to the longitudinal dimension of panels, we can explore factors that affect firms' changing abilities to innovate over time (fixed effects specification), as well as factors that can explain cross-firm variation in innovation abilities (random effects specification). The former, dynamic consideration of panels is of particular interest in this paper, as explained above. The fixed effects specification has a further appeal as its estimates are not based on the assumption of strict exogeneity of the independent variables (Greene, 2003; Baltagi, 1995), which is desirable, as decisions to form alliances may not be independent of past innovation performance (Colombo and Garrone, 1996; Gkypali et al., 2017). To explore further the nuanced factors underlying the development of alliance capabilities, we report on the findings of an in-depth case analysis that explores the link between frequency of engagement in alliances, and the development of practices and processes to manage alliances, an indication of alliance capabilities (Schreiner et al., 2009; Kale and Singh, 2007); this is a secondary analysis providing an illuminating complement. The case focusses on a UK bio-pharmaceutical firm that intensified its engagement in alliances particularly after the early 1990s when it started to invest in biotechnology R&D. Interviews were conducted in 2005 as the firm was forming a dedicated alliance function and entailed detailed reflection on the factors that underpin the creation of processes established to improve efficiency in alliance management. The processes identified were matched with those identified in existing literature on alliance capabilities (Kale et al., 2002; Kale and Singh, 2007, 2009; Heimeriks and Duysters, 2007). The case is based on interviews with directors of alliances and intellectual property, with a long working experience with the case firm and focussed on alliances that involve cooperation in research and that have led to patents. All materials and associated analyses were reviewed and approved by the interviewees. 5.3 Variables Patents (dependent variable) To test our hypotheses we create a dummy variable that is equal to 1 for each firm that is granted a patent at year t (and equal to zero otherwise). We use a patent dummy variable as we aim at capturing the impact of alliance experience in enabling firms to perform over a threshold, i.e. to innovate. To track the firm's innovation activities over time, we use patent filing dates. Filing dates have certain advantages over publication dates: first, they better reflect the originating time of innovations; and, second, they are not influenced by regulatory changes or fluctuations in resource availability in patent offices over time, as these can delay publication dates (e.g. Jaffe, 1986). Patents are widely used to capture firm returns from innovation alliances (Sampson, 2005, 2007; Shan et al., 1994; Demirkan, 2018). Alliance experience Research has explored the role of alliance experience in improving financial returns of firms from alliances (Kale et al., 2002), alliance project success (Hoang and Rothaermel, 2005), returns to firm innovation performance (Sampson, 2005, 2007; Belderbos et al., 2015; Di Guardo and Harrigan, 2016) and number of products under development (Deeds and Hill, 1996; Rothaermel and Deeds, 2006). Studies have operationalised alliance experience as a stock variable reflected either by the cumulative number of alliances formed throughout a firm's history (Kale and Singh, 2007; Hoang and Rothaermel, 2005; Kale et al., 2002; Rothaermel and Deeds, 2006), or by the cumulative number of years of experience in engaging in alliances (Rothaermel and Deeds, 2006). Only a handful of these studies consider diminishing (Sampson, 2005) or non-linear (Rothaermel and Deeds, 2006) relationships between alliance experience and firm returns from alliances and most studies base their estimates on cross-sectional data (an exception is Kale et al., 2002). Following existing literature, we operationalise alliance experience at each year t as the cumulative number of alliances throughout a firm's history. To aid interpretation of results, and following recent contributions (Lavie et al., 2011) we use one year lagged values of the above variable to more accurately capture the effect of cumulative alliance experience. Our variable is left-censored to 1991, the start year for our sampling. This is experienced elsewhere in alliance literature (e.g. Kale and Singh, 2007). Also, because in our study 87 per cent of total alliances from 1979 and 2001 fall within the 1991-2001 period, left censoring is not introducing considerable bias. As is customary in the literature (Sampson, 2005), we use the natural logarithmic transformation of cumulative alliance experience to test for diminishing returns (H2). As a robustness check, we test for a concave relationship (inverted U-shaped) by using both the linear and quadratic values of the cumulative number of alliances: again, this approach follows relevant literature (e.g. Deeds and Hill, 1996; Rothaermel and Deeds, 2006)[4]. Frequent engagement in alliances and expanding alliance portfolios To explore the effects of frequent engagement in alliances on firm abilities to innovate, we measure the total number of on-going alliances that a firm manages simultaneously at any point in time. As alliances last for several years, this requires information on alliance duration which is not available via the alliance data sources consulted in this study (RECap.com and BioScan). As lack of information on alliance duration has also been encountered in other work, we follow established tradition and rely on estimates of the average duration of alliances, i.e., five years (Kogut, 1988; Di Guardo and Harrigan, 2016). Therefore, we operationalise frequent engagement in alliances by calculating the number of alliances a firm manages at any given point in time by considering that alliances last on average for five years. Increases in this variable reflect the tendency of firms to engage more frequently in alliances over time, and on average, to initiate more alliances than they terminate, managing potentially an expanding portfolio of on-going alliances. As the use of a blanket measure for alliance duration might be considered simplistic, we cross-validated this information with alliance managers in the sector, and it was considered an acceptable approximation for the duration of the majority of alliances. We also performed robustness checks by using different estimates for alliance duration as detailed in our analysis. Control variables For parsimonious empirical models, we use the most appropriate control variables that have been employed in the literature on alliances and innovation. First, we account for the effects of well-established indicators of firm innovation, such as investments in R&D and firm size (Griliches, 1990) as they can influence firm abilities to innovate. We used a measure of R&D intensity that captures the amount of R&D expenditure per employee. Firm size is captured by annual turnover and we include a natural logarithm to overcome the possible problem of skewed distributions. Second, and following relevant research (Zollo et al., 2002), we account for differences in the content of alliances, as it influences innovation opportunities in alliances and affects firm performance (e.g. Anand and Khanna, 2000). Therefore, we include a dummy that takes the value of one when the focal firm establishes an alliance that involves R&D, and the value of zero for each year that a focal firm establishes alliances that do not involve R&D. Table I presents descriptive statistics and bivariate correlations. The panel is unbalanced, as a result of either missing information or the inclusion of less-mature dedicated biotechnology firms (i.e. a third of firms in the sample were established after 1995). The unbalanced data set includes firms observed for a maximum period of ten years and others for as little as four years. On average, firms are observed for approximately seven years (the range is from 6.6 to 7.4 years across the three estimated models) and form 8.28 alliances between 1991 and 2001. We estimated both the fixed and random effects specifications (FE and RE, respectively) of the Logit model for panel data for all three models (Greene, 2003; Baltagi, 1995)[5]. Results are obtained in STATA(c) 11. The r parameter is significant across all three models, indicating significant cross-firm heterogeneity. As Hausman's (1978) specification test has limited discerning power in discrete dependent variable models, we assess the relative merits of the two specifications on a conceptual basis and for reasons discussed in the methods section we present Logit FE estimates. Moreover, the residuals of the FE specification are corrected for potential heteroscedasticity (White, 1980; Baltagi, 1995). Based on model kh2 statistics in Table II, all three models are significant at 5 per cent levels and above. This suggests that our variables have a significant interpretative power with respect to the likelihood of firm innovation. We find support for all hypotheses. In Model 1, the cumulative number of alliances positively affects firm's likelihood of innovation over time, but levels of significance are low (10 per cent). Our results are consistent with studies showing a positive relationship between cumulative alliance experience and firm innovation, but at low significance levels (Hagedoorn and Schakenraad, 1994; Shan et al., 1994). The FE estimates of Model 2 provide support for diminishing returns to firm innovation from cumulative alliance experience, with the corresponding variable being positive and highly significant. The results of Model 2 are consistent with findings testing similar hypotheses in cross-sectional settings in the US bio-pharmaceutical and the telecom equipment industries (Hoang and Rothaermel, 2005; Sampson, 2005). Our results are robust to alternative operationalisations, as reflected by exploring concave relationships (inverted U-shaped), via use of both linear and quadratic values of cumulative alliance experience, with estimates (both FE and RE) confirming non-linear diminishing returns to likelihood of firm innovation. To unpack the extent of diminishing returns we estimate the inflection point based on estimates of the curvilinear relationship; this is equal to 127.515 points of cumulative experience, which corresponds to firms managing 18 alliances annually (average length of time per firm in the panel is seven years). Our results imply that at such a level of annual alliance activity, firms will accrue no additional benefits from gaining more experience in managing alliances. It should be noted that in our sample the average number of new alliances formed every year is 8.28, which suggests that the majority of firms in our sample are below the optimal level of alliance engagement. The concluding section discusses the implications of this. Model 3 provides support for our third hypothesis. The coefficient of the core independent variable is positive and significant (it marginally fails to pass the highest 1 per cent level of significance). Model 3 suggests that firms that establish alliances frequently and that continuously manage an expanding portfolio can enjoy improved likelihood of innovation. This provides fresh insights as, to date, there has been a scarcity of longitudinal studies in the literature (see Draulans et al., 2003; Rothaermel and Deeds, 2006). As a test of robustness, we estimated Model 3 by using four and six years as the average duration for alliances. The results remained robust to different operationalisations. Among the control variables, firm size is positively and significantly associated with firm's likelihood of innovation in Models 1 and 3. Investments in R&D per number of employees are negatively associated with likelihood of firm innovation but only in Model 2. Alliance content is insignificant across all three models. Finally, all three model estimates are robust to different measures of the control variables. For example, all models lead to the same outcomes in terms of significance when controls are collapsed to dichotomous variables indicating above and below average firm size and intensity of investments in R&D per number of employees. A further robustness check is undertaken to examine whether our results are sensitive to the statistical assumptions of the distribution followed by the error term in the Logit and Probit specifications. The Probit model only provides RE estimates for panel data (STATA(c) 11) and these further confirm our results. Moreover, we estimate our model specifications with the Negative Binomial model to explore differences between our dependent variable and a count variable capturing innovation performance with the number of patents. Both FE and RE estimates confirm our hypotheses. The results suggest that there are diminishing returns to cumulative alliance experience. This can signal that ageing experience contributes less to current outcomes and that recent experience may have a higher contribution (Shukla and Mital, 2018; Sampson, 2005). Following Sampson (2005), we explore the contributions of recent and past alliance experience. We develop a set of variables capturing alliance experience between 1 and 6 years prior to our year of observation. So, for example, in 2001 alliance experience of one year corresponds to the number of alliances initiated in 2000 and alliance experience of two years corresponds to those initiated in 1999 and so on. None of these variables appear to be significant, with the exception of alliance experience of four years prior to the year of observation, which appears with a negative and significant sign. Therefore, the diminishing returns to cumulative alliance experience identified in our paper cannot be attributed to decreasing contributions of distant experience. Our results most likely reflect that firms, by just forming more alliances and learning from experience how to improve alliance management and coordination, cannot experience improved efficiency ad infinitum. With regard to H3 we undertake further estimations to isolate the impact of frequent engagement in alliances from that of portfolio size. We break down the variable testing H3 into two components. First, capturing changes in the frequency of forming alliances over time, we construct a variable reflecting a decreasing, static and increasing trend in alliance portfolios (the variable takes the values of: -1, 0 and 1, respectively, and refers to comparisons between the current and the previous year). Second, we include a variable to capture the number of new alliances formed within every year (one year lagged values are used in the models). Our results show that the trend variable has a negative sign while the annual figures of newly formed alliances appear with a positive and significant sign (both for the Logit and the Negative Binomial models and the FE and RE specifications). It is worth noting that when the values of the trend variable are lagged by one year, their impact becomes insignificant (i.e. they retain a negative sign). This means that only the contemporaneous impact of alliance frequency on innovation is negative, implying that when firms expand their alliance portfolio there is an immediate adverse effect. This may result from immediate strains on their resources and cognitive abilities, which however, cease to have an adverse effect within a year's time. The results remain the same when the trend variable is substituted with a dummy variable capturing only increases in the frequency of forming alliances between current values and those of the previous year. Finally, we isolate partner-specific alliance experience from general (i.e. non-partner specific) experience to further explore the importance of alliance management learning within partner dyads and in alliance portfolios. None of these variables has a significant effect, suggesting that the separation lacks interpretative power. The paper contributes to the literature on the role of alliances in innovation and alliance capabilities by using a longitudinal analysis which enables capture of the impact of organisational learning in managing and coordinating alliances on innovation. As it is difficult to capture learning, the paper uses a longitudinal approach to capture dynamic changes within firms and it observes learning through its impact on outcome variables such as innovation. The paper adds to a slender body of work exploring antecedents to firm-level alliance capabilities and their impact on enhancing firm's abilities to innovate (for a review see Wang and Rajagopalan, 2015). First, we find that the impact of learning from gaining greater alliance experience in coordinating alliances and portfolios on firm innovation is positive but of weak significance. This is an important finding, as alliances in the bio-pharmaceutical sector can be characterised by power asymmetries and unbalanced resource dependence between partners, which are likely to create conflicts affecting the creation and appropriation of alliance value (Diestre and Rajagopalan, 2012; Caner and Tyler, 2013). The period of our study forms the first window of exploitation of biotechnologies via alliances which potentially creates disincentives to collaborating firms to behave opportunistically, as joining forces during this close to pre-competitive stage could be more beneficial than not (Powell et al., 2005). Second, our analysis proceeds by exploring the factors that affect learning in managing and coordinating alliances in this sector. The first arm to this exploration sheds light on the role of cumulative-historic experience in enhancing innovation performance, while the second arm focusses on the role of such experience as an antecedent to the creation of capabilities to manage interactions with alliance partners and to create synergies and minimise redundancies in alliance portfolios. In that endeavour, we first confirm diminishing returns to the benefits that can be accrued by gaining alliance experience for more effective management of alliances and capturing innovation returns. This supports a thin body of research which suggests that learning from gaining experience cannot confer benefits ad infinitum (e.g. Rothaermel and Deeds, 2006; Sampson, 2005). Our in-depth exploration shows that this cannot be attributed to diminishing contributions of ageing experience as identified in other work (Sampson, 2005; Shukla and Mital, 2018). It could be attributed to failure of firms to capture learning in a systematic way, or to costs associated with identifying more partners, establishing more agreements and coordinating increased diversity (White and Lui, 2005; Rothaermel and Deeds, 2006). Our last set of findings show that engaging in alliances more frequently over time, with firms managing an expanding portfolio, by forming more alliances than they terminate, enhances firms' likelihood of innovation. Our longitudinal approach is consistent with other work showing that persistence in collaboration has a positive impact on firm innovation as opposed to sporadic collaboration (Belderbos et al., 2015). Longitudinal explorations are currently almost absent in alliance literature, and their potential contribution has been particularly welcomed (Draulans et al., 2003; Rothaermel and Deeds, 2006), specifically as they might shed further light on the type of changes that can occur within firms, in particular, with respect to the dynamics of learning on alliance management and how it enhances innovation potential. Contributions within evolutionary theory (Winter, 2000, 2003) argue that frequent engagement in organisational tasks is linked to capability development, while the dynamic resource-based view (Helfat and Peteraf, 2003) argues that capabilities develop through several stages: experience accumulation compels firms to recognise the need to establish a capability, and this initiates a process of systematic learning and alliance management. Our results clearly pertain to this literature and specifically on how alliance experience forms an antecedent to alliance and alliance portfolio capabilities (Wang and Rajagopalan, 2015). To explore, in a situated and more granular fashion, whether frequent engagement in alliances and a dynamically expanding alliance portfolio contribute to the development of alliance capabilities, we conducted a detailed case analysis as a complement to our econometric analysis. The case analysis explores the link between an increasing engagement in alliances and the development of novel or changed practices and processes for alliance management, i.e., those factors that normally sit within the "black-box" of alliance capability studies (Anand and Khanna, 2000; Heimeriks and Duysters, 2007; Kale and Singh, 2007). The case analysis clearly reveals such a relationship. First, it shows that alliance management practices and routines are established after alliances become an integral part of firm R&D and innovation strategy and firms intensify and commit to alliance activity. As the Alliance Director interviewed for our study stated: "[...] we have standardised approaches for contractual compliance [with respect to alliances] [...] it's inconceivable to me that one wouldn't have, and yet five years ago we didn't [...] You've got one deal and you do it however you do it which largely depends on who you employ and who it's [the alliance] with [...] I am not sure how many deals you need to have before you get yourself a standardised process but it's probably substantially more than ten or twenty [...][7]". Second, several of the practices identified in the extant literature that are used to institutionalise and diffuse alliance management learning (namely, employment of alliance professionals, use of alliance metrics, creation of databases to solicit potential alliance partners and establishment of processes to monitor alliance performance) are deployed by the case firm (Heimeriks and Duysters, 2007; Kale and Singh, 2007). Ten years after intensifying alliance engagement, it developed a dedicated alliance management function to coordinate alliance activity and engage in monitoring and improving existing alliance management practices. The case analysis illustrates the links between frequent engagement in alliances and the ensuing expanding the alliance portfolio, with the motivation to invest resources in developing firm-specific capabilities for alliance management. It also reflects the time required to recognise that alliance management can be improved via the development of specific and specialist processes for the governance of cross-partner interaction (a finding that resonates strongly with existing literature, Kale and Singh, 2009). In addition, the case reflects the requirement for firms to continuously review and develop their capabilities in order to confront new challenges and intensified competition. Indeed, the case analysis, combined with the findings of our econometric exploration, suggest that renewing alliance experience and committing to forming more alliances over time are crucial in enabling firms to create and capture value in alliances. Our research contributes to the literature on the antecedents of alliance and alliance portfolio capabilities and on the conditions that shape superior firm outcomes from alliances, such as innovation. It suggests a need to delve into the antecedents to alliance capabilities, a thin body of research and to identify nascent factors that provide potential foundations for alliance capability development (Helfat and Peteraf, 2003), shifting attention away from the role played by alliance management practices (such as dedicated alliance functions) that have dominated existing research. This is particularly important as firms may establish such practices during the advanced stages of the alliance capability development process (Kale and Singh, 2009), and as such they may not appropriately reflect the foundational stages of alliance capability development. Here, we echo calls to delve deeper in understanding alliance management capabilities and the antecedents to alliance portfolio capabilities (Wang and Rajagopalan, 2015), especially in the context of alliances involving a higher learning potential (Heimeriks, 2010). Moreover, recent research shows that codification of alliance learning and systematic approaches to alliance management contribute to efficient partner selection and alliance termination, but may restrict flexibility and adaptability which are important for efficient management during the course of the alliance (Heimeriks et al., 2015; Wang and Rajagopalan, 2015). Our research has several implications for managerial practice and policy. First, we find that there are diminishing returns to benefits in terms of enhanced innovation accruing from gaining greater alliance experience. Specifically firms reach a plateau at 18 alliances in annual terms, suggesting that on average this is the optimal level of alliance engagement for enhancing firm likelihood of innovation. This has implications for managing alliance portfolio size for innovation efficiency (e.g. Hoffmann, 2007; Demirkan, 2018). On average firms in our sample are below the optimal level of alliance engagement (forming eight alliances annually). This suggests that innovation likelihood can be marginally enhanced by expanding alliance engagement. This is in line with findings of other research (Rothaermel and Deeds, 2006; Sampson, 2005; Hewitt-Dundas and Roper, 2018) which in the context of micro-businesses, for example, shows that sub-optimal engagement can be attributed to reluctance of firms to cooperate due to fear of trustworthiness and lack of knowledge on benefits of collaboration and partner capabilities (Hewitt-Dundas and Roper, 2018). This emphasises our findings with respect to the need to facilitate a culture of systematic management of alliances. It also supports research suggesting the need for policy interventions, for example in the form of intermediaries, which can support and encourage firms to establish collaborations and manage openness (Howells, 2006). This is not only due to the benefits of collaboration to firm innovation but also due to the positive externalities from openness, suggesting broader economic and societal benefits (Roper et al., 2013). At the firm level, our results suggest that it is important for firms to replenish their alliance portfolios and to gradually form more alliances than they terminate over time if they wish to enhance their innovation likelihood. In the bio-pharmaceuticals sector where there is high rate of alliance formation, it is important for firm competitiveness and future growth that they renew and expand their networks (Powell et al., 2005). This is linked to firms becoming more aware of the need to make alliance management learning more systematic. Firms should use such systematic practices in a discretionary manner and should also consider altering and upgrading such practices to respond to the challenges and demands of new alliances. Although our results are specific in terms of sector and period, they derive from the examination of a paradigmatic period in the development of new technologies, and address the implications of changing innovation processes in an established industry. As such, they provide useful insights that assist understanding of trajectories and dynamics in industries with similarities to biotech, for example, nanotechnology (e.g. Rothaermel and Thursby, 2007). Our aspiration is that our research will provide some re-focussing and clarity in an established field of study. We acknowledge that analysis of further sectors and periods will nuance our understanding of both the role of gaining experience in improving firm alliance management and in developing alliance capabilities, and hope that the ideas and methods explored above will assist in this on-going quest.
|
The paper shows that cumulative alliance experience has a marginally diminishing contribution to likelihood of firm innovation over time, while frequent engagement in alliances and the ensuing expansion of alliance portfolios enhance firm innovation. The exploratory case analysis demonstrates a link between frequent engagement in alliances and the development of processes for alliance management that could collectively reflect alliance capabilities.
|
[SECTION: Value] As the popularity of strategic alliances is increasing and such alliances are becoming an integral component in business development, attention in the research community has moved towards an exploration of their role in firm performance and innovation. Formal and informal interactions with external actors have long being argued to play a fundamental role in firm innovation (Von Hippel, 1988; Frankort et al., 2012; Rice et al., 2012; Colombo et al., 2011; Demirkan, 2018). However, it is accepted that alliances carry coordination costs and risks of misappropriation, and these frequently diminish chances of success or preclude full acquisition of anticipated benefits (e.g. De Man and Duysters, 2005; Gkypali et al., 2017; Faems et al., 2010). A substantial body of literature finds a positive relationship between the extent of alliances and firm innovation performance, whilst other work identifies diminishing and even negative returns (Hoang and Rothaermel, 2005; Sampson, 2005; Laursen and Salter, 2006; Rothaermel and Deeds, 2006). As a result, a growing strand in the literature has examined the factors enabling firms to generate and capture value from alliances. One such factor is alliance experience accumulation: as firms develop greater experience in managing alliances they become better at coordinating cross-organisational tasks and knowledge flows (e.g. Sampson, 2005). Another factor that can improve performance in alliances is developing formal and codified processes and routines for alliance management (e.g. the establishment of dedicated alliance functions), which are argued to capture, or to be reflective of, firm-specific alliance management capabilities (Kale et al., 2002; Kale and Singh, 2009, 2007; Heimeriks and Duysters, 2007; Sampson, 2005; Schreiner et al., 2009; Di Guardo and Harrigan, 2016; Shukla and Mital, 2018). Whilst these contributions offer useful insights, much the greater part of current theorising has been constructed via the use of cross-sectional data. Longitudinal explorations are scarce, thus we lack a nuanced understanding of the type of changes that occur within firms over time with respect to enhanced innovation potential from engaging in alliances. The contribution of this paper is that, on the basis of a longitudinal approach, it examines how frequent engagement in alliances and expanding alliance portfolios (a compilation of contemporaneous alliances) impact on the likelihood of firm innovation; this offers a longitudinal approach that is lacking in the current literature (Draulans et al., 2003; Rothaermel and Deeds, 2006) with few notable exceptions (Di Guardo and Harrigan, 2016; Belderbos et al., 2015). A dynamic approach is appealing as it captures the impact on innovation of firm learning with respect to management of alliances. Such learning is evident in the use of past alliance experience to support conflict avoidance or management, and the effective coordination of knowledge sharing and communication with partners. Improved management can affect value capture in alliances especially as firms manage expanding alliance portfolios due to both potential of cross-partner asymmetries and knowledge redundancies, which are common in the bio-pharmaceuticals sector (Diestre and Rajagopalan, 2012; Caner and Tyler, 2013). The paper uses longitudinal econometric panel data analysis to explore its hypotheses. Secondary to this, we deploy a case analysis to capture the nuanced changes due to learning, shedding more light on the role of frequent engagement in alliances in compelling firms to direct their attention and resources in developing practices and routines for alliance management. This unpacks the antecedents to alliance capabilities and travels beyond current understandings in relation to experience accumulation. As highlighted in a recent review (Wang and Rajagopalan, 2015), research on the antecedents to alliance and alliance portfolio capabilities is rather thin. This paper contributes to exploring this link further. Our arguments are explored in the context of a unique, bespoke and specially assembled data set, comprising the population of firms in the UK bio-pharmaceutical sector (110 firms "alive" in 2003) and active during the period 1991-2001. The UK bio-pharmaceutical sector presents an ideal setting for investigation. The UK accounts for a significant share of the worldwide bio-pharmaceutical sector (Mckinsey, 2014) and enjoys a revealed technological advantage (OECD, 2017) and comparative advantage (OECD, 2008) in this high-tech sector. Moreover, the UK is a prime example of the Liberal Market Economy variant detailed in the Varieties of Capitalism literature (Hall and Soskice, 2001; Hancke, 2009), where the socio-economic institutional configuration is associated with international competitiveness of industries that are characterised by radical innovations (such as the pharma and biotech industries). The period 1991 and 2001 is perceived to be particularly apposite for our analysis as it exhibits several critical features: a first instance and general upsurge in alliances (not only in the UK bio-pharmaceutical sector, but also in other sectors and territories, Kang and Sakai, 2000); the emergence of alliances as an integral component in firm R&D strategies in the bio-pharmaceutical industry (Hagedoorn, 2002; Demirkan, 2018; Powell et al., 2005) but also more generally (Rice et al., 2012); and the presence of significant variation - visible in terms of identifiable peaks and troughs in alliance activity - especially towards the end of the period (Hagedoorn, 2002). Moreover, this marks a paradigmatic era of new technology evolution, with new entrants and established firms developing capabilities in biotechnologies through alliance activity (Hopkins et al., 2007): thus it provides a historical backdrop that aids understanding of development of similar technologies, for example, nanotechnologies (Rothaermel and Thursby, 2007). The remainder of this paper is structured as follows. Section 2 sets-out the theoretical background to the study and reviews the literature on alliance experience and alliance/alliance portfolio capabilities. Sections 3 and 4 develop our hypotheses. Section 5 discusses our sample, data sources and methodological considerations and details the variables for our longitudinal econometric analysis. Section 6 is dedicated to estimation and results, and Section 7 provides results from post-estimation robustness checks. Section 8 discusses our findings together with the case study analysis and Section 9 highlights the implications of the study for management theory and practice. To explain heterogeneity with respect to firm's abilities to benefit from alliances, the alliance literature resonates strongly with knowledge and capability-based theories of the firm (e.g. Kogut and Zander, 1992; Helfat and Peteraf, 2003). Here, we adopt a perspective that is informed by both evolutionary theory of the firm (Nelson and Winter, 1982) and dynamic approaches to the resource-based view (RBV) (Helfat et al., 2007; Helfat and Peteraf, 2003). These approaches take a dynamic view of organisational development, emphasising the role of experience and knowledge accumulation in supporting improved management and coordination of organisational tasks and activities. Given their dynamic and longitudinal orientation, they are closely aligned with our own analytical approach, informing our exploration of the roles of alliance experience accumulation and frequent engagement in alliances in organisational learning and enhanced innovation. Evolutionary theory argues that by repetitively engaging in organisational tasks, or by gaining experience, organisations benefit from efficiency improvements as a result of "learning-by-doing". As firms follow different paths in accumulating and making sense of experience, they become heterogeneous in their abilities to carry out similar organisational activities. Experience assists firms in assessing which processes and practices are likely to perform better or to confer improved results, by judging on similar situations in the past. Organisations gradually create routines to efficiently coordinate organisational activities and develop capabilities to achieve outcomes important for competitive advantage (Dosi et al., 2000; Winter, 2000, 2003; Schilke and Goerzen, 2010; Nelson and Winter, 1982). Experience accumulation is an antecedent to developing capabilities; however, there can be a deliberate and conscious process of organisational learning as firms make investments in time, resources, training and knowledge to further improve their capabilities (Zollo and Winter, 2002), or to improve them through reflecting on newly acquired experience (Helfat and Peteraf, 2003). Dynamic approaches to RBV are illuminating on the process of capability development by suggesting that there is an evolutionary path to their creation whereby capabilities emerge and improve through several developmental stages (Helfat and Peteraf, 2003). First, there is a "capability founding" stage wherein organisations realise the need to develop a capability to meet desirable objectives and to orchestrate joint organisational action towards meeting these. Second, a "capability development" stage: here improvements to capabilities arise from a range of factors including further experience accumulation, investments in capital, organisational processes and learning mechanisms. Experience accumulation is crucial to both "founding" and "development" stages of a capability development. A third stage, "capability maturity" is also identified: in this phase capabilities may be replicated, retrenched, renewed, redeployed or re-combined with other capabilities. Even though alliances differ compared to other organisational tasks, in that firms collaborate with different partners in agreements with varying content, purpose and governance structure, gaining experience in managing alliances can lead to superior coordination, value creation and capture in alliances (Kale and Singh, 2007; Heimeriks and Duysters, 2007). Alliance experience is found to enhance the performance of alliances between recurrent partners, as firms use their experience and routines developed during their first interactions in managing subsequent collaborations within the dyad (Zollo et al., 2002). Moreover, firms mobilise experience and lessons gained across alliances with different partners (Anand and Khanna, 2000), with partner-specific alliance experience making a greater contribution to performance compared to general (i.e. non-partner specific) experience (Gulati et al., 2009). Alliance experience (a firm's cumulative number of alliances) is an antecedent to both superior coordination of alliance tasks (alliance capabilities), and to the efficient management of synergies and redundancies across alliances in alliance portfolios (alliance portfolio capabilities) (for a review see Wang and Rajagopalan, 2015). Kale and Singh (2007) were the first to detail how firms proactively learn from their alliance experience by externalising, codifying, diffusing and internalising individual and group-level alliance know-how. A stream within the alliance literature explores the link between cumulative alliance experience, the development of dedicated alliance management functions and other processes and practices connected with upgrading alliance management routines (Kale et al., 2002; Heimeriks and Duysters, 2007; Kale and Singh, 2007). Alliance functions reflect alliance capabilities and can lead to superior outcomes from alliances. Alliance experience indirectly affects firm performance via such capabilities, indicating that dedicated alliance functions embody and absorb the impact of alliance experience, suggesting that firms leverage all the learning gained from experience in their alliance management functions (Kale et al., 2002). Other research suggests that alliance experience directly influences both financial (Kale et al., 2002; Anand and Khanna, 2000) and non-financial outcomes such as innovation performance (Sampson, 2005), and that this is in addition to any impact conferred by dedicated alliance functions, suggesting a complementary relationship between the two (Heimeriks and Duysters, 2007). Indeed, systematic alliance practices are improved through training, mentoring programmes and external consulting. Such learning processes aim at introducing new alliance management practices and/or at improving existing ones in response to dissatisfaction with current performance (Zollo and Winter, 2002). The following section revisits existing but still contested hypotheses on the expected impact of alliance experience on firm innovation performance. We also develop our hypothesis on the impact of alliance frequency, which from a longitudinal perspective can be better positioned to capture nuancing and conditioning factors that reflect the impact of organisational learning from alliance experience on innovation. Firms with greater experience can draw from a greater pool of situations about what has worked in practice, when making decisions and inferences with respect to the performance of organisational practices (Levitt and March, 1988; Argote et al., 1990). Alliance experience (cumulative number of alliances) improves firms' abilities: to manage and coordinate alliances, to improve coordination of inter-organisational relationships and joint tasks, to form efficient arrangements for knowledge sharing, to deal effectively with unforeseen contingencies and to identify ways to overcome and resolve inter-partner conflict (Anand and Khanna, 2000; Sampson, 2005; Belderbos et al., 2015; Rothaermel and Deeds, 2006). Due to the link between alliance experience and organisational learning, experience is seen as a fundamental antecedent to both alliance and alliance portfolio capabilities (Wang and Rajagopalan, 2015; Kale and Singh, 2009; Shukla and Mital, 2018). Literature suggests that firms may not be in a position to benefit from learning from experience and superior coordination of alliances, when facing power asymmetries and resource dependence in alliances. Conflict is more frequent in such alliances which affects value creation and capture, especially for the weak partner as they are at a comparative disadvantage (Diestre and Rajagopalan, 2012). Power asymmetries are likely in the bio-pharmaceutical sector, as large pharmaceutical firms may be collaborating with small-dedicated biotech firms and due to their longer commitment to alliances and historic investments in downstream capabilities may be at a comparative advantage in deriving value from alliances (Caner and Tyler, 2013). We argue that by gaining greater alliance experience, firms gradually establish more fruitful conditions for effective knowledge sharing with their partners, and generate a basis for circumventing or resolving alliance conflicts. Because conflicts are more likely to occur in the bio-pharmaceuticals sector we expect that the higher the levels of experience the more likely the firms are to capture value from alliances and enhance their likelihood of innovation: H1. Alliance experience accumulation positively affects a firm's likelihood of innovation. A body of work suggests that efficiency improvements from experience accumulation make a gradually diminishing contribution to firm performance. There are two likely explanations for this. First, as firms form more alliances and accumulate greater experience, the contribution of any additional alliance to accumulated learning becomes increasingly reduced. This has been identified irrespective of type of partner and across upstream, downstream and horizontal alliances (Rothaermel and Deeds, 2006). Firms may experience diminishing returns when managing an increasing number of alliances due to the increased costs associated with identifying suitable partners, establishing and maintaining contracts, difficulties in absorbing knowledge from diverse sources, conflicts due to knowledge asymmetries and increased possibilities of knowledge redundancies (Diestre and Rajagopalan, 2012; Rothaermel and Deeds, 2006; White and Lui, 2005). A second likely explanation is that the value of lessons learned from cumulative experience depreciates over time as knowledge becomes increasingly obsolete (e.g. Argote et al., 1990). Past knowledge may depreciate as task requirements change over time. Further, the value of past lessons may dissipate gradually (or suddenly) as a result of employee turnover or a failure to store intelligence in organisational routines, or in a way that permits its re-application (Levitt and March, 1988; Dutton and Thomas, 1984). Unless firms make appropriate investments in retaining and leveraging lessons learned from their experience, its effects will increasingly dissipate. Research in the alliance literature has demonstrated a positive and linear relationship between alliance experience and firm-level indicators of financial (Anand and Khanna, 2000) and innovation performance (Shan et al., 1994). Only a slender body of work has explored diminishing returns to alliance experience. Hoang and Rothaermel (2005) find that, whilst alliance experience of small firms positively impacts the performance of their joint projects with large firms, the alliance experience of large firms makes no significant impact on their joint alliance project performance. Benefits stemming from alliance experience differ between small and large firms because they are positioned at different levels along the diminishing/concave alliance experience curve. In related research, Sampson (2005), commenting on patenting performance in the telecom equipment industry, identifies similarly diminishing returns to cumulative alliance experience, which she attributes to a diminishing contribution of distant experience, suggesting that distant alliance experience plays less important role than recent experience in efficient management of alliance portfolios (Shukla and Mital, 2018). As a consequence, we derive the following hypothesis: H2. The effect of cumulative alliance experience on likelihood of innovation exhibits diminishing marginal returns: as alliance experience accumulates, its contribution to the likelihood of firm innovation decreases. Deciphering, coding and measuring capabilities is notoriously difficult (Godfrey and Hill, 1995). As a result, the alliance literature has, in the main, relied on identifying alliance management practices (e.g. alliance functions) as a way of documenting alliance capabilities (Kale et al., 2002; Kale and Singh, 2007, 2009). An exception is found in the work of Rothaermel and Deeds (2006). They explore an inverted U-shaped relationship between cumulative alliance experience and new product development. They argue that, as the inflection point of the inverted U-shaped curve corresponds to the level of experience beyond which firms start experiencing inefficiencies in alliance management, it can reflect the level of their alliance capability. The notion that dedicated alliance management functions reflect alliance capabilities has being challenged in recent literature. First, anecdotal evidence suggests that firms establish dedicated alliance functions only after engaging in alliances over a protracted period, and after initiating a substantial number of such agreements (Kale and Singh, 2009). As a result such functions cannot reflect alliance capabilities, particularly when capabilities are at the initial stages of their development. Research indicates that firms - especially those motivated by perceived market signalling requirements - may establish dedicated alliance functions as a result of mimitisism or isomorphism (Heimeriks, 2010)[1]. Indeed, isolated mechanisms for institutionalising learning (e.g. dedicated alliance functions) from cumulative alliance experience may prove to be less effective compared to mechanisms designed to integrate and embed alliance knowledge throughout the organisation (Heimeriks et al., 2015). Despite some tentative and tangential work, explorations with respect to the underlying factors that lead firms to become aware of the need to leverage their alliance experience in developing a capability, and to invest efforts and resources in processes for deliberate learning from alliances are sparse in extant literature (see Wang and Rajagopalan, 2015). We contend that the first instances of frequent engagement in alliances can form such underlying factors and we turn to evolutionary theory and dynamic RBV to assist in elaborating our argument. Specifically, we suggest that frequent engagement in alliances may compel firms to found an alliance capability development lifecycle, as they both protract the period of engagement in alliances, and engage more substantially in this activity (Helfat and Peteraf, 2003). By engaging in alliances frequently, firms expand, augment and renew their stock of cumulative alliance experience. Frequent engagement in alliances induces firms to invest in developing higher order routines and practises to initiate purposeful and directed organisational learning. Developing processes for alliance management and practices for their further improvement requires managerial commitment, commitment that is highly demanding in terms of time and employee resources (Winter, 2000, 2003; Dosi et al., 2000). As a result, firms need to justify the allocation of resources to building dedicated routines for alliance management: frequent task reoccurrence can trigger and motivate firms to deploy resources in such a direction. It can also justify such investments to external stakeholders, building internal support and commitment to the purpose. Sparse empirical literature explores the role of frequent engagement in alliances in developing alliance capabilities (see Wang and Rajagopalan, 2015 for a review). There is a particular dearth of longitudinal studies in this field (Draulans et al., 2003; Rothaermel and Deeds, 2006). Research on CIS longitudinal Spanish data indicates that persistent and not sporadic collaboration exhibits a systematic positive relationship with firm innovativeness, which runs parallel to our argument (Belderbos et al., 2015). Moreover, firms that renew and expand their collaborations and alliances over time become more embedded in a network and experience higher growth over time (Powell et al., 2005). At the cross-sectional level, there is only limited research that links larger alliance portfolios to the development of alliance capability (as this is reflected in alliance management practices). Larger alliance portfolios are associated with the use of enhanced alliance management practices and processes, with the latter being linked to improved firm performance (Heimeriks et al., 2009). Our longitudinal approach, capturing the tendencies of firms to engage in alliances more frequently over time, complements such accounts which are based on the role of stock measures of alliance experience in firm performance (Kale et al., 2002; Heimeriks and Duysters, 2007; Kale and Singh, 2007). Therefore we propose that firms that engage in alliances frequently are more likely to innovate compared to those firms that engage in alliances irregularly or sporadically, or that retain a static engagement and a stagnating alliance portfolio over time: H3. Frequent engagement in alliances and the ensuing expanding alliance portfolios positively affect firm likelihood of firm innovation over time. 5.1 Sample We test our hypotheses on a panel data set observing the population of firms in the UK bio-pharma sector (110 firms in 2003) during the period 1991-2001. The empirical exploration in this paper relies fundamentally on econometric analysis. However, secondarily the paper employs a case analysis to shed further light on - and nuance - some of its arguments. To identify our sample, we used two editions of the UK biotechnology directory (Coombs and Alston, 2000, 2002). The directory lists all firms in the sector that undertake research using biotechnologies. It provides a more comprehensive account of the firms in the UK bio-pharma sector than that available via the use of SIC classifications, as biotechnologies are used by firms from a range of industries in the life sciences as broadly described (Walsh, 2004). To form our working sample, we used the "Who Owns Who" directory to identify all firms that were active in 2003. Out of that list, we identified 110 firms that publish accounts in FAME[2]. For the sample of 110 firms, we collected data on alliances from ReCap.com and BioScan, the two most popular sources of alliance data in empirical research on this sector (e.g. Deeds and Hill, 1996; Schilling, 2009). The databases report alliances established in the bio-pharmaceutical sector for innovation purposes, including R&D alliances, research alliances and alliances for technology licensing and product development[3]. Innovation is crucial to firm survival and prosperity in the bio-pharmaceutical sector and patents are perceived as an appropriate proxy for innovations due to the science-based nature of the sector (e.g. Pavitt, 1984). Moreover, patents are an appropriate indicator to capture firm-level returns from innovation alliances (Demirkan, 2018; Shan et al., 1994; Sampson, 2005). We collected data on patents granted to the 110 firms in our sample between 1991 and 2001 by the UKPTO through the publicly available database Esp@cenet. Via deployment of Boolean searches, it was possible to assemble a comprehensive list by matching both the name and the address of firms in our sample with those of the patent assignees (Arora et al., 2011). For firms in our sample that belong to MNEs with non-UK-based HQs, we examined information on inventor location, and allocated any patents assigned to the MNE HQ with UK-based inventors to the UK-based subsidiary in our sample. Since our research focusses on UK-based firms, the UKPTO is perceived to be an appropriate source for the collection of patent data. Filing patent applications at national patent offices is less costly (in terms of application and renewal fees) and less time consuming when compared with international patent applications. As international applications also involve costs of unfamiliar IP regulations and legal frameworks, many firms - at least those with constrained resources - are more likely to patent at national patent offices than internationally (Archambault, 2002). Finally, information on firm accounts was obtained from FAME, and information on investments in R&D was gathered from Thomson's Analytics and the UK DTI's R&D Scoreboard. 5.2 Methods Patents take positive integer values, making count dependent variable models, such as the Poisson and Negative Binomial, appropriate (Greene, 2003). Instead of count dependent variable models, we employ discrete dependent variable models as they permit the capture of factors that alter firm abilities enabling them to perform over a threshold level: successful filing of a patent (Long, 1997). This is better aligned to the core aim of this work, which is to explore whether frequent alliance engagement is linked to organisational learning in alliance management (and alliance capabilities), reflected in firms exhibiting over threshold performance (Hoang and Rothaermel, 2005). Due to the longitudinal dimension of panels, we can explore factors that affect firms' changing abilities to innovate over time (fixed effects specification), as well as factors that can explain cross-firm variation in innovation abilities (random effects specification). The former, dynamic consideration of panels is of particular interest in this paper, as explained above. The fixed effects specification has a further appeal as its estimates are not based on the assumption of strict exogeneity of the independent variables (Greene, 2003; Baltagi, 1995), which is desirable, as decisions to form alliances may not be independent of past innovation performance (Colombo and Garrone, 1996; Gkypali et al., 2017). To explore further the nuanced factors underlying the development of alliance capabilities, we report on the findings of an in-depth case analysis that explores the link between frequency of engagement in alliances, and the development of practices and processes to manage alliances, an indication of alliance capabilities (Schreiner et al., 2009; Kale and Singh, 2007); this is a secondary analysis providing an illuminating complement. The case focusses on a UK bio-pharmaceutical firm that intensified its engagement in alliances particularly after the early 1990s when it started to invest in biotechnology R&D. Interviews were conducted in 2005 as the firm was forming a dedicated alliance function and entailed detailed reflection on the factors that underpin the creation of processes established to improve efficiency in alliance management. The processes identified were matched with those identified in existing literature on alliance capabilities (Kale et al., 2002; Kale and Singh, 2007, 2009; Heimeriks and Duysters, 2007). The case is based on interviews with directors of alliances and intellectual property, with a long working experience with the case firm and focussed on alliances that involve cooperation in research and that have led to patents. All materials and associated analyses were reviewed and approved by the interviewees. 5.3 Variables Patents (dependent variable) To test our hypotheses we create a dummy variable that is equal to 1 for each firm that is granted a patent at year t (and equal to zero otherwise). We use a patent dummy variable as we aim at capturing the impact of alliance experience in enabling firms to perform over a threshold, i.e. to innovate. To track the firm's innovation activities over time, we use patent filing dates. Filing dates have certain advantages over publication dates: first, they better reflect the originating time of innovations; and, second, they are not influenced by regulatory changes or fluctuations in resource availability in patent offices over time, as these can delay publication dates (e.g. Jaffe, 1986). Patents are widely used to capture firm returns from innovation alliances (Sampson, 2005, 2007; Shan et al., 1994; Demirkan, 2018). Alliance experience Research has explored the role of alliance experience in improving financial returns of firms from alliances (Kale et al., 2002), alliance project success (Hoang and Rothaermel, 2005), returns to firm innovation performance (Sampson, 2005, 2007; Belderbos et al., 2015; Di Guardo and Harrigan, 2016) and number of products under development (Deeds and Hill, 1996; Rothaermel and Deeds, 2006). Studies have operationalised alliance experience as a stock variable reflected either by the cumulative number of alliances formed throughout a firm's history (Kale and Singh, 2007; Hoang and Rothaermel, 2005; Kale et al., 2002; Rothaermel and Deeds, 2006), or by the cumulative number of years of experience in engaging in alliances (Rothaermel and Deeds, 2006). Only a handful of these studies consider diminishing (Sampson, 2005) or non-linear (Rothaermel and Deeds, 2006) relationships between alliance experience and firm returns from alliances and most studies base their estimates on cross-sectional data (an exception is Kale et al., 2002). Following existing literature, we operationalise alliance experience at each year t as the cumulative number of alliances throughout a firm's history. To aid interpretation of results, and following recent contributions (Lavie et al., 2011) we use one year lagged values of the above variable to more accurately capture the effect of cumulative alliance experience. Our variable is left-censored to 1991, the start year for our sampling. This is experienced elsewhere in alliance literature (e.g. Kale and Singh, 2007). Also, because in our study 87 per cent of total alliances from 1979 and 2001 fall within the 1991-2001 period, left censoring is not introducing considerable bias. As is customary in the literature (Sampson, 2005), we use the natural logarithmic transformation of cumulative alliance experience to test for diminishing returns (H2). As a robustness check, we test for a concave relationship (inverted U-shaped) by using both the linear and quadratic values of the cumulative number of alliances: again, this approach follows relevant literature (e.g. Deeds and Hill, 1996; Rothaermel and Deeds, 2006)[4]. Frequent engagement in alliances and expanding alliance portfolios To explore the effects of frequent engagement in alliances on firm abilities to innovate, we measure the total number of on-going alliances that a firm manages simultaneously at any point in time. As alliances last for several years, this requires information on alliance duration which is not available via the alliance data sources consulted in this study (RECap.com and BioScan). As lack of information on alliance duration has also been encountered in other work, we follow established tradition and rely on estimates of the average duration of alliances, i.e., five years (Kogut, 1988; Di Guardo and Harrigan, 2016). Therefore, we operationalise frequent engagement in alliances by calculating the number of alliances a firm manages at any given point in time by considering that alliances last on average for five years. Increases in this variable reflect the tendency of firms to engage more frequently in alliances over time, and on average, to initiate more alliances than they terminate, managing potentially an expanding portfolio of on-going alliances. As the use of a blanket measure for alliance duration might be considered simplistic, we cross-validated this information with alliance managers in the sector, and it was considered an acceptable approximation for the duration of the majority of alliances. We also performed robustness checks by using different estimates for alliance duration as detailed in our analysis. Control variables For parsimonious empirical models, we use the most appropriate control variables that have been employed in the literature on alliances and innovation. First, we account for the effects of well-established indicators of firm innovation, such as investments in R&D and firm size (Griliches, 1990) as they can influence firm abilities to innovate. We used a measure of R&D intensity that captures the amount of R&D expenditure per employee. Firm size is captured by annual turnover and we include a natural logarithm to overcome the possible problem of skewed distributions. Second, and following relevant research (Zollo et al., 2002), we account for differences in the content of alliances, as it influences innovation opportunities in alliances and affects firm performance (e.g. Anand and Khanna, 2000). Therefore, we include a dummy that takes the value of one when the focal firm establishes an alliance that involves R&D, and the value of zero for each year that a focal firm establishes alliances that do not involve R&D. Table I presents descriptive statistics and bivariate correlations. The panel is unbalanced, as a result of either missing information or the inclusion of less-mature dedicated biotechnology firms (i.e. a third of firms in the sample were established after 1995). The unbalanced data set includes firms observed for a maximum period of ten years and others for as little as four years. On average, firms are observed for approximately seven years (the range is from 6.6 to 7.4 years across the three estimated models) and form 8.28 alliances between 1991 and 2001. We estimated both the fixed and random effects specifications (FE and RE, respectively) of the Logit model for panel data for all three models (Greene, 2003; Baltagi, 1995)[5]. Results are obtained in STATA(c) 11. The r parameter is significant across all three models, indicating significant cross-firm heterogeneity. As Hausman's (1978) specification test has limited discerning power in discrete dependent variable models, we assess the relative merits of the two specifications on a conceptual basis and for reasons discussed in the methods section we present Logit FE estimates. Moreover, the residuals of the FE specification are corrected for potential heteroscedasticity (White, 1980; Baltagi, 1995). Based on model kh2 statistics in Table II, all three models are significant at 5 per cent levels and above. This suggests that our variables have a significant interpretative power with respect to the likelihood of firm innovation. We find support for all hypotheses. In Model 1, the cumulative number of alliances positively affects firm's likelihood of innovation over time, but levels of significance are low (10 per cent). Our results are consistent with studies showing a positive relationship between cumulative alliance experience and firm innovation, but at low significance levels (Hagedoorn and Schakenraad, 1994; Shan et al., 1994). The FE estimates of Model 2 provide support for diminishing returns to firm innovation from cumulative alliance experience, with the corresponding variable being positive and highly significant. The results of Model 2 are consistent with findings testing similar hypotheses in cross-sectional settings in the US bio-pharmaceutical and the telecom equipment industries (Hoang and Rothaermel, 2005; Sampson, 2005). Our results are robust to alternative operationalisations, as reflected by exploring concave relationships (inverted U-shaped), via use of both linear and quadratic values of cumulative alliance experience, with estimates (both FE and RE) confirming non-linear diminishing returns to likelihood of firm innovation. To unpack the extent of diminishing returns we estimate the inflection point based on estimates of the curvilinear relationship; this is equal to 127.515 points of cumulative experience, which corresponds to firms managing 18 alliances annually (average length of time per firm in the panel is seven years). Our results imply that at such a level of annual alliance activity, firms will accrue no additional benefits from gaining more experience in managing alliances. It should be noted that in our sample the average number of new alliances formed every year is 8.28, which suggests that the majority of firms in our sample are below the optimal level of alliance engagement. The concluding section discusses the implications of this. Model 3 provides support for our third hypothesis. The coefficient of the core independent variable is positive and significant (it marginally fails to pass the highest 1 per cent level of significance). Model 3 suggests that firms that establish alliances frequently and that continuously manage an expanding portfolio can enjoy improved likelihood of innovation. This provides fresh insights as, to date, there has been a scarcity of longitudinal studies in the literature (see Draulans et al., 2003; Rothaermel and Deeds, 2006). As a test of robustness, we estimated Model 3 by using four and six years as the average duration for alliances. The results remained robust to different operationalisations. Among the control variables, firm size is positively and significantly associated with firm's likelihood of innovation in Models 1 and 3. Investments in R&D per number of employees are negatively associated with likelihood of firm innovation but only in Model 2. Alliance content is insignificant across all three models. Finally, all three model estimates are robust to different measures of the control variables. For example, all models lead to the same outcomes in terms of significance when controls are collapsed to dichotomous variables indicating above and below average firm size and intensity of investments in R&D per number of employees. A further robustness check is undertaken to examine whether our results are sensitive to the statistical assumptions of the distribution followed by the error term in the Logit and Probit specifications. The Probit model only provides RE estimates for panel data (STATA(c) 11) and these further confirm our results. Moreover, we estimate our model specifications with the Negative Binomial model to explore differences between our dependent variable and a count variable capturing innovation performance with the number of patents. Both FE and RE estimates confirm our hypotheses. The results suggest that there are diminishing returns to cumulative alliance experience. This can signal that ageing experience contributes less to current outcomes and that recent experience may have a higher contribution (Shukla and Mital, 2018; Sampson, 2005). Following Sampson (2005), we explore the contributions of recent and past alliance experience. We develop a set of variables capturing alliance experience between 1 and 6 years prior to our year of observation. So, for example, in 2001 alliance experience of one year corresponds to the number of alliances initiated in 2000 and alliance experience of two years corresponds to those initiated in 1999 and so on. None of these variables appear to be significant, with the exception of alliance experience of four years prior to the year of observation, which appears with a negative and significant sign. Therefore, the diminishing returns to cumulative alliance experience identified in our paper cannot be attributed to decreasing contributions of distant experience. Our results most likely reflect that firms, by just forming more alliances and learning from experience how to improve alliance management and coordination, cannot experience improved efficiency ad infinitum. With regard to H3 we undertake further estimations to isolate the impact of frequent engagement in alliances from that of portfolio size. We break down the variable testing H3 into two components. First, capturing changes in the frequency of forming alliances over time, we construct a variable reflecting a decreasing, static and increasing trend in alliance portfolios (the variable takes the values of: -1, 0 and 1, respectively, and refers to comparisons between the current and the previous year). Second, we include a variable to capture the number of new alliances formed within every year (one year lagged values are used in the models). Our results show that the trend variable has a negative sign while the annual figures of newly formed alliances appear with a positive and significant sign (both for the Logit and the Negative Binomial models and the FE and RE specifications). It is worth noting that when the values of the trend variable are lagged by one year, their impact becomes insignificant (i.e. they retain a negative sign). This means that only the contemporaneous impact of alliance frequency on innovation is negative, implying that when firms expand their alliance portfolio there is an immediate adverse effect. This may result from immediate strains on their resources and cognitive abilities, which however, cease to have an adverse effect within a year's time. The results remain the same when the trend variable is substituted with a dummy variable capturing only increases in the frequency of forming alliances between current values and those of the previous year. Finally, we isolate partner-specific alliance experience from general (i.e. non-partner specific) experience to further explore the importance of alliance management learning within partner dyads and in alliance portfolios. None of these variables has a significant effect, suggesting that the separation lacks interpretative power. The paper contributes to the literature on the role of alliances in innovation and alliance capabilities by using a longitudinal analysis which enables capture of the impact of organisational learning in managing and coordinating alliances on innovation. As it is difficult to capture learning, the paper uses a longitudinal approach to capture dynamic changes within firms and it observes learning through its impact on outcome variables such as innovation. The paper adds to a slender body of work exploring antecedents to firm-level alliance capabilities and their impact on enhancing firm's abilities to innovate (for a review see Wang and Rajagopalan, 2015). First, we find that the impact of learning from gaining greater alliance experience in coordinating alliances and portfolios on firm innovation is positive but of weak significance. This is an important finding, as alliances in the bio-pharmaceutical sector can be characterised by power asymmetries and unbalanced resource dependence between partners, which are likely to create conflicts affecting the creation and appropriation of alliance value (Diestre and Rajagopalan, 2012; Caner and Tyler, 2013). The period of our study forms the first window of exploitation of biotechnologies via alliances which potentially creates disincentives to collaborating firms to behave opportunistically, as joining forces during this close to pre-competitive stage could be more beneficial than not (Powell et al., 2005). Second, our analysis proceeds by exploring the factors that affect learning in managing and coordinating alliances in this sector. The first arm to this exploration sheds light on the role of cumulative-historic experience in enhancing innovation performance, while the second arm focusses on the role of such experience as an antecedent to the creation of capabilities to manage interactions with alliance partners and to create synergies and minimise redundancies in alliance portfolios. In that endeavour, we first confirm diminishing returns to the benefits that can be accrued by gaining alliance experience for more effective management of alliances and capturing innovation returns. This supports a thin body of research which suggests that learning from gaining experience cannot confer benefits ad infinitum (e.g. Rothaermel and Deeds, 2006; Sampson, 2005). Our in-depth exploration shows that this cannot be attributed to diminishing contributions of ageing experience as identified in other work (Sampson, 2005; Shukla and Mital, 2018). It could be attributed to failure of firms to capture learning in a systematic way, or to costs associated with identifying more partners, establishing more agreements and coordinating increased diversity (White and Lui, 2005; Rothaermel and Deeds, 2006). Our last set of findings show that engaging in alliances more frequently over time, with firms managing an expanding portfolio, by forming more alliances than they terminate, enhances firms' likelihood of innovation. Our longitudinal approach is consistent with other work showing that persistence in collaboration has a positive impact on firm innovation as opposed to sporadic collaboration (Belderbos et al., 2015). Longitudinal explorations are currently almost absent in alliance literature, and their potential contribution has been particularly welcomed (Draulans et al., 2003; Rothaermel and Deeds, 2006), specifically as they might shed further light on the type of changes that can occur within firms, in particular, with respect to the dynamics of learning on alliance management and how it enhances innovation potential. Contributions within evolutionary theory (Winter, 2000, 2003) argue that frequent engagement in organisational tasks is linked to capability development, while the dynamic resource-based view (Helfat and Peteraf, 2003) argues that capabilities develop through several stages: experience accumulation compels firms to recognise the need to establish a capability, and this initiates a process of systematic learning and alliance management. Our results clearly pertain to this literature and specifically on how alliance experience forms an antecedent to alliance and alliance portfolio capabilities (Wang and Rajagopalan, 2015). To explore, in a situated and more granular fashion, whether frequent engagement in alliances and a dynamically expanding alliance portfolio contribute to the development of alliance capabilities, we conducted a detailed case analysis as a complement to our econometric analysis. The case analysis explores the link between an increasing engagement in alliances and the development of novel or changed practices and processes for alliance management, i.e., those factors that normally sit within the "black-box" of alliance capability studies (Anand and Khanna, 2000; Heimeriks and Duysters, 2007; Kale and Singh, 2007). The case analysis clearly reveals such a relationship. First, it shows that alliance management practices and routines are established after alliances become an integral part of firm R&D and innovation strategy and firms intensify and commit to alliance activity. As the Alliance Director interviewed for our study stated: "[...] we have standardised approaches for contractual compliance [with respect to alliances] [...] it's inconceivable to me that one wouldn't have, and yet five years ago we didn't [...] You've got one deal and you do it however you do it which largely depends on who you employ and who it's [the alliance] with [...] I am not sure how many deals you need to have before you get yourself a standardised process but it's probably substantially more than ten or twenty [...][7]". Second, several of the practices identified in the extant literature that are used to institutionalise and diffuse alliance management learning (namely, employment of alliance professionals, use of alliance metrics, creation of databases to solicit potential alliance partners and establishment of processes to monitor alliance performance) are deployed by the case firm (Heimeriks and Duysters, 2007; Kale and Singh, 2007). Ten years after intensifying alliance engagement, it developed a dedicated alliance management function to coordinate alliance activity and engage in monitoring and improving existing alliance management practices. The case analysis illustrates the links between frequent engagement in alliances and the ensuing expanding the alliance portfolio, with the motivation to invest resources in developing firm-specific capabilities for alliance management. It also reflects the time required to recognise that alliance management can be improved via the development of specific and specialist processes for the governance of cross-partner interaction (a finding that resonates strongly with existing literature, Kale and Singh, 2009). In addition, the case reflects the requirement for firms to continuously review and develop their capabilities in order to confront new challenges and intensified competition. Indeed, the case analysis, combined with the findings of our econometric exploration, suggest that renewing alliance experience and committing to forming more alliances over time are crucial in enabling firms to create and capture value in alliances. Our research contributes to the literature on the antecedents of alliance and alliance portfolio capabilities and on the conditions that shape superior firm outcomes from alliances, such as innovation. It suggests a need to delve into the antecedents to alliance capabilities, a thin body of research and to identify nascent factors that provide potential foundations for alliance capability development (Helfat and Peteraf, 2003), shifting attention away from the role played by alliance management practices (such as dedicated alliance functions) that have dominated existing research. This is particularly important as firms may establish such practices during the advanced stages of the alliance capability development process (Kale and Singh, 2009), and as such they may not appropriately reflect the foundational stages of alliance capability development. Here, we echo calls to delve deeper in understanding alliance management capabilities and the antecedents to alliance portfolio capabilities (Wang and Rajagopalan, 2015), especially in the context of alliances involving a higher learning potential (Heimeriks, 2010). Moreover, recent research shows that codification of alliance learning and systematic approaches to alliance management contribute to efficient partner selection and alliance termination, but may restrict flexibility and adaptability which are important for efficient management during the course of the alliance (Heimeriks et al., 2015; Wang and Rajagopalan, 2015). Our research has several implications for managerial practice and policy. First, we find that there are diminishing returns to benefits in terms of enhanced innovation accruing from gaining greater alliance experience. Specifically firms reach a plateau at 18 alliances in annual terms, suggesting that on average this is the optimal level of alliance engagement for enhancing firm likelihood of innovation. This has implications for managing alliance portfolio size for innovation efficiency (e.g. Hoffmann, 2007; Demirkan, 2018). On average firms in our sample are below the optimal level of alliance engagement (forming eight alliances annually). This suggests that innovation likelihood can be marginally enhanced by expanding alliance engagement. This is in line with findings of other research (Rothaermel and Deeds, 2006; Sampson, 2005; Hewitt-Dundas and Roper, 2018) which in the context of micro-businesses, for example, shows that sub-optimal engagement can be attributed to reluctance of firms to cooperate due to fear of trustworthiness and lack of knowledge on benefits of collaboration and partner capabilities (Hewitt-Dundas and Roper, 2018). This emphasises our findings with respect to the need to facilitate a culture of systematic management of alliances. It also supports research suggesting the need for policy interventions, for example in the form of intermediaries, which can support and encourage firms to establish collaborations and manage openness (Howells, 2006). This is not only due to the benefits of collaboration to firm innovation but also due to the positive externalities from openness, suggesting broader economic and societal benefits (Roper et al., 2013). At the firm level, our results suggest that it is important for firms to replenish their alliance portfolios and to gradually form more alliances than they terminate over time if they wish to enhance their innovation likelihood. In the bio-pharmaceuticals sector where there is high rate of alliance formation, it is important for firm competitiveness and future growth that they renew and expand their networks (Powell et al., 2005). This is linked to firms becoming more aware of the need to make alliance management learning more systematic. Firms should use such systematic practices in a discretionary manner and should also consider altering and upgrading such practices to respond to the challenges and demands of new alliances. Although our results are specific in terms of sector and period, they derive from the examination of a paradigmatic period in the development of new technologies, and address the implications of changing innovation processes in an established industry. As such, they provide useful insights that assist understanding of trajectories and dynamics in industries with similarities to biotech, for example, nanotechnology (e.g. Rothaermel and Thursby, 2007). Our aspiration is that our research will provide some re-focussing and clarity in an established field of study. We acknowledge that analysis of further sectors and periods will nuance our understanding of both the role of gaining experience in improving firm alliance management and in developing alliance capabilities, and hope that the ideas and methods explored above will assist in this on-going quest.
|
Contribution derives from a longitudinal analysis of an original panel data set that maps the UK bio-pharmaceuticals sector over the initial period of its development. The paper sheds light on factors that can compel firms to form alliance capabilities, and extends a currently thin body of work on the foundations and antecedents to alliance and alliance portfolio capabilities.
|
[SECTION: Purpose] "Omnichannel" has become a buzzword in retail for good reason. New technologies, such as mobile devices and social media, combined with better data, bring the long-time dream of a unified cross-channel shopping experience within reach. In practice, however, most retailers still fall short of achieving this vision, especially as it applies to Generation Y ("Gen-Y" - those people born between 1980 and 1991) and their use of smartphones. Mobile users are increasingly accessing social media using mobile devices, whether via browsers or apps. A study by Adobe (2013) among mobile users in the USA, Canada, UK, France and Germany found that most had accessed social networks using a mobile device, ranging from 94 per cent for those 18-29 years of age to 75 per cent of those 50-64 years of age. In fact, Facebook was the second most visited Web site/application that was accessed by smartphones and was the top smartphone app in the USA in August 2013 (ComScore, 2013). In countries, such as Italy and Germany, penetration rates of mobile phones exceed 100 per cent, with some consumers owning more than one mobile phone (Kaplan, 2012). Mobile phones and devices are increasingly used in conducting mobile commerce (Venkatesh et al., 2003, Ngai and Gunasekaran, 2007). Reputation is one of the reasons why customers rely on particular Web sites, apps or Web-apps, and some studies have investigated trust and reputation issues in a mobile ad hoc network environment (Lax and Sarne, 2008). Newman (2010) found that some 700,000 people view wine-related videos every month; there are over 7,000 wine tweets per day and > 300 iPhone apps for wine. Breslin (2013) estimated that 90 per cent of wine drinkers use Facebook 6.2 hours per week. The wine industry is not excluded from this phenomenon; this is the problem we focus on in this study. The state of the art in the literature regarding wine industry during the past decade has seen an increasing series of studies of wine consumption and behavior (Celhay, 2008; Nowak et al., 2006; Ritchie et al., 2009; Thach and Olsen, 2004, 2006), but none of them has so far been interested in exploring the potential of mobile commerce (m-commerce) as it relates to the wine industry. Among the elements that contribute to this massive use of Internet devices, let us cite the growth in social media usage in recent years (Facebook, Twitter, and so on). Mobile social media is defined as "a group of mobile marketing applications that allow the creation and exchange of user-generated content" (Kaplan, 2012), such that a geo-social service like "Facebook Check-In" is a subset of mobile marketing. The instantaneous insights on consumer profiles, m-commerce and the development of mobile tools (smartphones, tablets and watches like the Samsung Gear) reinforce the appetite consumers have for this "Communicating Swiss-knife" style, which is also sometimes considered "habit forming". Consumers today are not passive shoppers. They seek out and disseminate information about a product using social media. They are not shy about providing their views and opinions, and they will abandon any company that does not serve their needs or is perceived as not providing value. Our question, therefore, relates to the coming of age of Generation Y. Because so many decisions related to wine are made on the move - in restaurants, at airports, in retail stores, in discussions with friends or during conferences - the smartphone or tablet people very often have in their hands can provide fast information on which wine to buy, although the subsequent information search related to a bottle of wine may or may not lead to purchase. Almost everyone in the developed world today uses a smartphone. Because Gen-Y is the most significant demographic cohort since the Baby Boomers and because this population is tech-savvy, it represents a significant market for wine marketers. This paper attempts to study and understand the wine shopping habits and smartphone use in Gen Y. In addition, this paper aims to provide insights on the perception of Gen-Y customers, their use of social media and the potential challenges and opportunities in terms of shopping behavior. We consider a potential m-commerce strategy and how it applies to the wine industry. In this article, we use the term m-commerce to refer to smartphones, social media, applications (or apps) and Web applications, despite the differences in usage patterns. We can access Facebook or any social media or blogs, Twitter, etc., from our smartphones, but we might then use apps/Web apps[1] to access, for example, eRobertParker, Wine Enthusiast or Wine Searcher, etc., or an online store or a restaurant wine list. Our research question is: "To what extent is consumer behavior shaped by the use of mobile devices and what are the Gen-Y cohort's expectations when they consider purchasing wine?" Until now, little has been written on the subject of smartphones and wine purchasing patterns. Such a study would provide academics and the wine industry with useful insights. Following Lockshin and Corsi's (2012) suggestions we strive to go deeper in one of the under-researched areas: determinants of online and social media influences on consumers. Our paper will therefore try to bring new insights into this growing market share represented by Gen-Y to enable wine professionals to understand and target them with even more acuity. To understand what wine marketers can do to leverage sales using mobile devices and social media, an empirical study based on a questionnaire was specifically designed with Gen-Y users in mind. Our research enabled us to highlight the importance of the effects of the design interface used on m-commerce Web sites on consumers' buying intention. It also shows that wine growers will need to adapt, respond and grow their digital business to suit their customers, who live in a state of constant interconnectedness. Growers must design strategies to fit the mobile contexts to sell more on these types of devices. We present the results of this study before concluding and presenting limits as well as future avenues of research. Who is a member of Gen-Y? Gen-Y is defined as those born between 1980 and 1991 (Chowdhury and Coulter, 2006; Lescohier, 2006). The literature in relation to this cohort is quite abundant as was identified by Lockshin and Corsi (2012). A number of articles compare wine preferences and/or behavior between the younger and older generations. For a review, the reader is invited to refer to articles by: Agnoli et al. (2011), Ritchie (2011), Fountain and Lamb (2011), de Magistris et al. (2011), Charters et al. (2011), Mueller et al. (2011), Qenani-Petrela et al. (2007) and Wolf et al. (2005). It is believed that Gen-Y has the potential of forming long-term loyalties with products which satisfy them at this vital stage (Paul, 2001). In 2014, Gen-Y, the largest generation since the Baby Boomers (Neuborne and Kerwin, 1999), range in age from 14 to 32 (Norum, 2008). For the purposes of this study, the age range of Gen-Y has been narrowed to ages 18 to 32 because, according to Djamasbi et al. (2008), increased financial independence is gained at the age of 18 and alcohol consumption is prohibited before the age of 18. Gen-Y grew up in the digital world. It is technically literate and continually wired, plugged in and connected to digitally streaming information, entertainment and contacts. It has mastered the technology in such a way that multitasking is a habit that carries into the workplace, where it tends to instant message its contacts while working (Lewis, 2003). Although studies into the general characteristics of this market are limited and none have been conducted related to wine consumption, most findings consistently highlight similar attributes for psychographics and purchase behavior across well- researched industries, such as fashion-related items (Dias, 2003; Paul, 2001) or more generally industries related to these five meaningful and distinct decision-making groups, namely, "recreational quality seekers", "recreational discount seekers", "trend setting loyals", "shopping and fashion uninterested" and "confused time/money conserving" (see Bakewell and Mitchell, 2003). Regarding the wine industry, attitude to wine and consumption behavior may vary, according to Fountain and Lamb (2011). Authors found that Gen-Y consumers drink more wine than Gen-X did at a similar age. However, behavioral differences between the two age cohorts are likely to be considerably smaller than differentials in their attitudes. Gen-Y is not a homogeneous cohort, according to Mueller et al. (2011) and de Magistris et al. (2011). By comparing views and behaviors of Gen-Y wine consumers in a number of countries, authors point to substantial cultural differences. Despite certain similarities between Gen-Y cohorts across a range of countries, it is very hard to generalize about the wine consumption behavior of the cohort across a range of countries (Mueller and Charter, 2011). M-commerce habits M-commerce, according to the Mobile Marketing Association (2013), is the one- or two-way exchange of value facilitated by a mobile consumer electronic device (e.g. mobile handset or smartphone), which is enabled by wireless technologies and communication networks. M-commerce, when used in conjunction with social media, shows considerable potential for generating sustainable profits; however, further demonstration is required. Articles about the use and effects of social media on consumer wine preference and behavior exist as identified by Lockshin and Corsi (2012); for a review, see works of Reyneke et al. (2011), Claster et al. (2010), Pitt et al. (2011) and Nicholls (2012). Unfortunately, due to the infancy of this topic, analysis of the combined use of smartphones and social media still seems almost nonexistent. To succeed in producing profits in the highly competitive m-commerce environment, organizations must not only develop the technical expertise and business strategy necessary for the creation of an effective "Social media-m-commerce website" tandem/partnership but also create a satisfying user experience (Pelet and Lecat, 2011; Lecat and Pelet, 2011). Today's competitive environment companies must evolve quickly and understand that the consumer shopping in a digital world is not bound by the old buying habits. A new kind of collaboration between sellers and buyers needs to happen. The speed of evolution of the Internet does not always allow companies to realize the implications of new consumer behavior trends or understand the value they expect to realize from investing in digital technologies. Indeed, the digital world has modified the way brands look at customers. Recent figures show an impressive growth of mobile use. Cell phone adoption is especially widespread in young populations, with 75 per cent of teenagers and 93 per cent of adults aged 18-29 years having a cell phone, with 55 per cent of the latter group accessing the Internet wirelessly from their cell phone (Lenhart et al., 2011). The evolution of mobile communications has triggered an increase in the use of mobile devices, such as mobile phones, to conduct m-commerce on the mobile Web (Venkatesh et al., 2003; Ngai and Gunasekaran, 2007). To support their digital business, many companies are turning to digital marketing. However, creating a well-conceived and executed digital marketing strategy requires clarity of concept and goals that few companies are willing or able to invest in. The value of digital marketing is, as yet, unclear because many consumers do not wish to be exposed to advertising on their smartphones. This is the challenge currently being faced by such giants as Google, Facebook and Amazon. Buying a product and sharing information in real time are two inherent aspects of the new consumption experience. Purchasing behaviors are increasingly affected by the way we use the Web to make purchasing decisions and how we influence, test, protest, praise, share, recommend or learn about products or services. As more consumers access social media via mobile devices, it changes the way they research and shop for products and services. M-commerce has not surpassed e-commerce yet, but the rapid growth of mobile telephony has fueled the expansion of the mobile Internet as a foundation for mobile commerce (Lee and Benbasat, 2004). The mobile Internet has unique strengths because users can connect to it wherever and whenever they want (Kakihara and Sorensen, 2001). The ability of the wine industry to attract a new customer base (Thach, 2005) requires a thorough understanding of the needs and wants of that group (Kotler, 2003). Given that there are > 10,000 wine labels on the market, buying wine can be a confusing experience, so wine buyers will rely on the advice of experts and friends to help them decide which wine to buy. This understanding is even more important when we recall that despite certain superficial similarities between Gen-Y cohorts across a range of countries, it is nonetheless very difficult to generalize about the wine consumption behavior of Gen-Y, as there are distinct cultural and market developmental differences that are much more significant than the age-related similarities according to Mueller and Charters (2011). As novice or potential wine consumers, they are becoming increasingly significant targets for wine marketers (Mueller and Charters, 2011), especially because they are very adept at using relatively new technologies such as smartphones. By providing a reliable location anywhere, anytime, mobile devices offer companies the opportunity to conduct marketing campaigns that aim to drive both traffic and sales, whether mobile or in-store, in ever expanding ways. Applications and Web-apps (especially when they are free) have become a way of life for consumers. Such apps and Web-apps are thus in a position to address each stage of the consumer purchase funnel - awareness, engagement, consideration, conversion and loyalty. Smartphone penetration could definitely favor the surge in mobile shopping for wine. Owners of such devices are more or less connected to social media with embedded location-based applications. Technology can help brands to locate their fans and target their mobile strategy to quickly interact with shoppers. Today, someone walking on a street can receive an SMS on her/his mobile phone, indicating that the restaurant of the X brand in the same street is offering her/him a free aperitif if he/she turns up within the next hour. The person who receives this alert is a fan of the X brand on a social media, such as Facebook. He/she may have left the location-based application of her/his favorite digital social network activated to locate her/him anywhere anytime. At the same time, the restaurant has learned that this person is also a fan of a particular wine and has friends who also seem to be fans of the food they cook and the wine they serve, so invitations are sent to these people as well. Schummer (2001) has shown that 90 per cent of shoppers prefer to communicate in real time while shopping. Consumer recommendations are the most trusted form of advertising around the world. Although this is true, it is also a double-edged sword, with consumers trashing a restaurant or store for personal reasons, thereby causing much damage to a store's reputation. Over 75 per cent of respondents from 47 markets across the world rated recommendations from consumers as a trusted form of advertising, compared to 63 per cent for newspapers, 56 per cent for TV and magazines and 34 per cent for search engine advertisements (Nielsen, 2007). This underscores the notion that people trust people with similar experiences. Thanks to their ease of use, social media can help shoppers provide positive ratings and recommendations about the goods of a vendor on its public social media's page/apps/Web-apps, an action which would ideally be rewarded. Any exploration of m-commerce marketing must thoroughly investigate how consumers position themselves on these aspects of the interaction with the brand. Research method The confirmatory study was implemented on the Internet using a questionnaire. Each participant has visited and completed it. By asking French and international consumers coming from > 10 countries how they perceive social media generally and m-commerce using social media, we obtained the following results. Participants The sample is composed of 190 valid questionnaires completed by international Gen-Y consumers born after 1980. It was selected to achieve a balance in terms of age and socio-professional background. We worked mainly with students (60 per cent) for this analysis because they are deemed suitable as a sample. Although the use of students has often been questioned in terms of their appropriateness, they share many characteristics with the profile of mobile and Internet users' population, such as age. As shown by several studies, Internet users tend to be young adults, while the Internet usage penetration within the age groups of 18-29 years reaches 95 per cent (Zickuhr, 2010; Pew Research Center, 2010). Hence, although our sample presents a bias towards younger subjects, it can arguably be acceptable as representative of Internet and mobile users. In addition, Gen-Y is an important group of online consumers (Delafrooz et al., 2010) and is useful as a sample for empirical studies in m-commerce, in line with previous research (Kim et al., 2008). Hypotheses Based on the assumption that wine purchase (irrespective of the distribution channel: physical store, e-commerce or m-commerce) should be positively influenced by their use of Internet, mobile devices and their relations with social media because m-commerce has removed all the barriers to get wine anywhere (Kakihara and Sorensen, 2001). Their age and income should also influence positively their frequency of purchase because of the potential rise of interest in wine and the money available to do so. Finally, frequency of wine purchase should also be positively influenced by wine purchase and consumption characteristics. We therefore pose the following hypothesis (Table I). Profile of the sample The table presented in Appendix summarizes the profile of the Gen-Y consumers who were surveyed. Among the 190 respondents, 68 per cent of them connect to the Internet using their mobile and 38 per cent spend between 10 and 19 hours a week on the Internet, whereas 41 per cent spend more than 20 hours a week. Nearly all of them (95 per cent) access the Internet every day. Only 43 per cent purchase on the Internet at least once a month. They mainly use the Internet to stay in touch with friends and relatives (95 per cent), but also to look for information on a product (69 per cent). They also access discussion groups (29 per cent) and chat rooms (21 per cent). In terms of social media, nearly all use Facebook (93.2 per cent), then Twitter (32.6 per cent), then Groon (18.9 per cent) and then Google+ (15.3 per cent). They are on social media mainly to visit friends' pages (88.4 per cent) and have been in touch with social media through their friends (90.5 per cent) and through their classmates (50 per cent). When they are connected, most of them spend one hour or more on social media (67.9 per cent). Most of them connect every day (77.4 per cent) to social media - 26.8 per cent declared they have already bought a good or a service based on friends' recommendations issued from social media (of importance: clothes, transport ticket, CD and then books). Regarding wine, only 5 of them were influenced by friends' recommendations. Regarding the wine consumption and habits, 7.4 per cent of our respondents are members of a group dedicated to wine. Wine purchases range from "don't buy wine" (32.6 per cent) to 36.8 per cent "buy wine several times a year". With regards to wine buyers only, they mainly buy it in supermarkets (56.3 per cent) and hypermarkets (19.5 per cent), but also in wine shops (19.5 per cent). They prefer to consume good wines: of the total 22.7 per cent buy wine below 5 , but 43.8 per cent buy wine between 5 and 10 ; 26 per cent between 11 and 20 ; and 17 per cent spend > 20 . They mainly buy (moderately) between half a bottle (32.8 per cent) and one bottle (23.4 per cent). Among the respondents purchasing whole bottles, 27.4 per cent drink one glass of wine every day; 3.2 per cent drink half a bottle; whereas 64.2 per cent consume occasionally (not every day). They mainly consume wine at home (26.3 per cent) or at a friend's place (32.1 per cent). Most of them never go to the Internet to look for information on wine (82.1 per cent). Of the total, 5.8 per cent were able to quote one Web site only. For 72.6 per cent of them, on-time delivery builds trust when they are buying wine through their mobile phone and is viewed as an important feature. The fact that the wine is delivered in good condition is an important aspect, with 76 per cent of respondents expecting it. The Gen-Y consumers (56/190) who frequently buy wine (several times a month or more) consider the usefulness of the information regarding a specific wine (3.2/5) as important when looking up information through their mobile. This generation also rates highly: their overall satisfaction regarding their ability to use a mobile (4/5); their learning about the use of a mobile (3.8/5); positive experiences linked to the use of the mobile (3.4/5); and their level of happiness at being able to use a mobile (3.4/5). Out of those 29 per cent who do often buy wine (56/190), 36 use their mobile to go to social media and 34 look for information on the Internet. On the other hand, only 10 of them purchase products through the Internet. Furthermore, 25 of them spend more than 1 hour per day on their mobile and 17 spend more than 2 hours a month (but it might mean anything because they are students with a lot of time to spare). Let us also note that 52 of them access the Internet daily, 38 look for information on the Internet and 22 spend more than 20 hours a week. In terms of projection, if they had to buy some wine through their mobile, they respectively would consider the following factors as important: on-time delivery (3.6); cost of transportation (3.6); delivery in good condition (3.5); and follow-up to the transaction (3.5). Results The sample of 190 respondents was cut into two groups: One composed of Gen-Y consumers who do not purchase alcohol at all (n = 62). The other (n = 128) composed of Gen-Y consumers who buy wine frequently. Only the group that had purchased alcohol was analysed. A linear regression was run with the frequency of wine purchase as dependent variable (Y) and the 15 independent variables (X) listed in Table I. Table III shows that X1, X2, X5, X8, X12 and X15 positively influence the frequency of wine purchase. It can also be observed that variables X7, X9 and X14 exert nearly no impact on the frequency of wine purchase (b coefficients are close to 0). And finally, X3, X4, X6 and X13 have a negative coefficient. Based on our hypotheses, the b coefficients should be positive but also significant (p < 0.05). Variables X1 and X2 are significant and positive, whereas X3 and X7 are negative but significant (see bold in Table II). We therefore conclude that only H1 and H2 are valid: H1. The frequency of wine purchase is positively influenced by the Gen-Y consumers who access the Internet through their mobile phone. This validation is very important because Gen-Ys are getting more and more mobile-friendly and can get easy access to the Internet with their mobile. H2. The frequency of wine purchase is positively influenced by the average hours spent by Gen-Y consumers on the Internet per week. Note that H7 and H3 are significant but have a negative b coefficient: H3. The frequency of wine purchase is negatively influenced by the frequency of purchase on the Internet (irrespective of the product category). This result needs further research because it may be due to the composition of the sample (majority of students) who does not earn money, and if they do buy goods or services, may not have enough remaining cash for buying wines. H7. The frequency of wine purchase is negatively influenced by the frequency of connection to social media, but with a small b coefficient (which limits the impact of this variable). Table III shows the importance of three factors when buying wine through the Internet. Quickness and reduced cost contribute to promote overall satisfaction and develop underlying loyalty, which is not surprising because e-commerce and m-commerce differentiate themselves from physical distribution in terms of quickness, cost reduction and therefore efficiency. Table IV on the other hand is showing the importance of specific factors when Gen-Y consumers are looking for information on wine online. In short, they want a lot of useful information which is rich in terms of content. Table V shows that if a wine purchase was planned through a mobile, cost and follow-up are not that important compared to delivery in good condition or delivery on time. Finally, if a comparison is made between wine buyers and non-buyers (see Table VI), we can observe that wine buyers are likelier to be male, older, foreign and wealthier. They are less likely to own a smartphone. Wine buyers also spend less time on their mobile and Internet, but they tend to buy more online. Their purchases are more influenced by their friends' recommendations. Finally, wine buyers are more likely to consume wine daily. As novice or potential wine consumers, members of Gen-Y are becoming increasingly significant targets for wine marketers (Mueller and Charters, 2011). This paper looks at the current state of m-commerce, the consumption of smartphones and social media and the transformation of the consumer into an omnichannel shopper. It also examines some responses to this emerging way of doing shopping which enriches the in-store experience with digital integrations, especially in relation to Gen-Y consumers. Armed with smartphones and tablets, wine shoppers go back and forth effortlessly between the real (whether in hypermarkets or supermarkets, convenience stores, wine shops, at the estate/winery or during a wine fair) and digital worlds (through the Internet, mobile app, wine club or by mail order). They are using their phones while in stores to research products and compare prices, and they order online and then pick up in person. At the same time, they consult friends near and far whenever they may find themselves contemplating a purchase, such as a nice bottle of wine. Every day, more of them come to expect a mobile or an "omnichannel" experience. Nearly all retailers aspire to create this type of experience for their customers because creating a consistent customer experience across channels is very important for their business. Indeed, most aspects of such an omnichannel strategy appear inherently important. Results from our confirmatory study show that the frequency of wine purchase is positively influenced by Gen-Y consumers who access the Internet through their mobile phone. This result is consistent with the Adobe (2013) study which showed that 94 per cent of 18-29-year-old consumers had accessed social media using a mobile device. Therefore, designing m-commerce Web sites and applications that stimulate the feeling of being in a real market or in a real store can give more confidence to users who are engaged in carrying out commercial transactions, especially those based on social media-derived peer advice. Current and potential purchasers emphasized the following three factors when buying wine through the Internet: Speed of delivery. Cost reduction. Efficiency. These factors should be taken into account when developing interfaces. In line with Atkin and Thach (2012) and Barber et al. (2008), we recommend that wine marketers take actions to help reduce perceived risks in wine purchasing, in particular, a specific focus on the delivery of wine in good condition or the delivery on time if a wine purchase was planned through a mobile. In addition to doing commerce, wine growers will need to develop entirely new and effective ways for people to share and communicate information especially because H2 shows the frequency of wine purchase is positively influenced by the average hours spent by Gen-Y consumers on the Internet per week. In this regard, Facebook (> 90 per cent of the surveyed Gen-Y members are members) or Twitter could contribute to improving sales. In addition, wine growers will have to find new partners to reach a wider audience of potential users. It is possible to imagine vineyards partnering with restaurants, retail chains, cooking enthusiasts, etc. Also important will be the need to simplify, optimize and streamline the design of the digital marketing strategy to create new experiences and behaviors. Authors found in previous studies that Gen-Y consumers drink more wine than Gen-X did at a similar age and behavioral differences between the two age cohorts are likely to be considerably smaller than differentials in their attitudes. Gen-Y is not a homogeneous cohort according to Mueller et al. (2011) and de Magistris et al. (2011). To better target young wine buyers, wine growers should have in mind that wine buyers are likelier to be male, older, foreign and have more money, although they are less likely to have the latest smartphone. Furthermore, they spend less time on their mobile and Internet (compared to non-wine buyers), but they tend to buy more online and are likelier to make purchase decisions based on friends' recommendation. As Kakihara and Sorensen (2001) underlined, the mobile Internet has unique strengths because users can connect to it wherever and whenever they want. This topic is promising because France AgriMer (2012) shows that wine consumption is composed of 45 per cent of occasional drinkers (once or twice a week) for consumers below the age of 25, and between 25 and 34 years, 50 per cent are occasional drinkers. The Wine Market Council has also published data showing that the millennials (Gen-Y) were consuming 24 per cent of the total volume. Finally, based on AC Nielsen (2011) in March, Internet and mail-order are representing 10 per cent of the total sales in UK. Those data show that the Internet is more and more important, that Gen-Y members are occasional drinkers, and therefore to link those occasional drinkers with wine, it is important to develop pleasant platforms and interactive social networks. The advantage of providing good service on an m-commerce Web site translates into satisfied customers who will become brand advocates: they can refer other people to the wine grower. Internet consumers talk to other consumers about a good customer service experience. For a service, such as the purchase of wine from a particular wine grower who is already selling online and thinking of expanding into m-commerce, providing good customer service is a must. When customers tell other people about a bad experience, they do it on social networks to reach a large audience. This is why social networks must also be taken into account when planning an m-commerce strategy to avoid any negative buzz and therefore leverage a good e-reputation. Indeed, wine growers now must navigate a very complex customer service landscape, and the stakes are high. Consumers can be quick to punish those who are slow to respond to their questions or fail to deliver their purchases on time, for example. Growers are also vulnerable to reputation damage if they offer unsatisfactory answers or ignore their customers altogether, behavior that is likely to make the customer mad and eager to exact revenge by writing negative reviews. Wine growers who stumble stand to lose business to rivals and become vulnerable to negative social buzz. Customer service is one of the few remaining sources of competitive differentiation for this type of business, especially in relation to Gen-Y consumers. Wine growers that understand how to use the latest digital channels to reach new customers and help them find, buy and consume the products they sell will be rewarded with loyal customers willing to spend more and advocate for the brand. Customers are loyal to businesses that provide good customer service in the brick and mortar world; this is not expected to change in the m-commerce environment. Further research should be carried on with a larger sample, and a specific look at the rejected H3, which showed that the frequency of wine purchase, is negatively influenced by the frequency of purchase on the Internet. Finally, further studies should focus on international comparisons because it is very hard to generalize about the wine consumption behavior of the cohort across a range of countries, as there are distinct cultural and market developmental differences that are much more significant than the age-related similarities (Mueller and Charters (2011)). Opens in a new window. Table I. Variables tested by linear regression and formulation of the hypotheses Opens in a new window. Table II. Linear regression for 128 respondents (who do buy) with frequency of wine purchase as dependent variable Y Opens in a new window. Table III. Necessary factors for Gen-Y consumers to develop satisfaction and loyalty when buying wine online Opens in a new window. Table IV. Necessary factors for Gen-Y consumers to develop satisfaction and loyalty when they are looking for information on wine online Opens in a new window. Table V. Importance of the following characteristics if a purchase was planned through your mobile Opens in a new window. Table VI. Importance of the following characteristics if a purchase was planned through your mobile Opens in a new window. Table AI. Characteristics of the sample of respondents
|
The purpose of this study is to explore the potential of m-commerce (mobile commerce) combined with social media for the wine industry in order to reach distinctive customer segments Differences about access to information and shopping habits on the Internet seem to exist according to the cohorts users belong to.
|
[SECTION: Method] "Omnichannel" has become a buzzword in retail for good reason. New technologies, such as mobile devices and social media, combined with better data, bring the long-time dream of a unified cross-channel shopping experience within reach. In practice, however, most retailers still fall short of achieving this vision, especially as it applies to Generation Y ("Gen-Y" - those people born between 1980 and 1991) and their use of smartphones. Mobile users are increasingly accessing social media using mobile devices, whether via browsers or apps. A study by Adobe (2013) among mobile users in the USA, Canada, UK, France and Germany found that most had accessed social networks using a mobile device, ranging from 94 per cent for those 18-29 years of age to 75 per cent of those 50-64 years of age. In fact, Facebook was the second most visited Web site/application that was accessed by smartphones and was the top smartphone app in the USA in August 2013 (ComScore, 2013). In countries, such as Italy and Germany, penetration rates of mobile phones exceed 100 per cent, with some consumers owning more than one mobile phone (Kaplan, 2012). Mobile phones and devices are increasingly used in conducting mobile commerce (Venkatesh et al., 2003, Ngai and Gunasekaran, 2007). Reputation is one of the reasons why customers rely on particular Web sites, apps or Web-apps, and some studies have investigated trust and reputation issues in a mobile ad hoc network environment (Lax and Sarne, 2008). Newman (2010) found that some 700,000 people view wine-related videos every month; there are over 7,000 wine tweets per day and > 300 iPhone apps for wine. Breslin (2013) estimated that 90 per cent of wine drinkers use Facebook 6.2 hours per week. The wine industry is not excluded from this phenomenon; this is the problem we focus on in this study. The state of the art in the literature regarding wine industry during the past decade has seen an increasing series of studies of wine consumption and behavior (Celhay, 2008; Nowak et al., 2006; Ritchie et al., 2009; Thach and Olsen, 2004, 2006), but none of them has so far been interested in exploring the potential of mobile commerce (m-commerce) as it relates to the wine industry. Among the elements that contribute to this massive use of Internet devices, let us cite the growth in social media usage in recent years (Facebook, Twitter, and so on). Mobile social media is defined as "a group of mobile marketing applications that allow the creation and exchange of user-generated content" (Kaplan, 2012), such that a geo-social service like "Facebook Check-In" is a subset of mobile marketing. The instantaneous insights on consumer profiles, m-commerce and the development of mobile tools (smartphones, tablets and watches like the Samsung Gear) reinforce the appetite consumers have for this "Communicating Swiss-knife" style, which is also sometimes considered "habit forming". Consumers today are not passive shoppers. They seek out and disseminate information about a product using social media. They are not shy about providing their views and opinions, and they will abandon any company that does not serve their needs or is perceived as not providing value. Our question, therefore, relates to the coming of age of Generation Y. Because so many decisions related to wine are made on the move - in restaurants, at airports, in retail stores, in discussions with friends or during conferences - the smartphone or tablet people very often have in their hands can provide fast information on which wine to buy, although the subsequent information search related to a bottle of wine may or may not lead to purchase. Almost everyone in the developed world today uses a smartphone. Because Gen-Y is the most significant demographic cohort since the Baby Boomers and because this population is tech-savvy, it represents a significant market for wine marketers. This paper attempts to study and understand the wine shopping habits and smartphone use in Gen Y. In addition, this paper aims to provide insights on the perception of Gen-Y customers, their use of social media and the potential challenges and opportunities in terms of shopping behavior. We consider a potential m-commerce strategy and how it applies to the wine industry. In this article, we use the term m-commerce to refer to smartphones, social media, applications (or apps) and Web applications, despite the differences in usage patterns. We can access Facebook or any social media or blogs, Twitter, etc., from our smartphones, but we might then use apps/Web apps[1] to access, for example, eRobertParker, Wine Enthusiast or Wine Searcher, etc., or an online store or a restaurant wine list. Our research question is: "To what extent is consumer behavior shaped by the use of mobile devices and what are the Gen-Y cohort's expectations when they consider purchasing wine?" Until now, little has been written on the subject of smartphones and wine purchasing patterns. Such a study would provide academics and the wine industry with useful insights. Following Lockshin and Corsi's (2012) suggestions we strive to go deeper in one of the under-researched areas: determinants of online and social media influences on consumers. Our paper will therefore try to bring new insights into this growing market share represented by Gen-Y to enable wine professionals to understand and target them with even more acuity. To understand what wine marketers can do to leverage sales using mobile devices and social media, an empirical study based on a questionnaire was specifically designed with Gen-Y users in mind. Our research enabled us to highlight the importance of the effects of the design interface used on m-commerce Web sites on consumers' buying intention. It also shows that wine growers will need to adapt, respond and grow their digital business to suit their customers, who live in a state of constant interconnectedness. Growers must design strategies to fit the mobile contexts to sell more on these types of devices. We present the results of this study before concluding and presenting limits as well as future avenues of research. Who is a member of Gen-Y? Gen-Y is defined as those born between 1980 and 1991 (Chowdhury and Coulter, 2006; Lescohier, 2006). The literature in relation to this cohort is quite abundant as was identified by Lockshin and Corsi (2012). A number of articles compare wine preferences and/or behavior between the younger and older generations. For a review, the reader is invited to refer to articles by: Agnoli et al. (2011), Ritchie (2011), Fountain and Lamb (2011), de Magistris et al. (2011), Charters et al. (2011), Mueller et al. (2011), Qenani-Petrela et al. (2007) and Wolf et al. (2005). It is believed that Gen-Y has the potential of forming long-term loyalties with products which satisfy them at this vital stage (Paul, 2001). In 2014, Gen-Y, the largest generation since the Baby Boomers (Neuborne and Kerwin, 1999), range in age from 14 to 32 (Norum, 2008). For the purposes of this study, the age range of Gen-Y has been narrowed to ages 18 to 32 because, according to Djamasbi et al. (2008), increased financial independence is gained at the age of 18 and alcohol consumption is prohibited before the age of 18. Gen-Y grew up in the digital world. It is technically literate and continually wired, plugged in and connected to digitally streaming information, entertainment and contacts. It has mastered the technology in such a way that multitasking is a habit that carries into the workplace, where it tends to instant message its contacts while working (Lewis, 2003). Although studies into the general characteristics of this market are limited and none have been conducted related to wine consumption, most findings consistently highlight similar attributes for psychographics and purchase behavior across well- researched industries, such as fashion-related items (Dias, 2003; Paul, 2001) or more generally industries related to these five meaningful and distinct decision-making groups, namely, "recreational quality seekers", "recreational discount seekers", "trend setting loyals", "shopping and fashion uninterested" and "confused time/money conserving" (see Bakewell and Mitchell, 2003). Regarding the wine industry, attitude to wine and consumption behavior may vary, according to Fountain and Lamb (2011). Authors found that Gen-Y consumers drink more wine than Gen-X did at a similar age. However, behavioral differences between the two age cohorts are likely to be considerably smaller than differentials in their attitudes. Gen-Y is not a homogeneous cohort, according to Mueller et al. (2011) and de Magistris et al. (2011). By comparing views and behaviors of Gen-Y wine consumers in a number of countries, authors point to substantial cultural differences. Despite certain similarities between Gen-Y cohorts across a range of countries, it is very hard to generalize about the wine consumption behavior of the cohort across a range of countries (Mueller and Charter, 2011). M-commerce habits M-commerce, according to the Mobile Marketing Association (2013), is the one- or two-way exchange of value facilitated by a mobile consumer electronic device (e.g. mobile handset or smartphone), which is enabled by wireless technologies and communication networks. M-commerce, when used in conjunction with social media, shows considerable potential for generating sustainable profits; however, further demonstration is required. Articles about the use and effects of social media on consumer wine preference and behavior exist as identified by Lockshin and Corsi (2012); for a review, see works of Reyneke et al. (2011), Claster et al. (2010), Pitt et al. (2011) and Nicholls (2012). Unfortunately, due to the infancy of this topic, analysis of the combined use of smartphones and social media still seems almost nonexistent. To succeed in producing profits in the highly competitive m-commerce environment, organizations must not only develop the technical expertise and business strategy necessary for the creation of an effective "Social media-m-commerce website" tandem/partnership but also create a satisfying user experience (Pelet and Lecat, 2011; Lecat and Pelet, 2011). Today's competitive environment companies must evolve quickly and understand that the consumer shopping in a digital world is not bound by the old buying habits. A new kind of collaboration between sellers and buyers needs to happen. The speed of evolution of the Internet does not always allow companies to realize the implications of new consumer behavior trends or understand the value they expect to realize from investing in digital technologies. Indeed, the digital world has modified the way brands look at customers. Recent figures show an impressive growth of mobile use. Cell phone adoption is especially widespread in young populations, with 75 per cent of teenagers and 93 per cent of adults aged 18-29 years having a cell phone, with 55 per cent of the latter group accessing the Internet wirelessly from their cell phone (Lenhart et al., 2011). The evolution of mobile communications has triggered an increase in the use of mobile devices, such as mobile phones, to conduct m-commerce on the mobile Web (Venkatesh et al., 2003; Ngai and Gunasekaran, 2007). To support their digital business, many companies are turning to digital marketing. However, creating a well-conceived and executed digital marketing strategy requires clarity of concept and goals that few companies are willing or able to invest in. The value of digital marketing is, as yet, unclear because many consumers do not wish to be exposed to advertising on their smartphones. This is the challenge currently being faced by such giants as Google, Facebook and Amazon. Buying a product and sharing information in real time are two inherent aspects of the new consumption experience. Purchasing behaviors are increasingly affected by the way we use the Web to make purchasing decisions and how we influence, test, protest, praise, share, recommend or learn about products or services. As more consumers access social media via mobile devices, it changes the way they research and shop for products and services. M-commerce has not surpassed e-commerce yet, but the rapid growth of mobile telephony has fueled the expansion of the mobile Internet as a foundation for mobile commerce (Lee and Benbasat, 2004). The mobile Internet has unique strengths because users can connect to it wherever and whenever they want (Kakihara and Sorensen, 2001). The ability of the wine industry to attract a new customer base (Thach, 2005) requires a thorough understanding of the needs and wants of that group (Kotler, 2003). Given that there are > 10,000 wine labels on the market, buying wine can be a confusing experience, so wine buyers will rely on the advice of experts and friends to help them decide which wine to buy. This understanding is even more important when we recall that despite certain superficial similarities between Gen-Y cohorts across a range of countries, it is nonetheless very difficult to generalize about the wine consumption behavior of Gen-Y, as there are distinct cultural and market developmental differences that are much more significant than the age-related similarities according to Mueller and Charters (2011). As novice or potential wine consumers, they are becoming increasingly significant targets for wine marketers (Mueller and Charters, 2011), especially because they are very adept at using relatively new technologies such as smartphones. By providing a reliable location anywhere, anytime, mobile devices offer companies the opportunity to conduct marketing campaigns that aim to drive both traffic and sales, whether mobile or in-store, in ever expanding ways. Applications and Web-apps (especially when they are free) have become a way of life for consumers. Such apps and Web-apps are thus in a position to address each stage of the consumer purchase funnel - awareness, engagement, consideration, conversion and loyalty. Smartphone penetration could definitely favor the surge in mobile shopping for wine. Owners of such devices are more or less connected to social media with embedded location-based applications. Technology can help brands to locate their fans and target their mobile strategy to quickly interact with shoppers. Today, someone walking on a street can receive an SMS on her/his mobile phone, indicating that the restaurant of the X brand in the same street is offering her/him a free aperitif if he/she turns up within the next hour. The person who receives this alert is a fan of the X brand on a social media, such as Facebook. He/she may have left the location-based application of her/his favorite digital social network activated to locate her/him anywhere anytime. At the same time, the restaurant has learned that this person is also a fan of a particular wine and has friends who also seem to be fans of the food they cook and the wine they serve, so invitations are sent to these people as well. Schummer (2001) has shown that 90 per cent of shoppers prefer to communicate in real time while shopping. Consumer recommendations are the most trusted form of advertising around the world. Although this is true, it is also a double-edged sword, with consumers trashing a restaurant or store for personal reasons, thereby causing much damage to a store's reputation. Over 75 per cent of respondents from 47 markets across the world rated recommendations from consumers as a trusted form of advertising, compared to 63 per cent for newspapers, 56 per cent for TV and magazines and 34 per cent for search engine advertisements (Nielsen, 2007). This underscores the notion that people trust people with similar experiences. Thanks to their ease of use, social media can help shoppers provide positive ratings and recommendations about the goods of a vendor on its public social media's page/apps/Web-apps, an action which would ideally be rewarded. Any exploration of m-commerce marketing must thoroughly investigate how consumers position themselves on these aspects of the interaction with the brand. Research method The confirmatory study was implemented on the Internet using a questionnaire. Each participant has visited and completed it. By asking French and international consumers coming from > 10 countries how they perceive social media generally and m-commerce using social media, we obtained the following results. Participants The sample is composed of 190 valid questionnaires completed by international Gen-Y consumers born after 1980. It was selected to achieve a balance in terms of age and socio-professional background. We worked mainly with students (60 per cent) for this analysis because they are deemed suitable as a sample. Although the use of students has often been questioned in terms of their appropriateness, they share many characteristics with the profile of mobile and Internet users' population, such as age. As shown by several studies, Internet users tend to be young adults, while the Internet usage penetration within the age groups of 18-29 years reaches 95 per cent (Zickuhr, 2010; Pew Research Center, 2010). Hence, although our sample presents a bias towards younger subjects, it can arguably be acceptable as representative of Internet and mobile users. In addition, Gen-Y is an important group of online consumers (Delafrooz et al., 2010) and is useful as a sample for empirical studies in m-commerce, in line with previous research (Kim et al., 2008). Hypotheses Based on the assumption that wine purchase (irrespective of the distribution channel: physical store, e-commerce or m-commerce) should be positively influenced by their use of Internet, mobile devices and their relations with social media because m-commerce has removed all the barriers to get wine anywhere (Kakihara and Sorensen, 2001). Their age and income should also influence positively their frequency of purchase because of the potential rise of interest in wine and the money available to do so. Finally, frequency of wine purchase should also be positively influenced by wine purchase and consumption characteristics. We therefore pose the following hypothesis (Table I). Profile of the sample The table presented in Appendix summarizes the profile of the Gen-Y consumers who were surveyed. Among the 190 respondents, 68 per cent of them connect to the Internet using their mobile and 38 per cent spend between 10 and 19 hours a week on the Internet, whereas 41 per cent spend more than 20 hours a week. Nearly all of them (95 per cent) access the Internet every day. Only 43 per cent purchase on the Internet at least once a month. They mainly use the Internet to stay in touch with friends and relatives (95 per cent), but also to look for information on a product (69 per cent). They also access discussion groups (29 per cent) and chat rooms (21 per cent). In terms of social media, nearly all use Facebook (93.2 per cent), then Twitter (32.6 per cent), then Groon (18.9 per cent) and then Google+ (15.3 per cent). They are on social media mainly to visit friends' pages (88.4 per cent) and have been in touch with social media through their friends (90.5 per cent) and through their classmates (50 per cent). When they are connected, most of them spend one hour or more on social media (67.9 per cent). Most of them connect every day (77.4 per cent) to social media - 26.8 per cent declared they have already bought a good or a service based on friends' recommendations issued from social media (of importance: clothes, transport ticket, CD and then books). Regarding wine, only 5 of them were influenced by friends' recommendations. Regarding the wine consumption and habits, 7.4 per cent of our respondents are members of a group dedicated to wine. Wine purchases range from "don't buy wine" (32.6 per cent) to 36.8 per cent "buy wine several times a year". With regards to wine buyers only, they mainly buy it in supermarkets (56.3 per cent) and hypermarkets (19.5 per cent), but also in wine shops (19.5 per cent). They prefer to consume good wines: of the total 22.7 per cent buy wine below 5 , but 43.8 per cent buy wine between 5 and 10 ; 26 per cent between 11 and 20 ; and 17 per cent spend > 20 . They mainly buy (moderately) between half a bottle (32.8 per cent) and one bottle (23.4 per cent). Among the respondents purchasing whole bottles, 27.4 per cent drink one glass of wine every day; 3.2 per cent drink half a bottle; whereas 64.2 per cent consume occasionally (not every day). They mainly consume wine at home (26.3 per cent) or at a friend's place (32.1 per cent). Most of them never go to the Internet to look for information on wine (82.1 per cent). Of the total, 5.8 per cent were able to quote one Web site only. For 72.6 per cent of them, on-time delivery builds trust when they are buying wine through their mobile phone and is viewed as an important feature. The fact that the wine is delivered in good condition is an important aspect, with 76 per cent of respondents expecting it. The Gen-Y consumers (56/190) who frequently buy wine (several times a month or more) consider the usefulness of the information regarding a specific wine (3.2/5) as important when looking up information through their mobile. This generation also rates highly: their overall satisfaction regarding their ability to use a mobile (4/5); their learning about the use of a mobile (3.8/5); positive experiences linked to the use of the mobile (3.4/5); and their level of happiness at being able to use a mobile (3.4/5). Out of those 29 per cent who do often buy wine (56/190), 36 use their mobile to go to social media and 34 look for information on the Internet. On the other hand, only 10 of them purchase products through the Internet. Furthermore, 25 of them spend more than 1 hour per day on their mobile and 17 spend more than 2 hours a month (but it might mean anything because they are students with a lot of time to spare). Let us also note that 52 of them access the Internet daily, 38 look for information on the Internet and 22 spend more than 20 hours a week. In terms of projection, if they had to buy some wine through their mobile, they respectively would consider the following factors as important: on-time delivery (3.6); cost of transportation (3.6); delivery in good condition (3.5); and follow-up to the transaction (3.5). Results The sample of 190 respondents was cut into two groups: One composed of Gen-Y consumers who do not purchase alcohol at all (n = 62). The other (n = 128) composed of Gen-Y consumers who buy wine frequently. Only the group that had purchased alcohol was analysed. A linear regression was run with the frequency of wine purchase as dependent variable (Y) and the 15 independent variables (X) listed in Table I. Table III shows that X1, X2, X5, X8, X12 and X15 positively influence the frequency of wine purchase. It can also be observed that variables X7, X9 and X14 exert nearly no impact on the frequency of wine purchase (b coefficients are close to 0). And finally, X3, X4, X6 and X13 have a negative coefficient. Based on our hypotheses, the b coefficients should be positive but also significant (p < 0.05). Variables X1 and X2 are significant and positive, whereas X3 and X7 are negative but significant (see bold in Table II). We therefore conclude that only H1 and H2 are valid: H1. The frequency of wine purchase is positively influenced by the Gen-Y consumers who access the Internet through their mobile phone. This validation is very important because Gen-Ys are getting more and more mobile-friendly and can get easy access to the Internet with their mobile. H2. The frequency of wine purchase is positively influenced by the average hours spent by Gen-Y consumers on the Internet per week. Note that H7 and H3 are significant but have a negative b coefficient: H3. The frequency of wine purchase is negatively influenced by the frequency of purchase on the Internet (irrespective of the product category). This result needs further research because it may be due to the composition of the sample (majority of students) who does not earn money, and if they do buy goods or services, may not have enough remaining cash for buying wines. H7. The frequency of wine purchase is negatively influenced by the frequency of connection to social media, but with a small b coefficient (which limits the impact of this variable). Table III shows the importance of three factors when buying wine through the Internet. Quickness and reduced cost contribute to promote overall satisfaction and develop underlying loyalty, which is not surprising because e-commerce and m-commerce differentiate themselves from physical distribution in terms of quickness, cost reduction and therefore efficiency. Table IV on the other hand is showing the importance of specific factors when Gen-Y consumers are looking for information on wine online. In short, they want a lot of useful information which is rich in terms of content. Table V shows that if a wine purchase was planned through a mobile, cost and follow-up are not that important compared to delivery in good condition or delivery on time. Finally, if a comparison is made between wine buyers and non-buyers (see Table VI), we can observe that wine buyers are likelier to be male, older, foreign and wealthier. They are less likely to own a smartphone. Wine buyers also spend less time on their mobile and Internet, but they tend to buy more online. Their purchases are more influenced by their friends' recommendations. Finally, wine buyers are more likely to consume wine daily. As novice or potential wine consumers, members of Gen-Y are becoming increasingly significant targets for wine marketers (Mueller and Charters, 2011). This paper looks at the current state of m-commerce, the consumption of smartphones and social media and the transformation of the consumer into an omnichannel shopper. It also examines some responses to this emerging way of doing shopping which enriches the in-store experience with digital integrations, especially in relation to Gen-Y consumers. Armed with smartphones and tablets, wine shoppers go back and forth effortlessly between the real (whether in hypermarkets or supermarkets, convenience stores, wine shops, at the estate/winery or during a wine fair) and digital worlds (through the Internet, mobile app, wine club or by mail order). They are using their phones while in stores to research products and compare prices, and they order online and then pick up in person. At the same time, they consult friends near and far whenever they may find themselves contemplating a purchase, such as a nice bottle of wine. Every day, more of them come to expect a mobile or an "omnichannel" experience. Nearly all retailers aspire to create this type of experience for their customers because creating a consistent customer experience across channels is very important for their business. Indeed, most aspects of such an omnichannel strategy appear inherently important. Results from our confirmatory study show that the frequency of wine purchase is positively influenced by Gen-Y consumers who access the Internet through their mobile phone. This result is consistent with the Adobe (2013) study which showed that 94 per cent of 18-29-year-old consumers had accessed social media using a mobile device. Therefore, designing m-commerce Web sites and applications that stimulate the feeling of being in a real market or in a real store can give more confidence to users who are engaged in carrying out commercial transactions, especially those based on social media-derived peer advice. Current and potential purchasers emphasized the following three factors when buying wine through the Internet: Speed of delivery. Cost reduction. Efficiency. These factors should be taken into account when developing interfaces. In line with Atkin and Thach (2012) and Barber et al. (2008), we recommend that wine marketers take actions to help reduce perceived risks in wine purchasing, in particular, a specific focus on the delivery of wine in good condition or the delivery on time if a wine purchase was planned through a mobile. In addition to doing commerce, wine growers will need to develop entirely new and effective ways for people to share and communicate information especially because H2 shows the frequency of wine purchase is positively influenced by the average hours spent by Gen-Y consumers on the Internet per week. In this regard, Facebook (> 90 per cent of the surveyed Gen-Y members are members) or Twitter could contribute to improving sales. In addition, wine growers will have to find new partners to reach a wider audience of potential users. It is possible to imagine vineyards partnering with restaurants, retail chains, cooking enthusiasts, etc. Also important will be the need to simplify, optimize and streamline the design of the digital marketing strategy to create new experiences and behaviors. Authors found in previous studies that Gen-Y consumers drink more wine than Gen-X did at a similar age and behavioral differences between the two age cohorts are likely to be considerably smaller than differentials in their attitudes. Gen-Y is not a homogeneous cohort according to Mueller et al. (2011) and de Magistris et al. (2011). To better target young wine buyers, wine growers should have in mind that wine buyers are likelier to be male, older, foreign and have more money, although they are less likely to have the latest smartphone. Furthermore, they spend less time on their mobile and Internet (compared to non-wine buyers), but they tend to buy more online and are likelier to make purchase decisions based on friends' recommendation. As Kakihara and Sorensen (2001) underlined, the mobile Internet has unique strengths because users can connect to it wherever and whenever they want. This topic is promising because France AgriMer (2012) shows that wine consumption is composed of 45 per cent of occasional drinkers (once or twice a week) for consumers below the age of 25, and between 25 and 34 years, 50 per cent are occasional drinkers. The Wine Market Council has also published data showing that the millennials (Gen-Y) were consuming 24 per cent of the total volume. Finally, based on AC Nielsen (2011) in March, Internet and mail-order are representing 10 per cent of the total sales in UK. Those data show that the Internet is more and more important, that Gen-Y members are occasional drinkers, and therefore to link those occasional drinkers with wine, it is important to develop pleasant platforms and interactive social networks. The advantage of providing good service on an m-commerce Web site translates into satisfied customers who will become brand advocates: they can refer other people to the wine grower. Internet consumers talk to other consumers about a good customer service experience. For a service, such as the purchase of wine from a particular wine grower who is already selling online and thinking of expanding into m-commerce, providing good customer service is a must. When customers tell other people about a bad experience, they do it on social networks to reach a large audience. This is why social networks must also be taken into account when planning an m-commerce strategy to avoid any negative buzz and therefore leverage a good e-reputation. Indeed, wine growers now must navigate a very complex customer service landscape, and the stakes are high. Consumers can be quick to punish those who are slow to respond to their questions or fail to deliver their purchases on time, for example. Growers are also vulnerable to reputation damage if they offer unsatisfactory answers or ignore their customers altogether, behavior that is likely to make the customer mad and eager to exact revenge by writing negative reviews. Wine growers who stumble stand to lose business to rivals and become vulnerable to negative social buzz. Customer service is one of the few remaining sources of competitive differentiation for this type of business, especially in relation to Gen-Y consumers. Wine growers that understand how to use the latest digital channels to reach new customers and help them find, buy and consume the products they sell will be rewarded with loyal customers willing to spend more and advocate for the brand. Customers are loyal to businesses that provide good customer service in the brick and mortar world; this is not expected to change in the m-commerce environment. Further research should be carried on with a larger sample, and a specific look at the rejected H3, which showed that the frequency of wine purchase, is negatively influenced by the frequency of purchase on the Internet. Finally, further studies should focus on international comparisons because it is very hard to generalize about the wine consumption behavior of the cohort across a range of countries, as there are distinct cultural and market developmental differences that are much more significant than the age-related similarities (Mueller and Charters (2011)). Opens in a new window. Table I. Variables tested by linear regression and formulation of the hypotheses Opens in a new window. Table II. Linear regression for 128 respondents (who do buy) with frequency of wine purchase as dependent variable Y Opens in a new window. Table III. Necessary factors for Gen-Y consumers to develop satisfaction and loyalty when buying wine online Opens in a new window. Table IV. Necessary factors for Gen-Y consumers to develop satisfaction and loyalty when they are looking for information on wine online Opens in a new window. Table V. Importance of the following characteristics if a purchase was planned through your mobile Opens in a new window. Table VI. Importance of the following characteristics if a purchase was planned through your mobile Opens in a new window. Table AI. Characteristics of the sample of respondents
|
Based on a study of 190 respondents, this paper describes the consumer behavior of the Y generation regarding several variables: level of Internet sophistication and orientation, brand loyalty, risk aversion, involvement, shopping behavior and perception of authenticity.
|
[SECTION: Findings] "Omnichannel" has become a buzzword in retail for good reason. New technologies, such as mobile devices and social media, combined with better data, bring the long-time dream of a unified cross-channel shopping experience within reach. In practice, however, most retailers still fall short of achieving this vision, especially as it applies to Generation Y ("Gen-Y" - those people born between 1980 and 1991) and their use of smartphones. Mobile users are increasingly accessing social media using mobile devices, whether via browsers or apps. A study by Adobe (2013) among mobile users in the USA, Canada, UK, France and Germany found that most had accessed social networks using a mobile device, ranging from 94 per cent for those 18-29 years of age to 75 per cent of those 50-64 years of age. In fact, Facebook was the second most visited Web site/application that was accessed by smartphones and was the top smartphone app in the USA in August 2013 (ComScore, 2013). In countries, such as Italy and Germany, penetration rates of mobile phones exceed 100 per cent, with some consumers owning more than one mobile phone (Kaplan, 2012). Mobile phones and devices are increasingly used in conducting mobile commerce (Venkatesh et al., 2003, Ngai and Gunasekaran, 2007). Reputation is one of the reasons why customers rely on particular Web sites, apps or Web-apps, and some studies have investigated trust and reputation issues in a mobile ad hoc network environment (Lax and Sarne, 2008). Newman (2010) found that some 700,000 people view wine-related videos every month; there are over 7,000 wine tweets per day and > 300 iPhone apps for wine. Breslin (2013) estimated that 90 per cent of wine drinkers use Facebook 6.2 hours per week. The wine industry is not excluded from this phenomenon; this is the problem we focus on in this study. The state of the art in the literature regarding wine industry during the past decade has seen an increasing series of studies of wine consumption and behavior (Celhay, 2008; Nowak et al., 2006; Ritchie et al., 2009; Thach and Olsen, 2004, 2006), but none of them has so far been interested in exploring the potential of mobile commerce (m-commerce) as it relates to the wine industry. Among the elements that contribute to this massive use of Internet devices, let us cite the growth in social media usage in recent years (Facebook, Twitter, and so on). Mobile social media is defined as "a group of mobile marketing applications that allow the creation and exchange of user-generated content" (Kaplan, 2012), such that a geo-social service like "Facebook Check-In" is a subset of mobile marketing. The instantaneous insights on consumer profiles, m-commerce and the development of mobile tools (smartphones, tablets and watches like the Samsung Gear) reinforce the appetite consumers have for this "Communicating Swiss-knife" style, which is also sometimes considered "habit forming". Consumers today are not passive shoppers. They seek out and disseminate information about a product using social media. They are not shy about providing their views and opinions, and they will abandon any company that does not serve their needs or is perceived as not providing value. Our question, therefore, relates to the coming of age of Generation Y. Because so many decisions related to wine are made on the move - in restaurants, at airports, in retail stores, in discussions with friends or during conferences - the smartphone or tablet people very often have in their hands can provide fast information on which wine to buy, although the subsequent information search related to a bottle of wine may or may not lead to purchase. Almost everyone in the developed world today uses a smartphone. Because Gen-Y is the most significant demographic cohort since the Baby Boomers and because this population is tech-savvy, it represents a significant market for wine marketers. This paper attempts to study and understand the wine shopping habits and smartphone use in Gen Y. In addition, this paper aims to provide insights on the perception of Gen-Y customers, their use of social media and the potential challenges and opportunities in terms of shopping behavior. We consider a potential m-commerce strategy and how it applies to the wine industry. In this article, we use the term m-commerce to refer to smartphones, social media, applications (or apps) and Web applications, despite the differences in usage patterns. We can access Facebook or any social media or blogs, Twitter, etc., from our smartphones, but we might then use apps/Web apps[1] to access, for example, eRobertParker, Wine Enthusiast or Wine Searcher, etc., or an online store or a restaurant wine list. Our research question is: "To what extent is consumer behavior shaped by the use of mobile devices and what are the Gen-Y cohort's expectations when they consider purchasing wine?" Until now, little has been written on the subject of smartphones and wine purchasing patterns. Such a study would provide academics and the wine industry with useful insights. Following Lockshin and Corsi's (2012) suggestions we strive to go deeper in one of the under-researched areas: determinants of online and social media influences on consumers. Our paper will therefore try to bring new insights into this growing market share represented by Gen-Y to enable wine professionals to understand and target them with even more acuity. To understand what wine marketers can do to leverage sales using mobile devices and social media, an empirical study based on a questionnaire was specifically designed with Gen-Y users in mind. Our research enabled us to highlight the importance of the effects of the design interface used on m-commerce Web sites on consumers' buying intention. It also shows that wine growers will need to adapt, respond and grow their digital business to suit their customers, who live in a state of constant interconnectedness. Growers must design strategies to fit the mobile contexts to sell more on these types of devices. We present the results of this study before concluding and presenting limits as well as future avenues of research. Who is a member of Gen-Y? Gen-Y is defined as those born between 1980 and 1991 (Chowdhury and Coulter, 2006; Lescohier, 2006). The literature in relation to this cohort is quite abundant as was identified by Lockshin and Corsi (2012). A number of articles compare wine preferences and/or behavior between the younger and older generations. For a review, the reader is invited to refer to articles by: Agnoli et al. (2011), Ritchie (2011), Fountain and Lamb (2011), de Magistris et al. (2011), Charters et al. (2011), Mueller et al. (2011), Qenani-Petrela et al. (2007) and Wolf et al. (2005). It is believed that Gen-Y has the potential of forming long-term loyalties with products which satisfy them at this vital stage (Paul, 2001). In 2014, Gen-Y, the largest generation since the Baby Boomers (Neuborne and Kerwin, 1999), range in age from 14 to 32 (Norum, 2008). For the purposes of this study, the age range of Gen-Y has been narrowed to ages 18 to 32 because, according to Djamasbi et al. (2008), increased financial independence is gained at the age of 18 and alcohol consumption is prohibited before the age of 18. Gen-Y grew up in the digital world. It is technically literate and continually wired, plugged in and connected to digitally streaming information, entertainment and contacts. It has mastered the technology in such a way that multitasking is a habit that carries into the workplace, where it tends to instant message its contacts while working (Lewis, 2003). Although studies into the general characteristics of this market are limited and none have been conducted related to wine consumption, most findings consistently highlight similar attributes for psychographics and purchase behavior across well- researched industries, such as fashion-related items (Dias, 2003; Paul, 2001) or more generally industries related to these five meaningful and distinct decision-making groups, namely, "recreational quality seekers", "recreational discount seekers", "trend setting loyals", "shopping and fashion uninterested" and "confused time/money conserving" (see Bakewell and Mitchell, 2003). Regarding the wine industry, attitude to wine and consumption behavior may vary, according to Fountain and Lamb (2011). Authors found that Gen-Y consumers drink more wine than Gen-X did at a similar age. However, behavioral differences between the two age cohorts are likely to be considerably smaller than differentials in their attitudes. Gen-Y is not a homogeneous cohort, according to Mueller et al. (2011) and de Magistris et al. (2011). By comparing views and behaviors of Gen-Y wine consumers in a number of countries, authors point to substantial cultural differences. Despite certain similarities between Gen-Y cohorts across a range of countries, it is very hard to generalize about the wine consumption behavior of the cohort across a range of countries (Mueller and Charter, 2011). M-commerce habits M-commerce, according to the Mobile Marketing Association (2013), is the one- or two-way exchange of value facilitated by a mobile consumer electronic device (e.g. mobile handset or smartphone), which is enabled by wireless technologies and communication networks. M-commerce, when used in conjunction with social media, shows considerable potential for generating sustainable profits; however, further demonstration is required. Articles about the use and effects of social media on consumer wine preference and behavior exist as identified by Lockshin and Corsi (2012); for a review, see works of Reyneke et al. (2011), Claster et al. (2010), Pitt et al. (2011) and Nicholls (2012). Unfortunately, due to the infancy of this topic, analysis of the combined use of smartphones and social media still seems almost nonexistent. To succeed in producing profits in the highly competitive m-commerce environment, organizations must not only develop the technical expertise and business strategy necessary for the creation of an effective "Social media-m-commerce website" tandem/partnership but also create a satisfying user experience (Pelet and Lecat, 2011; Lecat and Pelet, 2011). Today's competitive environment companies must evolve quickly and understand that the consumer shopping in a digital world is not bound by the old buying habits. A new kind of collaboration between sellers and buyers needs to happen. The speed of evolution of the Internet does not always allow companies to realize the implications of new consumer behavior trends or understand the value they expect to realize from investing in digital technologies. Indeed, the digital world has modified the way brands look at customers. Recent figures show an impressive growth of mobile use. Cell phone adoption is especially widespread in young populations, with 75 per cent of teenagers and 93 per cent of adults aged 18-29 years having a cell phone, with 55 per cent of the latter group accessing the Internet wirelessly from their cell phone (Lenhart et al., 2011). The evolution of mobile communications has triggered an increase in the use of mobile devices, such as mobile phones, to conduct m-commerce on the mobile Web (Venkatesh et al., 2003; Ngai and Gunasekaran, 2007). To support their digital business, many companies are turning to digital marketing. However, creating a well-conceived and executed digital marketing strategy requires clarity of concept and goals that few companies are willing or able to invest in. The value of digital marketing is, as yet, unclear because many consumers do not wish to be exposed to advertising on their smartphones. This is the challenge currently being faced by such giants as Google, Facebook and Amazon. Buying a product and sharing information in real time are two inherent aspects of the new consumption experience. Purchasing behaviors are increasingly affected by the way we use the Web to make purchasing decisions and how we influence, test, protest, praise, share, recommend or learn about products or services. As more consumers access social media via mobile devices, it changes the way they research and shop for products and services. M-commerce has not surpassed e-commerce yet, but the rapid growth of mobile telephony has fueled the expansion of the mobile Internet as a foundation for mobile commerce (Lee and Benbasat, 2004). The mobile Internet has unique strengths because users can connect to it wherever and whenever they want (Kakihara and Sorensen, 2001). The ability of the wine industry to attract a new customer base (Thach, 2005) requires a thorough understanding of the needs and wants of that group (Kotler, 2003). Given that there are > 10,000 wine labels on the market, buying wine can be a confusing experience, so wine buyers will rely on the advice of experts and friends to help them decide which wine to buy. This understanding is even more important when we recall that despite certain superficial similarities between Gen-Y cohorts across a range of countries, it is nonetheless very difficult to generalize about the wine consumption behavior of Gen-Y, as there are distinct cultural and market developmental differences that are much more significant than the age-related similarities according to Mueller and Charters (2011). As novice or potential wine consumers, they are becoming increasingly significant targets for wine marketers (Mueller and Charters, 2011), especially because they are very adept at using relatively new technologies such as smartphones. By providing a reliable location anywhere, anytime, mobile devices offer companies the opportunity to conduct marketing campaigns that aim to drive both traffic and sales, whether mobile or in-store, in ever expanding ways. Applications and Web-apps (especially when they are free) have become a way of life for consumers. Such apps and Web-apps are thus in a position to address each stage of the consumer purchase funnel - awareness, engagement, consideration, conversion and loyalty. Smartphone penetration could definitely favor the surge in mobile shopping for wine. Owners of such devices are more or less connected to social media with embedded location-based applications. Technology can help brands to locate their fans and target their mobile strategy to quickly interact with shoppers. Today, someone walking on a street can receive an SMS on her/his mobile phone, indicating that the restaurant of the X brand in the same street is offering her/him a free aperitif if he/she turns up within the next hour. The person who receives this alert is a fan of the X brand on a social media, such as Facebook. He/she may have left the location-based application of her/his favorite digital social network activated to locate her/him anywhere anytime. At the same time, the restaurant has learned that this person is also a fan of a particular wine and has friends who also seem to be fans of the food they cook and the wine they serve, so invitations are sent to these people as well. Schummer (2001) has shown that 90 per cent of shoppers prefer to communicate in real time while shopping. Consumer recommendations are the most trusted form of advertising around the world. Although this is true, it is also a double-edged sword, with consumers trashing a restaurant or store for personal reasons, thereby causing much damage to a store's reputation. Over 75 per cent of respondents from 47 markets across the world rated recommendations from consumers as a trusted form of advertising, compared to 63 per cent for newspapers, 56 per cent for TV and magazines and 34 per cent for search engine advertisements (Nielsen, 2007). This underscores the notion that people trust people with similar experiences. Thanks to their ease of use, social media can help shoppers provide positive ratings and recommendations about the goods of a vendor on its public social media's page/apps/Web-apps, an action which would ideally be rewarded. Any exploration of m-commerce marketing must thoroughly investigate how consumers position themselves on these aspects of the interaction with the brand. Research method The confirmatory study was implemented on the Internet using a questionnaire. Each participant has visited and completed it. By asking French and international consumers coming from > 10 countries how they perceive social media generally and m-commerce using social media, we obtained the following results. Participants The sample is composed of 190 valid questionnaires completed by international Gen-Y consumers born after 1980. It was selected to achieve a balance in terms of age and socio-professional background. We worked mainly with students (60 per cent) for this analysis because they are deemed suitable as a sample. Although the use of students has often been questioned in terms of their appropriateness, they share many characteristics with the profile of mobile and Internet users' population, such as age. As shown by several studies, Internet users tend to be young adults, while the Internet usage penetration within the age groups of 18-29 years reaches 95 per cent (Zickuhr, 2010; Pew Research Center, 2010). Hence, although our sample presents a bias towards younger subjects, it can arguably be acceptable as representative of Internet and mobile users. In addition, Gen-Y is an important group of online consumers (Delafrooz et al., 2010) and is useful as a sample for empirical studies in m-commerce, in line with previous research (Kim et al., 2008). Hypotheses Based on the assumption that wine purchase (irrespective of the distribution channel: physical store, e-commerce or m-commerce) should be positively influenced by their use of Internet, mobile devices and their relations with social media because m-commerce has removed all the barriers to get wine anywhere (Kakihara and Sorensen, 2001). Their age and income should also influence positively their frequency of purchase because of the potential rise of interest in wine and the money available to do so. Finally, frequency of wine purchase should also be positively influenced by wine purchase and consumption characteristics. We therefore pose the following hypothesis (Table I). Profile of the sample The table presented in Appendix summarizes the profile of the Gen-Y consumers who were surveyed. Among the 190 respondents, 68 per cent of them connect to the Internet using their mobile and 38 per cent spend between 10 and 19 hours a week on the Internet, whereas 41 per cent spend more than 20 hours a week. Nearly all of them (95 per cent) access the Internet every day. Only 43 per cent purchase on the Internet at least once a month. They mainly use the Internet to stay in touch with friends and relatives (95 per cent), but also to look for information on a product (69 per cent). They also access discussion groups (29 per cent) and chat rooms (21 per cent). In terms of social media, nearly all use Facebook (93.2 per cent), then Twitter (32.6 per cent), then Groon (18.9 per cent) and then Google+ (15.3 per cent). They are on social media mainly to visit friends' pages (88.4 per cent) and have been in touch with social media through their friends (90.5 per cent) and through their classmates (50 per cent). When they are connected, most of them spend one hour or more on social media (67.9 per cent). Most of them connect every day (77.4 per cent) to social media - 26.8 per cent declared they have already bought a good or a service based on friends' recommendations issued from social media (of importance: clothes, transport ticket, CD and then books). Regarding wine, only 5 of them were influenced by friends' recommendations. Regarding the wine consumption and habits, 7.4 per cent of our respondents are members of a group dedicated to wine. Wine purchases range from "don't buy wine" (32.6 per cent) to 36.8 per cent "buy wine several times a year". With regards to wine buyers only, they mainly buy it in supermarkets (56.3 per cent) and hypermarkets (19.5 per cent), but also in wine shops (19.5 per cent). They prefer to consume good wines: of the total 22.7 per cent buy wine below 5 , but 43.8 per cent buy wine between 5 and 10 ; 26 per cent between 11 and 20 ; and 17 per cent spend > 20 . They mainly buy (moderately) between half a bottle (32.8 per cent) and one bottle (23.4 per cent). Among the respondents purchasing whole bottles, 27.4 per cent drink one glass of wine every day; 3.2 per cent drink half a bottle; whereas 64.2 per cent consume occasionally (not every day). They mainly consume wine at home (26.3 per cent) or at a friend's place (32.1 per cent). Most of them never go to the Internet to look for information on wine (82.1 per cent). Of the total, 5.8 per cent were able to quote one Web site only. For 72.6 per cent of them, on-time delivery builds trust when they are buying wine through their mobile phone and is viewed as an important feature. The fact that the wine is delivered in good condition is an important aspect, with 76 per cent of respondents expecting it. The Gen-Y consumers (56/190) who frequently buy wine (several times a month or more) consider the usefulness of the information regarding a specific wine (3.2/5) as important when looking up information through their mobile. This generation also rates highly: their overall satisfaction regarding their ability to use a mobile (4/5); their learning about the use of a mobile (3.8/5); positive experiences linked to the use of the mobile (3.4/5); and their level of happiness at being able to use a mobile (3.4/5). Out of those 29 per cent who do often buy wine (56/190), 36 use their mobile to go to social media and 34 look for information on the Internet. On the other hand, only 10 of them purchase products through the Internet. Furthermore, 25 of them spend more than 1 hour per day on their mobile and 17 spend more than 2 hours a month (but it might mean anything because they are students with a lot of time to spare). Let us also note that 52 of them access the Internet daily, 38 look for information on the Internet and 22 spend more than 20 hours a week. In terms of projection, if they had to buy some wine through their mobile, they respectively would consider the following factors as important: on-time delivery (3.6); cost of transportation (3.6); delivery in good condition (3.5); and follow-up to the transaction (3.5). Results The sample of 190 respondents was cut into two groups: One composed of Gen-Y consumers who do not purchase alcohol at all (n = 62). The other (n = 128) composed of Gen-Y consumers who buy wine frequently. Only the group that had purchased alcohol was analysed. A linear regression was run with the frequency of wine purchase as dependent variable (Y) and the 15 independent variables (X) listed in Table I. Table III shows that X1, X2, X5, X8, X12 and X15 positively influence the frequency of wine purchase. It can also be observed that variables X7, X9 and X14 exert nearly no impact on the frequency of wine purchase (b coefficients are close to 0). And finally, X3, X4, X6 and X13 have a negative coefficient. Based on our hypotheses, the b coefficients should be positive but also significant (p < 0.05). Variables X1 and X2 are significant and positive, whereas X3 and X7 are negative but significant (see bold in Table II). We therefore conclude that only H1 and H2 are valid: H1. The frequency of wine purchase is positively influenced by the Gen-Y consumers who access the Internet through their mobile phone. This validation is very important because Gen-Ys are getting more and more mobile-friendly and can get easy access to the Internet with their mobile. H2. The frequency of wine purchase is positively influenced by the average hours spent by Gen-Y consumers on the Internet per week. Note that H7 and H3 are significant but have a negative b coefficient: H3. The frequency of wine purchase is negatively influenced by the frequency of purchase on the Internet (irrespective of the product category). This result needs further research because it may be due to the composition of the sample (majority of students) who does not earn money, and if they do buy goods or services, may not have enough remaining cash for buying wines. H7. The frequency of wine purchase is negatively influenced by the frequency of connection to social media, but with a small b coefficient (which limits the impact of this variable). Table III shows the importance of three factors when buying wine through the Internet. Quickness and reduced cost contribute to promote overall satisfaction and develop underlying loyalty, which is not surprising because e-commerce and m-commerce differentiate themselves from physical distribution in terms of quickness, cost reduction and therefore efficiency. Table IV on the other hand is showing the importance of specific factors when Gen-Y consumers are looking for information on wine online. In short, they want a lot of useful information which is rich in terms of content. Table V shows that if a wine purchase was planned through a mobile, cost and follow-up are not that important compared to delivery in good condition or delivery on time. Finally, if a comparison is made between wine buyers and non-buyers (see Table VI), we can observe that wine buyers are likelier to be male, older, foreign and wealthier. They are less likely to own a smartphone. Wine buyers also spend less time on their mobile and Internet, but they tend to buy more online. Their purchases are more influenced by their friends' recommendations. Finally, wine buyers are more likely to consume wine daily. As novice or potential wine consumers, members of Gen-Y are becoming increasingly significant targets for wine marketers (Mueller and Charters, 2011). This paper looks at the current state of m-commerce, the consumption of smartphones and social media and the transformation of the consumer into an omnichannel shopper. It also examines some responses to this emerging way of doing shopping which enriches the in-store experience with digital integrations, especially in relation to Gen-Y consumers. Armed with smartphones and tablets, wine shoppers go back and forth effortlessly between the real (whether in hypermarkets or supermarkets, convenience stores, wine shops, at the estate/winery or during a wine fair) and digital worlds (through the Internet, mobile app, wine club or by mail order). They are using their phones while in stores to research products and compare prices, and they order online and then pick up in person. At the same time, they consult friends near and far whenever they may find themselves contemplating a purchase, such as a nice bottle of wine. Every day, more of them come to expect a mobile or an "omnichannel" experience. Nearly all retailers aspire to create this type of experience for their customers because creating a consistent customer experience across channels is very important for their business. Indeed, most aspects of such an omnichannel strategy appear inherently important. Results from our confirmatory study show that the frequency of wine purchase is positively influenced by Gen-Y consumers who access the Internet through their mobile phone. This result is consistent with the Adobe (2013) study which showed that 94 per cent of 18-29-year-old consumers had accessed social media using a mobile device. Therefore, designing m-commerce Web sites and applications that stimulate the feeling of being in a real market or in a real store can give more confidence to users who are engaged in carrying out commercial transactions, especially those based on social media-derived peer advice. Current and potential purchasers emphasized the following three factors when buying wine through the Internet: Speed of delivery. Cost reduction. Efficiency. These factors should be taken into account when developing interfaces. In line with Atkin and Thach (2012) and Barber et al. (2008), we recommend that wine marketers take actions to help reduce perceived risks in wine purchasing, in particular, a specific focus on the delivery of wine in good condition or the delivery on time if a wine purchase was planned through a mobile. In addition to doing commerce, wine growers will need to develop entirely new and effective ways for people to share and communicate information especially because H2 shows the frequency of wine purchase is positively influenced by the average hours spent by Gen-Y consumers on the Internet per week. In this regard, Facebook (> 90 per cent of the surveyed Gen-Y members are members) or Twitter could contribute to improving sales. In addition, wine growers will have to find new partners to reach a wider audience of potential users. It is possible to imagine vineyards partnering with restaurants, retail chains, cooking enthusiasts, etc. Also important will be the need to simplify, optimize and streamline the design of the digital marketing strategy to create new experiences and behaviors. Authors found in previous studies that Gen-Y consumers drink more wine than Gen-X did at a similar age and behavioral differences between the two age cohorts are likely to be considerably smaller than differentials in their attitudes. Gen-Y is not a homogeneous cohort according to Mueller et al. (2011) and de Magistris et al. (2011). To better target young wine buyers, wine growers should have in mind that wine buyers are likelier to be male, older, foreign and have more money, although they are less likely to have the latest smartphone. Furthermore, they spend less time on their mobile and Internet (compared to non-wine buyers), but they tend to buy more online and are likelier to make purchase decisions based on friends' recommendation. As Kakihara and Sorensen (2001) underlined, the mobile Internet has unique strengths because users can connect to it wherever and whenever they want. This topic is promising because France AgriMer (2012) shows that wine consumption is composed of 45 per cent of occasional drinkers (once or twice a week) for consumers below the age of 25, and between 25 and 34 years, 50 per cent are occasional drinkers. The Wine Market Council has also published data showing that the millennials (Gen-Y) were consuming 24 per cent of the total volume. Finally, based on AC Nielsen (2011) in March, Internet and mail-order are representing 10 per cent of the total sales in UK. Those data show that the Internet is more and more important, that Gen-Y members are occasional drinkers, and therefore to link those occasional drinkers with wine, it is important to develop pleasant platforms and interactive social networks. The advantage of providing good service on an m-commerce Web site translates into satisfied customers who will become brand advocates: they can refer other people to the wine grower. Internet consumers talk to other consumers about a good customer service experience. For a service, such as the purchase of wine from a particular wine grower who is already selling online and thinking of expanding into m-commerce, providing good customer service is a must. When customers tell other people about a bad experience, they do it on social networks to reach a large audience. This is why social networks must also be taken into account when planning an m-commerce strategy to avoid any negative buzz and therefore leverage a good e-reputation. Indeed, wine growers now must navigate a very complex customer service landscape, and the stakes are high. Consumers can be quick to punish those who are slow to respond to their questions or fail to deliver their purchases on time, for example. Growers are also vulnerable to reputation damage if they offer unsatisfactory answers or ignore their customers altogether, behavior that is likely to make the customer mad and eager to exact revenge by writing negative reviews. Wine growers who stumble stand to lose business to rivals and become vulnerable to negative social buzz. Customer service is one of the few remaining sources of competitive differentiation for this type of business, especially in relation to Gen-Y consumers. Wine growers that understand how to use the latest digital channels to reach new customers and help them find, buy and consume the products they sell will be rewarded with loyal customers willing to spend more and advocate for the brand. Customers are loyal to businesses that provide good customer service in the brick and mortar world; this is not expected to change in the m-commerce environment. Further research should be carried on with a larger sample, and a specific look at the rejected H3, which showed that the frequency of wine purchase, is negatively influenced by the frequency of purchase on the Internet. Finally, further studies should focus on international comparisons because it is very hard to generalize about the wine consumption behavior of the cohort across a range of countries, as there are distinct cultural and market developmental differences that are much more significant than the age-related similarities (Mueller and Charters (2011)). Opens in a new window. Table I. Variables tested by linear regression and formulation of the hypotheses Opens in a new window. Table II. Linear regression for 128 respondents (who do buy) with frequency of wine purchase as dependent variable Y Opens in a new window. Table III. Necessary factors for Gen-Y consumers to develop satisfaction and loyalty when buying wine online Opens in a new window. Table IV. Necessary factors for Gen-Y consumers to develop satisfaction and loyalty when they are looking for information on wine online Opens in a new window. Table V. Importance of the following characteristics if a purchase was planned through your mobile Opens in a new window. Table VI. Importance of the following characteristics if a purchase was planned through your mobile Opens in a new window. Table AI. Characteristics of the sample of respondents
|
Results of the study show that m-commerce and e-commerce applications using social media platforms may prove greatly efficient in the foreseeable future for wine growers. Different ways of selling and disseminating information about their products are observed.
|
[SECTION: Value] "Omnichannel" has become a buzzword in retail for good reason. New technologies, such as mobile devices and social media, combined with better data, bring the long-time dream of a unified cross-channel shopping experience within reach. In practice, however, most retailers still fall short of achieving this vision, especially as it applies to Generation Y ("Gen-Y" - those people born between 1980 and 1991) and their use of smartphones. Mobile users are increasingly accessing social media using mobile devices, whether via browsers or apps. A study by Adobe (2013) among mobile users in the USA, Canada, UK, France and Germany found that most had accessed social networks using a mobile device, ranging from 94 per cent for those 18-29 years of age to 75 per cent of those 50-64 years of age. In fact, Facebook was the second most visited Web site/application that was accessed by smartphones and was the top smartphone app in the USA in August 2013 (ComScore, 2013). In countries, such as Italy and Germany, penetration rates of mobile phones exceed 100 per cent, with some consumers owning more than one mobile phone (Kaplan, 2012). Mobile phones and devices are increasingly used in conducting mobile commerce (Venkatesh et al., 2003, Ngai and Gunasekaran, 2007). Reputation is one of the reasons why customers rely on particular Web sites, apps or Web-apps, and some studies have investigated trust and reputation issues in a mobile ad hoc network environment (Lax and Sarne, 2008). Newman (2010) found that some 700,000 people view wine-related videos every month; there are over 7,000 wine tweets per day and > 300 iPhone apps for wine. Breslin (2013) estimated that 90 per cent of wine drinkers use Facebook 6.2 hours per week. The wine industry is not excluded from this phenomenon; this is the problem we focus on in this study. The state of the art in the literature regarding wine industry during the past decade has seen an increasing series of studies of wine consumption and behavior (Celhay, 2008; Nowak et al., 2006; Ritchie et al., 2009; Thach and Olsen, 2004, 2006), but none of them has so far been interested in exploring the potential of mobile commerce (m-commerce) as it relates to the wine industry. Among the elements that contribute to this massive use of Internet devices, let us cite the growth in social media usage in recent years (Facebook, Twitter, and so on). Mobile social media is defined as "a group of mobile marketing applications that allow the creation and exchange of user-generated content" (Kaplan, 2012), such that a geo-social service like "Facebook Check-In" is a subset of mobile marketing. The instantaneous insights on consumer profiles, m-commerce and the development of mobile tools (smartphones, tablets and watches like the Samsung Gear) reinforce the appetite consumers have for this "Communicating Swiss-knife" style, which is also sometimes considered "habit forming". Consumers today are not passive shoppers. They seek out and disseminate information about a product using social media. They are not shy about providing their views and opinions, and they will abandon any company that does not serve their needs or is perceived as not providing value. Our question, therefore, relates to the coming of age of Generation Y. Because so many decisions related to wine are made on the move - in restaurants, at airports, in retail stores, in discussions with friends or during conferences - the smartphone or tablet people very often have in their hands can provide fast information on which wine to buy, although the subsequent information search related to a bottle of wine may or may not lead to purchase. Almost everyone in the developed world today uses a smartphone. Because Gen-Y is the most significant demographic cohort since the Baby Boomers and because this population is tech-savvy, it represents a significant market for wine marketers. This paper attempts to study and understand the wine shopping habits and smartphone use in Gen Y. In addition, this paper aims to provide insights on the perception of Gen-Y customers, their use of social media and the potential challenges and opportunities in terms of shopping behavior. We consider a potential m-commerce strategy and how it applies to the wine industry. In this article, we use the term m-commerce to refer to smartphones, social media, applications (or apps) and Web applications, despite the differences in usage patterns. We can access Facebook or any social media or blogs, Twitter, etc., from our smartphones, but we might then use apps/Web apps[1] to access, for example, eRobertParker, Wine Enthusiast or Wine Searcher, etc., or an online store or a restaurant wine list. Our research question is: "To what extent is consumer behavior shaped by the use of mobile devices and what are the Gen-Y cohort's expectations when they consider purchasing wine?" Until now, little has been written on the subject of smartphones and wine purchasing patterns. Such a study would provide academics and the wine industry with useful insights. Following Lockshin and Corsi's (2012) suggestions we strive to go deeper in one of the under-researched areas: determinants of online and social media influences on consumers. Our paper will therefore try to bring new insights into this growing market share represented by Gen-Y to enable wine professionals to understand and target them with even more acuity. To understand what wine marketers can do to leverage sales using mobile devices and social media, an empirical study based on a questionnaire was specifically designed with Gen-Y users in mind. Our research enabled us to highlight the importance of the effects of the design interface used on m-commerce Web sites on consumers' buying intention. It also shows that wine growers will need to adapt, respond and grow their digital business to suit their customers, who live in a state of constant interconnectedness. Growers must design strategies to fit the mobile contexts to sell more on these types of devices. We present the results of this study before concluding and presenting limits as well as future avenues of research. Who is a member of Gen-Y? Gen-Y is defined as those born between 1980 and 1991 (Chowdhury and Coulter, 2006; Lescohier, 2006). The literature in relation to this cohort is quite abundant as was identified by Lockshin and Corsi (2012). A number of articles compare wine preferences and/or behavior between the younger and older generations. For a review, the reader is invited to refer to articles by: Agnoli et al. (2011), Ritchie (2011), Fountain and Lamb (2011), de Magistris et al. (2011), Charters et al. (2011), Mueller et al. (2011), Qenani-Petrela et al. (2007) and Wolf et al. (2005). It is believed that Gen-Y has the potential of forming long-term loyalties with products which satisfy them at this vital stage (Paul, 2001). In 2014, Gen-Y, the largest generation since the Baby Boomers (Neuborne and Kerwin, 1999), range in age from 14 to 32 (Norum, 2008). For the purposes of this study, the age range of Gen-Y has been narrowed to ages 18 to 32 because, according to Djamasbi et al. (2008), increased financial independence is gained at the age of 18 and alcohol consumption is prohibited before the age of 18. Gen-Y grew up in the digital world. It is technically literate and continually wired, plugged in and connected to digitally streaming information, entertainment and contacts. It has mastered the technology in such a way that multitasking is a habit that carries into the workplace, where it tends to instant message its contacts while working (Lewis, 2003). Although studies into the general characteristics of this market are limited and none have been conducted related to wine consumption, most findings consistently highlight similar attributes for psychographics and purchase behavior across well- researched industries, such as fashion-related items (Dias, 2003; Paul, 2001) or more generally industries related to these five meaningful and distinct decision-making groups, namely, "recreational quality seekers", "recreational discount seekers", "trend setting loyals", "shopping and fashion uninterested" and "confused time/money conserving" (see Bakewell and Mitchell, 2003). Regarding the wine industry, attitude to wine and consumption behavior may vary, according to Fountain and Lamb (2011). Authors found that Gen-Y consumers drink more wine than Gen-X did at a similar age. However, behavioral differences between the two age cohorts are likely to be considerably smaller than differentials in their attitudes. Gen-Y is not a homogeneous cohort, according to Mueller et al. (2011) and de Magistris et al. (2011). By comparing views and behaviors of Gen-Y wine consumers in a number of countries, authors point to substantial cultural differences. Despite certain similarities between Gen-Y cohorts across a range of countries, it is very hard to generalize about the wine consumption behavior of the cohort across a range of countries (Mueller and Charter, 2011). M-commerce habits M-commerce, according to the Mobile Marketing Association (2013), is the one- or two-way exchange of value facilitated by a mobile consumer electronic device (e.g. mobile handset or smartphone), which is enabled by wireless technologies and communication networks. M-commerce, when used in conjunction with social media, shows considerable potential for generating sustainable profits; however, further demonstration is required. Articles about the use and effects of social media on consumer wine preference and behavior exist as identified by Lockshin and Corsi (2012); for a review, see works of Reyneke et al. (2011), Claster et al. (2010), Pitt et al. (2011) and Nicholls (2012). Unfortunately, due to the infancy of this topic, analysis of the combined use of smartphones and social media still seems almost nonexistent. To succeed in producing profits in the highly competitive m-commerce environment, organizations must not only develop the technical expertise and business strategy necessary for the creation of an effective "Social media-m-commerce website" tandem/partnership but also create a satisfying user experience (Pelet and Lecat, 2011; Lecat and Pelet, 2011). Today's competitive environment companies must evolve quickly and understand that the consumer shopping in a digital world is not bound by the old buying habits. A new kind of collaboration between sellers and buyers needs to happen. The speed of evolution of the Internet does not always allow companies to realize the implications of new consumer behavior trends or understand the value they expect to realize from investing in digital technologies. Indeed, the digital world has modified the way brands look at customers. Recent figures show an impressive growth of mobile use. Cell phone adoption is especially widespread in young populations, with 75 per cent of teenagers and 93 per cent of adults aged 18-29 years having a cell phone, with 55 per cent of the latter group accessing the Internet wirelessly from their cell phone (Lenhart et al., 2011). The evolution of mobile communications has triggered an increase in the use of mobile devices, such as mobile phones, to conduct m-commerce on the mobile Web (Venkatesh et al., 2003; Ngai and Gunasekaran, 2007). To support their digital business, many companies are turning to digital marketing. However, creating a well-conceived and executed digital marketing strategy requires clarity of concept and goals that few companies are willing or able to invest in. The value of digital marketing is, as yet, unclear because many consumers do not wish to be exposed to advertising on their smartphones. This is the challenge currently being faced by such giants as Google, Facebook and Amazon. Buying a product and sharing information in real time are two inherent aspects of the new consumption experience. Purchasing behaviors are increasingly affected by the way we use the Web to make purchasing decisions and how we influence, test, protest, praise, share, recommend or learn about products or services. As more consumers access social media via mobile devices, it changes the way they research and shop for products and services. M-commerce has not surpassed e-commerce yet, but the rapid growth of mobile telephony has fueled the expansion of the mobile Internet as a foundation for mobile commerce (Lee and Benbasat, 2004). The mobile Internet has unique strengths because users can connect to it wherever and whenever they want (Kakihara and Sorensen, 2001). The ability of the wine industry to attract a new customer base (Thach, 2005) requires a thorough understanding of the needs and wants of that group (Kotler, 2003). Given that there are > 10,000 wine labels on the market, buying wine can be a confusing experience, so wine buyers will rely on the advice of experts and friends to help them decide which wine to buy. This understanding is even more important when we recall that despite certain superficial similarities between Gen-Y cohorts across a range of countries, it is nonetheless very difficult to generalize about the wine consumption behavior of Gen-Y, as there are distinct cultural and market developmental differences that are much more significant than the age-related similarities according to Mueller and Charters (2011). As novice or potential wine consumers, they are becoming increasingly significant targets for wine marketers (Mueller and Charters, 2011), especially because they are very adept at using relatively new technologies such as smartphones. By providing a reliable location anywhere, anytime, mobile devices offer companies the opportunity to conduct marketing campaigns that aim to drive both traffic and sales, whether mobile or in-store, in ever expanding ways. Applications and Web-apps (especially when they are free) have become a way of life for consumers. Such apps and Web-apps are thus in a position to address each stage of the consumer purchase funnel - awareness, engagement, consideration, conversion and loyalty. Smartphone penetration could definitely favor the surge in mobile shopping for wine. Owners of such devices are more or less connected to social media with embedded location-based applications. Technology can help brands to locate their fans and target their mobile strategy to quickly interact with shoppers. Today, someone walking on a street can receive an SMS on her/his mobile phone, indicating that the restaurant of the X brand in the same street is offering her/him a free aperitif if he/she turns up within the next hour. The person who receives this alert is a fan of the X brand on a social media, such as Facebook. He/she may have left the location-based application of her/his favorite digital social network activated to locate her/him anywhere anytime. At the same time, the restaurant has learned that this person is also a fan of a particular wine and has friends who also seem to be fans of the food they cook and the wine they serve, so invitations are sent to these people as well. Schummer (2001) has shown that 90 per cent of shoppers prefer to communicate in real time while shopping. Consumer recommendations are the most trusted form of advertising around the world. Although this is true, it is also a double-edged sword, with consumers trashing a restaurant or store for personal reasons, thereby causing much damage to a store's reputation. Over 75 per cent of respondents from 47 markets across the world rated recommendations from consumers as a trusted form of advertising, compared to 63 per cent for newspapers, 56 per cent for TV and magazines and 34 per cent for search engine advertisements (Nielsen, 2007). This underscores the notion that people trust people with similar experiences. Thanks to their ease of use, social media can help shoppers provide positive ratings and recommendations about the goods of a vendor on its public social media's page/apps/Web-apps, an action which would ideally be rewarded. Any exploration of m-commerce marketing must thoroughly investigate how consumers position themselves on these aspects of the interaction with the brand. Research method The confirmatory study was implemented on the Internet using a questionnaire. Each participant has visited and completed it. By asking French and international consumers coming from > 10 countries how they perceive social media generally and m-commerce using social media, we obtained the following results. Participants The sample is composed of 190 valid questionnaires completed by international Gen-Y consumers born after 1980. It was selected to achieve a balance in terms of age and socio-professional background. We worked mainly with students (60 per cent) for this analysis because they are deemed suitable as a sample. Although the use of students has often been questioned in terms of their appropriateness, they share many characteristics with the profile of mobile and Internet users' population, such as age. As shown by several studies, Internet users tend to be young adults, while the Internet usage penetration within the age groups of 18-29 years reaches 95 per cent (Zickuhr, 2010; Pew Research Center, 2010). Hence, although our sample presents a bias towards younger subjects, it can arguably be acceptable as representative of Internet and mobile users. In addition, Gen-Y is an important group of online consumers (Delafrooz et al., 2010) and is useful as a sample for empirical studies in m-commerce, in line with previous research (Kim et al., 2008). Hypotheses Based on the assumption that wine purchase (irrespective of the distribution channel: physical store, e-commerce or m-commerce) should be positively influenced by their use of Internet, mobile devices and their relations with social media because m-commerce has removed all the barriers to get wine anywhere (Kakihara and Sorensen, 2001). Their age and income should also influence positively their frequency of purchase because of the potential rise of interest in wine and the money available to do so. Finally, frequency of wine purchase should also be positively influenced by wine purchase and consumption characteristics. We therefore pose the following hypothesis (Table I). Profile of the sample The table presented in Appendix summarizes the profile of the Gen-Y consumers who were surveyed. Among the 190 respondents, 68 per cent of them connect to the Internet using their mobile and 38 per cent spend between 10 and 19 hours a week on the Internet, whereas 41 per cent spend more than 20 hours a week. Nearly all of them (95 per cent) access the Internet every day. Only 43 per cent purchase on the Internet at least once a month. They mainly use the Internet to stay in touch with friends and relatives (95 per cent), but also to look for information on a product (69 per cent). They also access discussion groups (29 per cent) and chat rooms (21 per cent). In terms of social media, nearly all use Facebook (93.2 per cent), then Twitter (32.6 per cent), then Groon (18.9 per cent) and then Google+ (15.3 per cent). They are on social media mainly to visit friends' pages (88.4 per cent) and have been in touch with social media through their friends (90.5 per cent) and through their classmates (50 per cent). When they are connected, most of them spend one hour or more on social media (67.9 per cent). Most of them connect every day (77.4 per cent) to social media - 26.8 per cent declared they have already bought a good or a service based on friends' recommendations issued from social media (of importance: clothes, transport ticket, CD and then books). Regarding wine, only 5 of them were influenced by friends' recommendations. Regarding the wine consumption and habits, 7.4 per cent of our respondents are members of a group dedicated to wine. Wine purchases range from "don't buy wine" (32.6 per cent) to 36.8 per cent "buy wine several times a year". With regards to wine buyers only, they mainly buy it in supermarkets (56.3 per cent) and hypermarkets (19.5 per cent), but also in wine shops (19.5 per cent). They prefer to consume good wines: of the total 22.7 per cent buy wine below 5 , but 43.8 per cent buy wine between 5 and 10 ; 26 per cent between 11 and 20 ; and 17 per cent spend > 20 . They mainly buy (moderately) between half a bottle (32.8 per cent) and one bottle (23.4 per cent). Among the respondents purchasing whole bottles, 27.4 per cent drink one glass of wine every day; 3.2 per cent drink half a bottle; whereas 64.2 per cent consume occasionally (not every day). They mainly consume wine at home (26.3 per cent) or at a friend's place (32.1 per cent). Most of them never go to the Internet to look for information on wine (82.1 per cent). Of the total, 5.8 per cent were able to quote one Web site only. For 72.6 per cent of them, on-time delivery builds trust when they are buying wine through their mobile phone and is viewed as an important feature. The fact that the wine is delivered in good condition is an important aspect, with 76 per cent of respondents expecting it. The Gen-Y consumers (56/190) who frequently buy wine (several times a month or more) consider the usefulness of the information regarding a specific wine (3.2/5) as important when looking up information through their mobile. This generation also rates highly: their overall satisfaction regarding their ability to use a mobile (4/5); their learning about the use of a mobile (3.8/5); positive experiences linked to the use of the mobile (3.4/5); and their level of happiness at being able to use a mobile (3.4/5). Out of those 29 per cent who do often buy wine (56/190), 36 use their mobile to go to social media and 34 look for information on the Internet. On the other hand, only 10 of them purchase products through the Internet. Furthermore, 25 of them spend more than 1 hour per day on their mobile and 17 spend more than 2 hours a month (but it might mean anything because they are students with a lot of time to spare). Let us also note that 52 of them access the Internet daily, 38 look for information on the Internet and 22 spend more than 20 hours a week. In terms of projection, if they had to buy some wine through their mobile, they respectively would consider the following factors as important: on-time delivery (3.6); cost of transportation (3.6); delivery in good condition (3.5); and follow-up to the transaction (3.5). Results The sample of 190 respondents was cut into two groups: One composed of Gen-Y consumers who do not purchase alcohol at all (n = 62). The other (n = 128) composed of Gen-Y consumers who buy wine frequently. Only the group that had purchased alcohol was analysed. A linear regression was run with the frequency of wine purchase as dependent variable (Y) and the 15 independent variables (X) listed in Table I. Table III shows that X1, X2, X5, X8, X12 and X15 positively influence the frequency of wine purchase. It can also be observed that variables X7, X9 and X14 exert nearly no impact on the frequency of wine purchase (b coefficients are close to 0). And finally, X3, X4, X6 and X13 have a negative coefficient. Based on our hypotheses, the b coefficients should be positive but also significant (p < 0.05). Variables X1 and X2 are significant and positive, whereas X3 and X7 are negative but significant (see bold in Table II). We therefore conclude that only H1 and H2 are valid: H1. The frequency of wine purchase is positively influenced by the Gen-Y consumers who access the Internet through their mobile phone. This validation is very important because Gen-Ys are getting more and more mobile-friendly and can get easy access to the Internet with their mobile. H2. The frequency of wine purchase is positively influenced by the average hours spent by Gen-Y consumers on the Internet per week. Note that H7 and H3 are significant but have a negative b coefficient: H3. The frequency of wine purchase is negatively influenced by the frequency of purchase on the Internet (irrespective of the product category). This result needs further research because it may be due to the composition of the sample (majority of students) who does not earn money, and if they do buy goods or services, may not have enough remaining cash for buying wines. H7. The frequency of wine purchase is negatively influenced by the frequency of connection to social media, but with a small b coefficient (which limits the impact of this variable). Table III shows the importance of three factors when buying wine through the Internet. Quickness and reduced cost contribute to promote overall satisfaction and develop underlying loyalty, which is not surprising because e-commerce and m-commerce differentiate themselves from physical distribution in terms of quickness, cost reduction and therefore efficiency. Table IV on the other hand is showing the importance of specific factors when Gen-Y consumers are looking for information on wine online. In short, they want a lot of useful information which is rich in terms of content. Table V shows that if a wine purchase was planned through a mobile, cost and follow-up are not that important compared to delivery in good condition or delivery on time. Finally, if a comparison is made between wine buyers and non-buyers (see Table VI), we can observe that wine buyers are likelier to be male, older, foreign and wealthier. They are less likely to own a smartphone. Wine buyers also spend less time on their mobile and Internet, but they tend to buy more online. Their purchases are more influenced by their friends' recommendations. Finally, wine buyers are more likely to consume wine daily. As novice or potential wine consumers, members of Gen-Y are becoming increasingly significant targets for wine marketers (Mueller and Charters, 2011). This paper looks at the current state of m-commerce, the consumption of smartphones and social media and the transformation of the consumer into an omnichannel shopper. It also examines some responses to this emerging way of doing shopping which enriches the in-store experience with digital integrations, especially in relation to Gen-Y consumers. Armed with smartphones and tablets, wine shoppers go back and forth effortlessly between the real (whether in hypermarkets or supermarkets, convenience stores, wine shops, at the estate/winery or during a wine fair) and digital worlds (through the Internet, mobile app, wine club or by mail order). They are using their phones while in stores to research products and compare prices, and they order online and then pick up in person. At the same time, they consult friends near and far whenever they may find themselves contemplating a purchase, such as a nice bottle of wine. Every day, more of them come to expect a mobile or an "omnichannel" experience. Nearly all retailers aspire to create this type of experience for their customers because creating a consistent customer experience across channels is very important for their business. Indeed, most aspects of such an omnichannel strategy appear inherently important. Results from our confirmatory study show that the frequency of wine purchase is positively influenced by Gen-Y consumers who access the Internet through their mobile phone. This result is consistent with the Adobe (2013) study which showed that 94 per cent of 18-29-year-old consumers had accessed social media using a mobile device. Therefore, designing m-commerce Web sites and applications that stimulate the feeling of being in a real market or in a real store can give more confidence to users who are engaged in carrying out commercial transactions, especially those based on social media-derived peer advice. Current and potential purchasers emphasized the following three factors when buying wine through the Internet: Speed of delivery. Cost reduction. Efficiency. These factors should be taken into account when developing interfaces. In line with Atkin and Thach (2012) and Barber et al. (2008), we recommend that wine marketers take actions to help reduce perceived risks in wine purchasing, in particular, a specific focus on the delivery of wine in good condition or the delivery on time if a wine purchase was planned through a mobile. In addition to doing commerce, wine growers will need to develop entirely new and effective ways for people to share and communicate information especially because H2 shows the frequency of wine purchase is positively influenced by the average hours spent by Gen-Y consumers on the Internet per week. In this regard, Facebook (> 90 per cent of the surveyed Gen-Y members are members) or Twitter could contribute to improving sales. In addition, wine growers will have to find new partners to reach a wider audience of potential users. It is possible to imagine vineyards partnering with restaurants, retail chains, cooking enthusiasts, etc. Also important will be the need to simplify, optimize and streamline the design of the digital marketing strategy to create new experiences and behaviors. Authors found in previous studies that Gen-Y consumers drink more wine than Gen-X did at a similar age and behavioral differences between the two age cohorts are likely to be considerably smaller than differentials in their attitudes. Gen-Y is not a homogeneous cohort according to Mueller et al. (2011) and de Magistris et al. (2011). To better target young wine buyers, wine growers should have in mind that wine buyers are likelier to be male, older, foreign and have more money, although they are less likely to have the latest smartphone. Furthermore, they spend less time on their mobile and Internet (compared to non-wine buyers), but they tend to buy more online and are likelier to make purchase decisions based on friends' recommendation. As Kakihara and Sorensen (2001) underlined, the mobile Internet has unique strengths because users can connect to it wherever and whenever they want. This topic is promising because France AgriMer (2012) shows that wine consumption is composed of 45 per cent of occasional drinkers (once or twice a week) for consumers below the age of 25, and between 25 and 34 years, 50 per cent are occasional drinkers. The Wine Market Council has also published data showing that the millennials (Gen-Y) were consuming 24 per cent of the total volume. Finally, based on AC Nielsen (2011) in March, Internet and mail-order are representing 10 per cent of the total sales in UK. Those data show that the Internet is more and more important, that Gen-Y members are occasional drinkers, and therefore to link those occasional drinkers with wine, it is important to develop pleasant platforms and interactive social networks. The advantage of providing good service on an m-commerce Web site translates into satisfied customers who will become brand advocates: they can refer other people to the wine grower. Internet consumers talk to other consumers about a good customer service experience. For a service, such as the purchase of wine from a particular wine grower who is already selling online and thinking of expanding into m-commerce, providing good customer service is a must. When customers tell other people about a bad experience, they do it on social networks to reach a large audience. This is why social networks must also be taken into account when planning an m-commerce strategy to avoid any negative buzz and therefore leverage a good e-reputation. Indeed, wine growers now must navigate a very complex customer service landscape, and the stakes are high. Consumers can be quick to punish those who are slow to respond to their questions or fail to deliver their purchases on time, for example. Growers are also vulnerable to reputation damage if they offer unsatisfactory answers or ignore their customers altogether, behavior that is likely to make the customer mad and eager to exact revenge by writing negative reviews. Wine growers who stumble stand to lose business to rivals and become vulnerable to negative social buzz. Customer service is one of the few remaining sources of competitive differentiation for this type of business, especially in relation to Gen-Y consumers. Wine growers that understand how to use the latest digital channels to reach new customers and help them find, buy and consume the products they sell will be rewarded with loyal customers willing to spend more and advocate for the brand. Customers are loyal to businesses that provide good customer service in the brick and mortar world; this is not expected to change in the m-commerce environment. Further research should be carried on with a larger sample, and a specific look at the rejected H3, which showed that the frequency of wine purchase, is negatively influenced by the frequency of purchase on the Internet. Finally, further studies should focus on international comparisons because it is very hard to generalize about the wine consumption behavior of the cohort across a range of countries, as there are distinct cultural and market developmental differences that are much more significant than the age-related similarities (Mueller and Charters (2011)). Opens in a new window. Table I. Variables tested by linear regression and formulation of the hypotheses Opens in a new window. Table II. Linear regression for 128 respondents (who do buy) with frequency of wine purchase as dependent variable Y Opens in a new window. Table III. Necessary factors for Gen-Y consumers to develop satisfaction and loyalty when buying wine online Opens in a new window. Table IV. Necessary factors for Gen-Y consumers to develop satisfaction and loyalty when they are looking for information on wine online Opens in a new window. Table V. Importance of the following characteristics if a purchase was planned through your mobile Opens in a new window. Table VI. Importance of the following characteristics if a purchase was planned through your mobile Opens in a new window. Table AI. Characteristics of the sample of respondents
|
Y-Gens are occasional drinkers and therefore, to link those occasional drinkers with wine, it is important to use social media as a communication tool and maybe a distribution channel to better reach this potential target.
|
[SECTION: Purpose] Music seeking and consumption are no longer confined by the boundaries of country, region or culture today (Lee et al., 2013). Music, as a cultural object, may be perceived differently by people from different cultural backgrounds, imposing a challenge on music digital libraries (MDL) in meeting the needs of a diverse audience (Weissenberger, 2015). Consequently, an increasing number of researchers have started to investigate various cross-cultural issues in the music information retrieval (MIR) and MDL fields. As people often seek music for emotional goals (Lavranos et al., 2015), music mood[1] has increasingly become a popular access point for music information in many MDL and online music services (Hu, 2010; Hu and Downie, 2007). This trend has raised questions regarding the applicability of music mood across cultural boundaries. Probably due to its subjective and context-based nature, music mood perception is often regarded as culture-dependent (Wong et al., 2009). A number of previous studies have compared mood perceptions of music in various cultures by listeners from different cultural backgrounds (e.g. Balkwill and Thompson, 1999; Fritz et al., 2009; Hu and Lee, 2012; Lee et al., 2013; Singhi and Brown, 2014; Egermann et al., 2014). Although specific findings vary, a general trend found in these existing studies is that listeners' perceptions of music mood can be influenced by their cultural backgrounds. However, most existing studies have focussed on western (classical or pop) music or ethnic music with small audiences (e.g. Congolese music produced and consumed by Congolese Pygmies (Egermann et al., 2014)). Such studies mainly aim to answer questions on music psychology and/or the social, developmental aspects of music. The purpose of this study is different; we are interested in exploring how music mood could be used as an access point or metadata type that can transcend the cultural boundaries of users in the context of MDL. To date, the question of how people with different cultural backgrounds would perceive the mood of Chinese music remains under-investigated. Considering its appeal to one of the largest music listener populations in the world (IFPI, 2014), Chinese music should not be left out of today's global scope of MDL/MIR studies. As Chinese music possesses unique characteristics that are remarkably different from western music (Fung, 2013), it is of theoretical and practical value to investigate the case of Chinese music. Therefore, in this study, we aim to fill this gap and explore how people with different cultural backgrounds, specifically those in the USA and Hong Kong (representing western and eastern listeners, respectively), perceive the mood of Chinese music, including traditional Chinese music and contemporary Chinese pop music that infuses western pop characteristics and local Chinese elements. We specifically answer the following research questions: RQ1. How differently or similarly do Hong Kong and US music listeners perceive the mood of Chinese music? RQ2. How differently or similarly do Hong Kong and US music listeners perceive the mood of different genres/styles of Chinese music? RQ3. Do the listeners' demographic characteristics and familiarity with the music affect their mood perceptions of Chinese music? Answers to these questions have important implications for MDL that use mood as an access point to facilitate MIR and discovery in a global environment. These answers will help determine whether it is necessary to develop different mood-related labels or tags that take their suitability for various user groups into consideration. Implications of the results in this study will be discussed. The goal of this study is to contribute to the literature of cross-cultural MDL/MIR and provide empirical evidence for designing cross-cultural MDL/MIR systems. 2.1 A brief history and types of Chinese music Chinese music broadly refers to music that uses traditional Chinese instruments or is in the Chinese language, and is the main music consumed in the region of Greater China (i.e. regions where Chinese is the official language, including mainland China, Hong Kong, Macau, Taiwan and Singapore). However, the consumption of Chinese music is not limited to that region. People around of the world are accessing Chinese music, and all major online music services, such as Spotify and Last.fm, contain Chinese music to cater to the demand. There is a broad range of genres and styles in Chinese music. Traditional Chinese folk music dates back to approximately 7,000 years ago, played mostly in the pentatonic scale with its own set of instruments. Contemporary Chinese pop music (also referred to as C-pop) was developed in the middle of the twentieth century with the influence of western music. The subgenre "Cantopop" (Hong Kong pop music sung in Cantonese) was developed in the 1970s, followed by the rise of another subgenre, "Mandopop" (Mandarin Chinese pop music), in the 1980s (Fung, 2013). This study includes six genres/styles of Chinese music that are widely appreciated in the region and around the world. The first two are traditional Chinese music, whereas the rest are the most popular C-pop styles featured in online music services in China (e.g. Baidu, Xiami, Kugou): Traditional Instrumental Folk Music of China (qiyue): on the pentatonic scale, musicians perform using traditional instrument(s) of woodwind and percussion, bowed strings, plucked and struck strings in solo, small ensembles or orchestra (Randel, 2003). Canto(nese) folk songs: Han Chinese folk songs are divided into northern and southern styles geographically. Provinces south of the Yangzi River have mild weather and abundant rain, and therefore the folk songs tend to be more "lyrical, gentle and conjunct" (Han, 1989). Near the Canton province and Pearl River Delta, Cantonese folk music is especially light-spirited (China Culture, 2006). Contemporary Chinese instrumental music: this style of music is also called "New Music of China." It mixes the techniques, instrumentation and composition of both Chinese traditional music and the European/western music (Liu and Mason, 2010). The Girls 12 is a famous mainland Chinese all-female group who plays this type of music with traditional Chinese instruments. Cantopop (or Cantonese popular music): this is a subgenre of C-pop. Sung in Cantonese, Cantopop usually refers to popular songs produced in Hong Kong. Blended with traditional Chinese elements, as well as western-influenced harmonies and melodies, Cantopop was a dominant music genre in Asia in the 1980s' and 1990s. Its iconic star, Anita Mui, was recognized as "Asia's Madonna." China wind (zhongguofeng): this is a subgenre of Mandopop and is also called "Chinese Style." It features traditional Chinese instruments and classical Chinese melody juxtaposed with popular western music styles and cultural references to Chinese history, classics and folklore in the lyrics (Xinhuanet, 2006). Taiwanese singer Jay Chou is regarded as the "godfather" of this type of music, bringing forth the wave of China Wind pop. Chinese rock: as another subgenre of Mandopop, Chinese rock is locally produced in mainland China and is the most notable sub-category of Mandopop to the world outside of China. Chinese rock combines the loud and forceful folk song traditions of Northern China and characteristics of western-style rock such as a fast tempo, a strong beat and aggressive bass lines (Fung, 2013). The singer/songwriter Cui Jian from mainland China has been widely recognized as the father of Chinese rock (Fung, 2013). It is noteworthy that all C-pop styles are, to varying extents, influenced by western pop or Rock music. Compared to Traditional Instrumental Chinese Folk Music and Canto Folk Songs, C-pop may be easier to understand and appreciate for listeners familiar with western music. In addition, because of the mixture of western music elements, listeners from both cultural groups in this study might share more common perceptions on C-pop than the other two genres/styles. Therefore, in addition to analyzing aggregated perceptions across all songs, this study also compares and contrasts listeners' perceptions on these six music styles. 2.2 Cross-cultural studies on music mood perceptions A number of studies have investigated the effects of culture on music mood perception. Although fundamental emotions such as happiness, sadness, anger or fear were found to be generally agreed upon across cultures, cultural traditions and background were found to affect individuals' music mood perception, in addition to the inherent qualities of the music (e.g. Gregory and Varney, 1996; Kosta et al., 2013; Lee et al., 2013). For example, Wong et al. (2009) compared mood descriptors used by European and Indian communities and discovered that cultural tradition played a more crucial role than the inherent qualities of the music in users' mood perception. Fritz et al. (2009) found that mood perceptions of western music by African listeners partially agreed with those by western listeners. In a similar vein, Hu and Lee (2012) discovered that discrepancies existed between Chinese and American music listeners in the way they determined the mood of western music pieces. More recently, Lee and Hu (2014) also compared perceptions of western music mood among American, Korean and Chinese listeners, and found that the results of Korean listeners' mood perceptions were situated in between the two other groups. There are also studies investigating the cultural backgrounds of listeners in relation to their mood perceptions of non-western music. Kosta et al. (2013) found greater agreement on the mood of Greek songs among Greek listeners compared to non-Greek listeners with varying backgrounds. Similar results were found by Hu et al. (2014) in comparing American and Korean listeners' mood perceptions of a set of Korean pop songs. Egermann et al. (2014) compared subjective and psychophysiological emotional responses by Canadians and Congolese Pygmies to both western and Congolese music. They found that the two groups had very different responses to Pygmies music but similar responses to some western music. To the best of our knowledge, there has been little work focussed on comparing the mood perceptions of various genres/styles of Chinese music across different cultural groups. In addition, many of the previous studies were conducted in the context of music psychology or ethnomusicology. This study, however, reflects the perspective of information organization and access, with the goal of finding out how music mood can be organized to best meet the information needs of users from different cultural groups. There are mainly two kinds of models to represent mood in music psychology and MIR: categorical and dimensional. The former uses a set of discrete terms (e.g. "passionate," "cheerful") to represent the mood of a piece of music. The most classical model of this kind is Hevner's (1936) adjective circle where eight mood categories placed in a circle, with a set of terms in each category. The dimensional models, in contrast, represent mood with continuous values in a low dimensional space. Different models may have different dimensions, yet valence (i.e. level of pleasure) and arousal (i.e. level of energy) are among the most popular dimensions. The dimensional model used very often in MIR is Russell's (1980) model. The categorical and dimensional models have their own pros and cons. Yang and Chen (2012) summarize that categorical models are more user-friendly as they only consist of terms in natural language; whereas dimensional models are advantageous in terms of quantifying the intensity of moods. Notwithstanding the importance of dimensional models, categorical ones are more suitable for this current study, for the purpose of validating and comparing the values of mood metadata in the context of cross-cultural MDLs. The models from music psychology have solid theoretical ground and can be borrowed into MIR studies. However, one criticism of adopting these models directly in MIR is that they were developed in experimental settings and thus might not reflect users' preferences and behaviors in today's music listening environment (Hu, 2010). Therefore, MIR researchers developed mood models based on domain analysis of widely used music repositories and streaming services. One of the influential mood models in MIR is the five-cluster model proposed in the audio music mood classification task (Hu et al., 2008) of Music Information Retrieval Evaluation eXchange (MIREX), a community-wide MIR evaluation campaign held annually since 2005 (Downie et al., 2014). This model was developed from more than 180 editorial mood labels on AllMusic.com, a very popular online music repository. It has been used in many music mood classification studies (e.g. Laurier et al., 2008; Bischoff et al., 2009; Yang and Chen, 2012) as well as cross-cultural user studies (Hu and Lee, 2012; Lee and Hu, 2014; Hu et al., 2014; Yang and Hu, 2012). The five mood clusters are: C1: passionate, rousing, confident, boisterous, rowdy; C2: rollicking, cheerful, fun, sweet, amiable/good natured; C3: literate, poignant, wistful, bittersweet, autumnal, brooding; C4: humorous, silly, campy, quirky, whimsical, witty, wry; and C5: aggressive, fiery, tense/anxious, intense, volatile, visceral. In this five-cluster model, each cluster is collectively represented by a set of mood terms that share similar semantics. As the terms are describing affect, which can be somewhat vague (as opposed to terms referring to concrete objects), it is argued that using multiple terms with similar meanings to collectively describe one kind of mood is clearer than single-word representations (Hu and Downie, 2007). A more recent categorical mood model was also used in MIREX, in the mood subtask of the audio tag classification task (Hu et al., 2009). It consists of 18 mood categories derived computationally from mood-related social tags from Last.fm, arguably the largest music tagging and sharing website. This 18-class model consists of 135 tags. Besides the automated tag classification task in MIREX, a simplified version of this model has been used in a cross-cultural user study on Korean music (Hu et al., 2014), together with the five-cluster model. As the MIREX five-cluster mood model has been used in a number of cross-cultural MIR studies, for comparison purposes, the current study adopts this model and attempts to make suggestions on modifications and refinement based on the results. 4.1 The participant groups The user groups recruited in this study consist of music listeners from Hong Kong and the USA. Hong Kong plays a special role in the history of Chinese music. Located adjacent to the Canton province, Hong Kong was the origin of Cantopop, which highly influenced the entire region of East Asia in the 1980s and 1990s. Furthermore, for historical and political reasons, Hong Kong was able to preserve the traditional Chinese culture to a large extent. Canto folk music and even traditional Canto opera are often appreciated and practiced by Hongkongers. Since the government handover in 1997, people in Hong Kong also have had increasing exposure to Mandopop. Therefore, it makes an interesting case to investigate Hongkongers' perceptions of a variety of Chinese music types, since they are culturally exposed to all of them to certain degrees. On the other hand, US listeners can represent people with the western cultural background, and, additionally, they have been found to exhibit different music consumption behaviors than Hongkongers in a prior study (Nettamo et al., 2006). We would like to explore whether differences also exist in the two user groups' mood perceptions of non-western music. Invitations were distributed to a number of departmental student mailing lists at the University of Hong Kong and University of Washington, as well as via the authors' various social network websites, such as Twitter, Facebook and Google Plus. Participation was completely voluntary. 4.2 Song selection This study used an online questionnaire of 29 clips of Chinese music to solicit listeners' music mood perceptions. The clips were manually selected from the authors' personal music collections, with the goal of reflecting the broad range of Chinese music. As the goal of this study is to propose suggestions for global MIR systems, two kinds of Chinese music were considered: traditional Chinese music and contemporary Chinese pop music. The former features distinct cultural origin and musical characteristics, and thus can be attractive to users who do not have a Chinese background but are curious about Chinese or Asian music. The latter (C-pop) is widely consumed by Chinese-speaking listeners around the world. As it is heavily influenced by western pop music, yet has Chinese flavor, it is potentially of interest to users in all regions who like pop music in general. Specifically, the selected songs represented the following six genres or styles described in Section 2.1: traditional Chinese instrumental music, traditional Cantonese (Canto) folk music, contemporary Chinese instrumental music, Cantopop and two unique styles of Mandopop: China wind and Chinese rock. Each type contains five representative pieces except for Chine wind which contains four. The list of titles can be found at: http://ccmir.cite.hku.hk/pubs/moodCH/List_of_Songs.xlsx 4.3 The questionnaire At the beginning of the questionnaire, participants were asked about their gender, year of birth and the place where they grew up. The last question was used for screening purposes so as to ensure that only participants with Hong Kong and US cultural backgrounds were included. We believe that people are greatly affected by the culture of the place they resided when they grew up, regardless of their nationality. All participants were 18 years of age or above and agreed with the informed consent statement before proceeding to the questionnaire. Next, participants were asked to listen to the selected music clips, with one song on each page. The questionnaire consisted of five questions for each clip. The first question asked the participant to listen to the clip and decide the most appropriate mood expressed by the song. The response options given were the mood clusters of the MIREX five-cluster model. In case the participants found that the mood of the song did not fit into any of the five default clusters, they could select the option "Other," and use their own words to describe the appropriate mood of the music clip. The participants were also asked about their familiarity with the music clips. Because people may interpret "familiarity" differently, instead of directly asking how familiar they are with the given songs, we used two questions: whether they had heard the song before, and if they could provide the artist name and the title of the song. Responses to the two questions were deemed more objective than self-rated familiarity because the questions are fact-based. The questions were presented in English in the US questionnaire and in both English and Chinese in the Hong Kong questionnaire. In the Hong Kong questionnaire, the five mood clusters were also translated into Chinese, and both English and Chinese terms were presented. We used the same translation of mood terms as in Yang and Hu (2012). Even though English is one of the official languages in Hong Kong, people's English proficiency varies. Therefore, both Chinese translations and original terms in English were deemed necessary for the respondents to have an accurate interpretation of the clusters. Figure 1 illustrates the survey interface of the Hong Kong version. 4.4 Data analysis Responses to the first question for each song (i.e. mood judgment) were compared quantitatively. The kh2 test was used for comparing the distributions of music mood judgments between the two groups of listeners (Hinkle et al., 2003). Significant results would indicate that the judgment distributions are dependent on cultural groups. The agreement between groups of listeners in various combinations of music and listener characteristics was measured by the raw ratio of agreed pairs to all possible pairs. For example, for calculating the agreement ratio among 32 Hong Kong listeners on five Cantopop songs, we first calculated the total number of judgment pairs as (32x(32-1)/2)x5=2,480, and counted the number of pairs with the same mood judgments. Suppose there were 248 matched pairs; the agreement ratio would be 10 percent. Raw agreement measures are recommended for cases where the annotations are dominated by negative values (i.e. "not chosen") (Cicchetti and Feinstein, 1990). This applies to the mood judgment data collected in this study. For instance, "Other" was chosen only 102 times out of 1,856 times it could have been selected. Text responses to the "Other" option were coded with a scheme of categories that emerged from the analysis process. The codebook was developed through an iterative process involving test-coding and revising the codes for clarity. The consensus model (Hill et al., 1997) was applied to coding the responses: two independent coders coded the responses and then discussed on disagreed cases until a consensus was reached. The results were compared across the two groups. Responses to demographic questions and song familiarity questions were used to group the users with a range of criteria, for the purpose of testing factors that affect users' mood perceptions. 4.5 Limitations As each music piece needs to have a minimum length (typically 15-30 seconds) for listeners to establish a confident perception of its mood (Speck et al., 2011), the questionnaire can contain only a limited number of song clips so that it can be completed within a reasonable amount of time. We used a set of 29 clips in this study, which may seem rather limited given there are millions of Chinese songs available in the market. To mitigate this limitation, we tried our best to carefully select a good variety of songs based on demands for them by users with various cultural backgrounds, and representativeness in each category. In addition, the sample size of each cultural group (32) is limited, partially due to the length of the survey and the lack of monetary incentive. Nonetheless, it is sufficient for statistical tests and comparisons. The online venues used to recruit participants (i.e. mailing lists and social network sites) may also have limited the participants to frequent internet users. Future studies are warranted to further verify the evidence found in this exploratory study in order to generalize findings to broader contexts. 5.1 Characteristics of participants The Hong Kong survey had a total of 70 responses, among which 34 were completed. The US survey had 72 responses, of which 34 were completed. However, two listeners in each group answered that they were raised in mainland China (not Hong Kong or the USA), leaving 32 valid responses for each group. Table I shows the demographic information of the listeners. 5.2 Mood judgments 5.2.1 Cultural backgrounds and mood judgments Table II presents the distribution of mood judgments of the two groups on all 29 songs. Both groups chose Cluster 3 (literate, poignant, wistful, etc.) most often, indicating that many Chinese songs may have some traits associated with the mood of Cluster 3, and these traits can be perceived by both Hong Kong and US listeners. The kh2 test shows that the distribution and cultural group are not independent (kh2=52.54, df=5, p<0.001). In other words, it provides strong evidence that Hong Kong and US listeners tend to perceive the mood of given Chinese music differently, within the framework of five mood clusters. From Table II, we can observe that Hong Kong listeners tended to select Cluster 1 (passionate, rousing, confident, etc.) and Cluster 3 while US listeners tended to spread their choices equally across Clusters 1, 2 (rollicking, cheerful, fun, etc.) and 3. The US listeners were also more likely to select the "Other" category than their Hong Kong counterparts. In all, 26 out of the 29 songs were labeled with "Other" at least once by the US group. In comparison, the Hong Kong group chose the "Other" option for 17 songs. On average, 2.93 US listeners chose "Other" for each of these songs, whereas the number for Hong Kong listeners was 1.35. This is similar to the findings in Lee and Hu (2014), where US listeners chose the "Other" category more often than Korean and Chinese listeners when they made mood judgments on western songs. Even though the objects/stimuli have been changed to Chinese music, the observation still holds. The consistent pattern regardless of the cultural background of the music further suggests that this phenomenon might be related to the individualism-collectivism difference between western and eastern cultures as discussed in previous studies (Boer and Fischer, 2010; Hu and Lee, 2012; Lee and Hu, 2014): western users tend to hold their opinions regardless of the given options, whereas Eastern people are more likely to conform to existing norms. We conducted a content analysis on the free-text responses provided in the "Other" option. The two independent coders reached an excellent reliability level (Cohen's k=0.858). Table III shows the seven resultant categories and corresponding counts in each of the groups. The largest category of the self-supplied terms is new mood terms that are not included in the five-cluster scheme. The new mood terms provided by both groups are quite similar, with, "relaxing," "calm" and "longing" being the most frequent, appearing six, four and four times, respectively. Users in both cultural groups also filled in terms that were not mood related. Some of them were music specific (e.g. "fast pace") while others were not (e.g. "focus"). Interestingly, the US group also provided responses that were either a mixture of terms from multiple clusters (e.g. "passionate" from Cluster 1 and "intense" from Cluster 5) or part of the terms in one cluster (e.g. "poignant, wistful, autumnal" in Cluster 3). The US participants also filled in descriptions of scenes they imagined when hearing the songs, although the instructions specifically asked for mood-related terms. Perhaps this group of users thought that scenes inspired by the songs also reflected the mood to some extent. The Hong Kong listeners did not provide any terms in these categories, possibly because: they tried harder to follow the instructions that asked respondents to "choose one dominant cluster when a song sounds to have moods in more than one cluster," and people from individualistic (mostly western) cultures were found to be more used to divergent thinking which involves the generation of novel ideas (Schmidt et al., 2013). 5.2.2 Agreement on mood judgments within and across culture groups The agreement ratios were calculated for pairs of judgments between two listeners from the same cultural group and across cultural groups. The results are shown in Table IV. In particular, in the cross-cultural case, the number of total pairs with one judgment from each cultural group is calculated as ((32x32)/2)x29=29,696. The results and non-paired t tests reveal that the level of intra-agreement among Hong Kong listeners is significantly higher than that of US listeners (p<0.001). A similar result was reported in Kosta et al. (2013), where Greek listeners were found to agree more on the mood of Greek music than non-Greek listeners. In comparing Korean and American listeners' perceptions of the mood of Korean pop songs, Hu et al. (2014) also found that Korean listeners reached a higher agreement level than their US counterparts. However, the result is partially different from Hu and Lee (2012), where both US and Chinese listeners had the same agreement ratio on the mood of western pop songs. We conjecture that this difference in results between Hu et al. (2014) and Hu and Lee (2012) might be due to Chinese listeners' wide exposure to western pop songs. For non-western songs that have not yet spread as widely, sharing the same cultural background as the music stimuli does seem to help listeners reach a higher agreement on music mood. It is also noteworthy that the agreement level among US listeners was not statistically different from the cross-cultural agreement (p=0.54). This is similar to the result in Hu et al. (2014), where the agreement level among American listeners on the mood of Korean pop songs was the same as that between American and Korean listeners. 5.2.3 Agreement on mood clusters To examine which moods are easier or more difficult to agree upon, we counted the number and percentage of agreed pairs across the five mood clusters (Table V). The kh2 test indicates that the two cultural groups showed a statistically significant difference in the agreement pattern across the mood clusters (kh2=603.55, df=10, p<0.001). Both within and across cultures, Cluster 3 (literate, poignant, wistful, etc.) got the most agreements, which again evidences that many Chinese pop songs have traits of Cluster 3 and these traits can transcend the cultural boundaries of listeners to some extent. The difference between the two cultural groups observed here is similar to the findings in Hu and Lee (2012), where American listeners agreed more on Cluster 2 (rollicking, cheerful, fun, etc.), Cluster 4 (humorous, silly, campy, etc.) and Cluster 5 (aggressive, fiery, tense/anxious, etc.) compared to Chinese listeners, who agreed more on Cluster 1 (passionate, rousing, confident, etc.) and Cluster 3 (literate, poignant, wistful, etc.). As western music was used as the stimuli in Hu and Lee (2012), it is interesting to see that the patterns of mood perception are similar to this study even though the stimuli here consist of completely different, non-western music. 5.3 Mood perception on different types of music 5.3.1 Instrumental vs vocal The music pieces in this study could be divided into instrumental and vocal sets. The former includes 15 pieces in instrumental folk, Canto folk or contemporary instrumental styles, and the latter includes the rest of the three types with 14 pieces total. We can observe from Table VI that overall, Hong Kong listeners agreed among themselves more often than US listeners on both instrumental and vocal pieces. Listeners also agreed more often on vocal pieces than instrumental music within each cultural group, as well as cross-culturally. This is in accordance with the literature on both music psychology and user studies in MDL/MIR. Juslin (2005) argued that the verbal (lyric) and non-verbal (melody) channels of music complemented each other in communicating emotions. Vocal delivery and the lyrics of the song were also found to be among the various factors that influence people's perception of music mood (Lee et al., 2012). In comparing the effects of melodies and lyrics on mood perceptions, studies found that, although melodies were more dominant than lyrics, lyrics were particularly useful in strengthening emotional responses to music with negative (Ali and Peynircioglu, 2006) and low-arousal emotions (Kazuma, 2009), which roughly correspond to Clusters 5 and 3, respectively, in this study. Even though the US listeners probably did not understand the lyrics in those songs, they might still have been able to get some emotional clues from the vocal delivery. 5.3.2 Genres/styles of Chinese music Table VII shows the agreement ratios within and across cultural groups with regard to the six types of music. The Hong Kong listeners showed an agreement of 60 percent on Cantopop, followed by 54 percent on China wind. The US listeners agreed more on Canto folk and China wind (43 percent), followed by Cantopop (41 percent). Cross-culturally, kh2 tests show that the two groups significantly differed in their mood judgments on all styles except for Cantopop (kh2=9.22, df=10, p=0.10). We conjecture two possible reasons for Cantopop to reach a high cross-cultural agreement ratio: as introduced before, Cantopop was developed with heavy influence from western pop music, and thus the western elements could have helped to bridge the differences between the two user groups; and all Cantopop songs are lyrical and thus the mood cues in the vocal delivery may have helped strengthen mood perceptions. The category with the least agreement both within and across cultures was contemporary instrumental music. This may be due to the complex musical compositions of this style and the lack of lyrical cues. In fact, the song with the least agreement was of this type and will be analyzed in more detail in the next subsection. To facilitate further examination of the mood judgments on different types of music stimuli, Figure 2 shows the distribution of mood judgments for each of the six music types included in this study. It can be seen that the first three types of music (i.e. instrumental folk, Canto folk and contemporary instrumental) were labeled with a mix of diverse mood clusters without a clear, dominant cluster, whereas the last 3 (i.e. Cantopop, China wind and Chinese rock) have one dominant cluster from both cultural groups: China wind and Cantopop were mostly labeled with Cluster 3 (literate, poignant, wistful, etc.) and Chinese rock with Cluster 1 (passionate, rousing, confident, etc.). Instrumental folk music seems to be the most controversial, in that the pieces got similar numbers of judgments across all mood clusters, especially from the US group. The two cultural groups seemed to disagree most on Canto folk and contemporary instrumental music. Hong Kong listeners categorized Canto folk music mostly in Clusters and 2 (rollicking, cheerful, fun, etc.) but US listeners tended to label those pieces more often with Cluster 2 and Cluster 4 (humorous, silly, campy, etc.). For contemporary instrumental music, US listeners chose Cluster 2 the most, followed by similar numbers for Cluster 3, Cluster 1 and "Other," while Hong Kong listeners used Cluster 3 more often, followed by Cluster 2, Cluster 1 and Cluster 4. 5.3.3 The songs with highest and lowest agreement The song that reached the highest agreement on mood judgment was the same within each cultural group and across cultural groups. In fact, the US listeners agreed more than Hong Kong listeners, with agreement ratios of 0.94 and 0.88, respectively. This was a China wind song titled Pearl-decorated Curtain Rolled Up by a Chinese male singer, Huozun. In all, 30 Hong Kong listeners and 31 US listeners selected Cluster 3 (literate, poignant, wistful, etc.) for this song. Released in 2014, the song features a slow melody and extensive use of falsettos. It can be characterized as mellow, ethereal and graceful, similar to the style of ancient Chinese music (and thus recognized as China wind). Although the lyrics describe a beautiful young woman missing a lost love in a poetic way, the US listeners probably could not understand them. Yet they reached near consensus on the mood of this song, possibly attributable to the emotionally powerful delivery of the song. The song with the lowest agreement among US listeners as well as cross-culturally was Shangri-la by Girls 12, a female band famous for using traditional Chinese instruments to play both traditional Chinese and western music. This song had the second-lowest agreement among Hong Kong listeners, and their lowest agreement was for another piece by the same band. Many pieces by Girls 12 are modern compositions using ancient instruments, including Shangri-la, and thus are classified as contemporary instrumental music, the type that has the least agreement both within and across cultures (Table VII). The song Shangri-la received only 16 percent agreement from US listeners and 21 percent from Hong Kong listeners. Figure 3 illustrates the distribution of mood judgments of this song. We can observe that the judgments from Hong Kong listeners were distributed fairly evenly across Cluster 1 (passionate, rousing, confident, etc.), Cluster 2 (rollicking, cheerful, fun, etc.), Cluster 3 (literate, poignant, wistful, etc.) and Cluster 4 (humorous, silly, campy, etc.), and about equal numbers of US listeners labeled it with Cluster 1, Cluster 2, Cluster 5 (aggressive, fiery, tense/anxious, intense, etc.) and "Other." The piece was performed with traditional Chinese instruments: bamboo flute, erhu (bowed strings), pipa and guzheng (both plucked strings), mixed with western-sounding percussion and electronica fusions. Just as the song title suggests, the piece conjures the mystic and mysterious aura of the utopia named Shangri-la. The mixed mood perceptions of this song can be at least partially explained by the mixed musical nature of the song, as well as its fictional, mystical theme. 5.4 Listener characteristics 5.4.1 Gender Table VIII shows the distribution of mood judgments based on gender across all mood clusters. Both genders in both cultural groups selected Cluster 3 the most. The kh2 test indicates that gender difference in mood judgment does exist within the US group (kh2=24.62, df=5, p <0.001), but not in the Hong Kong group (kh2=7.53, df=5, p<0.184). This was also the case in Lee and Hu (2014), where American listeners' mood judgments on western music differed based on gender, but Chinese listeners' did not. In order to see the influence of gender difference across cultural groups, we compared the judgments from US and Hong Kong listeners of the same gender. The comparison shows female listeners in the two cultural groups did have different mood judgments (kh 2=46.51, df=5, p<0.001). However, the judgments of male listeners from the two cultural groups did not have a significant difference (kh2=7.99, df=5, p=0.157). The results suggest that cultural differences played a significant role in mood judgments among female listeners, but not among male listeners. In contrast, Hu and Lee (2012) found significant differences for both genders between US and Chinese listeners when the music stimuli were western pop songs. This discrepancy might have been affected by the small sample size of male participants in the US group (n=8) in this study. Future studies with a larger and more balanced sample across gender groups are warranted to further investigate the extent to which gender and cultural groups affect music mood perception. 5.4.2 Age The ages of the participants vary between the two groups, but both groups have listeners between 21 and 60 years old. We divided those listeners into two age groups, 21-33 and 34-60, so as to balance the sample sizes in both age groups. The kh2 tests on the distribution of judgments revealed a statistically significant difference between age groups among Hong Kong listeners (kh2=39.86, df=5, p<0.001) but no significant difference among US listeners. (kh2=4.72, df=5, p=0.45). This is, in fact, consistent with the findings from Lee and Hu (2014), where the authors found a significant difference between the age groups for Korean and Chinese music listeners but not for American listeners. This could be possibly attributed to the fast-changing environment in Hong Kong and other Asian countries compared to the relatively stable environment in countries such as the USA (Lee and Hu, 2014). 5.4.3 Familiarity with the music Listeners' familiarity with the songs was categorized into three levels: "high familiarity" if they had listened to the song before and could also identify the artist's name and the song title; "medium familiarity" if they had heard the song but could not name it; and no familiarity if they had not heard the song. There was one US listener who was highly familiar with two songs, and seven who, collectively, had heard a total of 15 songs before. In the Hong Kong group, 29 out 32 listeners indicated high familiarity with at least one song. Unsurprisingly, more than half of the highly familiar songs were Cantopop. Due to the sparseness of data in the US group, we examined how the levels of agreement related to familiarity with the music only in the Hong Kong group. Table IX lists the agreement ratios among judgment pairs with regard to various levels of music familiarity. When both participants were highly familiar with a song, their agreement was the highest (ratio=0.57). While a higher familiarity level helped increase the level of agreement when both mood judgments were made for songs with which participants were familiar, it did not help when one participant was familiar with the song and the other was not. We conjecture that when users have heard a song multiple times or know the artist name and the song title, they may have developed and/or gained some contextual knowledge of the song, which may affect their selection of mood clusters. On the other hand, when listeners hear the song for the first time, they may naturally rely more on the musical characteristics for judging the mood of the song. As advocated by Weissenberger (2015), a flexible organization system is needed for MIR/MDL to meet the needs of different musical traditions. The results of this study have important implications for designing MDL of this kind. As users in different cultural groups had significantly different mood judgments overall (Section 5.2.1), it is better to differentiate user-generated mood labels and/or mood-related social tags according to the users' cultural background. In particular, as the agreement ratio among Hong Kong listeners was higher than the cross-cultural one (Section 5.2.2, Table IV), Hong Kong listeners would benefit more from mood tags supplied by listeners in the same cultural group. In contrast, as the within-group agreement ratio of US listeners was similar to the cross-cultural agreement level (0.38 vs 0.39), US listeners could make use of mood tags generated by Hong Kong listeners, especially when those generated by US listeners are sparse. The analysis on the textual input in the "Other" option (Section 5.2.1) suggests that the MIREX five-cluster model may not be sufficient and more terms are needed to describe the mood of Chinese music. Particularly, terms like "calm," "relaxing," "peaceful" and "longing" appeared most frequently (Table III, Section 5.2.1). In fact, a previous study on automated mood classification on Chinese pop songs (Yang and Hu, 2012) also used mood terms in the MIREX five-cluster model, and ended up with adding additional mood labels that were deemed as important to represent moods in Chinese pop songs: "calm/peaceful," "dreamy," "encouraging," "nostalgic," "relaxed," "soothing" and "tender." Based on this finding, it is suggested a sixth cluster can be added to the MIREX five-cluster model: (calm, relaxing, peaceful, etc.). In addition, some of the suggested terms in the current study were mixtures of terms from different clusters, which implies that the mood clusters may not be mutually exclusive, at least for US listeners. Lee and Hu (2014) also found that some songs may be perceived to be in two or more clusters at the same time. Therefore, it is suggested that the MIREX five-cluster model should allow multiple mood clusters to be applied to each song. As US listeners specified terms that describe visual scenes, it might be a good idea to use scenes as a new access point in MDL. The association between music and images has been studied in music psychology (Ali and Peynircioglu, 2006) and MIR (Mattek and Casey, 2011), but how it can be used in MDL warrants future research to explore. The fact that both cultural groups chose Cluster 3 (literate, poignant, wistful, etc.) most frequently (Table II) and agreed more on songs labeled with Cluster 3 (Table V) suggests that Chinese music may tend to have musical traits that are associated with moods in Cluster 3, and that such musical traits can transcend the cultural boundaries of the audience. MDL using automated mood classification algorithms (Yang and Chen, 2012) could exploit this finding and assign more weight to the classifier of Cluster 3, allowing songs to have a greater chance of being classified in Cluster 3. Similarly, when recommending songs based on mood labels, those labeled with Cluster 3 could be assigned more confidence. As different styles had different agreement levels within and across cultures (Table VII), music in different styles can be treated differently in MDL. For example, contemporary instrumental pieces had low agreement levels between audiences, and thus classification and recommendations of pieces in this style could be given less confidence or provide more candidates to improve the chance of hits. The results on gender and age group differences can be exploited, too. For instance, Hong Kong listeners would benefit more from recommendations and tags made by other Hong Kong listeners in the same age group, whereas for US listeners there is no need to differentiate the age group of other listeners. The effect of song familiarity on mood perception suggests that MDL should allow users to change their mood labels and tags over time, because when they get more familiar with a song, they may perceive the mood of the song differently. This study compares Hong Kong and US listeners' mood perceptions of 29 Chinese music pieces, with the goal of investigating whether and how mood perceptions of the two groups of listeners differ, and how the differences can inform the design of cross-cultural MDL. Music mood was modeled with the MIREX five mood clusters and the results suggested further refinement of this model. The selected music pieces included six genres and styles of Chinese music, ranging from traditional folk music to several sub-genres of C-pop. Overall, the patterns of mood judgments by listeners from Hong Kong and the USA (e.g. commonly agreed-upon clusters, differences based on gender and age) resonated with the findings in previous studies on cross-cultural mood perceptions of western music, even though this study used Chinese music, which has a contrasting cultural background to that of western music. This has important implications for designing global MDL: for both western and Chinese pop music, listeners from different cultures may need different mood labels. For systems that use social tags to facilitate access, it is recommended that tags given by users in different cultural groups be differentiated. Those given by users in the same cultural groups can be more helpful. This study also confirms that listeners do exhibit cross-cultural differences in mood judgment for all but one type (Cantopop) of Chinese music considered in this study. Therefore, it may be desirable for MDL to differentiate music types when classifying or recommending music by mood. Future work will include further investigation on other cultural groups interested in Chinese music, such as mainland Chinese, Taiwanese and Singaporeans. In addition, we plan to conduct similar studies on music from other cultural backgrounds, such as Korean and Japanese music. Findings of such studies will further contribute to the goal of building effective and user-centered global MDL.
|
The purpose of this paper is to compare music mood perceptions of people with diverse cultural backgrounds when they interact with Chinese music. It also discusses how the results can inform the design of global music digital libraries (MDL).
|
[SECTION: Method] Music seeking and consumption are no longer confined by the boundaries of country, region or culture today (Lee et al., 2013). Music, as a cultural object, may be perceived differently by people from different cultural backgrounds, imposing a challenge on music digital libraries (MDL) in meeting the needs of a diverse audience (Weissenberger, 2015). Consequently, an increasing number of researchers have started to investigate various cross-cultural issues in the music information retrieval (MIR) and MDL fields. As people often seek music for emotional goals (Lavranos et al., 2015), music mood[1] has increasingly become a popular access point for music information in many MDL and online music services (Hu, 2010; Hu and Downie, 2007). This trend has raised questions regarding the applicability of music mood across cultural boundaries. Probably due to its subjective and context-based nature, music mood perception is often regarded as culture-dependent (Wong et al., 2009). A number of previous studies have compared mood perceptions of music in various cultures by listeners from different cultural backgrounds (e.g. Balkwill and Thompson, 1999; Fritz et al., 2009; Hu and Lee, 2012; Lee et al., 2013; Singhi and Brown, 2014; Egermann et al., 2014). Although specific findings vary, a general trend found in these existing studies is that listeners' perceptions of music mood can be influenced by their cultural backgrounds. However, most existing studies have focussed on western (classical or pop) music or ethnic music with small audiences (e.g. Congolese music produced and consumed by Congolese Pygmies (Egermann et al., 2014)). Such studies mainly aim to answer questions on music psychology and/or the social, developmental aspects of music. The purpose of this study is different; we are interested in exploring how music mood could be used as an access point or metadata type that can transcend the cultural boundaries of users in the context of MDL. To date, the question of how people with different cultural backgrounds would perceive the mood of Chinese music remains under-investigated. Considering its appeal to one of the largest music listener populations in the world (IFPI, 2014), Chinese music should not be left out of today's global scope of MDL/MIR studies. As Chinese music possesses unique characteristics that are remarkably different from western music (Fung, 2013), it is of theoretical and practical value to investigate the case of Chinese music. Therefore, in this study, we aim to fill this gap and explore how people with different cultural backgrounds, specifically those in the USA and Hong Kong (representing western and eastern listeners, respectively), perceive the mood of Chinese music, including traditional Chinese music and contemporary Chinese pop music that infuses western pop characteristics and local Chinese elements. We specifically answer the following research questions: RQ1. How differently or similarly do Hong Kong and US music listeners perceive the mood of Chinese music? RQ2. How differently or similarly do Hong Kong and US music listeners perceive the mood of different genres/styles of Chinese music? RQ3. Do the listeners' demographic characteristics and familiarity with the music affect their mood perceptions of Chinese music? Answers to these questions have important implications for MDL that use mood as an access point to facilitate MIR and discovery in a global environment. These answers will help determine whether it is necessary to develop different mood-related labels or tags that take their suitability for various user groups into consideration. Implications of the results in this study will be discussed. The goal of this study is to contribute to the literature of cross-cultural MDL/MIR and provide empirical evidence for designing cross-cultural MDL/MIR systems. 2.1 A brief history and types of Chinese music Chinese music broadly refers to music that uses traditional Chinese instruments or is in the Chinese language, and is the main music consumed in the region of Greater China (i.e. regions where Chinese is the official language, including mainland China, Hong Kong, Macau, Taiwan and Singapore). However, the consumption of Chinese music is not limited to that region. People around of the world are accessing Chinese music, and all major online music services, such as Spotify and Last.fm, contain Chinese music to cater to the demand. There is a broad range of genres and styles in Chinese music. Traditional Chinese folk music dates back to approximately 7,000 years ago, played mostly in the pentatonic scale with its own set of instruments. Contemporary Chinese pop music (also referred to as C-pop) was developed in the middle of the twentieth century with the influence of western music. The subgenre "Cantopop" (Hong Kong pop music sung in Cantonese) was developed in the 1970s, followed by the rise of another subgenre, "Mandopop" (Mandarin Chinese pop music), in the 1980s (Fung, 2013). This study includes six genres/styles of Chinese music that are widely appreciated in the region and around the world. The first two are traditional Chinese music, whereas the rest are the most popular C-pop styles featured in online music services in China (e.g. Baidu, Xiami, Kugou): Traditional Instrumental Folk Music of China (qiyue): on the pentatonic scale, musicians perform using traditional instrument(s) of woodwind and percussion, bowed strings, plucked and struck strings in solo, small ensembles or orchestra (Randel, 2003). Canto(nese) folk songs: Han Chinese folk songs are divided into northern and southern styles geographically. Provinces south of the Yangzi River have mild weather and abundant rain, and therefore the folk songs tend to be more "lyrical, gentle and conjunct" (Han, 1989). Near the Canton province and Pearl River Delta, Cantonese folk music is especially light-spirited (China Culture, 2006). Contemporary Chinese instrumental music: this style of music is also called "New Music of China." It mixes the techniques, instrumentation and composition of both Chinese traditional music and the European/western music (Liu and Mason, 2010). The Girls 12 is a famous mainland Chinese all-female group who plays this type of music with traditional Chinese instruments. Cantopop (or Cantonese popular music): this is a subgenre of C-pop. Sung in Cantonese, Cantopop usually refers to popular songs produced in Hong Kong. Blended with traditional Chinese elements, as well as western-influenced harmonies and melodies, Cantopop was a dominant music genre in Asia in the 1980s' and 1990s. Its iconic star, Anita Mui, was recognized as "Asia's Madonna." China wind (zhongguofeng): this is a subgenre of Mandopop and is also called "Chinese Style." It features traditional Chinese instruments and classical Chinese melody juxtaposed with popular western music styles and cultural references to Chinese history, classics and folklore in the lyrics (Xinhuanet, 2006). Taiwanese singer Jay Chou is regarded as the "godfather" of this type of music, bringing forth the wave of China Wind pop. Chinese rock: as another subgenre of Mandopop, Chinese rock is locally produced in mainland China and is the most notable sub-category of Mandopop to the world outside of China. Chinese rock combines the loud and forceful folk song traditions of Northern China and characteristics of western-style rock such as a fast tempo, a strong beat and aggressive bass lines (Fung, 2013). The singer/songwriter Cui Jian from mainland China has been widely recognized as the father of Chinese rock (Fung, 2013). It is noteworthy that all C-pop styles are, to varying extents, influenced by western pop or Rock music. Compared to Traditional Instrumental Chinese Folk Music and Canto Folk Songs, C-pop may be easier to understand and appreciate for listeners familiar with western music. In addition, because of the mixture of western music elements, listeners from both cultural groups in this study might share more common perceptions on C-pop than the other two genres/styles. Therefore, in addition to analyzing aggregated perceptions across all songs, this study also compares and contrasts listeners' perceptions on these six music styles. 2.2 Cross-cultural studies on music mood perceptions A number of studies have investigated the effects of culture on music mood perception. Although fundamental emotions such as happiness, sadness, anger or fear were found to be generally agreed upon across cultures, cultural traditions and background were found to affect individuals' music mood perception, in addition to the inherent qualities of the music (e.g. Gregory and Varney, 1996; Kosta et al., 2013; Lee et al., 2013). For example, Wong et al. (2009) compared mood descriptors used by European and Indian communities and discovered that cultural tradition played a more crucial role than the inherent qualities of the music in users' mood perception. Fritz et al. (2009) found that mood perceptions of western music by African listeners partially agreed with those by western listeners. In a similar vein, Hu and Lee (2012) discovered that discrepancies existed between Chinese and American music listeners in the way they determined the mood of western music pieces. More recently, Lee and Hu (2014) also compared perceptions of western music mood among American, Korean and Chinese listeners, and found that the results of Korean listeners' mood perceptions were situated in between the two other groups. There are also studies investigating the cultural backgrounds of listeners in relation to their mood perceptions of non-western music. Kosta et al. (2013) found greater agreement on the mood of Greek songs among Greek listeners compared to non-Greek listeners with varying backgrounds. Similar results were found by Hu et al. (2014) in comparing American and Korean listeners' mood perceptions of a set of Korean pop songs. Egermann et al. (2014) compared subjective and psychophysiological emotional responses by Canadians and Congolese Pygmies to both western and Congolese music. They found that the two groups had very different responses to Pygmies music but similar responses to some western music. To the best of our knowledge, there has been little work focussed on comparing the mood perceptions of various genres/styles of Chinese music across different cultural groups. In addition, many of the previous studies were conducted in the context of music psychology or ethnomusicology. This study, however, reflects the perspective of information organization and access, with the goal of finding out how music mood can be organized to best meet the information needs of users from different cultural groups. There are mainly two kinds of models to represent mood in music psychology and MIR: categorical and dimensional. The former uses a set of discrete terms (e.g. "passionate," "cheerful") to represent the mood of a piece of music. The most classical model of this kind is Hevner's (1936) adjective circle where eight mood categories placed in a circle, with a set of terms in each category. The dimensional models, in contrast, represent mood with continuous values in a low dimensional space. Different models may have different dimensions, yet valence (i.e. level of pleasure) and arousal (i.e. level of energy) are among the most popular dimensions. The dimensional model used very often in MIR is Russell's (1980) model. The categorical and dimensional models have their own pros and cons. Yang and Chen (2012) summarize that categorical models are more user-friendly as they only consist of terms in natural language; whereas dimensional models are advantageous in terms of quantifying the intensity of moods. Notwithstanding the importance of dimensional models, categorical ones are more suitable for this current study, for the purpose of validating and comparing the values of mood metadata in the context of cross-cultural MDLs. The models from music psychology have solid theoretical ground and can be borrowed into MIR studies. However, one criticism of adopting these models directly in MIR is that they were developed in experimental settings and thus might not reflect users' preferences and behaviors in today's music listening environment (Hu, 2010). Therefore, MIR researchers developed mood models based on domain analysis of widely used music repositories and streaming services. One of the influential mood models in MIR is the five-cluster model proposed in the audio music mood classification task (Hu et al., 2008) of Music Information Retrieval Evaluation eXchange (MIREX), a community-wide MIR evaluation campaign held annually since 2005 (Downie et al., 2014). This model was developed from more than 180 editorial mood labels on AllMusic.com, a very popular online music repository. It has been used in many music mood classification studies (e.g. Laurier et al., 2008; Bischoff et al., 2009; Yang and Chen, 2012) as well as cross-cultural user studies (Hu and Lee, 2012; Lee and Hu, 2014; Hu et al., 2014; Yang and Hu, 2012). The five mood clusters are: C1: passionate, rousing, confident, boisterous, rowdy; C2: rollicking, cheerful, fun, sweet, amiable/good natured; C3: literate, poignant, wistful, bittersweet, autumnal, brooding; C4: humorous, silly, campy, quirky, whimsical, witty, wry; and C5: aggressive, fiery, tense/anxious, intense, volatile, visceral. In this five-cluster model, each cluster is collectively represented by a set of mood terms that share similar semantics. As the terms are describing affect, which can be somewhat vague (as opposed to terms referring to concrete objects), it is argued that using multiple terms with similar meanings to collectively describe one kind of mood is clearer than single-word representations (Hu and Downie, 2007). A more recent categorical mood model was also used in MIREX, in the mood subtask of the audio tag classification task (Hu et al., 2009). It consists of 18 mood categories derived computationally from mood-related social tags from Last.fm, arguably the largest music tagging and sharing website. This 18-class model consists of 135 tags. Besides the automated tag classification task in MIREX, a simplified version of this model has been used in a cross-cultural user study on Korean music (Hu et al., 2014), together with the five-cluster model. As the MIREX five-cluster mood model has been used in a number of cross-cultural MIR studies, for comparison purposes, the current study adopts this model and attempts to make suggestions on modifications and refinement based on the results. 4.1 The participant groups The user groups recruited in this study consist of music listeners from Hong Kong and the USA. Hong Kong plays a special role in the history of Chinese music. Located adjacent to the Canton province, Hong Kong was the origin of Cantopop, which highly influenced the entire region of East Asia in the 1980s and 1990s. Furthermore, for historical and political reasons, Hong Kong was able to preserve the traditional Chinese culture to a large extent. Canto folk music and even traditional Canto opera are often appreciated and practiced by Hongkongers. Since the government handover in 1997, people in Hong Kong also have had increasing exposure to Mandopop. Therefore, it makes an interesting case to investigate Hongkongers' perceptions of a variety of Chinese music types, since they are culturally exposed to all of them to certain degrees. On the other hand, US listeners can represent people with the western cultural background, and, additionally, they have been found to exhibit different music consumption behaviors than Hongkongers in a prior study (Nettamo et al., 2006). We would like to explore whether differences also exist in the two user groups' mood perceptions of non-western music. Invitations were distributed to a number of departmental student mailing lists at the University of Hong Kong and University of Washington, as well as via the authors' various social network websites, such as Twitter, Facebook and Google Plus. Participation was completely voluntary. 4.2 Song selection This study used an online questionnaire of 29 clips of Chinese music to solicit listeners' music mood perceptions. The clips were manually selected from the authors' personal music collections, with the goal of reflecting the broad range of Chinese music. As the goal of this study is to propose suggestions for global MIR systems, two kinds of Chinese music were considered: traditional Chinese music and contemporary Chinese pop music. The former features distinct cultural origin and musical characteristics, and thus can be attractive to users who do not have a Chinese background but are curious about Chinese or Asian music. The latter (C-pop) is widely consumed by Chinese-speaking listeners around the world. As it is heavily influenced by western pop music, yet has Chinese flavor, it is potentially of interest to users in all regions who like pop music in general. Specifically, the selected songs represented the following six genres or styles described in Section 2.1: traditional Chinese instrumental music, traditional Cantonese (Canto) folk music, contemporary Chinese instrumental music, Cantopop and two unique styles of Mandopop: China wind and Chinese rock. Each type contains five representative pieces except for Chine wind which contains four. The list of titles can be found at: http://ccmir.cite.hku.hk/pubs/moodCH/List_of_Songs.xlsx 4.3 The questionnaire At the beginning of the questionnaire, participants were asked about their gender, year of birth and the place where they grew up. The last question was used for screening purposes so as to ensure that only participants with Hong Kong and US cultural backgrounds were included. We believe that people are greatly affected by the culture of the place they resided when they grew up, regardless of their nationality. All participants were 18 years of age or above and agreed with the informed consent statement before proceeding to the questionnaire. Next, participants were asked to listen to the selected music clips, with one song on each page. The questionnaire consisted of five questions for each clip. The first question asked the participant to listen to the clip and decide the most appropriate mood expressed by the song. The response options given were the mood clusters of the MIREX five-cluster model. In case the participants found that the mood of the song did not fit into any of the five default clusters, they could select the option "Other," and use their own words to describe the appropriate mood of the music clip. The participants were also asked about their familiarity with the music clips. Because people may interpret "familiarity" differently, instead of directly asking how familiar they are with the given songs, we used two questions: whether they had heard the song before, and if they could provide the artist name and the title of the song. Responses to the two questions were deemed more objective than self-rated familiarity because the questions are fact-based. The questions were presented in English in the US questionnaire and in both English and Chinese in the Hong Kong questionnaire. In the Hong Kong questionnaire, the five mood clusters were also translated into Chinese, and both English and Chinese terms were presented. We used the same translation of mood terms as in Yang and Hu (2012). Even though English is one of the official languages in Hong Kong, people's English proficiency varies. Therefore, both Chinese translations and original terms in English were deemed necessary for the respondents to have an accurate interpretation of the clusters. Figure 1 illustrates the survey interface of the Hong Kong version. 4.4 Data analysis Responses to the first question for each song (i.e. mood judgment) were compared quantitatively. The kh2 test was used for comparing the distributions of music mood judgments between the two groups of listeners (Hinkle et al., 2003). Significant results would indicate that the judgment distributions are dependent on cultural groups. The agreement between groups of listeners in various combinations of music and listener characteristics was measured by the raw ratio of agreed pairs to all possible pairs. For example, for calculating the agreement ratio among 32 Hong Kong listeners on five Cantopop songs, we first calculated the total number of judgment pairs as (32x(32-1)/2)x5=2,480, and counted the number of pairs with the same mood judgments. Suppose there were 248 matched pairs; the agreement ratio would be 10 percent. Raw agreement measures are recommended for cases where the annotations are dominated by negative values (i.e. "not chosen") (Cicchetti and Feinstein, 1990). This applies to the mood judgment data collected in this study. For instance, "Other" was chosen only 102 times out of 1,856 times it could have been selected. Text responses to the "Other" option were coded with a scheme of categories that emerged from the analysis process. The codebook was developed through an iterative process involving test-coding and revising the codes for clarity. The consensus model (Hill et al., 1997) was applied to coding the responses: two independent coders coded the responses and then discussed on disagreed cases until a consensus was reached. The results were compared across the two groups. Responses to demographic questions and song familiarity questions were used to group the users with a range of criteria, for the purpose of testing factors that affect users' mood perceptions. 4.5 Limitations As each music piece needs to have a minimum length (typically 15-30 seconds) for listeners to establish a confident perception of its mood (Speck et al., 2011), the questionnaire can contain only a limited number of song clips so that it can be completed within a reasonable amount of time. We used a set of 29 clips in this study, which may seem rather limited given there are millions of Chinese songs available in the market. To mitigate this limitation, we tried our best to carefully select a good variety of songs based on demands for them by users with various cultural backgrounds, and representativeness in each category. In addition, the sample size of each cultural group (32) is limited, partially due to the length of the survey and the lack of monetary incentive. Nonetheless, it is sufficient for statistical tests and comparisons. The online venues used to recruit participants (i.e. mailing lists and social network sites) may also have limited the participants to frequent internet users. Future studies are warranted to further verify the evidence found in this exploratory study in order to generalize findings to broader contexts. 5.1 Characteristics of participants The Hong Kong survey had a total of 70 responses, among which 34 were completed. The US survey had 72 responses, of which 34 were completed. However, two listeners in each group answered that they were raised in mainland China (not Hong Kong or the USA), leaving 32 valid responses for each group. Table I shows the demographic information of the listeners. 5.2 Mood judgments 5.2.1 Cultural backgrounds and mood judgments Table II presents the distribution of mood judgments of the two groups on all 29 songs. Both groups chose Cluster 3 (literate, poignant, wistful, etc.) most often, indicating that many Chinese songs may have some traits associated with the mood of Cluster 3, and these traits can be perceived by both Hong Kong and US listeners. The kh2 test shows that the distribution and cultural group are not independent (kh2=52.54, df=5, p<0.001). In other words, it provides strong evidence that Hong Kong and US listeners tend to perceive the mood of given Chinese music differently, within the framework of five mood clusters. From Table II, we can observe that Hong Kong listeners tended to select Cluster 1 (passionate, rousing, confident, etc.) and Cluster 3 while US listeners tended to spread their choices equally across Clusters 1, 2 (rollicking, cheerful, fun, etc.) and 3. The US listeners were also more likely to select the "Other" category than their Hong Kong counterparts. In all, 26 out of the 29 songs were labeled with "Other" at least once by the US group. In comparison, the Hong Kong group chose the "Other" option for 17 songs. On average, 2.93 US listeners chose "Other" for each of these songs, whereas the number for Hong Kong listeners was 1.35. This is similar to the findings in Lee and Hu (2014), where US listeners chose the "Other" category more often than Korean and Chinese listeners when they made mood judgments on western songs. Even though the objects/stimuli have been changed to Chinese music, the observation still holds. The consistent pattern regardless of the cultural background of the music further suggests that this phenomenon might be related to the individualism-collectivism difference between western and eastern cultures as discussed in previous studies (Boer and Fischer, 2010; Hu and Lee, 2012; Lee and Hu, 2014): western users tend to hold their opinions regardless of the given options, whereas Eastern people are more likely to conform to existing norms. We conducted a content analysis on the free-text responses provided in the "Other" option. The two independent coders reached an excellent reliability level (Cohen's k=0.858). Table III shows the seven resultant categories and corresponding counts in each of the groups. The largest category of the self-supplied terms is new mood terms that are not included in the five-cluster scheme. The new mood terms provided by both groups are quite similar, with, "relaxing," "calm" and "longing" being the most frequent, appearing six, four and four times, respectively. Users in both cultural groups also filled in terms that were not mood related. Some of them were music specific (e.g. "fast pace") while others were not (e.g. "focus"). Interestingly, the US group also provided responses that were either a mixture of terms from multiple clusters (e.g. "passionate" from Cluster 1 and "intense" from Cluster 5) or part of the terms in one cluster (e.g. "poignant, wistful, autumnal" in Cluster 3). The US participants also filled in descriptions of scenes they imagined when hearing the songs, although the instructions specifically asked for mood-related terms. Perhaps this group of users thought that scenes inspired by the songs also reflected the mood to some extent. The Hong Kong listeners did not provide any terms in these categories, possibly because: they tried harder to follow the instructions that asked respondents to "choose one dominant cluster when a song sounds to have moods in more than one cluster," and people from individualistic (mostly western) cultures were found to be more used to divergent thinking which involves the generation of novel ideas (Schmidt et al., 2013). 5.2.2 Agreement on mood judgments within and across culture groups The agreement ratios were calculated for pairs of judgments between two listeners from the same cultural group and across cultural groups. The results are shown in Table IV. In particular, in the cross-cultural case, the number of total pairs with one judgment from each cultural group is calculated as ((32x32)/2)x29=29,696. The results and non-paired t tests reveal that the level of intra-agreement among Hong Kong listeners is significantly higher than that of US listeners (p<0.001). A similar result was reported in Kosta et al. (2013), where Greek listeners were found to agree more on the mood of Greek music than non-Greek listeners. In comparing Korean and American listeners' perceptions of the mood of Korean pop songs, Hu et al. (2014) also found that Korean listeners reached a higher agreement level than their US counterparts. However, the result is partially different from Hu and Lee (2012), where both US and Chinese listeners had the same agreement ratio on the mood of western pop songs. We conjecture that this difference in results between Hu et al. (2014) and Hu and Lee (2012) might be due to Chinese listeners' wide exposure to western pop songs. For non-western songs that have not yet spread as widely, sharing the same cultural background as the music stimuli does seem to help listeners reach a higher agreement on music mood. It is also noteworthy that the agreement level among US listeners was not statistically different from the cross-cultural agreement (p=0.54). This is similar to the result in Hu et al. (2014), where the agreement level among American listeners on the mood of Korean pop songs was the same as that between American and Korean listeners. 5.2.3 Agreement on mood clusters To examine which moods are easier or more difficult to agree upon, we counted the number and percentage of agreed pairs across the five mood clusters (Table V). The kh2 test indicates that the two cultural groups showed a statistically significant difference in the agreement pattern across the mood clusters (kh2=603.55, df=10, p<0.001). Both within and across cultures, Cluster 3 (literate, poignant, wistful, etc.) got the most agreements, which again evidences that many Chinese pop songs have traits of Cluster 3 and these traits can transcend the cultural boundaries of listeners to some extent. The difference between the two cultural groups observed here is similar to the findings in Hu and Lee (2012), where American listeners agreed more on Cluster 2 (rollicking, cheerful, fun, etc.), Cluster 4 (humorous, silly, campy, etc.) and Cluster 5 (aggressive, fiery, tense/anxious, etc.) compared to Chinese listeners, who agreed more on Cluster 1 (passionate, rousing, confident, etc.) and Cluster 3 (literate, poignant, wistful, etc.). As western music was used as the stimuli in Hu and Lee (2012), it is interesting to see that the patterns of mood perception are similar to this study even though the stimuli here consist of completely different, non-western music. 5.3 Mood perception on different types of music 5.3.1 Instrumental vs vocal The music pieces in this study could be divided into instrumental and vocal sets. The former includes 15 pieces in instrumental folk, Canto folk or contemporary instrumental styles, and the latter includes the rest of the three types with 14 pieces total. We can observe from Table VI that overall, Hong Kong listeners agreed among themselves more often than US listeners on both instrumental and vocal pieces. Listeners also agreed more often on vocal pieces than instrumental music within each cultural group, as well as cross-culturally. This is in accordance with the literature on both music psychology and user studies in MDL/MIR. Juslin (2005) argued that the verbal (lyric) and non-verbal (melody) channels of music complemented each other in communicating emotions. Vocal delivery and the lyrics of the song were also found to be among the various factors that influence people's perception of music mood (Lee et al., 2012). In comparing the effects of melodies and lyrics on mood perceptions, studies found that, although melodies were more dominant than lyrics, lyrics were particularly useful in strengthening emotional responses to music with negative (Ali and Peynircioglu, 2006) and low-arousal emotions (Kazuma, 2009), which roughly correspond to Clusters 5 and 3, respectively, in this study. Even though the US listeners probably did not understand the lyrics in those songs, they might still have been able to get some emotional clues from the vocal delivery. 5.3.2 Genres/styles of Chinese music Table VII shows the agreement ratios within and across cultural groups with regard to the six types of music. The Hong Kong listeners showed an agreement of 60 percent on Cantopop, followed by 54 percent on China wind. The US listeners agreed more on Canto folk and China wind (43 percent), followed by Cantopop (41 percent). Cross-culturally, kh2 tests show that the two groups significantly differed in their mood judgments on all styles except for Cantopop (kh2=9.22, df=10, p=0.10). We conjecture two possible reasons for Cantopop to reach a high cross-cultural agreement ratio: as introduced before, Cantopop was developed with heavy influence from western pop music, and thus the western elements could have helped to bridge the differences between the two user groups; and all Cantopop songs are lyrical and thus the mood cues in the vocal delivery may have helped strengthen mood perceptions. The category with the least agreement both within and across cultures was contemporary instrumental music. This may be due to the complex musical compositions of this style and the lack of lyrical cues. In fact, the song with the least agreement was of this type and will be analyzed in more detail in the next subsection. To facilitate further examination of the mood judgments on different types of music stimuli, Figure 2 shows the distribution of mood judgments for each of the six music types included in this study. It can be seen that the first three types of music (i.e. instrumental folk, Canto folk and contemporary instrumental) were labeled with a mix of diverse mood clusters without a clear, dominant cluster, whereas the last 3 (i.e. Cantopop, China wind and Chinese rock) have one dominant cluster from both cultural groups: China wind and Cantopop were mostly labeled with Cluster 3 (literate, poignant, wistful, etc.) and Chinese rock with Cluster 1 (passionate, rousing, confident, etc.). Instrumental folk music seems to be the most controversial, in that the pieces got similar numbers of judgments across all mood clusters, especially from the US group. The two cultural groups seemed to disagree most on Canto folk and contemporary instrumental music. Hong Kong listeners categorized Canto folk music mostly in Clusters and 2 (rollicking, cheerful, fun, etc.) but US listeners tended to label those pieces more often with Cluster 2 and Cluster 4 (humorous, silly, campy, etc.). For contemporary instrumental music, US listeners chose Cluster 2 the most, followed by similar numbers for Cluster 3, Cluster 1 and "Other," while Hong Kong listeners used Cluster 3 more often, followed by Cluster 2, Cluster 1 and Cluster 4. 5.3.3 The songs with highest and lowest agreement The song that reached the highest agreement on mood judgment was the same within each cultural group and across cultural groups. In fact, the US listeners agreed more than Hong Kong listeners, with agreement ratios of 0.94 and 0.88, respectively. This was a China wind song titled Pearl-decorated Curtain Rolled Up by a Chinese male singer, Huozun. In all, 30 Hong Kong listeners and 31 US listeners selected Cluster 3 (literate, poignant, wistful, etc.) for this song. Released in 2014, the song features a slow melody and extensive use of falsettos. It can be characterized as mellow, ethereal and graceful, similar to the style of ancient Chinese music (and thus recognized as China wind). Although the lyrics describe a beautiful young woman missing a lost love in a poetic way, the US listeners probably could not understand them. Yet they reached near consensus on the mood of this song, possibly attributable to the emotionally powerful delivery of the song. The song with the lowest agreement among US listeners as well as cross-culturally was Shangri-la by Girls 12, a female band famous for using traditional Chinese instruments to play both traditional Chinese and western music. This song had the second-lowest agreement among Hong Kong listeners, and their lowest agreement was for another piece by the same band. Many pieces by Girls 12 are modern compositions using ancient instruments, including Shangri-la, and thus are classified as contemporary instrumental music, the type that has the least agreement both within and across cultures (Table VII). The song Shangri-la received only 16 percent agreement from US listeners and 21 percent from Hong Kong listeners. Figure 3 illustrates the distribution of mood judgments of this song. We can observe that the judgments from Hong Kong listeners were distributed fairly evenly across Cluster 1 (passionate, rousing, confident, etc.), Cluster 2 (rollicking, cheerful, fun, etc.), Cluster 3 (literate, poignant, wistful, etc.) and Cluster 4 (humorous, silly, campy, etc.), and about equal numbers of US listeners labeled it with Cluster 1, Cluster 2, Cluster 5 (aggressive, fiery, tense/anxious, intense, etc.) and "Other." The piece was performed with traditional Chinese instruments: bamboo flute, erhu (bowed strings), pipa and guzheng (both plucked strings), mixed with western-sounding percussion and electronica fusions. Just as the song title suggests, the piece conjures the mystic and mysterious aura of the utopia named Shangri-la. The mixed mood perceptions of this song can be at least partially explained by the mixed musical nature of the song, as well as its fictional, mystical theme. 5.4 Listener characteristics 5.4.1 Gender Table VIII shows the distribution of mood judgments based on gender across all mood clusters. Both genders in both cultural groups selected Cluster 3 the most. The kh2 test indicates that gender difference in mood judgment does exist within the US group (kh2=24.62, df=5, p <0.001), but not in the Hong Kong group (kh2=7.53, df=5, p<0.184). This was also the case in Lee and Hu (2014), where American listeners' mood judgments on western music differed based on gender, but Chinese listeners' did not. In order to see the influence of gender difference across cultural groups, we compared the judgments from US and Hong Kong listeners of the same gender. The comparison shows female listeners in the two cultural groups did have different mood judgments (kh 2=46.51, df=5, p<0.001). However, the judgments of male listeners from the two cultural groups did not have a significant difference (kh2=7.99, df=5, p=0.157). The results suggest that cultural differences played a significant role in mood judgments among female listeners, but not among male listeners. In contrast, Hu and Lee (2012) found significant differences for both genders between US and Chinese listeners when the music stimuli were western pop songs. This discrepancy might have been affected by the small sample size of male participants in the US group (n=8) in this study. Future studies with a larger and more balanced sample across gender groups are warranted to further investigate the extent to which gender and cultural groups affect music mood perception. 5.4.2 Age The ages of the participants vary between the two groups, but both groups have listeners between 21 and 60 years old. We divided those listeners into two age groups, 21-33 and 34-60, so as to balance the sample sizes in both age groups. The kh2 tests on the distribution of judgments revealed a statistically significant difference between age groups among Hong Kong listeners (kh2=39.86, df=5, p<0.001) but no significant difference among US listeners. (kh2=4.72, df=5, p=0.45). This is, in fact, consistent with the findings from Lee and Hu (2014), where the authors found a significant difference between the age groups for Korean and Chinese music listeners but not for American listeners. This could be possibly attributed to the fast-changing environment in Hong Kong and other Asian countries compared to the relatively stable environment in countries such as the USA (Lee and Hu, 2014). 5.4.3 Familiarity with the music Listeners' familiarity with the songs was categorized into three levels: "high familiarity" if they had listened to the song before and could also identify the artist's name and the song title; "medium familiarity" if they had heard the song but could not name it; and no familiarity if they had not heard the song. There was one US listener who was highly familiar with two songs, and seven who, collectively, had heard a total of 15 songs before. In the Hong Kong group, 29 out 32 listeners indicated high familiarity with at least one song. Unsurprisingly, more than half of the highly familiar songs were Cantopop. Due to the sparseness of data in the US group, we examined how the levels of agreement related to familiarity with the music only in the Hong Kong group. Table IX lists the agreement ratios among judgment pairs with regard to various levels of music familiarity. When both participants were highly familiar with a song, their agreement was the highest (ratio=0.57). While a higher familiarity level helped increase the level of agreement when both mood judgments were made for songs with which participants were familiar, it did not help when one participant was familiar with the song and the other was not. We conjecture that when users have heard a song multiple times or know the artist name and the song title, they may have developed and/or gained some contextual knowledge of the song, which may affect their selection of mood clusters. On the other hand, when listeners hear the song for the first time, they may naturally rely more on the musical characteristics for judging the mood of the song. As advocated by Weissenberger (2015), a flexible organization system is needed for MIR/MDL to meet the needs of different musical traditions. The results of this study have important implications for designing MDL of this kind. As users in different cultural groups had significantly different mood judgments overall (Section 5.2.1), it is better to differentiate user-generated mood labels and/or mood-related social tags according to the users' cultural background. In particular, as the agreement ratio among Hong Kong listeners was higher than the cross-cultural one (Section 5.2.2, Table IV), Hong Kong listeners would benefit more from mood tags supplied by listeners in the same cultural group. In contrast, as the within-group agreement ratio of US listeners was similar to the cross-cultural agreement level (0.38 vs 0.39), US listeners could make use of mood tags generated by Hong Kong listeners, especially when those generated by US listeners are sparse. The analysis on the textual input in the "Other" option (Section 5.2.1) suggests that the MIREX five-cluster model may not be sufficient and more terms are needed to describe the mood of Chinese music. Particularly, terms like "calm," "relaxing," "peaceful" and "longing" appeared most frequently (Table III, Section 5.2.1). In fact, a previous study on automated mood classification on Chinese pop songs (Yang and Hu, 2012) also used mood terms in the MIREX five-cluster model, and ended up with adding additional mood labels that were deemed as important to represent moods in Chinese pop songs: "calm/peaceful," "dreamy," "encouraging," "nostalgic," "relaxed," "soothing" and "tender." Based on this finding, it is suggested a sixth cluster can be added to the MIREX five-cluster model: (calm, relaxing, peaceful, etc.). In addition, some of the suggested terms in the current study were mixtures of terms from different clusters, which implies that the mood clusters may not be mutually exclusive, at least for US listeners. Lee and Hu (2014) also found that some songs may be perceived to be in two or more clusters at the same time. Therefore, it is suggested that the MIREX five-cluster model should allow multiple mood clusters to be applied to each song. As US listeners specified terms that describe visual scenes, it might be a good idea to use scenes as a new access point in MDL. The association between music and images has been studied in music psychology (Ali and Peynircioglu, 2006) and MIR (Mattek and Casey, 2011), but how it can be used in MDL warrants future research to explore. The fact that both cultural groups chose Cluster 3 (literate, poignant, wistful, etc.) most frequently (Table II) and agreed more on songs labeled with Cluster 3 (Table V) suggests that Chinese music may tend to have musical traits that are associated with moods in Cluster 3, and that such musical traits can transcend the cultural boundaries of the audience. MDL using automated mood classification algorithms (Yang and Chen, 2012) could exploit this finding and assign more weight to the classifier of Cluster 3, allowing songs to have a greater chance of being classified in Cluster 3. Similarly, when recommending songs based on mood labels, those labeled with Cluster 3 could be assigned more confidence. As different styles had different agreement levels within and across cultures (Table VII), music in different styles can be treated differently in MDL. For example, contemporary instrumental pieces had low agreement levels between audiences, and thus classification and recommendations of pieces in this style could be given less confidence or provide more candidates to improve the chance of hits. The results on gender and age group differences can be exploited, too. For instance, Hong Kong listeners would benefit more from recommendations and tags made by other Hong Kong listeners in the same age group, whereas for US listeners there is no need to differentiate the age group of other listeners. The effect of song familiarity on mood perception suggests that MDL should allow users to change their mood labels and tags over time, because when they get more familiar with a song, they may perceive the mood of the song differently. This study compares Hong Kong and US listeners' mood perceptions of 29 Chinese music pieces, with the goal of investigating whether and how mood perceptions of the two groups of listeners differ, and how the differences can inform the design of cross-cultural MDL. Music mood was modeled with the MIREX five mood clusters and the results suggested further refinement of this model. The selected music pieces included six genres and styles of Chinese music, ranging from traditional folk music to several sub-genres of C-pop. Overall, the patterns of mood judgments by listeners from Hong Kong and the USA (e.g. commonly agreed-upon clusters, differences based on gender and age) resonated with the findings in previous studies on cross-cultural mood perceptions of western music, even though this study used Chinese music, which has a contrasting cultural background to that of western music. This has important implications for designing global MDL: for both western and Chinese pop music, listeners from different cultures may need different mood labels. For systems that use social tags to facilitate access, it is recommended that tags given by users in different cultural groups be differentiated. Those given by users in the same cultural groups can be more helpful. This study also confirms that listeners do exhibit cross-cultural differences in mood judgment for all but one type (Cantopop) of Chinese music considered in this study. Therefore, it may be desirable for MDL to differentiate music types when classifying or recommending music by mood. Future work will include further investigation on other cultural groups interested in Chinese music, such as mainland Chinese, Taiwanese and Singaporeans. In addition, we plan to conduct similar studies on music from other cultural backgrounds, such as Korean and Japanese music. Findings of such studies will further contribute to the goal of building effective and user-centered global MDL.
|
An online survey was designed based on the Music Information Retrieval Evaluation eXchange (MIREX) five-cluster mood model, to solicit mood perceptions of listeners in Hong Kong and the USA on a diverse set of Chinese music. Statistical analysis was applied to compare responses from the two user groups, with consideration of different music types and characteristics of listeners. Listeners' textual responses were also analyzed with content coding.
|
[SECTION: Findings] Music seeking and consumption are no longer confined by the boundaries of country, region or culture today (Lee et al., 2013). Music, as a cultural object, may be perceived differently by people from different cultural backgrounds, imposing a challenge on music digital libraries (MDL) in meeting the needs of a diverse audience (Weissenberger, 2015). Consequently, an increasing number of researchers have started to investigate various cross-cultural issues in the music information retrieval (MIR) and MDL fields. As people often seek music for emotional goals (Lavranos et al., 2015), music mood[1] has increasingly become a popular access point for music information in many MDL and online music services (Hu, 2010; Hu and Downie, 2007). This trend has raised questions regarding the applicability of music mood across cultural boundaries. Probably due to its subjective and context-based nature, music mood perception is often regarded as culture-dependent (Wong et al., 2009). A number of previous studies have compared mood perceptions of music in various cultures by listeners from different cultural backgrounds (e.g. Balkwill and Thompson, 1999; Fritz et al., 2009; Hu and Lee, 2012; Lee et al., 2013; Singhi and Brown, 2014; Egermann et al., 2014). Although specific findings vary, a general trend found in these existing studies is that listeners' perceptions of music mood can be influenced by their cultural backgrounds. However, most existing studies have focussed on western (classical or pop) music or ethnic music with small audiences (e.g. Congolese music produced and consumed by Congolese Pygmies (Egermann et al., 2014)). Such studies mainly aim to answer questions on music psychology and/or the social, developmental aspects of music. The purpose of this study is different; we are interested in exploring how music mood could be used as an access point or metadata type that can transcend the cultural boundaries of users in the context of MDL. To date, the question of how people with different cultural backgrounds would perceive the mood of Chinese music remains under-investigated. Considering its appeal to one of the largest music listener populations in the world (IFPI, 2014), Chinese music should not be left out of today's global scope of MDL/MIR studies. As Chinese music possesses unique characteristics that are remarkably different from western music (Fung, 2013), it is of theoretical and practical value to investigate the case of Chinese music. Therefore, in this study, we aim to fill this gap and explore how people with different cultural backgrounds, specifically those in the USA and Hong Kong (representing western and eastern listeners, respectively), perceive the mood of Chinese music, including traditional Chinese music and contemporary Chinese pop music that infuses western pop characteristics and local Chinese elements. We specifically answer the following research questions: RQ1. How differently or similarly do Hong Kong and US music listeners perceive the mood of Chinese music? RQ2. How differently or similarly do Hong Kong and US music listeners perceive the mood of different genres/styles of Chinese music? RQ3. Do the listeners' demographic characteristics and familiarity with the music affect their mood perceptions of Chinese music? Answers to these questions have important implications for MDL that use mood as an access point to facilitate MIR and discovery in a global environment. These answers will help determine whether it is necessary to develop different mood-related labels or tags that take their suitability for various user groups into consideration. Implications of the results in this study will be discussed. The goal of this study is to contribute to the literature of cross-cultural MDL/MIR and provide empirical evidence for designing cross-cultural MDL/MIR systems. 2.1 A brief history and types of Chinese music Chinese music broadly refers to music that uses traditional Chinese instruments or is in the Chinese language, and is the main music consumed in the region of Greater China (i.e. regions where Chinese is the official language, including mainland China, Hong Kong, Macau, Taiwan and Singapore). However, the consumption of Chinese music is not limited to that region. People around of the world are accessing Chinese music, and all major online music services, such as Spotify and Last.fm, contain Chinese music to cater to the demand. There is a broad range of genres and styles in Chinese music. Traditional Chinese folk music dates back to approximately 7,000 years ago, played mostly in the pentatonic scale with its own set of instruments. Contemporary Chinese pop music (also referred to as C-pop) was developed in the middle of the twentieth century with the influence of western music. The subgenre "Cantopop" (Hong Kong pop music sung in Cantonese) was developed in the 1970s, followed by the rise of another subgenre, "Mandopop" (Mandarin Chinese pop music), in the 1980s (Fung, 2013). This study includes six genres/styles of Chinese music that are widely appreciated in the region and around the world. The first two are traditional Chinese music, whereas the rest are the most popular C-pop styles featured in online music services in China (e.g. Baidu, Xiami, Kugou): Traditional Instrumental Folk Music of China (qiyue): on the pentatonic scale, musicians perform using traditional instrument(s) of woodwind and percussion, bowed strings, plucked and struck strings in solo, small ensembles or orchestra (Randel, 2003). Canto(nese) folk songs: Han Chinese folk songs are divided into northern and southern styles geographically. Provinces south of the Yangzi River have mild weather and abundant rain, and therefore the folk songs tend to be more "lyrical, gentle and conjunct" (Han, 1989). Near the Canton province and Pearl River Delta, Cantonese folk music is especially light-spirited (China Culture, 2006). Contemporary Chinese instrumental music: this style of music is also called "New Music of China." It mixes the techniques, instrumentation and composition of both Chinese traditional music and the European/western music (Liu and Mason, 2010). The Girls 12 is a famous mainland Chinese all-female group who plays this type of music with traditional Chinese instruments. Cantopop (or Cantonese popular music): this is a subgenre of C-pop. Sung in Cantonese, Cantopop usually refers to popular songs produced in Hong Kong. Blended with traditional Chinese elements, as well as western-influenced harmonies and melodies, Cantopop was a dominant music genre in Asia in the 1980s' and 1990s. Its iconic star, Anita Mui, was recognized as "Asia's Madonna." China wind (zhongguofeng): this is a subgenre of Mandopop and is also called "Chinese Style." It features traditional Chinese instruments and classical Chinese melody juxtaposed with popular western music styles and cultural references to Chinese history, classics and folklore in the lyrics (Xinhuanet, 2006). Taiwanese singer Jay Chou is regarded as the "godfather" of this type of music, bringing forth the wave of China Wind pop. Chinese rock: as another subgenre of Mandopop, Chinese rock is locally produced in mainland China and is the most notable sub-category of Mandopop to the world outside of China. Chinese rock combines the loud and forceful folk song traditions of Northern China and characteristics of western-style rock such as a fast tempo, a strong beat and aggressive bass lines (Fung, 2013). The singer/songwriter Cui Jian from mainland China has been widely recognized as the father of Chinese rock (Fung, 2013). It is noteworthy that all C-pop styles are, to varying extents, influenced by western pop or Rock music. Compared to Traditional Instrumental Chinese Folk Music and Canto Folk Songs, C-pop may be easier to understand and appreciate for listeners familiar with western music. In addition, because of the mixture of western music elements, listeners from both cultural groups in this study might share more common perceptions on C-pop than the other two genres/styles. Therefore, in addition to analyzing aggregated perceptions across all songs, this study also compares and contrasts listeners' perceptions on these six music styles. 2.2 Cross-cultural studies on music mood perceptions A number of studies have investigated the effects of culture on music mood perception. Although fundamental emotions such as happiness, sadness, anger or fear were found to be generally agreed upon across cultures, cultural traditions and background were found to affect individuals' music mood perception, in addition to the inherent qualities of the music (e.g. Gregory and Varney, 1996; Kosta et al., 2013; Lee et al., 2013). For example, Wong et al. (2009) compared mood descriptors used by European and Indian communities and discovered that cultural tradition played a more crucial role than the inherent qualities of the music in users' mood perception. Fritz et al. (2009) found that mood perceptions of western music by African listeners partially agreed with those by western listeners. In a similar vein, Hu and Lee (2012) discovered that discrepancies existed between Chinese and American music listeners in the way they determined the mood of western music pieces. More recently, Lee and Hu (2014) also compared perceptions of western music mood among American, Korean and Chinese listeners, and found that the results of Korean listeners' mood perceptions were situated in between the two other groups. There are also studies investigating the cultural backgrounds of listeners in relation to their mood perceptions of non-western music. Kosta et al. (2013) found greater agreement on the mood of Greek songs among Greek listeners compared to non-Greek listeners with varying backgrounds. Similar results were found by Hu et al. (2014) in comparing American and Korean listeners' mood perceptions of a set of Korean pop songs. Egermann et al. (2014) compared subjective and psychophysiological emotional responses by Canadians and Congolese Pygmies to both western and Congolese music. They found that the two groups had very different responses to Pygmies music but similar responses to some western music. To the best of our knowledge, there has been little work focussed on comparing the mood perceptions of various genres/styles of Chinese music across different cultural groups. In addition, many of the previous studies were conducted in the context of music psychology or ethnomusicology. This study, however, reflects the perspective of information organization and access, with the goal of finding out how music mood can be organized to best meet the information needs of users from different cultural groups. There are mainly two kinds of models to represent mood in music psychology and MIR: categorical and dimensional. The former uses a set of discrete terms (e.g. "passionate," "cheerful") to represent the mood of a piece of music. The most classical model of this kind is Hevner's (1936) adjective circle where eight mood categories placed in a circle, with a set of terms in each category. The dimensional models, in contrast, represent mood with continuous values in a low dimensional space. Different models may have different dimensions, yet valence (i.e. level of pleasure) and arousal (i.e. level of energy) are among the most popular dimensions. The dimensional model used very often in MIR is Russell's (1980) model. The categorical and dimensional models have their own pros and cons. Yang and Chen (2012) summarize that categorical models are more user-friendly as they only consist of terms in natural language; whereas dimensional models are advantageous in terms of quantifying the intensity of moods. Notwithstanding the importance of dimensional models, categorical ones are more suitable for this current study, for the purpose of validating and comparing the values of mood metadata in the context of cross-cultural MDLs. The models from music psychology have solid theoretical ground and can be borrowed into MIR studies. However, one criticism of adopting these models directly in MIR is that they were developed in experimental settings and thus might not reflect users' preferences and behaviors in today's music listening environment (Hu, 2010). Therefore, MIR researchers developed mood models based on domain analysis of widely used music repositories and streaming services. One of the influential mood models in MIR is the five-cluster model proposed in the audio music mood classification task (Hu et al., 2008) of Music Information Retrieval Evaluation eXchange (MIREX), a community-wide MIR evaluation campaign held annually since 2005 (Downie et al., 2014). This model was developed from more than 180 editorial mood labels on AllMusic.com, a very popular online music repository. It has been used in many music mood classification studies (e.g. Laurier et al., 2008; Bischoff et al., 2009; Yang and Chen, 2012) as well as cross-cultural user studies (Hu and Lee, 2012; Lee and Hu, 2014; Hu et al., 2014; Yang and Hu, 2012). The five mood clusters are: C1: passionate, rousing, confident, boisterous, rowdy; C2: rollicking, cheerful, fun, sweet, amiable/good natured; C3: literate, poignant, wistful, bittersweet, autumnal, brooding; C4: humorous, silly, campy, quirky, whimsical, witty, wry; and C5: aggressive, fiery, tense/anxious, intense, volatile, visceral. In this five-cluster model, each cluster is collectively represented by a set of mood terms that share similar semantics. As the terms are describing affect, which can be somewhat vague (as opposed to terms referring to concrete objects), it is argued that using multiple terms with similar meanings to collectively describe one kind of mood is clearer than single-word representations (Hu and Downie, 2007). A more recent categorical mood model was also used in MIREX, in the mood subtask of the audio tag classification task (Hu et al., 2009). It consists of 18 mood categories derived computationally from mood-related social tags from Last.fm, arguably the largest music tagging and sharing website. This 18-class model consists of 135 tags. Besides the automated tag classification task in MIREX, a simplified version of this model has been used in a cross-cultural user study on Korean music (Hu et al., 2014), together with the five-cluster model. As the MIREX five-cluster mood model has been used in a number of cross-cultural MIR studies, for comparison purposes, the current study adopts this model and attempts to make suggestions on modifications and refinement based on the results. 4.1 The participant groups The user groups recruited in this study consist of music listeners from Hong Kong and the USA. Hong Kong plays a special role in the history of Chinese music. Located adjacent to the Canton province, Hong Kong was the origin of Cantopop, which highly influenced the entire region of East Asia in the 1980s and 1990s. Furthermore, for historical and political reasons, Hong Kong was able to preserve the traditional Chinese culture to a large extent. Canto folk music and even traditional Canto opera are often appreciated and practiced by Hongkongers. Since the government handover in 1997, people in Hong Kong also have had increasing exposure to Mandopop. Therefore, it makes an interesting case to investigate Hongkongers' perceptions of a variety of Chinese music types, since they are culturally exposed to all of them to certain degrees. On the other hand, US listeners can represent people with the western cultural background, and, additionally, they have been found to exhibit different music consumption behaviors than Hongkongers in a prior study (Nettamo et al., 2006). We would like to explore whether differences also exist in the two user groups' mood perceptions of non-western music. Invitations were distributed to a number of departmental student mailing lists at the University of Hong Kong and University of Washington, as well as via the authors' various social network websites, such as Twitter, Facebook and Google Plus. Participation was completely voluntary. 4.2 Song selection This study used an online questionnaire of 29 clips of Chinese music to solicit listeners' music mood perceptions. The clips were manually selected from the authors' personal music collections, with the goal of reflecting the broad range of Chinese music. As the goal of this study is to propose suggestions for global MIR systems, two kinds of Chinese music were considered: traditional Chinese music and contemporary Chinese pop music. The former features distinct cultural origin and musical characteristics, and thus can be attractive to users who do not have a Chinese background but are curious about Chinese or Asian music. The latter (C-pop) is widely consumed by Chinese-speaking listeners around the world. As it is heavily influenced by western pop music, yet has Chinese flavor, it is potentially of interest to users in all regions who like pop music in general. Specifically, the selected songs represented the following six genres or styles described in Section 2.1: traditional Chinese instrumental music, traditional Cantonese (Canto) folk music, contemporary Chinese instrumental music, Cantopop and two unique styles of Mandopop: China wind and Chinese rock. Each type contains five representative pieces except for Chine wind which contains four. The list of titles can be found at: http://ccmir.cite.hku.hk/pubs/moodCH/List_of_Songs.xlsx 4.3 The questionnaire At the beginning of the questionnaire, participants were asked about their gender, year of birth and the place where they grew up. The last question was used for screening purposes so as to ensure that only participants with Hong Kong and US cultural backgrounds were included. We believe that people are greatly affected by the culture of the place they resided when they grew up, regardless of their nationality. All participants were 18 years of age or above and agreed with the informed consent statement before proceeding to the questionnaire. Next, participants were asked to listen to the selected music clips, with one song on each page. The questionnaire consisted of five questions for each clip. The first question asked the participant to listen to the clip and decide the most appropriate mood expressed by the song. The response options given were the mood clusters of the MIREX five-cluster model. In case the participants found that the mood of the song did not fit into any of the five default clusters, they could select the option "Other," and use their own words to describe the appropriate mood of the music clip. The participants were also asked about their familiarity with the music clips. Because people may interpret "familiarity" differently, instead of directly asking how familiar they are with the given songs, we used two questions: whether they had heard the song before, and if they could provide the artist name and the title of the song. Responses to the two questions were deemed more objective than self-rated familiarity because the questions are fact-based. The questions were presented in English in the US questionnaire and in both English and Chinese in the Hong Kong questionnaire. In the Hong Kong questionnaire, the five mood clusters were also translated into Chinese, and both English and Chinese terms were presented. We used the same translation of mood terms as in Yang and Hu (2012). Even though English is one of the official languages in Hong Kong, people's English proficiency varies. Therefore, both Chinese translations and original terms in English were deemed necessary for the respondents to have an accurate interpretation of the clusters. Figure 1 illustrates the survey interface of the Hong Kong version. 4.4 Data analysis Responses to the first question for each song (i.e. mood judgment) were compared quantitatively. The kh2 test was used for comparing the distributions of music mood judgments between the two groups of listeners (Hinkle et al., 2003). Significant results would indicate that the judgment distributions are dependent on cultural groups. The agreement between groups of listeners in various combinations of music and listener characteristics was measured by the raw ratio of agreed pairs to all possible pairs. For example, for calculating the agreement ratio among 32 Hong Kong listeners on five Cantopop songs, we first calculated the total number of judgment pairs as (32x(32-1)/2)x5=2,480, and counted the number of pairs with the same mood judgments. Suppose there were 248 matched pairs; the agreement ratio would be 10 percent. Raw agreement measures are recommended for cases where the annotations are dominated by negative values (i.e. "not chosen") (Cicchetti and Feinstein, 1990). This applies to the mood judgment data collected in this study. For instance, "Other" was chosen only 102 times out of 1,856 times it could have been selected. Text responses to the "Other" option were coded with a scheme of categories that emerged from the analysis process. The codebook was developed through an iterative process involving test-coding and revising the codes for clarity. The consensus model (Hill et al., 1997) was applied to coding the responses: two independent coders coded the responses and then discussed on disagreed cases until a consensus was reached. The results were compared across the two groups. Responses to demographic questions and song familiarity questions were used to group the users with a range of criteria, for the purpose of testing factors that affect users' mood perceptions. 4.5 Limitations As each music piece needs to have a minimum length (typically 15-30 seconds) for listeners to establish a confident perception of its mood (Speck et al., 2011), the questionnaire can contain only a limited number of song clips so that it can be completed within a reasonable amount of time. We used a set of 29 clips in this study, which may seem rather limited given there are millions of Chinese songs available in the market. To mitigate this limitation, we tried our best to carefully select a good variety of songs based on demands for them by users with various cultural backgrounds, and representativeness in each category. In addition, the sample size of each cultural group (32) is limited, partially due to the length of the survey and the lack of monetary incentive. Nonetheless, it is sufficient for statistical tests and comparisons. The online venues used to recruit participants (i.e. mailing lists and social network sites) may also have limited the participants to frequent internet users. Future studies are warranted to further verify the evidence found in this exploratory study in order to generalize findings to broader contexts. 5.1 Characteristics of participants The Hong Kong survey had a total of 70 responses, among which 34 were completed. The US survey had 72 responses, of which 34 were completed. However, two listeners in each group answered that they were raised in mainland China (not Hong Kong or the USA), leaving 32 valid responses for each group. Table I shows the demographic information of the listeners. 5.2 Mood judgments 5.2.1 Cultural backgrounds and mood judgments Table II presents the distribution of mood judgments of the two groups on all 29 songs. Both groups chose Cluster 3 (literate, poignant, wistful, etc.) most often, indicating that many Chinese songs may have some traits associated with the mood of Cluster 3, and these traits can be perceived by both Hong Kong and US listeners. The kh2 test shows that the distribution and cultural group are not independent (kh2=52.54, df=5, p<0.001). In other words, it provides strong evidence that Hong Kong and US listeners tend to perceive the mood of given Chinese music differently, within the framework of five mood clusters. From Table II, we can observe that Hong Kong listeners tended to select Cluster 1 (passionate, rousing, confident, etc.) and Cluster 3 while US listeners tended to spread their choices equally across Clusters 1, 2 (rollicking, cheerful, fun, etc.) and 3. The US listeners were also more likely to select the "Other" category than their Hong Kong counterparts. In all, 26 out of the 29 songs were labeled with "Other" at least once by the US group. In comparison, the Hong Kong group chose the "Other" option for 17 songs. On average, 2.93 US listeners chose "Other" for each of these songs, whereas the number for Hong Kong listeners was 1.35. This is similar to the findings in Lee and Hu (2014), where US listeners chose the "Other" category more often than Korean and Chinese listeners when they made mood judgments on western songs. Even though the objects/stimuli have been changed to Chinese music, the observation still holds. The consistent pattern regardless of the cultural background of the music further suggests that this phenomenon might be related to the individualism-collectivism difference between western and eastern cultures as discussed in previous studies (Boer and Fischer, 2010; Hu and Lee, 2012; Lee and Hu, 2014): western users tend to hold their opinions regardless of the given options, whereas Eastern people are more likely to conform to existing norms. We conducted a content analysis on the free-text responses provided in the "Other" option. The two independent coders reached an excellent reliability level (Cohen's k=0.858). Table III shows the seven resultant categories and corresponding counts in each of the groups. The largest category of the self-supplied terms is new mood terms that are not included in the five-cluster scheme. The new mood terms provided by both groups are quite similar, with, "relaxing," "calm" and "longing" being the most frequent, appearing six, four and four times, respectively. Users in both cultural groups also filled in terms that were not mood related. Some of them were music specific (e.g. "fast pace") while others were not (e.g. "focus"). Interestingly, the US group also provided responses that were either a mixture of terms from multiple clusters (e.g. "passionate" from Cluster 1 and "intense" from Cluster 5) or part of the terms in one cluster (e.g. "poignant, wistful, autumnal" in Cluster 3). The US participants also filled in descriptions of scenes they imagined when hearing the songs, although the instructions specifically asked for mood-related terms. Perhaps this group of users thought that scenes inspired by the songs also reflected the mood to some extent. The Hong Kong listeners did not provide any terms in these categories, possibly because: they tried harder to follow the instructions that asked respondents to "choose one dominant cluster when a song sounds to have moods in more than one cluster," and people from individualistic (mostly western) cultures were found to be more used to divergent thinking which involves the generation of novel ideas (Schmidt et al., 2013). 5.2.2 Agreement on mood judgments within and across culture groups The agreement ratios were calculated for pairs of judgments between two listeners from the same cultural group and across cultural groups. The results are shown in Table IV. In particular, in the cross-cultural case, the number of total pairs with one judgment from each cultural group is calculated as ((32x32)/2)x29=29,696. The results and non-paired t tests reveal that the level of intra-agreement among Hong Kong listeners is significantly higher than that of US listeners (p<0.001). A similar result was reported in Kosta et al. (2013), where Greek listeners were found to agree more on the mood of Greek music than non-Greek listeners. In comparing Korean and American listeners' perceptions of the mood of Korean pop songs, Hu et al. (2014) also found that Korean listeners reached a higher agreement level than their US counterparts. However, the result is partially different from Hu and Lee (2012), where both US and Chinese listeners had the same agreement ratio on the mood of western pop songs. We conjecture that this difference in results between Hu et al. (2014) and Hu and Lee (2012) might be due to Chinese listeners' wide exposure to western pop songs. For non-western songs that have not yet spread as widely, sharing the same cultural background as the music stimuli does seem to help listeners reach a higher agreement on music mood. It is also noteworthy that the agreement level among US listeners was not statistically different from the cross-cultural agreement (p=0.54). This is similar to the result in Hu et al. (2014), where the agreement level among American listeners on the mood of Korean pop songs was the same as that between American and Korean listeners. 5.2.3 Agreement on mood clusters To examine which moods are easier or more difficult to agree upon, we counted the number and percentage of agreed pairs across the five mood clusters (Table V). The kh2 test indicates that the two cultural groups showed a statistically significant difference in the agreement pattern across the mood clusters (kh2=603.55, df=10, p<0.001). Both within and across cultures, Cluster 3 (literate, poignant, wistful, etc.) got the most agreements, which again evidences that many Chinese pop songs have traits of Cluster 3 and these traits can transcend the cultural boundaries of listeners to some extent. The difference between the two cultural groups observed here is similar to the findings in Hu and Lee (2012), where American listeners agreed more on Cluster 2 (rollicking, cheerful, fun, etc.), Cluster 4 (humorous, silly, campy, etc.) and Cluster 5 (aggressive, fiery, tense/anxious, etc.) compared to Chinese listeners, who agreed more on Cluster 1 (passionate, rousing, confident, etc.) and Cluster 3 (literate, poignant, wistful, etc.). As western music was used as the stimuli in Hu and Lee (2012), it is interesting to see that the patterns of mood perception are similar to this study even though the stimuli here consist of completely different, non-western music. 5.3 Mood perception on different types of music 5.3.1 Instrumental vs vocal The music pieces in this study could be divided into instrumental and vocal sets. The former includes 15 pieces in instrumental folk, Canto folk or contemporary instrumental styles, and the latter includes the rest of the three types with 14 pieces total. We can observe from Table VI that overall, Hong Kong listeners agreed among themselves more often than US listeners on both instrumental and vocal pieces. Listeners also agreed more often on vocal pieces than instrumental music within each cultural group, as well as cross-culturally. This is in accordance with the literature on both music psychology and user studies in MDL/MIR. Juslin (2005) argued that the verbal (lyric) and non-verbal (melody) channels of music complemented each other in communicating emotions. Vocal delivery and the lyrics of the song were also found to be among the various factors that influence people's perception of music mood (Lee et al., 2012). In comparing the effects of melodies and lyrics on mood perceptions, studies found that, although melodies were more dominant than lyrics, lyrics were particularly useful in strengthening emotional responses to music with negative (Ali and Peynircioglu, 2006) and low-arousal emotions (Kazuma, 2009), which roughly correspond to Clusters 5 and 3, respectively, in this study. Even though the US listeners probably did not understand the lyrics in those songs, they might still have been able to get some emotional clues from the vocal delivery. 5.3.2 Genres/styles of Chinese music Table VII shows the agreement ratios within and across cultural groups with regard to the six types of music. The Hong Kong listeners showed an agreement of 60 percent on Cantopop, followed by 54 percent on China wind. The US listeners agreed more on Canto folk and China wind (43 percent), followed by Cantopop (41 percent). Cross-culturally, kh2 tests show that the two groups significantly differed in their mood judgments on all styles except for Cantopop (kh2=9.22, df=10, p=0.10). We conjecture two possible reasons for Cantopop to reach a high cross-cultural agreement ratio: as introduced before, Cantopop was developed with heavy influence from western pop music, and thus the western elements could have helped to bridge the differences between the two user groups; and all Cantopop songs are lyrical and thus the mood cues in the vocal delivery may have helped strengthen mood perceptions. The category with the least agreement both within and across cultures was contemporary instrumental music. This may be due to the complex musical compositions of this style and the lack of lyrical cues. In fact, the song with the least agreement was of this type and will be analyzed in more detail in the next subsection. To facilitate further examination of the mood judgments on different types of music stimuli, Figure 2 shows the distribution of mood judgments for each of the six music types included in this study. It can be seen that the first three types of music (i.e. instrumental folk, Canto folk and contemporary instrumental) were labeled with a mix of diverse mood clusters without a clear, dominant cluster, whereas the last 3 (i.e. Cantopop, China wind and Chinese rock) have one dominant cluster from both cultural groups: China wind and Cantopop were mostly labeled with Cluster 3 (literate, poignant, wistful, etc.) and Chinese rock with Cluster 1 (passionate, rousing, confident, etc.). Instrumental folk music seems to be the most controversial, in that the pieces got similar numbers of judgments across all mood clusters, especially from the US group. The two cultural groups seemed to disagree most on Canto folk and contemporary instrumental music. Hong Kong listeners categorized Canto folk music mostly in Clusters and 2 (rollicking, cheerful, fun, etc.) but US listeners tended to label those pieces more often with Cluster 2 and Cluster 4 (humorous, silly, campy, etc.). For contemporary instrumental music, US listeners chose Cluster 2 the most, followed by similar numbers for Cluster 3, Cluster 1 and "Other," while Hong Kong listeners used Cluster 3 more often, followed by Cluster 2, Cluster 1 and Cluster 4. 5.3.3 The songs with highest and lowest agreement The song that reached the highest agreement on mood judgment was the same within each cultural group and across cultural groups. In fact, the US listeners agreed more than Hong Kong listeners, with agreement ratios of 0.94 and 0.88, respectively. This was a China wind song titled Pearl-decorated Curtain Rolled Up by a Chinese male singer, Huozun. In all, 30 Hong Kong listeners and 31 US listeners selected Cluster 3 (literate, poignant, wistful, etc.) for this song. Released in 2014, the song features a slow melody and extensive use of falsettos. It can be characterized as mellow, ethereal and graceful, similar to the style of ancient Chinese music (and thus recognized as China wind). Although the lyrics describe a beautiful young woman missing a lost love in a poetic way, the US listeners probably could not understand them. Yet they reached near consensus on the mood of this song, possibly attributable to the emotionally powerful delivery of the song. The song with the lowest agreement among US listeners as well as cross-culturally was Shangri-la by Girls 12, a female band famous for using traditional Chinese instruments to play both traditional Chinese and western music. This song had the second-lowest agreement among Hong Kong listeners, and their lowest agreement was for another piece by the same band. Many pieces by Girls 12 are modern compositions using ancient instruments, including Shangri-la, and thus are classified as contemporary instrumental music, the type that has the least agreement both within and across cultures (Table VII). The song Shangri-la received only 16 percent agreement from US listeners and 21 percent from Hong Kong listeners. Figure 3 illustrates the distribution of mood judgments of this song. We can observe that the judgments from Hong Kong listeners were distributed fairly evenly across Cluster 1 (passionate, rousing, confident, etc.), Cluster 2 (rollicking, cheerful, fun, etc.), Cluster 3 (literate, poignant, wistful, etc.) and Cluster 4 (humorous, silly, campy, etc.), and about equal numbers of US listeners labeled it with Cluster 1, Cluster 2, Cluster 5 (aggressive, fiery, tense/anxious, intense, etc.) and "Other." The piece was performed with traditional Chinese instruments: bamboo flute, erhu (bowed strings), pipa and guzheng (both plucked strings), mixed with western-sounding percussion and electronica fusions. Just as the song title suggests, the piece conjures the mystic and mysterious aura of the utopia named Shangri-la. The mixed mood perceptions of this song can be at least partially explained by the mixed musical nature of the song, as well as its fictional, mystical theme. 5.4 Listener characteristics 5.4.1 Gender Table VIII shows the distribution of mood judgments based on gender across all mood clusters. Both genders in both cultural groups selected Cluster 3 the most. The kh2 test indicates that gender difference in mood judgment does exist within the US group (kh2=24.62, df=5, p <0.001), but not in the Hong Kong group (kh2=7.53, df=5, p<0.184). This was also the case in Lee and Hu (2014), where American listeners' mood judgments on western music differed based on gender, but Chinese listeners' did not. In order to see the influence of gender difference across cultural groups, we compared the judgments from US and Hong Kong listeners of the same gender. The comparison shows female listeners in the two cultural groups did have different mood judgments (kh 2=46.51, df=5, p<0.001). However, the judgments of male listeners from the two cultural groups did not have a significant difference (kh2=7.99, df=5, p=0.157). The results suggest that cultural differences played a significant role in mood judgments among female listeners, but not among male listeners. In contrast, Hu and Lee (2012) found significant differences for both genders between US and Chinese listeners when the music stimuli were western pop songs. This discrepancy might have been affected by the small sample size of male participants in the US group (n=8) in this study. Future studies with a larger and more balanced sample across gender groups are warranted to further investigate the extent to which gender and cultural groups affect music mood perception. 5.4.2 Age The ages of the participants vary between the two groups, but both groups have listeners between 21 and 60 years old. We divided those listeners into two age groups, 21-33 and 34-60, so as to balance the sample sizes in both age groups. The kh2 tests on the distribution of judgments revealed a statistically significant difference between age groups among Hong Kong listeners (kh2=39.86, df=5, p<0.001) but no significant difference among US listeners. (kh2=4.72, df=5, p=0.45). This is, in fact, consistent with the findings from Lee and Hu (2014), where the authors found a significant difference between the age groups for Korean and Chinese music listeners but not for American listeners. This could be possibly attributed to the fast-changing environment in Hong Kong and other Asian countries compared to the relatively stable environment in countries such as the USA (Lee and Hu, 2014). 5.4.3 Familiarity with the music Listeners' familiarity with the songs was categorized into three levels: "high familiarity" if they had listened to the song before and could also identify the artist's name and the song title; "medium familiarity" if they had heard the song but could not name it; and no familiarity if they had not heard the song. There was one US listener who was highly familiar with two songs, and seven who, collectively, had heard a total of 15 songs before. In the Hong Kong group, 29 out 32 listeners indicated high familiarity with at least one song. Unsurprisingly, more than half of the highly familiar songs were Cantopop. Due to the sparseness of data in the US group, we examined how the levels of agreement related to familiarity with the music only in the Hong Kong group. Table IX lists the agreement ratios among judgment pairs with regard to various levels of music familiarity. When both participants were highly familiar with a song, their agreement was the highest (ratio=0.57). While a higher familiarity level helped increase the level of agreement when both mood judgments were made for songs with which participants were familiar, it did not help when one participant was familiar with the song and the other was not. We conjecture that when users have heard a song multiple times or know the artist name and the song title, they may have developed and/or gained some contextual knowledge of the song, which may affect their selection of mood clusters. On the other hand, when listeners hear the song for the first time, they may naturally rely more on the musical characteristics for judging the mood of the song. As advocated by Weissenberger (2015), a flexible organization system is needed for MIR/MDL to meet the needs of different musical traditions. The results of this study have important implications for designing MDL of this kind. As users in different cultural groups had significantly different mood judgments overall (Section 5.2.1), it is better to differentiate user-generated mood labels and/or mood-related social tags according to the users' cultural background. In particular, as the agreement ratio among Hong Kong listeners was higher than the cross-cultural one (Section 5.2.2, Table IV), Hong Kong listeners would benefit more from mood tags supplied by listeners in the same cultural group. In contrast, as the within-group agreement ratio of US listeners was similar to the cross-cultural agreement level (0.38 vs 0.39), US listeners could make use of mood tags generated by Hong Kong listeners, especially when those generated by US listeners are sparse. The analysis on the textual input in the "Other" option (Section 5.2.1) suggests that the MIREX five-cluster model may not be sufficient and more terms are needed to describe the mood of Chinese music. Particularly, terms like "calm," "relaxing," "peaceful" and "longing" appeared most frequently (Table III, Section 5.2.1). In fact, a previous study on automated mood classification on Chinese pop songs (Yang and Hu, 2012) also used mood terms in the MIREX five-cluster model, and ended up with adding additional mood labels that were deemed as important to represent moods in Chinese pop songs: "calm/peaceful," "dreamy," "encouraging," "nostalgic," "relaxed," "soothing" and "tender." Based on this finding, it is suggested a sixth cluster can be added to the MIREX five-cluster model: (calm, relaxing, peaceful, etc.). In addition, some of the suggested terms in the current study were mixtures of terms from different clusters, which implies that the mood clusters may not be mutually exclusive, at least for US listeners. Lee and Hu (2014) also found that some songs may be perceived to be in two or more clusters at the same time. Therefore, it is suggested that the MIREX five-cluster model should allow multiple mood clusters to be applied to each song. As US listeners specified terms that describe visual scenes, it might be a good idea to use scenes as a new access point in MDL. The association between music and images has been studied in music psychology (Ali and Peynircioglu, 2006) and MIR (Mattek and Casey, 2011), but how it can be used in MDL warrants future research to explore. The fact that both cultural groups chose Cluster 3 (literate, poignant, wistful, etc.) most frequently (Table II) and agreed more on songs labeled with Cluster 3 (Table V) suggests that Chinese music may tend to have musical traits that are associated with moods in Cluster 3, and that such musical traits can transcend the cultural boundaries of the audience. MDL using automated mood classification algorithms (Yang and Chen, 2012) could exploit this finding and assign more weight to the classifier of Cluster 3, allowing songs to have a greater chance of being classified in Cluster 3. Similarly, when recommending songs based on mood labels, those labeled with Cluster 3 could be assigned more confidence. As different styles had different agreement levels within and across cultures (Table VII), music in different styles can be treated differently in MDL. For example, contemporary instrumental pieces had low agreement levels between audiences, and thus classification and recommendations of pieces in this style could be given less confidence or provide more candidates to improve the chance of hits. The results on gender and age group differences can be exploited, too. For instance, Hong Kong listeners would benefit more from recommendations and tags made by other Hong Kong listeners in the same age group, whereas for US listeners there is no need to differentiate the age group of other listeners. The effect of song familiarity on mood perception suggests that MDL should allow users to change their mood labels and tags over time, because when they get more familiar with a song, they may perceive the mood of the song differently. This study compares Hong Kong and US listeners' mood perceptions of 29 Chinese music pieces, with the goal of investigating whether and how mood perceptions of the two groups of listeners differ, and how the differences can inform the design of cross-cultural MDL. Music mood was modeled with the MIREX five mood clusters and the results suggested further refinement of this model. The selected music pieces included six genres and styles of Chinese music, ranging from traditional folk music to several sub-genres of C-pop. Overall, the patterns of mood judgments by listeners from Hong Kong and the USA (e.g. commonly agreed-upon clusters, differences based on gender and age) resonated with the findings in previous studies on cross-cultural mood perceptions of western music, even though this study used Chinese music, which has a contrasting cultural background to that of western music. This has important implications for designing global MDL: for both western and Chinese pop music, listeners from different cultures may need different mood labels. For systems that use social tags to facilitate access, it is recommended that tags given by users in different cultural groups be differentiated. Those given by users in the same cultural groups can be more helpful. This study also confirms that listeners do exhibit cross-cultural differences in mood judgment for all but one type (Cantopop) of Chinese music considered in this study. Therefore, it may be desirable for MDL to differentiate music types when classifying or recommending music by mood. Future work will include further investigation on other cultural groups interested in Chinese music, such as mainland Chinese, Taiwanese and Singaporeans. In addition, we plan to conduct similar studies on music from other cultural backgrounds, such as Korean and Japanese music. Findings of such studies will further contribute to the goal of building effective and user-centered global MDL.
|
Listeners from the two cultural groups made different mood judgments on all but one type of Chinese music. Hong Kong listeners reached higher levels of agreement on mood judgments than their US counterparts. Gender, age and familiarity with the songs were related to listeners' mood judgment to some extent.
|
[SECTION: Value] Music seeking and consumption are no longer confined by the boundaries of country, region or culture today (Lee et al., 2013). Music, as a cultural object, may be perceived differently by people from different cultural backgrounds, imposing a challenge on music digital libraries (MDL) in meeting the needs of a diverse audience (Weissenberger, 2015). Consequently, an increasing number of researchers have started to investigate various cross-cultural issues in the music information retrieval (MIR) and MDL fields. As people often seek music for emotional goals (Lavranos et al., 2015), music mood[1] has increasingly become a popular access point for music information in many MDL and online music services (Hu, 2010; Hu and Downie, 2007). This trend has raised questions regarding the applicability of music mood across cultural boundaries. Probably due to its subjective and context-based nature, music mood perception is often regarded as culture-dependent (Wong et al., 2009). A number of previous studies have compared mood perceptions of music in various cultures by listeners from different cultural backgrounds (e.g. Balkwill and Thompson, 1999; Fritz et al., 2009; Hu and Lee, 2012; Lee et al., 2013; Singhi and Brown, 2014; Egermann et al., 2014). Although specific findings vary, a general trend found in these existing studies is that listeners' perceptions of music mood can be influenced by their cultural backgrounds. However, most existing studies have focussed on western (classical or pop) music or ethnic music with small audiences (e.g. Congolese music produced and consumed by Congolese Pygmies (Egermann et al., 2014)). Such studies mainly aim to answer questions on music psychology and/or the social, developmental aspects of music. The purpose of this study is different; we are interested in exploring how music mood could be used as an access point or metadata type that can transcend the cultural boundaries of users in the context of MDL. To date, the question of how people with different cultural backgrounds would perceive the mood of Chinese music remains under-investigated. Considering its appeal to one of the largest music listener populations in the world (IFPI, 2014), Chinese music should not be left out of today's global scope of MDL/MIR studies. As Chinese music possesses unique characteristics that are remarkably different from western music (Fung, 2013), it is of theoretical and practical value to investigate the case of Chinese music. Therefore, in this study, we aim to fill this gap and explore how people with different cultural backgrounds, specifically those in the USA and Hong Kong (representing western and eastern listeners, respectively), perceive the mood of Chinese music, including traditional Chinese music and contemporary Chinese pop music that infuses western pop characteristics and local Chinese elements. We specifically answer the following research questions: RQ1. How differently or similarly do Hong Kong and US music listeners perceive the mood of Chinese music? RQ2. How differently or similarly do Hong Kong and US music listeners perceive the mood of different genres/styles of Chinese music? RQ3. Do the listeners' demographic characteristics and familiarity with the music affect their mood perceptions of Chinese music? Answers to these questions have important implications for MDL that use mood as an access point to facilitate MIR and discovery in a global environment. These answers will help determine whether it is necessary to develop different mood-related labels or tags that take their suitability for various user groups into consideration. Implications of the results in this study will be discussed. The goal of this study is to contribute to the literature of cross-cultural MDL/MIR and provide empirical evidence for designing cross-cultural MDL/MIR systems. 2.1 A brief history and types of Chinese music Chinese music broadly refers to music that uses traditional Chinese instruments or is in the Chinese language, and is the main music consumed in the region of Greater China (i.e. regions where Chinese is the official language, including mainland China, Hong Kong, Macau, Taiwan and Singapore). However, the consumption of Chinese music is not limited to that region. People around of the world are accessing Chinese music, and all major online music services, such as Spotify and Last.fm, contain Chinese music to cater to the demand. There is a broad range of genres and styles in Chinese music. Traditional Chinese folk music dates back to approximately 7,000 years ago, played mostly in the pentatonic scale with its own set of instruments. Contemporary Chinese pop music (also referred to as C-pop) was developed in the middle of the twentieth century with the influence of western music. The subgenre "Cantopop" (Hong Kong pop music sung in Cantonese) was developed in the 1970s, followed by the rise of another subgenre, "Mandopop" (Mandarin Chinese pop music), in the 1980s (Fung, 2013). This study includes six genres/styles of Chinese music that are widely appreciated in the region and around the world. The first two are traditional Chinese music, whereas the rest are the most popular C-pop styles featured in online music services in China (e.g. Baidu, Xiami, Kugou): Traditional Instrumental Folk Music of China (qiyue): on the pentatonic scale, musicians perform using traditional instrument(s) of woodwind and percussion, bowed strings, plucked and struck strings in solo, small ensembles or orchestra (Randel, 2003). Canto(nese) folk songs: Han Chinese folk songs are divided into northern and southern styles geographically. Provinces south of the Yangzi River have mild weather and abundant rain, and therefore the folk songs tend to be more "lyrical, gentle and conjunct" (Han, 1989). Near the Canton province and Pearl River Delta, Cantonese folk music is especially light-spirited (China Culture, 2006). Contemporary Chinese instrumental music: this style of music is also called "New Music of China." It mixes the techniques, instrumentation and composition of both Chinese traditional music and the European/western music (Liu and Mason, 2010). The Girls 12 is a famous mainland Chinese all-female group who plays this type of music with traditional Chinese instruments. Cantopop (or Cantonese popular music): this is a subgenre of C-pop. Sung in Cantonese, Cantopop usually refers to popular songs produced in Hong Kong. Blended with traditional Chinese elements, as well as western-influenced harmonies and melodies, Cantopop was a dominant music genre in Asia in the 1980s' and 1990s. Its iconic star, Anita Mui, was recognized as "Asia's Madonna." China wind (zhongguofeng): this is a subgenre of Mandopop and is also called "Chinese Style." It features traditional Chinese instruments and classical Chinese melody juxtaposed with popular western music styles and cultural references to Chinese history, classics and folklore in the lyrics (Xinhuanet, 2006). Taiwanese singer Jay Chou is regarded as the "godfather" of this type of music, bringing forth the wave of China Wind pop. Chinese rock: as another subgenre of Mandopop, Chinese rock is locally produced in mainland China and is the most notable sub-category of Mandopop to the world outside of China. Chinese rock combines the loud and forceful folk song traditions of Northern China and characteristics of western-style rock such as a fast tempo, a strong beat and aggressive bass lines (Fung, 2013). The singer/songwriter Cui Jian from mainland China has been widely recognized as the father of Chinese rock (Fung, 2013). It is noteworthy that all C-pop styles are, to varying extents, influenced by western pop or Rock music. Compared to Traditional Instrumental Chinese Folk Music and Canto Folk Songs, C-pop may be easier to understand and appreciate for listeners familiar with western music. In addition, because of the mixture of western music elements, listeners from both cultural groups in this study might share more common perceptions on C-pop than the other two genres/styles. Therefore, in addition to analyzing aggregated perceptions across all songs, this study also compares and contrasts listeners' perceptions on these six music styles. 2.2 Cross-cultural studies on music mood perceptions A number of studies have investigated the effects of culture on music mood perception. Although fundamental emotions such as happiness, sadness, anger or fear were found to be generally agreed upon across cultures, cultural traditions and background were found to affect individuals' music mood perception, in addition to the inherent qualities of the music (e.g. Gregory and Varney, 1996; Kosta et al., 2013; Lee et al., 2013). For example, Wong et al. (2009) compared mood descriptors used by European and Indian communities and discovered that cultural tradition played a more crucial role than the inherent qualities of the music in users' mood perception. Fritz et al. (2009) found that mood perceptions of western music by African listeners partially agreed with those by western listeners. In a similar vein, Hu and Lee (2012) discovered that discrepancies existed between Chinese and American music listeners in the way they determined the mood of western music pieces. More recently, Lee and Hu (2014) also compared perceptions of western music mood among American, Korean and Chinese listeners, and found that the results of Korean listeners' mood perceptions were situated in between the two other groups. There are also studies investigating the cultural backgrounds of listeners in relation to their mood perceptions of non-western music. Kosta et al. (2013) found greater agreement on the mood of Greek songs among Greek listeners compared to non-Greek listeners with varying backgrounds. Similar results were found by Hu et al. (2014) in comparing American and Korean listeners' mood perceptions of a set of Korean pop songs. Egermann et al. (2014) compared subjective and psychophysiological emotional responses by Canadians and Congolese Pygmies to both western and Congolese music. They found that the two groups had very different responses to Pygmies music but similar responses to some western music. To the best of our knowledge, there has been little work focussed on comparing the mood perceptions of various genres/styles of Chinese music across different cultural groups. In addition, many of the previous studies were conducted in the context of music psychology or ethnomusicology. This study, however, reflects the perspective of information organization and access, with the goal of finding out how music mood can be organized to best meet the information needs of users from different cultural groups. There are mainly two kinds of models to represent mood in music psychology and MIR: categorical and dimensional. The former uses a set of discrete terms (e.g. "passionate," "cheerful") to represent the mood of a piece of music. The most classical model of this kind is Hevner's (1936) adjective circle where eight mood categories placed in a circle, with a set of terms in each category. The dimensional models, in contrast, represent mood with continuous values in a low dimensional space. Different models may have different dimensions, yet valence (i.e. level of pleasure) and arousal (i.e. level of energy) are among the most popular dimensions. The dimensional model used very often in MIR is Russell's (1980) model. The categorical and dimensional models have their own pros and cons. Yang and Chen (2012) summarize that categorical models are more user-friendly as they only consist of terms in natural language; whereas dimensional models are advantageous in terms of quantifying the intensity of moods. Notwithstanding the importance of dimensional models, categorical ones are more suitable for this current study, for the purpose of validating and comparing the values of mood metadata in the context of cross-cultural MDLs. The models from music psychology have solid theoretical ground and can be borrowed into MIR studies. However, one criticism of adopting these models directly in MIR is that they were developed in experimental settings and thus might not reflect users' preferences and behaviors in today's music listening environment (Hu, 2010). Therefore, MIR researchers developed mood models based on domain analysis of widely used music repositories and streaming services. One of the influential mood models in MIR is the five-cluster model proposed in the audio music mood classification task (Hu et al., 2008) of Music Information Retrieval Evaluation eXchange (MIREX), a community-wide MIR evaluation campaign held annually since 2005 (Downie et al., 2014). This model was developed from more than 180 editorial mood labels on AllMusic.com, a very popular online music repository. It has been used in many music mood classification studies (e.g. Laurier et al., 2008; Bischoff et al., 2009; Yang and Chen, 2012) as well as cross-cultural user studies (Hu and Lee, 2012; Lee and Hu, 2014; Hu et al., 2014; Yang and Hu, 2012). The five mood clusters are: C1: passionate, rousing, confident, boisterous, rowdy; C2: rollicking, cheerful, fun, sweet, amiable/good natured; C3: literate, poignant, wistful, bittersweet, autumnal, brooding; C4: humorous, silly, campy, quirky, whimsical, witty, wry; and C5: aggressive, fiery, tense/anxious, intense, volatile, visceral. In this five-cluster model, each cluster is collectively represented by a set of mood terms that share similar semantics. As the terms are describing affect, which can be somewhat vague (as opposed to terms referring to concrete objects), it is argued that using multiple terms with similar meanings to collectively describe one kind of mood is clearer than single-word representations (Hu and Downie, 2007). A more recent categorical mood model was also used in MIREX, in the mood subtask of the audio tag classification task (Hu et al., 2009). It consists of 18 mood categories derived computationally from mood-related social tags from Last.fm, arguably the largest music tagging and sharing website. This 18-class model consists of 135 tags. Besides the automated tag classification task in MIREX, a simplified version of this model has been used in a cross-cultural user study on Korean music (Hu et al., 2014), together with the five-cluster model. As the MIREX five-cluster mood model has been used in a number of cross-cultural MIR studies, for comparison purposes, the current study adopts this model and attempts to make suggestions on modifications and refinement based on the results. 4.1 The participant groups The user groups recruited in this study consist of music listeners from Hong Kong and the USA. Hong Kong plays a special role in the history of Chinese music. Located adjacent to the Canton province, Hong Kong was the origin of Cantopop, which highly influenced the entire region of East Asia in the 1980s and 1990s. Furthermore, for historical and political reasons, Hong Kong was able to preserve the traditional Chinese culture to a large extent. Canto folk music and even traditional Canto opera are often appreciated and practiced by Hongkongers. Since the government handover in 1997, people in Hong Kong also have had increasing exposure to Mandopop. Therefore, it makes an interesting case to investigate Hongkongers' perceptions of a variety of Chinese music types, since they are culturally exposed to all of them to certain degrees. On the other hand, US listeners can represent people with the western cultural background, and, additionally, they have been found to exhibit different music consumption behaviors than Hongkongers in a prior study (Nettamo et al., 2006). We would like to explore whether differences also exist in the two user groups' mood perceptions of non-western music. Invitations were distributed to a number of departmental student mailing lists at the University of Hong Kong and University of Washington, as well as via the authors' various social network websites, such as Twitter, Facebook and Google Plus. Participation was completely voluntary. 4.2 Song selection This study used an online questionnaire of 29 clips of Chinese music to solicit listeners' music mood perceptions. The clips were manually selected from the authors' personal music collections, with the goal of reflecting the broad range of Chinese music. As the goal of this study is to propose suggestions for global MIR systems, two kinds of Chinese music were considered: traditional Chinese music and contemporary Chinese pop music. The former features distinct cultural origin and musical characteristics, and thus can be attractive to users who do not have a Chinese background but are curious about Chinese or Asian music. The latter (C-pop) is widely consumed by Chinese-speaking listeners around the world. As it is heavily influenced by western pop music, yet has Chinese flavor, it is potentially of interest to users in all regions who like pop music in general. Specifically, the selected songs represented the following six genres or styles described in Section 2.1: traditional Chinese instrumental music, traditional Cantonese (Canto) folk music, contemporary Chinese instrumental music, Cantopop and two unique styles of Mandopop: China wind and Chinese rock. Each type contains five representative pieces except for Chine wind which contains four. The list of titles can be found at: http://ccmir.cite.hku.hk/pubs/moodCH/List_of_Songs.xlsx 4.3 The questionnaire At the beginning of the questionnaire, participants were asked about their gender, year of birth and the place where they grew up. The last question was used for screening purposes so as to ensure that only participants with Hong Kong and US cultural backgrounds were included. We believe that people are greatly affected by the culture of the place they resided when they grew up, regardless of their nationality. All participants were 18 years of age or above and agreed with the informed consent statement before proceeding to the questionnaire. Next, participants were asked to listen to the selected music clips, with one song on each page. The questionnaire consisted of five questions for each clip. The first question asked the participant to listen to the clip and decide the most appropriate mood expressed by the song. The response options given were the mood clusters of the MIREX five-cluster model. In case the participants found that the mood of the song did not fit into any of the five default clusters, they could select the option "Other," and use their own words to describe the appropriate mood of the music clip. The participants were also asked about their familiarity with the music clips. Because people may interpret "familiarity" differently, instead of directly asking how familiar they are with the given songs, we used two questions: whether they had heard the song before, and if they could provide the artist name and the title of the song. Responses to the two questions were deemed more objective than self-rated familiarity because the questions are fact-based. The questions were presented in English in the US questionnaire and in both English and Chinese in the Hong Kong questionnaire. In the Hong Kong questionnaire, the five mood clusters were also translated into Chinese, and both English and Chinese terms were presented. We used the same translation of mood terms as in Yang and Hu (2012). Even though English is one of the official languages in Hong Kong, people's English proficiency varies. Therefore, both Chinese translations and original terms in English were deemed necessary for the respondents to have an accurate interpretation of the clusters. Figure 1 illustrates the survey interface of the Hong Kong version. 4.4 Data analysis Responses to the first question for each song (i.e. mood judgment) were compared quantitatively. The kh2 test was used for comparing the distributions of music mood judgments between the two groups of listeners (Hinkle et al., 2003). Significant results would indicate that the judgment distributions are dependent on cultural groups. The agreement between groups of listeners in various combinations of music and listener characteristics was measured by the raw ratio of agreed pairs to all possible pairs. For example, for calculating the agreement ratio among 32 Hong Kong listeners on five Cantopop songs, we first calculated the total number of judgment pairs as (32x(32-1)/2)x5=2,480, and counted the number of pairs with the same mood judgments. Suppose there were 248 matched pairs; the agreement ratio would be 10 percent. Raw agreement measures are recommended for cases where the annotations are dominated by negative values (i.e. "not chosen") (Cicchetti and Feinstein, 1990). This applies to the mood judgment data collected in this study. For instance, "Other" was chosen only 102 times out of 1,856 times it could have been selected. Text responses to the "Other" option were coded with a scheme of categories that emerged from the analysis process. The codebook was developed through an iterative process involving test-coding and revising the codes for clarity. The consensus model (Hill et al., 1997) was applied to coding the responses: two independent coders coded the responses and then discussed on disagreed cases until a consensus was reached. The results were compared across the two groups. Responses to demographic questions and song familiarity questions were used to group the users with a range of criteria, for the purpose of testing factors that affect users' mood perceptions. 4.5 Limitations As each music piece needs to have a minimum length (typically 15-30 seconds) for listeners to establish a confident perception of its mood (Speck et al., 2011), the questionnaire can contain only a limited number of song clips so that it can be completed within a reasonable amount of time. We used a set of 29 clips in this study, which may seem rather limited given there are millions of Chinese songs available in the market. To mitigate this limitation, we tried our best to carefully select a good variety of songs based on demands for them by users with various cultural backgrounds, and representativeness in each category. In addition, the sample size of each cultural group (32) is limited, partially due to the length of the survey and the lack of monetary incentive. Nonetheless, it is sufficient for statistical tests and comparisons. The online venues used to recruit participants (i.e. mailing lists and social network sites) may also have limited the participants to frequent internet users. Future studies are warranted to further verify the evidence found in this exploratory study in order to generalize findings to broader contexts. 5.1 Characteristics of participants The Hong Kong survey had a total of 70 responses, among which 34 were completed. The US survey had 72 responses, of which 34 were completed. However, two listeners in each group answered that they were raised in mainland China (not Hong Kong or the USA), leaving 32 valid responses for each group. Table I shows the demographic information of the listeners. 5.2 Mood judgments 5.2.1 Cultural backgrounds and mood judgments Table II presents the distribution of mood judgments of the two groups on all 29 songs. Both groups chose Cluster 3 (literate, poignant, wistful, etc.) most often, indicating that many Chinese songs may have some traits associated with the mood of Cluster 3, and these traits can be perceived by both Hong Kong and US listeners. The kh2 test shows that the distribution and cultural group are not independent (kh2=52.54, df=5, p<0.001). In other words, it provides strong evidence that Hong Kong and US listeners tend to perceive the mood of given Chinese music differently, within the framework of five mood clusters. From Table II, we can observe that Hong Kong listeners tended to select Cluster 1 (passionate, rousing, confident, etc.) and Cluster 3 while US listeners tended to spread their choices equally across Clusters 1, 2 (rollicking, cheerful, fun, etc.) and 3. The US listeners were also more likely to select the "Other" category than their Hong Kong counterparts. In all, 26 out of the 29 songs were labeled with "Other" at least once by the US group. In comparison, the Hong Kong group chose the "Other" option for 17 songs. On average, 2.93 US listeners chose "Other" for each of these songs, whereas the number for Hong Kong listeners was 1.35. This is similar to the findings in Lee and Hu (2014), where US listeners chose the "Other" category more often than Korean and Chinese listeners when they made mood judgments on western songs. Even though the objects/stimuli have been changed to Chinese music, the observation still holds. The consistent pattern regardless of the cultural background of the music further suggests that this phenomenon might be related to the individualism-collectivism difference between western and eastern cultures as discussed in previous studies (Boer and Fischer, 2010; Hu and Lee, 2012; Lee and Hu, 2014): western users tend to hold their opinions regardless of the given options, whereas Eastern people are more likely to conform to existing norms. We conducted a content analysis on the free-text responses provided in the "Other" option. The two independent coders reached an excellent reliability level (Cohen's k=0.858). Table III shows the seven resultant categories and corresponding counts in each of the groups. The largest category of the self-supplied terms is new mood terms that are not included in the five-cluster scheme. The new mood terms provided by both groups are quite similar, with, "relaxing," "calm" and "longing" being the most frequent, appearing six, four and four times, respectively. Users in both cultural groups also filled in terms that were not mood related. Some of them were music specific (e.g. "fast pace") while others were not (e.g. "focus"). Interestingly, the US group also provided responses that were either a mixture of terms from multiple clusters (e.g. "passionate" from Cluster 1 and "intense" from Cluster 5) or part of the terms in one cluster (e.g. "poignant, wistful, autumnal" in Cluster 3). The US participants also filled in descriptions of scenes they imagined when hearing the songs, although the instructions specifically asked for mood-related terms. Perhaps this group of users thought that scenes inspired by the songs also reflected the mood to some extent. The Hong Kong listeners did not provide any terms in these categories, possibly because: they tried harder to follow the instructions that asked respondents to "choose one dominant cluster when a song sounds to have moods in more than one cluster," and people from individualistic (mostly western) cultures were found to be more used to divergent thinking which involves the generation of novel ideas (Schmidt et al., 2013). 5.2.2 Agreement on mood judgments within and across culture groups The agreement ratios were calculated for pairs of judgments between two listeners from the same cultural group and across cultural groups. The results are shown in Table IV. In particular, in the cross-cultural case, the number of total pairs with one judgment from each cultural group is calculated as ((32x32)/2)x29=29,696. The results and non-paired t tests reveal that the level of intra-agreement among Hong Kong listeners is significantly higher than that of US listeners (p<0.001). A similar result was reported in Kosta et al. (2013), where Greek listeners were found to agree more on the mood of Greek music than non-Greek listeners. In comparing Korean and American listeners' perceptions of the mood of Korean pop songs, Hu et al. (2014) also found that Korean listeners reached a higher agreement level than their US counterparts. However, the result is partially different from Hu and Lee (2012), where both US and Chinese listeners had the same agreement ratio on the mood of western pop songs. We conjecture that this difference in results between Hu et al. (2014) and Hu and Lee (2012) might be due to Chinese listeners' wide exposure to western pop songs. For non-western songs that have not yet spread as widely, sharing the same cultural background as the music stimuli does seem to help listeners reach a higher agreement on music mood. It is also noteworthy that the agreement level among US listeners was not statistically different from the cross-cultural agreement (p=0.54). This is similar to the result in Hu et al. (2014), where the agreement level among American listeners on the mood of Korean pop songs was the same as that between American and Korean listeners. 5.2.3 Agreement on mood clusters To examine which moods are easier or more difficult to agree upon, we counted the number and percentage of agreed pairs across the five mood clusters (Table V). The kh2 test indicates that the two cultural groups showed a statistically significant difference in the agreement pattern across the mood clusters (kh2=603.55, df=10, p<0.001). Both within and across cultures, Cluster 3 (literate, poignant, wistful, etc.) got the most agreements, which again evidences that many Chinese pop songs have traits of Cluster 3 and these traits can transcend the cultural boundaries of listeners to some extent. The difference between the two cultural groups observed here is similar to the findings in Hu and Lee (2012), where American listeners agreed more on Cluster 2 (rollicking, cheerful, fun, etc.), Cluster 4 (humorous, silly, campy, etc.) and Cluster 5 (aggressive, fiery, tense/anxious, etc.) compared to Chinese listeners, who agreed more on Cluster 1 (passionate, rousing, confident, etc.) and Cluster 3 (literate, poignant, wistful, etc.). As western music was used as the stimuli in Hu and Lee (2012), it is interesting to see that the patterns of mood perception are similar to this study even though the stimuli here consist of completely different, non-western music. 5.3 Mood perception on different types of music 5.3.1 Instrumental vs vocal The music pieces in this study could be divided into instrumental and vocal sets. The former includes 15 pieces in instrumental folk, Canto folk or contemporary instrumental styles, and the latter includes the rest of the three types with 14 pieces total. We can observe from Table VI that overall, Hong Kong listeners agreed among themselves more often than US listeners on both instrumental and vocal pieces. Listeners also agreed more often on vocal pieces than instrumental music within each cultural group, as well as cross-culturally. This is in accordance with the literature on both music psychology and user studies in MDL/MIR. Juslin (2005) argued that the verbal (lyric) and non-verbal (melody) channels of music complemented each other in communicating emotions. Vocal delivery and the lyrics of the song were also found to be among the various factors that influence people's perception of music mood (Lee et al., 2012). In comparing the effects of melodies and lyrics on mood perceptions, studies found that, although melodies were more dominant than lyrics, lyrics were particularly useful in strengthening emotional responses to music with negative (Ali and Peynircioglu, 2006) and low-arousal emotions (Kazuma, 2009), which roughly correspond to Clusters 5 and 3, respectively, in this study. Even though the US listeners probably did not understand the lyrics in those songs, they might still have been able to get some emotional clues from the vocal delivery. 5.3.2 Genres/styles of Chinese music Table VII shows the agreement ratios within and across cultural groups with regard to the six types of music. The Hong Kong listeners showed an agreement of 60 percent on Cantopop, followed by 54 percent on China wind. The US listeners agreed more on Canto folk and China wind (43 percent), followed by Cantopop (41 percent). Cross-culturally, kh2 tests show that the two groups significantly differed in their mood judgments on all styles except for Cantopop (kh2=9.22, df=10, p=0.10). We conjecture two possible reasons for Cantopop to reach a high cross-cultural agreement ratio: as introduced before, Cantopop was developed with heavy influence from western pop music, and thus the western elements could have helped to bridge the differences between the two user groups; and all Cantopop songs are lyrical and thus the mood cues in the vocal delivery may have helped strengthen mood perceptions. The category with the least agreement both within and across cultures was contemporary instrumental music. This may be due to the complex musical compositions of this style and the lack of lyrical cues. In fact, the song with the least agreement was of this type and will be analyzed in more detail in the next subsection. To facilitate further examination of the mood judgments on different types of music stimuli, Figure 2 shows the distribution of mood judgments for each of the six music types included in this study. It can be seen that the first three types of music (i.e. instrumental folk, Canto folk and contemporary instrumental) were labeled with a mix of diverse mood clusters without a clear, dominant cluster, whereas the last 3 (i.e. Cantopop, China wind and Chinese rock) have one dominant cluster from both cultural groups: China wind and Cantopop were mostly labeled with Cluster 3 (literate, poignant, wistful, etc.) and Chinese rock with Cluster 1 (passionate, rousing, confident, etc.). Instrumental folk music seems to be the most controversial, in that the pieces got similar numbers of judgments across all mood clusters, especially from the US group. The two cultural groups seemed to disagree most on Canto folk and contemporary instrumental music. Hong Kong listeners categorized Canto folk music mostly in Clusters and 2 (rollicking, cheerful, fun, etc.) but US listeners tended to label those pieces more often with Cluster 2 and Cluster 4 (humorous, silly, campy, etc.). For contemporary instrumental music, US listeners chose Cluster 2 the most, followed by similar numbers for Cluster 3, Cluster 1 and "Other," while Hong Kong listeners used Cluster 3 more often, followed by Cluster 2, Cluster 1 and Cluster 4. 5.3.3 The songs with highest and lowest agreement The song that reached the highest agreement on mood judgment was the same within each cultural group and across cultural groups. In fact, the US listeners agreed more than Hong Kong listeners, with agreement ratios of 0.94 and 0.88, respectively. This was a China wind song titled Pearl-decorated Curtain Rolled Up by a Chinese male singer, Huozun. In all, 30 Hong Kong listeners and 31 US listeners selected Cluster 3 (literate, poignant, wistful, etc.) for this song. Released in 2014, the song features a slow melody and extensive use of falsettos. It can be characterized as mellow, ethereal and graceful, similar to the style of ancient Chinese music (and thus recognized as China wind). Although the lyrics describe a beautiful young woman missing a lost love in a poetic way, the US listeners probably could not understand them. Yet they reached near consensus on the mood of this song, possibly attributable to the emotionally powerful delivery of the song. The song with the lowest agreement among US listeners as well as cross-culturally was Shangri-la by Girls 12, a female band famous for using traditional Chinese instruments to play both traditional Chinese and western music. This song had the second-lowest agreement among Hong Kong listeners, and their lowest agreement was for another piece by the same band. Many pieces by Girls 12 are modern compositions using ancient instruments, including Shangri-la, and thus are classified as contemporary instrumental music, the type that has the least agreement both within and across cultures (Table VII). The song Shangri-la received only 16 percent agreement from US listeners and 21 percent from Hong Kong listeners. Figure 3 illustrates the distribution of mood judgments of this song. We can observe that the judgments from Hong Kong listeners were distributed fairly evenly across Cluster 1 (passionate, rousing, confident, etc.), Cluster 2 (rollicking, cheerful, fun, etc.), Cluster 3 (literate, poignant, wistful, etc.) and Cluster 4 (humorous, silly, campy, etc.), and about equal numbers of US listeners labeled it with Cluster 1, Cluster 2, Cluster 5 (aggressive, fiery, tense/anxious, intense, etc.) and "Other." The piece was performed with traditional Chinese instruments: bamboo flute, erhu (bowed strings), pipa and guzheng (both plucked strings), mixed with western-sounding percussion and electronica fusions. Just as the song title suggests, the piece conjures the mystic and mysterious aura of the utopia named Shangri-la. The mixed mood perceptions of this song can be at least partially explained by the mixed musical nature of the song, as well as its fictional, mystical theme. 5.4 Listener characteristics 5.4.1 Gender Table VIII shows the distribution of mood judgments based on gender across all mood clusters. Both genders in both cultural groups selected Cluster 3 the most. The kh2 test indicates that gender difference in mood judgment does exist within the US group (kh2=24.62, df=5, p <0.001), but not in the Hong Kong group (kh2=7.53, df=5, p<0.184). This was also the case in Lee and Hu (2014), where American listeners' mood judgments on western music differed based on gender, but Chinese listeners' did not. In order to see the influence of gender difference across cultural groups, we compared the judgments from US and Hong Kong listeners of the same gender. The comparison shows female listeners in the two cultural groups did have different mood judgments (kh 2=46.51, df=5, p<0.001). However, the judgments of male listeners from the two cultural groups did not have a significant difference (kh2=7.99, df=5, p=0.157). The results suggest that cultural differences played a significant role in mood judgments among female listeners, but not among male listeners. In contrast, Hu and Lee (2012) found significant differences for both genders between US and Chinese listeners when the music stimuli were western pop songs. This discrepancy might have been affected by the small sample size of male participants in the US group (n=8) in this study. Future studies with a larger and more balanced sample across gender groups are warranted to further investigate the extent to which gender and cultural groups affect music mood perception. 5.4.2 Age The ages of the participants vary between the two groups, but both groups have listeners between 21 and 60 years old. We divided those listeners into two age groups, 21-33 and 34-60, so as to balance the sample sizes in both age groups. The kh2 tests on the distribution of judgments revealed a statistically significant difference between age groups among Hong Kong listeners (kh2=39.86, df=5, p<0.001) but no significant difference among US listeners. (kh2=4.72, df=5, p=0.45). This is, in fact, consistent with the findings from Lee and Hu (2014), where the authors found a significant difference between the age groups for Korean and Chinese music listeners but not for American listeners. This could be possibly attributed to the fast-changing environment in Hong Kong and other Asian countries compared to the relatively stable environment in countries such as the USA (Lee and Hu, 2014). 5.4.3 Familiarity with the music Listeners' familiarity with the songs was categorized into three levels: "high familiarity" if they had listened to the song before and could also identify the artist's name and the song title; "medium familiarity" if they had heard the song but could not name it; and no familiarity if they had not heard the song. There was one US listener who was highly familiar with two songs, and seven who, collectively, had heard a total of 15 songs before. In the Hong Kong group, 29 out 32 listeners indicated high familiarity with at least one song. Unsurprisingly, more than half of the highly familiar songs were Cantopop. Due to the sparseness of data in the US group, we examined how the levels of agreement related to familiarity with the music only in the Hong Kong group. Table IX lists the agreement ratios among judgment pairs with regard to various levels of music familiarity. When both participants were highly familiar with a song, their agreement was the highest (ratio=0.57). While a higher familiarity level helped increase the level of agreement when both mood judgments were made for songs with which participants were familiar, it did not help when one participant was familiar with the song and the other was not. We conjecture that when users have heard a song multiple times or know the artist name and the song title, they may have developed and/or gained some contextual knowledge of the song, which may affect their selection of mood clusters. On the other hand, when listeners hear the song for the first time, they may naturally rely more on the musical characteristics for judging the mood of the song. As advocated by Weissenberger (2015), a flexible organization system is needed for MIR/MDL to meet the needs of different musical traditions. The results of this study have important implications for designing MDL of this kind. As users in different cultural groups had significantly different mood judgments overall (Section 5.2.1), it is better to differentiate user-generated mood labels and/or mood-related social tags according to the users' cultural background. In particular, as the agreement ratio among Hong Kong listeners was higher than the cross-cultural one (Section 5.2.2, Table IV), Hong Kong listeners would benefit more from mood tags supplied by listeners in the same cultural group. In contrast, as the within-group agreement ratio of US listeners was similar to the cross-cultural agreement level (0.38 vs 0.39), US listeners could make use of mood tags generated by Hong Kong listeners, especially when those generated by US listeners are sparse. The analysis on the textual input in the "Other" option (Section 5.2.1) suggests that the MIREX five-cluster model may not be sufficient and more terms are needed to describe the mood of Chinese music. Particularly, terms like "calm," "relaxing," "peaceful" and "longing" appeared most frequently (Table III, Section 5.2.1). In fact, a previous study on automated mood classification on Chinese pop songs (Yang and Hu, 2012) also used mood terms in the MIREX five-cluster model, and ended up with adding additional mood labels that were deemed as important to represent moods in Chinese pop songs: "calm/peaceful," "dreamy," "encouraging," "nostalgic," "relaxed," "soothing" and "tender." Based on this finding, it is suggested a sixth cluster can be added to the MIREX five-cluster model: (calm, relaxing, peaceful, etc.). In addition, some of the suggested terms in the current study were mixtures of terms from different clusters, which implies that the mood clusters may not be mutually exclusive, at least for US listeners. Lee and Hu (2014) also found that some songs may be perceived to be in two or more clusters at the same time. Therefore, it is suggested that the MIREX five-cluster model should allow multiple mood clusters to be applied to each song. As US listeners specified terms that describe visual scenes, it might be a good idea to use scenes as a new access point in MDL. The association between music and images has been studied in music psychology (Ali and Peynircioglu, 2006) and MIR (Mattek and Casey, 2011), but how it can be used in MDL warrants future research to explore. The fact that both cultural groups chose Cluster 3 (literate, poignant, wistful, etc.) most frequently (Table II) and agreed more on songs labeled with Cluster 3 (Table V) suggests that Chinese music may tend to have musical traits that are associated with moods in Cluster 3, and that such musical traits can transcend the cultural boundaries of the audience. MDL using automated mood classification algorithms (Yang and Chen, 2012) could exploit this finding and assign more weight to the classifier of Cluster 3, allowing songs to have a greater chance of being classified in Cluster 3. Similarly, when recommending songs based on mood labels, those labeled with Cluster 3 could be assigned more confidence. As different styles had different agreement levels within and across cultures (Table VII), music in different styles can be treated differently in MDL. For example, contemporary instrumental pieces had low agreement levels between audiences, and thus classification and recommendations of pieces in this style could be given less confidence or provide more candidates to improve the chance of hits. The results on gender and age group differences can be exploited, too. For instance, Hong Kong listeners would benefit more from recommendations and tags made by other Hong Kong listeners in the same age group, whereas for US listeners there is no need to differentiate the age group of other listeners. The effect of song familiarity on mood perception suggests that MDL should allow users to change their mood labels and tags over time, because when they get more familiar with a song, they may perceive the mood of the song differently. This study compares Hong Kong and US listeners' mood perceptions of 29 Chinese music pieces, with the goal of investigating whether and how mood perceptions of the two groups of listeners differ, and how the differences can inform the design of cross-cultural MDL. Music mood was modeled with the MIREX five mood clusters and the results suggested further refinement of this model. The selected music pieces included six genres and styles of Chinese music, ranging from traditional folk music to several sub-genres of C-pop. Overall, the patterns of mood judgments by listeners from Hong Kong and the USA (e.g. commonly agreed-upon clusters, differences based on gender and age) resonated with the findings in previous studies on cross-cultural mood perceptions of western music, even though this study used Chinese music, which has a contrasting cultural background to that of western music. This has important implications for designing global MDL: for both western and Chinese pop music, listeners from different cultures may need different mood labels. For systems that use social tags to facilitate access, it is recommended that tags given by users in different cultural groups be differentiated. Those given by users in the same cultural groups can be more helpful. This study also confirms that listeners do exhibit cross-cultural differences in mood judgment for all but one type (Cantopop) of Chinese music considered in this study. Therefore, it may be desirable for MDL to differentiate music types when classifying or recommending music by mood. Future work will include further investigation on other cultural groups interested in Chinese music, such as mainland Chinese, Taiwanese and Singaporeans. In addition, we plan to conduct similar studies on music from other cultural backgrounds, such as Korean and Japanese music. Findings of such studies will further contribute to the goal of building effective and user-centered global MDL.
|
It is the first study on cross-cultural access to Chinese music in MDL. Methods and the refined mood model can be applied to cross-cultural access to other music types and information objects.
|
[SECTION: Purpose] Today schools operate in dynamic environment, each struggling to gain a competitive edge (Orr and Orphanos, 2011). This environment reinforces the understanding that schools should strive to employ teachers who are willing to go the extra mile, namely to engage in organizational citizenship behavior (OCB). OCBs are behaviors and actions that go beyond the formal role and what is laid down in the teacher's job description, and contribute to the achievement of the school's objectives (Oplatka, 2006). OCB in schools is commonly taken to be an individual phenomenon, namely a personality trait or a social response to the behavior or attitude of superiors or to other motivation-based mechanisms in the workplace (Zeinabadi, 2010). This individual approach seems wanting: teachers perform or fail to perform OCBs not in a vacuum but in an organizational context, which most probably serves to encourage or discourage them (Somech and Drach-Zahavy, 2004). Only in the last two decades have some scholars studied OCB as a team or organizational-level concept (e.g. Ehrhart, 2004). Put otherwise, the individual approach probes the possibility that teacher OCB can be better understood as a team or an organizational feature that thrives in a context. This development is important because the aggregate OCB level, not sporadic actions by some individuals, influences organizational effectiveness (Organ, 1988). Moreover, the several studies that have captured OCB as a collective phenomenon examined either its antecedents (e.g. Hu and Liden, 2011) or its consequences (e.g. Yen et al., 2008). Very few researchers, if any, have explored the mediating role of team OCB in relation to specific antecedents and outcomes. The present research seeks to address this deficiency by examining OCB as a team phenomenon. The first premise of the research model is that it is important to measure OCB as a collective structure. Next, the mediating role of team OCB is investigated. Specifically, the research model posits that the contextual variable of a team's justice climate, and the team's collective psychological state (psychological capital), will be positively related to team OCB. Furthermore, team OCB will be positively related to the outcome of team innovation. The model also suggests that team OCB will mediate the relation between the antecedents and the outcome. Teachers' OCB is defined as "[...] teachers voluntarily going out of their way to help their students, colleagues, and others as they engage in the work of teaching and learning" (DiPaola and Hoy, 2005, p. 390). OCBs are essential because formal in-role job descriptions cannot cover the entire array of behaviors needed for achieving school goals (Zeinabadi, 2010). They operate indirectly: they influence schools' social and psychological environment, enhance school effectiveness by freeing up resources for more productive purposes, help coordinate activities within the organization, and enable teachers to adapt more effectively to environmental changes (Sesen and Basim, 2012). Teachers manifest OCB, for instance, by staying after school hours to help a student with learning materials, aid a colleague with a heavy workload, volunteer for unpaid tasks, or make innovative suggestions to improve the school (Somech and Oplatka, 2014). Scholars have identified two broad dimensions of OCB (Sesen and Basim, 2012; Williams and Anderson, 1991). The first is OCBI (OCBs toward individuals in school), namely behaviors that immediately benefit particular individuals, such as colleagues and/or students, hence indirectly contribute to the organization. This dimension includes such behaviors as taking a personal interest in a novice teacher or helping a student after working hours. The second dimension is OCBO (OCB toward the organization), namely behaviors that benefit the school as a whole, for example, organizing special social activities for the school or making innovative suggestions to improve it. According to Podsakoff et al. (2009), these two broad categories capture most previous taxonomies effectively. Although OCBs are acts by individuals, our main argument is that these extra-role behaviors grow in a context, therefore can be extended to the team level. Put simply, because OCBs are interactive or "social" in nature, teams differ in their teachers' tendency to engage in them (Ehrhart, 2004). Social psychology research has shown that teams are powerful instruments of social influence and significantly affect individuals' perceptions, attitudes and behaviors (Mathieu et al., 2000). Two theoretical models may provide the rationale for considering OCB a team phenomenon: the social learning theory and the social information processing theory. The first posits that people learn by observing the behavior of others (Bandura, 1986). The more OCB is modeled by the members of one's team, the more likely one is to behave in a manner consistent with those models, in particular when that behavior is associated with positive social consequences (Ehrhart, 2004). The second the social information processing theory (Salancik and Pfeffer, 1978) underlines the importance of social cues in shaping attitudes. When OCB levels are high among team members, an individual member will be more likely to view such behavior as acceptable and expected. Ultimately this recognition results in higher levels of individual OCB. George (1991), for example, found high similarities among team members in the level of positive and/or negative mood, as well as in the extent they exhibited citizenship behaviors. Conceptualizing OCB as a team phenomenon implies that teachers who do or do not perform OCB do not do so in a vacuum; the team context in which these behaviors are performed undoubtedly serves to encourage or discourage them (Vigoda-Gadot et al., 2007). In other words, this perspective suggests that contextual variables determine when, and sometimes why, teachers do or do not engage in OCBs (Ehrhart, 2004). In this vein, the present model focused on two team-level variables as antecedents of team OCB: justice climate and team psychological capital, and team innovation as a consequence. Furthermore, the model suggested that team OCB serves as a mediator in the relation of justice climate to team psychological capital and team innovation. The relation of justice climate to team psychological capital and team OCB We focused on two elements of team characteristics: team justice climate and team psychological capital. Our choice of these two variables was grounded in recent OCB research emphasizing the critical role of positive team atmosphere in enhancing positive attitudes and behaviors, as expressed by team citizenship behaviors (Lau and Lam, 2008). Team justice climate and team OCB Team climate derives from the array of the work-contextual properties that the team emphasizes and evaluates; these are perceived as critical in shaping team members' attitudes and behaviors (Schneider et al., 2013). The main claim of climate research is the idea of shared perceptions constructed through social interactions among individual employees engaged in teamwork (Spell and Arnold, 2007). Although there are several kinds of climate in every workplace, justice climate is crucial for promoting a positive atmosphere prevailing over team members, as well as between them and their supervisor and the organization as a whole (Colquitt et al., 2002). Building on the perception of a "social" creation of justice perceptions, justice climate is defined as "'shared perceptions' of team fairness in outcomes, treatment, and supervisory interaction" (Moliner et al., 2005, p. 101). Scholars have identified three distinct dimensions of justice climate: distributive justice, procedural justice and interpersonal justice (Li and Cropanzano, 2008). Distributive justice climate addresses the organizational reward system and reflects fairness of outcomes, namely fairness in distribution of tangible or intangible rewards and/or resources (Li and Cropanzano, 2008); procedural justice climate is conceptualized as fairness of procedures and of the decision-making process (i.e. the degree to which decisions on reward allocation are made according to fair methods and guidelines) (Niehoff and Moorman, 1993); interpersonal justice climate refers to the treatment accorded to team members when decisions are made. It marks the extent to which team members are treated with respect and dignity during the decision-making process (Li and Cropanzano, 2008). The link between justice climate and team OCB can be perceived through social exchange theory (Blau, 1964) and through the norm of reciprocity (Gouldner, 1960). These two models suggest that team members seek to balance their inputs and outcomes in relation to others (Elstad et al., 2012). Generally, with high levels of justice climate, team members share the perception that their organization and/or its representatives can be trusted to protect their interests; this engenders an obligation to reciprocate in positive, beneficial ways, namely OCBs (Lambert, 2000). More specifically, when the collective experience of the team is that they are rewarded well (distributive justice), the team as the whole is treated with respect and dignity (interpersonal justice), and the process leading to outcomes is fairly applied (procedural justice); team members will seek to reciprocate by investing in citizenship behaviors (Parzefall and Salin, 2010). OCB is especially suitable for reciprocity because whereas traditional in-role behaviors can be enforced on the basis of formal role obligations (e.g. Konovsky and Pugh, 1994), OCB consists of informal contributions that team members can choose to exhibit or withhold (Somech and Oplatka, 2014). For example, Ehrhart (2004), who found a positive association between procedural justice climate and team OCB, suggested: "If group members perceive they are treated fairly and this perspective is shared among them [...] their social interactions will also result in similar responses to that assigned meaning, such as the performance of higher levels of OCB" (p. 67). Accordingly, we propose a positive link between justice climate and team OCB: H1. A positive relationship will be found between team justice climate and team OCB. Team psychological capital and team OCB Team psychological capital (team PsyCap) is a team's positive psychological state of development (Luthans et al., 2008), and is characterized by four psychological components: hope, optimism, collective efficacy and resilience (Avey et al., 2009; Heled et al., 2015). Hope is defined as a positive motivational state of the team that is based on a feeling that is fueled by successful agency (goal-oriented energy) and pathways (planning to meet goals) (Snyder et al., 1996). Optimism is the team's ability to create a positive attribution with respect to success in both the present and the future. Collective efficacy means that "a group shares a common belief in regards to the standardization of the capacities of the group members" (Bandura, 1997, p. 477). And resilience is a positive psychological ability of the team "to respond productively to significant change that disrupts the expected pattern of events without engaging in an extended period of regressive behavior" (Horne and Orr, 1997, p. 31). Although each of these four components has the potential to influence teamwork, studies support the notion that together, they have a much greater impact than each variable individually (Luthans et al., 2007). Team PsyCap is a construct of shared psychological state of team members, and can be considered an emergent state (Heled et al., 2015). According to Marks et al. (2001), emergent states are "constructs that characterize properties of the team that are typically dynamic in nature and vary as a function of team context, inputs, processes, and outcomes. Emergent states describe cognitive, motivational, and affective states of teams [...]" (p. 357). Put simply, team PsyCap is a team resource that can be developed and managed, and may thereby influence team processes and outcomes (Mathe-Soulek et al., 2014). The present study posits a positive link between team PsyCap and team OCB. Generally, teams with high levels of PsyCap are confident that good things happen at work; they believe they create their success and can overcome failure (Avey et al., 2011). Accordingly, we argue that team PsyCap plays an essential role in promoting positive attitudes and behaviors among team members, as expressed by team OCBs. The affective approach, which derives from the social-psychological literature (e.g. Williams and Shiaw, 1999), may provide the theoretical foundation for the relationship (George, 1991). This approach suggests that when people are in a positive mood, and hold positive attitudes; they are more likely to be helpful, namely to exhibit high levels of citizenship behaviors. Put simply, team members who experience a good mood, a sense of optimism and hope tend to perceive other people and situations in a more positive light (George, 1996), which may predispose individuals to exhibit behaviors that benefit team members, enhance organizational outcomes, or both (Heled et al., 2015). In other words, teams characterized by optimism, hope and resilience will tend to demonstrate pro-social behaviors that preserve their positive emotional state (Srivastava et al., 2006). Furthermore, this approach suggests that teams in a good mood are more helpful because being so enables them to maintain their positive mood (George, 1991). For example, Heled et al. (2015), who examined 82 school management teams, found positive relations between team PsyCap and the individual's job satisfaction and team OCB. Accordingly, we propose a positive link between Team PsyCap and team OCB: H2. A positive relation will be found between team PsyCap and team OCB. The relation between team OCB and team innovation Educational innovation is perceived as a key factor for promoting school effectiveness (Gilad-Hai and Somech, 2016). Innovation enables schools to cope with a complex and changing environment (Chen, 2010; Tubin, 2009). Team innovation is defined as "the intentional introduction and application within a team, of ideas, processes, products or procedures new to the team, designed to significantly benefit the individual, the team, the organization, or wider society" (West and Wallace, 1991, p. 303). Similarly, educational innovation is defined as "an idea or practice that reflects an emergent trend and offers an optimistic alternative for a current situation of dissatisfaction" (Tubin, 2009, p. 406). Teams may offer an alternative approach to learning, teaching and/or administrative routines (Gilad-Hai and Somech, 2016). Overall, innovation undergoes two phases: the creativity phase, namely generation or production of ideas that are both novel and useful (George, 2007), and the implementation phase, namely "the process of gaining targeted employees' appropriate and committed use of an innovation" (Klein and Sorra, 1996, p. 1055). According to the present model, team OCB can be linked to team innovation. There are logical reasons to expect a positive relation between team OCB and team innovation. According to Organ (1997), OCBs support the social and psychological environment within which the technical core functions. Accordingly, teachers who are willing to collaborate with their team members, to actively participate in meetings, to share knowledge, to explore new alternatives, or to learn new skills, contribute to the team's ability to be innovative (Jimmieson et al., 2010). Moreover, at the stage of innovation implementation, when job definitions are ambiguous, schools depend more on teams that are willing to contribute to successful change, regardless of formal job requirements (Oplatka, 2006). Thus, team OCB is assumed to help schools adopt changes and innovations that are necessary in the consistently changing environment (Savelyeva and Lee, 2012): H3. A positive relationship will be found between team OCB and team innovation. The mediating role of team psychological capital Finally, we propose that team OCB will serve as a mediator between the antecedents of team justice climate and team PsyCap and the outcomes: team innovation (Figure 1). This argument can be supported by the classic team effectiveness models of input-process-outcomes (Hackman and Morris, 1975) wherein team effectiveness involves a three-stage process. In the first stage, team and job characteristics (inputs) that structure teamwork influence the work processes in stage 2 (processes), which in turn lead to positive results in stage 3 (outcomes). Accordingly, we suggest that justice climate and team PsyCap set the stage for encouraging team members to invest in the team above a beyond the call of duty, namely to exhibit high levels of citizenship behaviors, which in turn enhance team innovation: H4a. Team OCB will mediate the relationship between justice climate and team innovation. H4b. Team OCB will mediate the relationship between team PsyCap and team innovation. Sample and procedure This study covered 78 disciplinary teams sampled from 13 junior-high schools (grades 7-9), located in the Northern and Central parts of Israel. In each school, we sampled six disciplinary teams (linguistics, civics, history, English, mathematics, science), totaling 214 team members, 78 coordinators and 13 school principals. We conducted pre-assessment interviews with the principal of each school, and learned that all team members interacted regularly to achieve shared goals regarding teaching issues and other school matters. They also depended on one another for knowledge and effort by means of several permanent structures such as scheduled staff meetings, and joint refresher workshops. Team size ranged from three to six members, with an average of 3.26 (SD=1.73). Of the team members, 181 were women (84.5 percent) and 33 were men. Average age was 45.66 years (SD=7.85), with average job tenure 6.73 (SD=4.42). In education, 80 percent of participants had a bachelor's degree, 20 percent had a master's degree. Of the coordinators, 56 were women (71.8 percent), average age 45 years (SD=7.35). They had served an average of 5.42 years (SD=4.39) in their coordinating role. In education, 62 percent had a Bachelor's degree and 38 percent a Master's degree. Of the school principals, 69 percent were women. Their average age was 48.82 years (SD=6.53), and they had an average of 10.42 years' experience directing their current school (SD=9.57). In education, 69 percent had a Bachelor's degree, and 31 percent had a master's degree. Data were obtained through a survey, and gathered from three different sources to avoid one-source bias. Team members were asked to fill out the scales of team justice climate and team psychological capital; team coordinators were asked to fill out the scale of team OCB; and school principals evaluated team innovation. All participants completed a demographic questionnaire. Data were collected on site at each school and all respondents were assured anonymity of responses. Measures All scales were measured on a five-point Likert scale from 1 (strongly disagree) to 5 (strongly agree). Team justice climate Team justice climate was measured by the 14-item scale of Li, Cropanzano and Benson (2007). The scale consists of three dimensions: first, team distributive justice climate (five items: e.g. "Some of my teammates have received more praise for the team projects than they deserved" (r)), a=0.71. Second, team procedural justice climate (five items: e.g. "The way my teammates make decisions is free from personal bias"), a=0.75. Third, team interpersonal justice climate (four items: e.g. "My teammates treat each other with respect"), a=0.34. Due to the low internal reliability of the interpersonal justice sub-scale (a=0.34), the model's analyses covered only the two sub-scales of team justice climate: distributive justice and procedural justice. Team PsyCap Team psychological capital was measured by the collective PsyCap scale developed by Mathe-Soulek et al. (2014), and adapted to the school context by Heled et al. (2015). The scale has eight items that measure the four dimensions of PsyCap: self-efficacy (two items: e.g. "Members of this team confidently contribute to discussions about the team's strategy"); hope (two items: e.g. "Members of this team think of many ways to attain work goals"); resilience (two items: e.g. "Members of this team usually take stressful things at work in their stride"); and optimism (two items: e.g. "Members of this team are optimistic about what will happen to them in the future as it pertains to work"). The questionnaire's reliability was 0.78. Team OCB Team OCB was measured on Vigoda-Gadot et al.'s (2007) scale, which was developed in the educational context. The scale has 18 items that measure two dimensions: OCBI, namely OCB directed at and contributing to a certain person at school (nine items: e.g. "Members of this team help teachers who have heavy workloads") (a=0.72), and OCBO, namely OCB directed to the school as a whole (nine items: e.g. "Our team makes innovative suggestions to improve school life") (a=0.81). The questionnaire's reliability was 0.80. Team innovation Team innovation was measured on a four-item scale adapted from West and Wallace (1991). The items reflected the extent to which in the previous six months the team had initiated changes in each of four job areas: work objectives, working methods, teaching methods and development of skills (e.g. "The team initiated new teaching methods) (a=0.93). Control variable Team size was chosen as a control variable. Previous studies showed that team size affects the team's process and its outcomes (Ancona and Caldwell, 1992; Heled et al., 2015; Keller, 2001). Level of analysis The unit of theory in the present study was the team. Accordingly, the hypotheses were posited at the team level, and the study variables (justice climate and team PsyCap) were aggregates of individual responses to the team level of analysis. The variables: team OCB, and team innovation were measured at the team level, and were evaluated by the team coordinator and the school principal, respectively. To justify aggregation of the variables team PsyCap and team OCB, it was critical to demonstrate high within-team agreement (rwg: James et al., 1993). A value of 0.70 or above is suggested as a "good" amount of within-group interrater agreement (James et al., 1993). All scales exceeded this criterion: average scores were 0.96 for team PsyCap, 0.90 for distributive justice and 0.90 for procedural justice. We also obtained the following ICC(1) and ICC(2) values: team PsyCap 0.42 and 0.71; distributive justice 0.64 and 0.83; procedural justice 0.45 and 0.70. As indicated by Bliese (2000), ICC(1) generally ranges from 0 to 0.50 with a median of 0.12. There are, however, no definite guidelines for determining acceptable values. All the above values were comparable to the median or to recommended ICC values so we concluded that aggregation was justified for these variables. Table I shows the means, standard deviations and correlations for the study variables. To test our model we used PROCESS, a computational tool that applies a bootstrapping technique (5,000 boots) to estimate the confidence interval (CI) of the mediation effect (indirect relationships) (Hayes, 2013). Bootstrapping is a non-parametric method based on re-sampling with replacement which is done many times, e.g. 5,000 times. From each of these samples, the indirect effect is computed and a sampling distribution can be empirically generated. Two mediation models were tested. The first examined the mediating role of team OCB (OCBI, OCBO) in the relation of justice climate (distributive justice and procedural justice) to team innovation. The second examined the mediating role of team OCB in the team PsyCap - team innovation relation. To analyze the two models we conducted hierarchical linear modeling analyses because teams (level 1) were nested within schools (level 2). The results are presented in Tables II and III. H1 proposed a positive relation between justice climate (distributive justice and procedural justice) and team OCB (OCBI, OCBO). A positive and significant relationship was indicated between procedural justice and team OCBI (b=0.78, t=2.97, p<0.01) and between procedural justice and team OCBO (b=0.79, t=2.73, p<0.01) (see Table II). However, no significant relation was found between distributive justice and OCBI and OCBO (p>0.05). H2 proposed a positive relationship between team PsyCap and team OCB (OCBI, OCBO): a positive and significant relation was indicated between PsyCap and team OCBI (b=0.55, t=2.69, p<0.01) and between PsyCap and team OCBO (b=0.99, t=3.21, p<0.01) (see Table II). Accordingly, H1 and H2 were supported. H3 proposed a positive relation between OCB (OCBI, OCBO) and team innovation. The results supported the hypothesis, showing a positive and significant relation between team OCBI and team innovation (b=0.25, t=5.95, p<0.001) and between team OCBI and team innovation (b=0.39, t=8.62, p<0.01) (see Table II). H4a proposed that team OCB (OCBI, OCBO) mediated the relation between justice climate and team innovation. The mediation analyses, based on bootstrapping, supported the hypothesis for the mediating role of OCBI (ab=0.21, CI 95% (0.04, 0.47)) and of OCBO (ab=0.34, CI 95% (0.13, 0.61)) in the relation of procedural justice to team innovation (see Table III). Finally, H4b proposed that team OCB (OCBI, OCBO) mediated the relation between team PsyCap and team innovation. The results supported the hypothesis for OCBI (ab=0.13, CI 95% (0.03, 0.45)), and for OCBO (ab=0.31, CI 95% (0.12, 0.60)). To summarize, regarding the role of justice climate's two dimensions - distributive justice and procedural justice - in promoting OCB, and in turn team innovation, only the latter was positively and significantly associated with team OCBO and OCBI. Furthermore, team OCBI and team OCBO alike fully mediated the relationship between procedural justice and team innovation. Regarding the role of team PsyCap in advancing team OCB and team innovation, the results showed positively and significantly associations between team PsyCap and team OCBO and OCBI. Also, team OCBI as well as team OCBO fully mediated the relationship between team PsyCap and team innovation. The present study explored team OCB from a context perspective. This course is important because to date most studies on OCB in schools have focused on teacher OCB as an individual feature, without regard for its contextual nature (Somech and Oplatka, 2014). Put simply, although OCB is performed by individuals, willingness to engage in OCB is not generated in a vacuum, and the team context most likely serves to encourage or discourage people in respect of exerting these extra-role behaviors (Vigoda-Gadot et al., 2007). In this regard, our findings highlighted the importance of team-level antecedents and outcomes for better understanding how between team differences in OCB are developed and what their consequences are. The results contribute to the educational administration literature in several respects. In general, the results support the premise that citizenship behaviors grow in a context. The OCB construct is "social" in nature, so the extent teams exhibit citizenship behaviors depends chiefly on teams' characteristics and context (Jackson, 2009). Specifically, the two contextual variables of team justice climate and team PsyCap proved positively associated with team OCB. This is important because it emphasizes the idea that leaders exercise crucial influence on teachers' tendency to contribute above and beyond their formal obligations (Somech and Oplatka, 2014). Leaders can create a team environment that determines when, and why, team members do or do not engage in OCBs (e.g. Ehrhart, 2004). Regarding the link between justice climate and OCB, the relationships between procedural justice and OCBI and OCBO were found positive. This suggests that when team members feel they are treated fairly they tend to exhibit higher levels of helping behaviors toward their colleagues (OCBI), as well as investing extra effort in promoting team and organization goals (OCBO). These findings may further support the exchange approach (Blau, 1964), as the cognitive mechanism sustaining these relationships. Team members who experience the decision-making processes and procedures in their team as fair will want to reciprocate by engaging in citizenship behaviors that may contribute the team and the school as a whole. However, unlike the procedural justice climate, we did not find significant relationships between distributive justice and the two dimensions of team OCB (OCBI and OCBO). These contrasting results accord with the pattern of relations found in previous studies on the links between perceived organizational justice and OCB at the individual level. For example, Moorman (1991) found a positive relation between perceptions of procedural justice and four out of five dimensions of OCB. Conversely, perceptions of distributive justice failed to influence any dimension of citizenship behavior. One possible explanation is that the distributive justice component refers not to social norms and procedures but to allocation of rewards and resources (Masterson et al., 2000). It does not possess the "social aspect" that team procedural justice climate does (Ambrose et al., 2002). Accordingly, distributive justice is less vital for establishing social exchange relationships, making it especially relevant in the context of parties that do not share social bonds. Thus, it does not contribute to the social exchange outcomes, as the procedural justice climate does (Ben-Sasson and Somech, 2015). Regarding the antecedent of team PsyCap, the results indicated, as expected, positive associations between team PsyCap and team OCBI and OCBO. In team PsyCap, high levels of optimism, hope, collective efficacy and resilience are positive psychological resources available to team members to share. They enable them to develop their team's strengths, as well as positive coping methods in daily life. All these create a positive atmosphere in the team and promote positive interactions and dynamics among its members, enhancing their willingness to exhibit citizenship behaviors to advance team and organizational effectiveness (Heled et al., 2015). This finding may provide additional support for the position that positive emotions reinforce individuals' and groups' tendency to engage in pro-social behaviors (George, 1996). Teams that perceive themselves as competent, in an upbeat mood, seek to maintain their positive emotional state by exhibiting positive social behaviors such as OCBs, which contribute to their social and psychological environment (Srivastava et al., 2006). This perspective may also suggest that the relationship between team PsyCap and team OCB is spiral: that is, teams high in PsyCap tend to exhibit high OCB levels, which in turn may further strengthen their team PsyCap levels. Luthans et al. (2011) proposed: "[...] overall PsyCap and its facets can be developed through intentional activities [...]. These intentional activities can increase levels of positivity and well-being in general (Lyubomirsky, 2007) [...]. In other words, positivity can beget more positivity" (p. 340). Furthermore, with respect to outcomes at the team level, the study findings support the H3 of a positive relation between team OCB (OCBI and OCBO) and team innovation. These findings promote the idea that teams that exhibit behaviors of helping, collaboration and support (OCBI), as well as exerting extra effort for the team and the school as a whole (OCBO), are more willing to seek challenges, to exhibit "out-of-the-box" thinking and exploration, namely to be innovative (Oldham and Cummings, 1996). This positive link between team OCB and team innovation may also be supported by Fredrickson's (1998, 2009) broaden-and-build theory, holding that a positive environment broadens thought-action tendencies. According to the theory, positivity broadens psychological, cognitive and physical resources that may motivate team members to exhibit high levels of innovation. The present results also support the mediating role of team OCB in the relationship between the two antecedents of procedural justice climate and team PsyCap and team innovation. Specifically, the findings demonstrated that procedural justice climate was positively related to team OCB, which in turn was positively related to team innovation. The results also demonstrated a positive link between team PsyCap and team OCB, which in turn was positively related to team innovation. These findings support the classic model of input-process-outcomes (Hackman and Morris, 1975), which emphasizes that team OCB serves as a vehicle whereby the contextual variables of team procedural justice climate and team PsyCap enhance team innovation. Accordingly, the present results demonstrate that team procedural justice climate and team PsyCap "set the stage" for encouraging the team to invest above and beyond the call of duty, namely to engage in citizenship behaviors, which create a positive and supporting psychological and social environment, in order to maximize their innovation (Yi, 2005). Beyond this study's theoretical contribution, its findings also have practical implications. By taking a team perspective, the findings offer evidence that despite the usual approach treating OCB as a personal or individual phenomenon; it may be regarded as shared norms of behaviors at the team level. In other words, interactions among team members over time, in a certain context, determine the extent team members exhibit citizenship behaviors, which are important for promoting team effectiveness (Podsakoff et al., 2009). This contextual approach may urge principals and other educational leaders to realize that teachers' willingness to invest extra effort in school is mainly the result of an appropriate team context, which can be shaped and developed. Specifically, by creating an atmosphere of procedural justice and by enhancing a sense of team's competence, resilience, optimism and hope (team PsyCap), leaders may motivate educational team members to exhibit high levels of OCBs, which in turn may improve those outcomes that promote the success of the individual, the team, and the entire school. Finally, as in any research, this study has some limitations as well. First, data were gathered using questionnaires so certain potential biases may be present, such as one-source bias or subjectivity of the responses. Nevertheless, recall that in this study data were gathered from three different sources (team members, team coordinators, principals) to avoid one-source bias (Avolio et al., 1991). It is, however, recommended that in future studies researchers employ more objective measures to further validate the present results. Second, the nature of the data collecting limits the extent our logical explanations can be conclusively supported by the data. The causal direction may be the reverse, and teams that are more innovative may exhibit higher levels of OCBs; alternately, engaging in citizenship behaviors may affect the levels of justice climate and/or team PsyCap. Future longitudinal research should be conducted before causal explanations. Finally, the present research focused on two variables as antecedents: team justice climate and team PsyCap. We recommend examining team OCB with respect to additional context variables, such as team composition as promoting or inhibiting team OCB. Similarly, although in the present study, we focused on a crucial team-level outcome of innovation, we encourage researchers to expand future research and examine additional variables as outcomes at the individual, team and school levels.
|
The purpose of this paper is to use a model to broaden the understanding of the organizational citizenship behavior (OCB) phenomenon in educational teams and examines team OCB's mediating role in the relation of the contextual variables of team justice climate (distributive justice, procedural justice, interpersonal justice) to team psychological capital (PsyCap) and team innovation.
|
[SECTION: Method] Today schools operate in dynamic environment, each struggling to gain a competitive edge (Orr and Orphanos, 2011). This environment reinforces the understanding that schools should strive to employ teachers who are willing to go the extra mile, namely to engage in organizational citizenship behavior (OCB). OCBs are behaviors and actions that go beyond the formal role and what is laid down in the teacher's job description, and contribute to the achievement of the school's objectives (Oplatka, 2006). OCB in schools is commonly taken to be an individual phenomenon, namely a personality trait or a social response to the behavior or attitude of superiors or to other motivation-based mechanisms in the workplace (Zeinabadi, 2010). This individual approach seems wanting: teachers perform or fail to perform OCBs not in a vacuum but in an organizational context, which most probably serves to encourage or discourage them (Somech and Drach-Zahavy, 2004). Only in the last two decades have some scholars studied OCB as a team or organizational-level concept (e.g. Ehrhart, 2004). Put otherwise, the individual approach probes the possibility that teacher OCB can be better understood as a team or an organizational feature that thrives in a context. This development is important because the aggregate OCB level, not sporadic actions by some individuals, influences organizational effectiveness (Organ, 1988). Moreover, the several studies that have captured OCB as a collective phenomenon examined either its antecedents (e.g. Hu and Liden, 2011) or its consequences (e.g. Yen et al., 2008). Very few researchers, if any, have explored the mediating role of team OCB in relation to specific antecedents and outcomes. The present research seeks to address this deficiency by examining OCB as a team phenomenon. The first premise of the research model is that it is important to measure OCB as a collective structure. Next, the mediating role of team OCB is investigated. Specifically, the research model posits that the contextual variable of a team's justice climate, and the team's collective psychological state (psychological capital), will be positively related to team OCB. Furthermore, team OCB will be positively related to the outcome of team innovation. The model also suggests that team OCB will mediate the relation between the antecedents and the outcome. Teachers' OCB is defined as "[...] teachers voluntarily going out of their way to help their students, colleagues, and others as they engage in the work of teaching and learning" (DiPaola and Hoy, 2005, p. 390). OCBs are essential because formal in-role job descriptions cannot cover the entire array of behaviors needed for achieving school goals (Zeinabadi, 2010). They operate indirectly: they influence schools' social and psychological environment, enhance school effectiveness by freeing up resources for more productive purposes, help coordinate activities within the organization, and enable teachers to adapt more effectively to environmental changes (Sesen and Basim, 2012). Teachers manifest OCB, for instance, by staying after school hours to help a student with learning materials, aid a colleague with a heavy workload, volunteer for unpaid tasks, or make innovative suggestions to improve the school (Somech and Oplatka, 2014). Scholars have identified two broad dimensions of OCB (Sesen and Basim, 2012; Williams and Anderson, 1991). The first is OCBI (OCBs toward individuals in school), namely behaviors that immediately benefit particular individuals, such as colleagues and/or students, hence indirectly contribute to the organization. This dimension includes such behaviors as taking a personal interest in a novice teacher or helping a student after working hours. The second dimension is OCBO (OCB toward the organization), namely behaviors that benefit the school as a whole, for example, organizing special social activities for the school or making innovative suggestions to improve it. According to Podsakoff et al. (2009), these two broad categories capture most previous taxonomies effectively. Although OCBs are acts by individuals, our main argument is that these extra-role behaviors grow in a context, therefore can be extended to the team level. Put simply, because OCBs are interactive or "social" in nature, teams differ in their teachers' tendency to engage in them (Ehrhart, 2004). Social psychology research has shown that teams are powerful instruments of social influence and significantly affect individuals' perceptions, attitudes and behaviors (Mathieu et al., 2000). Two theoretical models may provide the rationale for considering OCB a team phenomenon: the social learning theory and the social information processing theory. The first posits that people learn by observing the behavior of others (Bandura, 1986). The more OCB is modeled by the members of one's team, the more likely one is to behave in a manner consistent with those models, in particular when that behavior is associated with positive social consequences (Ehrhart, 2004). The second the social information processing theory (Salancik and Pfeffer, 1978) underlines the importance of social cues in shaping attitudes. When OCB levels are high among team members, an individual member will be more likely to view such behavior as acceptable and expected. Ultimately this recognition results in higher levels of individual OCB. George (1991), for example, found high similarities among team members in the level of positive and/or negative mood, as well as in the extent they exhibited citizenship behaviors. Conceptualizing OCB as a team phenomenon implies that teachers who do or do not perform OCB do not do so in a vacuum; the team context in which these behaviors are performed undoubtedly serves to encourage or discourage them (Vigoda-Gadot et al., 2007). In other words, this perspective suggests that contextual variables determine when, and sometimes why, teachers do or do not engage in OCBs (Ehrhart, 2004). In this vein, the present model focused on two team-level variables as antecedents of team OCB: justice climate and team psychological capital, and team innovation as a consequence. Furthermore, the model suggested that team OCB serves as a mediator in the relation of justice climate to team psychological capital and team innovation. The relation of justice climate to team psychological capital and team OCB We focused on two elements of team characteristics: team justice climate and team psychological capital. Our choice of these two variables was grounded in recent OCB research emphasizing the critical role of positive team atmosphere in enhancing positive attitudes and behaviors, as expressed by team citizenship behaviors (Lau and Lam, 2008). Team justice climate and team OCB Team climate derives from the array of the work-contextual properties that the team emphasizes and evaluates; these are perceived as critical in shaping team members' attitudes and behaviors (Schneider et al., 2013). The main claim of climate research is the idea of shared perceptions constructed through social interactions among individual employees engaged in teamwork (Spell and Arnold, 2007). Although there are several kinds of climate in every workplace, justice climate is crucial for promoting a positive atmosphere prevailing over team members, as well as between them and their supervisor and the organization as a whole (Colquitt et al., 2002). Building on the perception of a "social" creation of justice perceptions, justice climate is defined as "'shared perceptions' of team fairness in outcomes, treatment, and supervisory interaction" (Moliner et al., 2005, p. 101). Scholars have identified three distinct dimensions of justice climate: distributive justice, procedural justice and interpersonal justice (Li and Cropanzano, 2008). Distributive justice climate addresses the organizational reward system and reflects fairness of outcomes, namely fairness in distribution of tangible or intangible rewards and/or resources (Li and Cropanzano, 2008); procedural justice climate is conceptualized as fairness of procedures and of the decision-making process (i.e. the degree to which decisions on reward allocation are made according to fair methods and guidelines) (Niehoff and Moorman, 1993); interpersonal justice climate refers to the treatment accorded to team members when decisions are made. It marks the extent to which team members are treated with respect and dignity during the decision-making process (Li and Cropanzano, 2008). The link between justice climate and team OCB can be perceived through social exchange theory (Blau, 1964) and through the norm of reciprocity (Gouldner, 1960). These two models suggest that team members seek to balance their inputs and outcomes in relation to others (Elstad et al., 2012). Generally, with high levels of justice climate, team members share the perception that their organization and/or its representatives can be trusted to protect their interests; this engenders an obligation to reciprocate in positive, beneficial ways, namely OCBs (Lambert, 2000). More specifically, when the collective experience of the team is that they are rewarded well (distributive justice), the team as the whole is treated with respect and dignity (interpersonal justice), and the process leading to outcomes is fairly applied (procedural justice); team members will seek to reciprocate by investing in citizenship behaviors (Parzefall and Salin, 2010). OCB is especially suitable for reciprocity because whereas traditional in-role behaviors can be enforced on the basis of formal role obligations (e.g. Konovsky and Pugh, 1994), OCB consists of informal contributions that team members can choose to exhibit or withhold (Somech and Oplatka, 2014). For example, Ehrhart (2004), who found a positive association between procedural justice climate and team OCB, suggested: "If group members perceive they are treated fairly and this perspective is shared among them [...] their social interactions will also result in similar responses to that assigned meaning, such as the performance of higher levels of OCB" (p. 67). Accordingly, we propose a positive link between justice climate and team OCB: H1. A positive relationship will be found between team justice climate and team OCB. Team psychological capital and team OCB Team psychological capital (team PsyCap) is a team's positive psychological state of development (Luthans et al., 2008), and is characterized by four psychological components: hope, optimism, collective efficacy and resilience (Avey et al., 2009; Heled et al., 2015). Hope is defined as a positive motivational state of the team that is based on a feeling that is fueled by successful agency (goal-oriented energy) and pathways (planning to meet goals) (Snyder et al., 1996). Optimism is the team's ability to create a positive attribution with respect to success in both the present and the future. Collective efficacy means that "a group shares a common belief in regards to the standardization of the capacities of the group members" (Bandura, 1997, p. 477). And resilience is a positive psychological ability of the team "to respond productively to significant change that disrupts the expected pattern of events without engaging in an extended period of regressive behavior" (Horne and Orr, 1997, p. 31). Although each of these four components has the potential to influence teamwork, studies support the notion that together, they have a much greater impact than each variable individually (Luthans et al., 2007). Team PsyCap is a construct of shared psychological state of team members, and can be considered an emergent state (Heled et al., 2015). According to Marks et al. (2001), emergent states are "constructs that characterize properties of the team that are typically dynamic in nature and vary as a function of team context, inputs, processes, and outcomes. Emergent states describe cognitive, motivational, and affective states of teams [...]" (p. 357). Put simply, team PsyCap is a team resource that can be developed and managed, and may thereby influence team processes and outcomes (Mathe-Soulek et al., 2014). The present study posits a positive link between team PsyCap and team OCB. Generally, teams with high levels of PsyCap are confident that good things happen at work; they believe they create their success and can overcome failure (Avey et al., 2011). Accordingly, we argue that team PsyCap plays an essential role in promoting positive attitudes and behaviors among team members, as expressed by team OCBs. The affective approach, which derives from the social-psychological literature (e.g. Williams and Shiaw, 1999), may provide the theoretical foundation for the relationship (George, 1991). This approach suggests that when people are in a positive mood, and hold positive attitudes; they are more likely to be helpful, namely to exhibit high levels of citizenship behaviors. Put simply, team members who experience a good mood, a sense of optimism and hope tend to perceive other people and situations in a more positive light (George, 1996), which may predispose individuals to exhibit behaviors that benefit team members, enhance organizational outcomes, or both (Heled et al., 2015). In other words, teams characterized by optimism, hope and resilience will tend to demonstrate pro-social behaviors that preserve their positive emotional state (Srivastava et al., 2006). Furthermore, this approach suggests that teams in a good mood are more helpful because being so enables them to maintain their positive mood (George, 1991). For example, Heled et al. (2015), who examined 82 school management teams, found positive relations between team PsyCap and the individual's job satisfaction and team OCB. Accordingly, we propose a positive link between Team PsyCap and team OCB: H2. A positive relation will be found between team PsyCap and team OCB. The relation between team OCB and team innovation Educational innovation is perceived as a key factor for promoting school effectiveness (Gilad-Hai and Somech, 2016). Innovation enables schools to cope with a complex and changing environment (Chen, 2010; Tubin, 2009). Team innovation is defined as "the intentional introduction and application within a team, of ideas, processes, products or procedures new to the team, designed to significantly benefit the individual, the team, the organization, or wider society" (West and Wallace, 1991, p. 303). Similarly, educational innovation is defined as "an idea or practice that reflects an emergent trend and offers an optimistic alternative for a current situation of dissatisfaction" (Tubin, 2009, p. 406). Teams may offer an alternative approach to learning, teaching and/or administrative routines (Gilad-Hai and Somech, 2016). Overall, innovation undergoes two phases: the creativity phase, namely generation or production of ideas that are both novel and useful (George, 2007), and the implementation phase, namely "the process of gaining targeted employees' appropriate and committed use of an innovation" (Klein and Sorra, 1996, p. 1055). According to the present model, team OCB can be linked to team innovation. There are logical reasons to expect a positive relation between team OCB and team innovation. According to Organ (1997), OCBs support the social and psychological environment within which the technical core functions. Accordingly, teachers who are willing to collaborate with their team members, to actively participate in meetings, to share knowledge, to explore new alternatives, or to learn new skills, contribute to the team's ability to be innovative (Jimmieson et al., 2010). Moreover, at the stage of innovation implementation, when job definitions are ambiguous, schools depend more on teams that are willing to contribute to successful change, regardless of formal job requirements (Oplatka, 2006). Thus, team OCB is assumed to help schools adopt changes and innovations that are necessary in the consistently changing environment (Savelyeva and Lee, 2012): H3. A positive relationship will be found between team OCB and team innovation. The mediating role of team psychological capital Finally, we propose that team OCB will serve as a mediator between the antecedents of team justice climate and team PsyCap and the outcomes: team innovation (Figure 1). This argument can be supported by the classic team effectiveness models of input-process-outcomes (Hackman and Morris, 1975) wherein team effectiveness involves a three-stage process. In the first stage, team and job characteristics (inputs) that structure teamwork influence the work processes in stage 2 (processes), which in turn lead to positive results in stage 3 (outcomes). Accordingly, we suggest that justice climate and team PsyCap set the stage for encouraging team members to invest in the team above a beyond the call of duty, namely to exhibit high levels of citizenship behaviors, which in turn enhance team innovation: H4a. Team OCB will mediate the relationship between justice climate and team innovation. H4b. Team OCB will mediate the relationship between team PsyCap and team innovation. Sample and procedure This study covered 78 disciplinary teams sampled from 13 junior-high schools (grades 7-9), located in the Northern and Central parts of Israel. In each school, we sampled six disciplinary teams (linguistics, civics, history, English, mathematics, science), totaling 214 team members, 78 coordinators and 13 school principals. We conducted pre-assessment interviews with the principal of each school, and learned that all team members interacted regularly to achieve shared goals regarding teaching issues and other school matters. They also depended on one another for knowledge and effort by means of several permanent structures such as scheduled staff meetings, and joint refresher workshops. Team size ranged from three to six members, with an average of 3.26 (SD=1.73). Of the team members, 181 were women (84.5 percent) and 33 were men. Average age was 45.66 years (SD=7.85), with average job tenure 6.73 (SD=4.42). In education, 80 percent of participants had a bachelor's degree, 20 percent had a master's degree. Of the coordinators, 56 were women (71.8 percent), average age 45 years (SD=7.35). They had served an average of 5.42 years (SD=4.39) in their coordinating role. In education, 62 percent had a Bachelor's degree and 38 percent a Master's degree. Of the school principals, 69 percent were women. Their average age was 48.82 years (SD=6.53), and they had an average of 10.42 years' experience directing their current school (SD=9.57). In education, 69 percent had a Bachelor's degree, and 31 percent had a master's degree. Data were obtained through a survey, and gathered from three different sources to avoid one-source bias. Team members were asked to fill out the scales of team justice climate and team psychological capital; team coordinators were asked to fill out the scale of team OCB; and school principals evaluated team innovation. All participants completed a demographic questionnaire. Data were collected on site at each school and all respondents were assured anonymity of responses. Measures All scales were measured on a five-point Likert scale from 1 (strongly disagree) to 5 (strongly agree). Team justice climate Team justice climate was measured by the 14-item scale of Li, Cropanzano and Benson (2007). The scale consists of three dimensions: first, team distributive justice climate (five items: e.g. "Some of my teammates have received more praise for the team projects than they deserved" (r)), a=0.71. Second, team procedural justice climate (five items: e.g. "The way my teammates make decisions is free from personal bias"), a=0.75. Third, team interpersonal justice climate (four items: e.g. "My teammates treat each other with respect"), a=0.34. Due to the low internal reliability of the interpersonal justice sub-scale (a=0.34), the model's analyses covered only the two sub-scales of team justice climate: distributive justice and procedural justice. Team PsyCap Team psychological capital was measured by the collective PsyCap scale developed by Mathe-Soulek et al. (2014), and adapted to the school context by Heled et al. (2015). The scale has eight items that measure the four dimensions of PsyCap: self-efficacy (two items: e.g. "Members of this team confidently contribute to discussions about the team's strategy"); hope (two items: e.g. "Members of this team think of many ways to attain work goals"); resilience (two items: e.g. "Members of this team usually take stressful things at work in their stride"); and optimism (two items: e.g. "Members of this team are optimistic about what will happen to them in the future as it pertains to work"). The questionnaire's reliability was 0.78. Team OCB Team OCB was measured on Vigoda-Gadot et al.'s (2007) scale, which was developed in the educational context. The scale has 18 items that measure two dimensions: OCBI, namely OCB directed at and contributing to a certain person at school (nine items: e.g. "Members of this team help teachers who have heavy workloads") (a=0.72), and OCBO, namely OCB directed to the school as a whole (nine items: e.g. "Our team makes innovative suggestions to improve school life") (a=0.81). The questionnaire's reliability was 0.80. Team innovation Team innovation was measured on a four-item scale adapted from West and Wallace (1991). The items reflected the extent to which in the previous six months the team had initiated changes in each of four job areas: work objectives, working methods, teaching methods and development of skills (e.g. "The team initiated new teaching methods) (a=0.93). Control variable Team size was chosen as a control variable. Previous studies showed that team size affects the team's process and its outcomes (Ancona and Caldwell, 1992; Heled et al., 2015; Keller, 2001). Level of analysis The unit of theory in the present study was the team. Accordingly, the hypotheses were posited at the team level, and the study variables (justice climate and team PsyCap) were aggregates of individual responses to the team level of analysis. The variables: team OCB, and team innovation were measured at the team level, and were evaluated by the team coordinator and the school principal, respectively. To justify aggregation of the variables team PsyCap and team OCB, it was critical to demonstrate high within-team agreement (rwg: James et al., 1993). A value of 0.70 or above is suggested as a "good" amount of within-group interrater agreement (James et al., 1993). All scales exceeded this criterion: average scores were 0.96 for team PsyCap, 0.90 for distributive justice and 0.90 for procedural justice. We also obtained the following ICC(1) and ICC(2) values: team PsyCap 0.42 and 0.71; distributive justice 0.64 and 0.83; procedural justice 0.45 and 0.70. As indicated by Bliese (2000), ICC(1) generally ranges from 0 to 0.50 with a median of 0.12. There are, however, no definite guidelines for determining acceptable values. All the above values were comparable to the median or to recommended ICC values so we concluded that aggregation was justified for these variables. Table I shows the means, standard deviations and correlations for the study variables. To test our model we used PROCESS, a computational tool that applies a bootstrapping technique (5,000 boots) to estimate the confidence interval (CI) of the mediation effect (indirect relationships) (Hayes, 2013). Bootstrapping is a non-parametric method based on re-sampling with replacement which is done many times, e.g. 5,000 times. From each of these samples, the indirect effect is computed and a sampling distribution can be empirically generated. Two mediation models were tested. The first examined the mediating role of team OCB (OCBI, OCBO) in the relation of justice climate (distributive justice and procedural justice) to team innovation. The second examined the mediating role of team OCB in the team PsyCap - team innovation relation. To analyze the two models we conducted hierarchical linear modeling analyses because teams (level 1) were nested within schools (level 2). The results are presented in Tables II and III. H1 proposed a positive relation between justice climate (distributive justice and procedural justice) and team OCB (OCBI, OCBO). A positive and significant relationship was indicated between procedural justice and team OCBI (b=0.78, t=2.97, p<0.01) and between procedural justice and team OCBO (b=0.79, t=2.73, p<0.01) (see Table II). However, no significant relation was found between distributive justice and OCBI and OCBO (p>0.05). H2 proposed a positive relationship between team PsyCap and team OCB (OCBI, OCBO): a positive and significant relation was indicated between PsyCap and team OCBI (b=0.55, t=2.69, p<0.01) and between PsyCap and team OCBO (b=0.99, t=3.21, p<0.01) (see Table II). Accordingly, H1 and H2 were supported. H3 proposed a positive relation between OCB (OCBI, OCBO) and team innovation. The results supported the hypothesis, showing a positive and significant relation between team OCBI and team innovation (b=0.25, t=5.95, p<0.001) and between team OCBI and team innovation (b=0.39, t=8.62, p<0.01) (see Table II). H4a proposed that team OCB (OCBI, OCBO) mediated the relation between justice climate and team innovation. The mediation analyses, based on bootstrapping, supported the hypothesis for the mediating role of OCBI (ab=0.21, CI 95% (0.04, 0.47)) and of OCBO (ab=0.34, CI 95% (0.13, 0.61)) in the relation of procedural justice to team innovation (see Table III). Finally, H4b proposed that team OCB (OCBI, OCBO) mediated the relation between team PsyCap and team innovation. The results supported the hypothesis for OCBI (ab=0.13, CI 95% (0.03, 0.45)), and for OCBO (ab=0.31, CI 95% (0.12, 0.60)). To summarize, regarding the role of justice climate's two dimensions - distributive justice and procedural justice - in promoting OCB, and in turn team innovation, only the latter was positively and significantly associated with team OCBO and OCBI. Furthermore, team OCBI and team OCBO alike fully mediated the relationship between procedural justice and team innovation. Regarding the role of team PsyCap in advancing team OCB and team innovation, the results showed positively and significantly associations between team PsyCap and team OCBO and OCBI. Also, team OCBI as well as team OCBO fully mediated the relationship between team PsyCap and team innovation. The present study explored team OCB from a context perspective. This course is important because to date most studies on OCB in schools have focused on teacher OCB as an individual feature, without regard for its contextual nature (Somech and Oplatka, 2014). Put simply, although OCB is performed by individuals, willingness to engage in OCB is not generated in a vacuum, and the team context most likely serves to encourage or discourage people in respect of exerting these extra-role behaviors (Vigoda-Gadot et al., 2007). In this regard, our findings highlighted the importance of team-level antecedents and outcomes for better understanding how between team differences in OCB are developed and what their consequences are. The results contribute to the educational administration literature in several respects. In general, the results support the premise that citizenship behaviors grow in a context. The OCB construct is "social" in nature, so the extent teams exhibit citizenship behaviors depends chiefly on teams' characteristics and context (Jackson, 2009). Specifically, the two contextual variables of team justice climate and team PsyCap proved positively associated with team OCB. This is important because it emphasizes the idea that leaders exercise crucial influence on teachers' tendency to contribute above and beyond their formal obligations (Somech and Oplatka, 2014). Leaders can create a team environment that determines when, and why, team members do or do not engage in OCBs (e.g. Ehrhart, 2004). Regarding the link between justice climate and OCB, the relationships between procedural justice and OCBI and OCBO were found positive. This suggests that when team members feel they are treated fairly they tend to exhibit higher levels of helping behaviors toward their colleagues (OCBI), as well as investing extra effort in promoting team and organization goals (OCBO). These findings may further support the exchange approach (Blau, 1964), as the cognitive mechanism sustaining these relationships. Team members who experience the decision-making processes and procedures in their team as fair will want to reciprocate by engaging in citizenship behaviors that may contribute the team and the school as a whole. However, unlike the procedural justice climate, we did not find significant relationships between distributive justice and the two dimensions of team OCB (OCBI and OCBO). These contrasting results accord with the pattern of relations found in previous studies on the links between perceived organizational justice and OCB at the individual level. For example, Moorman (1991) found a positive relation between perceptions of procedural justice and four out of five dimensions of OCB. Conversely, perceptions of distributive justice failed to influence any dimension of citizenship behavior. One possible explanation is that the distributive justice component refers not to social norms and procedures but to allocation of rewards and resources (Masterson et al., 2000). It does not possess the "social aspect" that team procedural justice climate does (Ambrose et al., 2002). Accordingly, distributive justice is less vital for establishing social exchange relationships, making it especially relevant in the context of parties that do not share social bonds. Thus, it does not contribute to the social exchange outcomes, as the procedural justice climate does (Ben-Sasson and Somech, 2015). Regarding the antecedent of team PsyCap, the results indicated, as expected, positive associations between team PsyCap and team OCBI and OCBO. In team PsyCap, high levels of optimism, hope, collective efficacy and resilience are positive psychological resources available to team members to share. They enable them to develop their team's strengths, as well as positive coping methods in daily life. All these create a positive atmosphere in the team and promote positive interactions and dynamics among its members, enhancing their willingness to exhibit citizenship behaviors to advance team and organizational effectiveness (Heled et al., 2015). This finding may provide additional support for the position that positive emotions reinforce individuals' and groups' tendency to engage in pro-social behaviors (George, 1996). Teams that perceive themselves as competent, in an upbeat mood, seek to maintain their positive emotional state by exhibiting positive social behaviors such as OCBs, which contribute to their social and psychological environment (Srivastava et al., 2006). This perspective may also suggest that the relationship between team PsyCap and team OCB is spiral: that is, teams high in PsyCap tend to exhibit high OCB levels, which in turn may further strengthen their team PsyCap levels. Luthans et al. (2011) proposed: "[...] overall PsyCap and its facets can be developed through intentional activities [...]. These intentional activities can increase levels of positivity and well-being in general (Lyubomirsky, 2007) [...]. In other words, positivity can beget more positivity" (p. 340). Furthermore, with respect to outcomes at the team level, the study findings support the H3 of a positive relation between team OCB (OCBI and OCBO) and team innovation. These findings promote the idea that teams that exhibit behaviors of helping, collaboration and support (OCBI), as well as exerting extra effort for the team and the school as a whole (OCBO), are more willing to seek challenges, to exhibit "out-of-the-box" thinking and exploration, namely to be innovative (Oldham and Cummings, 1996). This positive link between team OCB and team innovation may also be supported by Fredrickson's (1998, 2009) broaden-and-build theory, holding that a positive environment broadens thought-action tendencies. According to the theory, positivity broadens psychological, cognitive and physical resources that may motivate team members to exhibit high levels of innovation. The present results also support the mediating role of team OCB in the relationship between the two antecedents of procedural justice climate and team PsyCap and team innovation. Specifically, the findings demonstrated that procedural justice climate was positively related to team OCB, which in turn was positively related to team innovation. The results also demonstrated a positive link between team PsyCap and team OCB, which in turn was positively related to team innovation. These findings support the classic model of input-process-outcomes (Hackman and Morris, 1975), which emphasizes that team OCB serves as a vehicle whereby the contextual variables of team procedural justice climate and team PsyCap enhance team innovation. Accordingly, the present results demonstrate that team procedural justice climate and team PsyCap "set the stage" for encouraging the team to invest above and beyond the call of duty, namely to engage in citizenship behaviors, which create a positive and supporting psychological and social environment, in order to maximize their innovation (Yi, 2005). Beyond this study's theoretical contribution, its findings also have practical implications. By taking a team perspective, the findings offer evidence that despite the usual approach treating OCB as a personal or individual phenomenon; it may be regarded as shared norms of behaviors at the team level. In other words, interactions among team members over time, in a certain context, determine the extent team members exhibit citizenship behaviors, which are important for promoting team effectiveness (Podsakoff et al., 2009). This contextual approach may urge principals and other educational leaders to realize that teachers' willingness to invest extra effort in school is mainly the result of an appropriate team context, which can be shaped and developed. Specifically, by creating an atmosphere of procedural justice and by enhancing a sense of team's competence, resilience, optimism and hope (team PsyCap), leaders may motivate educational team members to exhibit high levels of OCBs, which in turn may improve those outcomes that promote the success of the individual, the team, and the entire school. Finally, as in any research, this study has some limitations as well. First, data were gathered using questionnaires so certain potential biases may be present, such as one-source bias or subjectivity of the responses. Nevertheless, recall that in this study data were gathered from three different sources (team members, team coordinators, principals) to avoid one-source bias (Avolio et al., 1991). It is, however, recommended that in future studies researchers employ more objective measures to further validate the present results. Second, the nature of the data collecting limits the extent our logical explanations can be conclusively supported by the data. The causal direction may be the reverse, and teams that are more innovative may exhibit higher levels of OCBs; alternately, engaging in citizenship behaviors may affect the levels of justice climate and/or team PsyCap. Future longitudinal research should be conducted before causal explanations. Finally, the present research focused on two variables as antecedents: team justice climate and team PsyCap. We recommend examining team OCB with respect to additional context variables, such as team composition as promoting or inhibiting team OCB. Similarly, although in the present study, we focused on a crucial team-level outcome of innovation, we encourage researchers to expand future research and examine additional variables as outcomes at the individual, team and school levels.
|
Data were collected through a survey consisting of validated scales. The sample covered 78 disciplinary teams, 78 coordinators and 13 school principals at 13 junior-high schools.
|
[SECTION: Findings] Today schools operate in dynamic environment, each struggling to gain a competitive edge (Orr and Orphanos, 2011). This environment reinforces the understanding that schools should strive to employ teachers who are willing to go the extra mile, namely to engage in organizational citizenship behavior (OCB). OCBs are behaviors and actions that go beyond the formal role and what is laid down in the teacher's job description, and contribute to the achievement of the school's objectives (Oplatka, 2006). OCB in schools is commonly taken to be an individual phenomenon, namely a personality trait or a social response to the behavior or attitude of superiors or to other motivation-based mechanisms in the workplace (Zeinabadi, 2010). This individual approach seems wanting: teachers perform or fail to perform OCBs not in a vacuum but in an organizational context, which most probably serves to encourage or discourage them (Somech and Drach-Zahavy, 2004). Only in the last two decades have some scholars studied OCB as a team or organizational-level concept (e.g. Ehrhart, 2004). Put otherwise, the individual approach probes the possibility that teacher OCB can be better understood as a team or an organizational feature that thrives in a context. This development is important because the aggregate OCB level, not sporadic actions by some individuals, influences organizational effectiveness (Organ, 1988). Moreover, the several studies that have captured OCB as a collective phenomenon examined either its antecedents (e.g. Hu and Liden, 2011) or its consequences (e.g. Yen et al., 2008). Very few researchers, if any, have explored the mediating role of team OCB in relation to specific antecedents and outcomes. The present research seeks to address this deficiency by examining OCB as a team phenomenon. The first premise of the research model is that it is important to measure OCB as a collective structure. Next, the mediating role of team OCB is investigated. Specifically, the research model posits that the contextual variable of a team's justice climate, and the team's collective psychological state (psychological capital), will be positively related to team OCB. Furthermore, team OCB will be positively related to the outcome of team innovation. The model also suggests that team OCB will mediate the relation between the antecedents and the outcome. Teachers' OCB is defined as "[...] teachers voluntarily going out of their way to help their students, colleagues, and others as they engage in the work of teaching and learning" (DiPaola and Hoy, 2005, p. 390). OCBs are essential because formal in-role job descriptions cannot cover the entire array of behaviors needed for achieving school goals (Zeinabadi, 2010). They operate indirectly: they influence schools' social and psychological environment, enhance school effectiveness by freeing up resources for more productive purposes, help coordinate activities within the organization, and enable teachers to adapt more effectively to environmental changes (Sesen and Basim, 2012). Teachers manifest OCB, for instance, by staying after school hours to help a student with learning materials, aid a colleague with a heavy workload, volunteer for unpaid tasks, or make innovative suggestions to improve the school (Somech and Oplatka, 2014). Scholars have identified two broad dimensions of OCB (Sesen and Basim, 2012; Williams and Anderson, 1991). The first is OCBI (OCBs toward individuals in school), namely behaviors that immediately benefit particular individuals, such as colleagues and/or students, hence indirectly contribute to the organization. This dimension includes such behaviors as taking a personal interest in a novice teacher or helping a student after working hours. The second dimension is OCBO (OCB toward the organization), namely behaviors that benefit the school as a whole, for example, organizing special social activities for the school or making innovative suggestions to improve it. According to Podsakoff et al. (2009), these two broad categories capture most previous taxonomies effectively. Although OCBs are acts by individuals, our main argument is that these extra-role behaviors grow in a context, therefore can be extended to the team level. Put simply, because OCBs are interactive or "social" in nature, teams differ in their teachers' tendency to engage in them (Ehrhart, 2004). Social psychology research has shown that teams are powerful instruments of social influence and significantly affect individuals' perceptions, attitudes and behaviors (Mathieu et al., 2000). Two theoretical models may provide the rationale for considering OCB a team phenomenon: the social learning theory and the social information processing theory. The first posits that people learn by observing the behavior of others (Bandura, 1986). The more OCB is modeled by the members of one's team, the more likely one is to behave in a manner consistent with those models, in particular when that behavior is associated with positive social consequences (Ehrhart, 2004). The second the social information processing theory (Salancik and Pfeffer, 1978) underlines the importance of social cues in shaping attitudes. When OCB levels are high among team members, an individual member will be more likely to view such behavior as acceptable and expected. Ultimately this recognition results in higher levels of individual OCB. George (1991), for example, found high similarities among team members in the level of positive and/or negative mood, as well as in the extent they exhibited citizenship behaviors. Conceptualizing OCB as a team phenomenon implies that teachers who do or do not perform OCB do not do so in a vacuum; the team context in which these behaviors are performed undoubtedly serves to encourage or discourage them (Vigoda-Gadot et al., 2007). In other words, this perspective suggests that contextual variables determine when, and sometimes why, teachers do or do not engage in OCBs (Ehrhart, 2004). In this vein, the present model focused on two team-level variables as antecedents of team OCB: justice climate and team psychological capital, and team innovation as a consequence. Furthermore, the model suggested that team OCB serves as a mediator in the relation of justice climate to team psychological capital and team innovation. The relation of justice climate to team psychological capital and team OCB We focused on two elements of team characteristics: team justice climate and team psychological capital. Our choice of these two variables was grounded in recent OCB research emphasizing the critical role of positive team atmosphere in enhancing positive attitudes and behaviors, as expressed by team citizenship behaviors (Lau and Lam, 2008). Team justice climate and team OCB Team climate derives from the array of the work-contextual properties that the team emphasizes and evaluates; these are perceived as critical in shaping team members' attitudes and behaviors (Schneider et al., 2013). The main claim of climate research is the idea of shared perceptions constructed through social interactions among individual employees engaged in teamwork (Spell and Arnold, 2007). Although there are several kinds of climate in every workplace, justice climate is crucial for promoting a positive atmosphere prevailing over team members, as well as between them and their supervisor and the organization as a whole (Colquitt et al., 2002). Building on the perception of a "social" creation of justice perceptions, justice climate is defined as "'shared perceptions' of team fairness in outcomes, treatment, and supervisory interaction" (Moliner et al., 2005, p. 101). Scholars have identified three distinct dimensions of justice climate: distributive justice, procedural justice and interpersonal justice (Li and Cropanzano, 2008). Distributive justice climate addresses the organizational reward system and reflects fairness of outcomes, namely fairness in distribution of tangible or intangible rewards and/or resources (Li and Cropanzano, 2008); procedural justice climate is conceptualized as fairness of procedures and of the decision-making process (i.e. the degree to which decisions on reward allocation are made according to fair methods and guidelines) (Niehoff and Moorman, 1993); interpersonal justice climate refers to the treatment accorded to team members when decisions are made. It marks the extent to which team members are treated with respect and dignity during the decision-making process (Li and Cropanzano, 2008). The link between justice climate and team OCB can be perceived through social exchange theory (Blau, 1964) and through the norm of reciprocity (Gouldner, 1960). These two models suggest that team members seek to balance their inputs and outcomes in relation to others (Elstad et al., 2012). Generally, with high levels of justice climate, team members share the perception that their organization and/or its representatives can be trusted to protect their interests; this engenders an obligation to reciprocate in positive, beneficial ways, namely OCBs (Lambert, 2000). More specifically, when the collective experience of the team is that they are rewarded well (distributive justice), the team as the whole is treated with respect and dignity (interpersonal justice), and the process leading to outcomes is fairly applied (procedural justice); team members will seek to reciprocate by investing in citizenship behaviors (Parzefall and Salin, 2010). OCB is especially suitable for reciprocity because whereas traditional in-role behaviors can be enforced on the basis of formal role obligations (e.g. Konovsky and Pugh, 1994), OCB consists of informal contributions that team members can choose to exhibit or withhold (Somech and Oplatka, 2014). For example, Ehrhart (2004), who found a positive association between procedural justice climate and team OCB, suggested: "If group members perceive they are treated fairly and this perspective is shared among them [...] their social interactions will also result in similar responses to that assigned meaning, such as the performance of higher levels of OCB" (p. 67). Accordingly, we propose a positive link between justice climate and team OCB: H1. A positive relationship will be found between team justice climate and team OCB. Team psychological capital and team OCB Team psychological capital (team PsyCap) is a team's positive psychological state of development (Luthans et al., 2008), and is characterized by four psychological components: hope, optimism, collective efficacy and resilience (Avey et al., 2009; Heled et al., 2015). Hope is defined as a positive motivational state of the team that is based on a feeling that is fueled by successful agency (goal-oriented energy) and pathways (planning to meet goals) (Snyder et al., 1996). Optimism is the team's ability to create a positive attribution with respect to success in both the present and the future. Collective efficacy means that "a group shares a common belief in regards to the standardization of the capacities of the group members" (Bandura, 1997, p. 477). And resilience is a positive psychological ability of the team "to respond productively to significant change that disrupts the expected pattern of events without engaging in an extended period of regressive behavior" (Horne and Orr, 1997, p. 31). Although each of these four components has the potential to influence teamwork, studies support the notion that together, they have a much greater impact than each variable individually (Luthans et al., 2007). Team PsyCap is a construct of shared psychological state of team members, and can be considered an emergent state (Heled et al., 2015). According to Marks et al. (2001), emergent states are "constructs that characterize properties of the team that are typically dynamic in nature and vary as a function of team context, inputs, processes, and outcomes. Emergent states describe cognitive, motivational, and affective states of teams [...]" (p. 357). Put simply, team PsyCap is a team resource that can be developed and managed, and may thereby influence team processes and outcomes (Mathe-Soulek et al., 2014). The present study posits a positive link between team PsyCap and team OCB. Generally, teams with high levels of PsyCap are confident that good things happen at work; they believe they create their success and can overcome failure (Avey et al., 2011). Accordingly, we argue that team PsyCap plays an essential role in promoting positive attitudes and behaviors among team members, as expressed by team OCBs. The affective approach, which derives from the social-psychological literature (e.g. Williams and Shiaw, 1999), may provide the theoretical foundation for the relationship (George, 1991). This approach suggests that when people are in a positive mood, and hold positive attitudes; they are more likely to be helpful, namely to exhibit high levels of citizenship behaviors. Put simply, team members who experience a good mood, a sense of optimism and hope tend to perceive other people and situations in a more positive light (George, 1996), which may predispose individuals to exhibit behaviors that benefit team members, enhance organizational outcomes, or both (Heled et al., 2015). In other words, teams characterized by optimism, hope and resilience will tend to demonstrate pro-social behaviors that preserve their positive emotional state (Srivastava et al., 2006). Furthermore, this approach suggests that teams in a good mood are more helpful because being so enables them to maintain their positive mood (George, 1991). For example, Heled et al. (2015), who examined 82 school management teams, found positive relations between team PsyCap and the individual's job satisfaction and team OCB. Accordingly, we propose a positive link between Team PsyCap and team OCB: H2. A positive relation will be found between team PsyCap and team OCB. The relation between team OCB and team innovation Educational innovation is perceived as a key factor for promoting school effectiveness (Gilad-Hai and Somech, 2016). Innovation enables schools to cope with a complex and changing environment (Chen, 2010; Tubin, 2009). Team innovation is defined as "the intentional introduction and application within a team, of ideas, processes, products or procedures new to the team, designed to significantly benefit the individual, the team, the organization, or wider society" (West and Wallace, 1991, p. 303). Similarly, educational innovation is defined as "an idea or practice that reflects an emergent trend and offers an optimistic alternative for a current situation of dissatisfaction" (Tubin, 2009, p. 406). Teams may offer an alternative approach to learning, teaching and/or administrative routines (Gilad-Hai and Somech, 2016). Overall, innovation undergoes two phases: the creativity phase, namely generation or production of ideas that are both novel and useful (George, 2007), and the implementation phase, namely "the process of gaining targeted employees' appropriate and committed use of an innovation" (Klein and Sorra, 1996, p. 1055). According to the present model, team OCB can be linked to team innovation. There are logical reasons to expect a positive relation between team OCB and team innovation. According to Organ (1997), OCBs support the social and psychological environment within which the technical core functions. Accordingly, teachers who are willing to collaborate with their team members, to actively participate in meetings, to share knowledge, to explore new alternatives, or to learn new skills, contribute to the team's ability to be innovative (Jimmieson et al., 2010). Moreover, at the stage of innovation implementation, when job definitions are ambiguous, schools depend more on teams that are willing to contribute to successful change, regardless of formal job requirements (Oplatka, 2006). Thus, team OCB is assumed to help schools adopt changes and innovations that are necessary in the consistently changing environment (Savelyeva and Lee, 2012): H3. A positive relationship will be found between team OCB and team innovation. The mediating role of team psychological capital Finally, we propose that team OCB will serve as a mediator between the antecedents of team justice climate and team PsyCap and the outcomes: team innovation (Figure 1). This argument can be supported by the classic team effectiveness models of input-process-outcomes (Hackman and Morris, 1975) wherein team effectiveness involves a three-stage process. In the first stage, team and job characteristics (inputs) that structure teamwork influence the work processes in stage 2 (processes), which in turn lead to positive results in stage 3 (outcomes). Accordingly, we suggest that justice climate and team PsyCap set the stage for encouraging team members to invest in the team above a beyond the call of duty, namely to exhibit high levels of citizenship behaviors, which in turn enhance team innovation: H4a. Team OCB will mediate the relationship between justice climate and team innovation. H4b. Team OCB will mediate the relationship between team PsyCap and team innovation. Sample and procedure This study covered 78 disciplinary teams sampled from 13 junior-high schools (grades 7-9), located in the Northern and Central parts of Israel. In each school, we sampled six disciplinary teams (linguistics, civics, history, English, mathematics, science), totaling 214 team members, 78 coordinators and 13 school principals. We conducted pre-assessment interviews with the principal of each school, and learned that all team members interacted regularly to achieve shared goals regarding teaching issues and other school matters. They also depended on one another for knowledge and effort by means of several permanent structures such as scheduled staff meetings, and joint refresher workshops. Team size ranged from three to six members, with an average of 3.26 (SD=1.73). Of the team members, 181 were women (84.5 percent) and 33 were men. Average age was 45.66 years (SD=7.85), with average job tenure 6.73 (SD=4.42). In education, 80 percent of participants had a bachelor's degree, 20 percent had a master's degree. Of the coordinators, 56 were women (71.8 percent), average age 45 years (SD=7.35). They had served an average of 5.42 years (SD=4.39) in their coordinating role. In education, 62 percent had a Bachelor's degree and 38 percent a Master's degree. Of the school principals, 69 percent were women. Their average age was 48.82 years (SD=6.53), and they had an average of 10.42 years' experience directing their current school (SD=9.57). In education, 69 percent had a Bachelor's degree, and 31 percent had a master's degree. Data were obtained through a survey, and gathered from three different sources to avoid one-source bias. Team members were asked to fill out the scales of team justice climate and team psychological capital; team coordinators were asked to fill out the scale of team OCB; and school principals evaluated team innovation. All participants completed a demographic questionnaire. Data were collected on site at each school and all respondents were assured anonymity of responses. Measures All scales were measured on a five-point Likert scale from 1 (strongly disagree) to 5 (strongly agree). Team justice climate Team justice climate was measured by the 14-item scale of Li, Cropanzano and Benson (2007). The scale consists of three dimensions: first, team distributive justice climate (five items: e.g. "Some of my teammates have received more praise for the team projects than they deserved" (r)), a=0.71. Second, team procedural justice climate (five items: e.g. "The way my teammates make decisions is free from personal bias"), a=0.75. Third, team interpersonal justice climate (four items: e.g. "My teammates treat each other with respect"), a=0.34. Due to the low internal reliability of the interpersonal justice sub-scale (a=0.34), the model's analyses covered only the two sub-scales of team justice climate: distributive justice and procedural justice. Team PsyCap Team psychological capital was measured by the collective PsyCap scale developed by Mathe-Soulek et al. (2014), and adapted to the school context by Heled et al. (2015). The scale has eight items that measure the four dimensions of PsyCap: self-efficacy (two items: e.g. "Members of this team confidently contribute to discussions about the team's strategy"); hope (two items: e.g. "Members of this team think of many ways to attain work goals"); resilience (two items: e.g. "Members of this team usually take stressful things at work in their stride"); and optimism (two items: e.g. "Members of this team are optimistic about what will happen to them in the future as it pertains to work"). The questionnaire's reliability was 0.78. Team OCB Team OCB was measured on Vigoda-Gadot et al.'s (2007) scale, which was developed in the educational context. The scale has 18 items that measure two dimensions: OCBI, namely OCB directed at and contributing to a certain person at school (nine items: e.g. "Members of this team help teachers who have heavy workloads") (a=0.72), and OCBO, namely OCB directed to the school as a whole (nine items: e.g. "Our team makes innovative suggestions to improve school life") (a=0.81). The questionnaire's reliability was 0.80. Team innovation Team innovation was measured on a four-item scale adapted from West and Wallace (1991). The items reflected the extent to which in the previous six months the team had initiated changes in each of four job areas: work objectives, working methods, teaching methods and development of skills (e.g. "The team initiated new teaching methods) (a=0.93). Control variable Team size was chosen as a control variable. Previous studies showed that team size affects the team's process and its outcomes (Ancona and Caldwell, 1992; Heled et al., 2015; Keller, 2001). Level of analysis The unit of theory in the present study was the team. Accordingly, the hypotheses were posited at the team level, and the study variables (justice climate and team PsyCap) were aggregates of individual responses to the team level of analysis. The variables: team OCB, and team innovation were measured at the team level, and were evaluated by the team coordinator and the school principal, respectively. To justify aggregation of the variables team PsyCap and team OCB, it was critical to demonstrate high within-team agreement (rwg: James et al., 1993). A value of 0.70 or above is suggested as a "good" amount of within-group interrater agreement (James et al., 1993). All scales exceeded this criterion: average scores were 0.96 for team PsyCap, 0.90 for distributive justice and 0.90 for procedural justice. We also obtained the following ICC(1) and ICC(2) values: team PsyCap 0.42 and 0.71; distributive justice 0.64 and 0.83; procedural justice 0.45 and 0.70. As indicated by Bliese (2000), ICC(1) generally ranges from 0 to 0.50 with a median of 0.12. There are, however, no definite guidelines for determining acceptable values. All the above values were comparable to the median or to recommended ICC values so we concluded that aggregation was justified for these variables. Table I shows the means, standard deviations and correlations for the study variables. To test our model we used PROCESS, a computational tool that applies a bootstrapping technique (5,000 boots) to estimate the confidence interval (CI) of the mediation effect (indirect relationships) (Hayes, 2013). Bootstrapping is a non-parametric method based on re-sampling with replacement which is done many times, e.g. 5,000 times. From each of these samples, the indirect effect is computed and a sampling distribution can be empirically generated. Two mediation models were tested. The first examined the mediating role of team OCB (OCBI, OCBO) in the relation of justice climate (distributive justice and procedural justice) to team innovation. The second examined the mediating role of team OCB in the team PsyCap - team innovation relation. To analyze the two models we conducted hierarchical linear modeling analyses because teams (level 1) were nested within schools (level 2). The results are presented in Tables II and III. H1 proposed a positive relation between justice climate (distributive justice and procedural justice) and team OCB (OCBI, OCBO). A positive and significant relationship was indicated between procedural justice and team OCBI (b=0.78, t=2.97, p<0.01) and between procedural justice and team OCBO (b=0.79, t=2.73, p<0.01) (see Table II). However, no significant relation was found between distributive justice and OCBI and OCBO (p>0.05). H2 proposed a positive relationship between team PsyCap and team OCB (OCBI, OCBO): a positive and significant relation was indicated between PsyCap and team OCBI (b=0.55, t=2.69, p<0.01) and between PsyCap and team OCBO (b=0.99, t=3.21, p<0.01) (see Table II). Accordingly, H1 and H2 were supported. H3 proposed a positive relation between OCB (OCBI, OCBO) and team innovation. The results supported the hypothesis, showing a positive and significant relation between team OCBI and team innovation (b=0.25, t=5.95, p<0.001) and between team OCBI and team innovation (b=0.39, t=8.62, p<0.01) (see Table II). H4a proposed that team OCB (OCBI, OCBO) mediated the relation between justice climate and team innovation. The mediation analyses, based on bootstrapping, supported the hypothesis for the mediating role of OCBI (ab=0.21, CI 95% (0.04, 0.47)) and of OCBO (ab=0.34, CI 95% (0.13, 0.61)) in the relation of procedural justice to team innovation (see Table III). Finally, H4b proposed that team OCB (OCBI, OCBO) mediated the relation between team PsyCap and team innovation. The results supported the hypothesis for OCBI (ab=0.13, CI 95% (0.03, 0.45)), and for OCBO (ab=0.31, CI 95% (0.12, 0.60)). To summarize, regarding the role of justice climate's two dimensions - distributive justice and procedural justice - in promoting OCB, and in turn team innovation, only the latter was positively and significantly associated with team OCBO and OCBI. Furthermore, team OCBI and team OCBO alike fully mediated the relationship between procedural justice and team innovation. Regarding the role of team PsyCap in advancing team OCB and team innovation, the results showed positively and significantly associations between team PsyCap and team OCBO and OCBI. Also, team OCBI as well as team OCBO fully mediated the relationship between team PsyCap and team innovation. The present study explored team OCB from a context perspective. This course is important because to date most studies on OCB in schools have focused on teacher OCB as an individual feature, without regard for its contextual nature (Somech and Oplatka, 2014). Put simply, although OCB is performed by individuals, willingness to engage in OCB is not generated in a vacuum, and the team context most likely serves to encourage or discourage people in respect of exerting these extra-role behaviors (Vigoda-Gadot et al., 2007). In this regard, our findings highlighted the importance of team-level antecedents and outcomes for better understanding how between team differences in OCB are developed and what their consequences are. The results contribute to the educational administration literature in several respects. In general, the results support the premise that citizenship behaviors grow in a context. The OCB construct is "social" in nature, so the extent teams exhibit citizenship behaviors depends chiefly on teams' characteristics and context (Jackson, 2009). Specifically, the two contextual variables of team justice climate and team PsyCap proved positively associated with team OCB. This is important because it emphasizes the idea that leaders exercise crucial influence on teachers' tendency to contribute above and beyond their formal obligations (Somech and Oplatka, 2014). Leaders can create a team environment that determines when, and why, team members do or do not engage in OCBs (e.g. Ehrhart, 2004). Regarding the link between justice climate and OCB, the relationships between procedural justice and OCBI and OCBO were found positive. This suggests that when team members feel they are treated fairly they tend to exhibit higher levels of helping behaviors toward their colleagues (OCBI), as well as investing extra effort in promoting team and organization goals (OCBO). These findings may further support the exchange approach (Blau, 1964), as the cognitive mechanism sustaining these relationships. Team members who experience the decision-making processes and procedures in their team as fair will want to reciprocate by engaging in citizenship behaviors that may contribute the team and the school as a whole. However, unlike the procedural justice climate, we did not find significant relationships between distributive justice and the two dimensions of team OCB (OCBI and OCBO). These contrasting results accord with the pattern of relations found in previous studies on the links between perceived organizational justice and OCB at the individual level. For example, Moorman (1991) found a positive relation between perceptions of procedural justice and four out of five dimensions of OCB. Conversely, perceptions of distributive justice failed to influence any dimension of citizenship behavior. One possible explanation is that the distributive justice component refers not to social norms and procedures but to allocation of rewards and resources (Masterson et al., 2000). It does not possess the "social aspect" that team procedural justice climate does (Ambrose et al., 2002). Accordingly, distributive justice is less vital for establishing social exchange relationships, making it especially relevant in the context of parties that do not share social bonds. Thus, it does not contribute to the social exchange outcomes, as the procedural justice climate does (Ben-Sasson and Somech, 2015). Regarding the antecedent of team PsyCap, the results indicated, as expected, positive associations between team PsyCap and team OCBI and OCBO. In team PsyCap, high levels of optimism, hope, collective efficacy and resilience are positive psychological resources available to team members to share. They enable them to develop their team's strengths, as well as positive coping methods in daily life. All these create a positive atmosphere in the team and promote positive interactions and dynamics among its members, enhancing their willingness to exhibit citizenship behaviors to advance team and organizational effectiveness (Heled et al., 2015). This finding may provide additional support for the position that positive emotions reinforce individuals' and groups' tendency to engage in pro-social behaviors (George, 1996). Teams that perceive themselves as competent, in an upbeat mood, seek to maintain their positive emotional state by exhibiting positive social behaviors such as OCBs, which contribute to their social and psychological environment (Srivastava et al., 2006). This perspective may also suggest that the relationship between team PsyCap and team OCB is spiral: that is, teams high in PsyCap tend to exhibit high OCB levels, which in turn may further strengthen their team PsyCap levels. Luthans et al. (2011) proposed: "[...] overall PsyCap and its facets can be developed through intentional activities [...]. These intentional activities can increase levels of positivity and well-being in general (Lyubomirsky, 2007) [...]. In other words, positivity can beget more positivity" (p. 340). Furthermore, with respect to outcomes at the team level, the study findings support the H3 of a positive relation between team OCB (OCBI and OCBO) and team innovation. These findings promote the idea that teams that exhibit behaviors of helping, collaboration and support (OCBI), as well as exerting extra effort for the team and the school as a whole (OCBO), are more willing to seek challenges, to exhibit "out-of-the-box" thinking and exploration, namely to be innovative (Oldham and Cummings, 1996). This positive link between team OCB and team innovation may also be supported by Fredrickson's (1998, 2009) broaden-and-build theory, holding that a positive environment broadens thought-action tendencies. According to the theory, positivity broadens psychological, cognitive and physical resources that may motivate team members to exhibit high levels of innovation. The present results also support the mediating role of team OCB in the relationship between the two antecedents of procedural justice climate and team PsyCap and team innovation. Specifically, the findings demonstrated that procedural justice climate was positively related to team OCB, which in turn was positively related to team innovation. The results also demonstrated a positive link between team PsyCap and team OCB, which in turn was positively related to team innovation. These findings support the classic model of input-process-outcomes (Hackman and Morris, 1975), which emphasizes that team OCB serves as a vehicle whereby the contextual variables of team procedural justice climate and team PsyCap enhance team innovation. Accordingly, the present results demonstrate that team procedural justice climate and team PsyCap "set the stage" for encouraging the team to invest above and beyond the call of duty, namely to engage in citizenship behaviors, which create a positive and supporting psychological and social environment, in order to maximize their innovation (Yi, 2005). Beyond this study's theoretical contribution, its findings also have practical implications. By taking a team perspective, the findings offer evidence that despite the usual approach treating OCB as a personal or individual phenomenon; it may be regarded as shared norms of behaviors at the team level. In other words, interactions among team members over time, in a certain context, determine the extent team members exhibit citizenship behaviors, which are important for promoting team effectiveness (Podsakoff et al., 2009). This contextual approach may urge principals and other educational leaders to realize that teachers' willingness to invest extra effort in school is mainly the result of an appropriate team context, which can be shaped and developed. Specifically, by creating an atmosphere of procedural justice and by enhancing a sense of team's competence, resilience, optimism and hope (team PsyCap), leaders may motivate educational team members to exhibit high levels of OCBs, which in turn may improve those outcomes that promote the success of the individual, the team, and the entire school. Finally, as in any research, this study has some limitations as well. First, data were gathered using questionnaires so certain potential biases may be present, such as one-source bias or subjectivity of the responses. Nevertheless, recall that in this study data were gathered from three different sources (team members, team coordinators, principals) to avoid one-source bias (Avolio et al., 1991). It is, however, recommended that in future studies researchers employ more objective measures to further validate the present results. Second, the nature of the data collecting limits the extent our logical explanations can be conclusively supported by the data. The causal direction may be the reverse, and teams that are more innovative may exhibit higher levels of OCBs; alternately, engaging in citizenship behaviors may affect the levels of justice climate and/or team PsyCap. Future longitudinal research should be conducted before causal explanations. Finally, the present research focused on two variables as antecedents: team justice climate and team PsyCap. We recommend examining team OCB with respect to additional context variables, such as team composition as promoting or inhibiting team OCB. Similarly, although in the present study, we focused on a crucial team-level outcome of innovation, we encourage researchers to expand future research and examine additional variables as outcomes at the individual, team and school levels.
|
The PROCESS test confirmed the mediating role of team OCB, showing positive relations between team procedural justice and team PsyCap and team OCB, and also between team OCB and team innovation.
|
[SECTION: Value] Today schools operate in dynamic environment, each struggling to gain a competitive edge (Orr and Orphanos, 2011). This environment reinforces the understanding that schools should strive to employ teachers who are willing to go the extra mile, namely to engage in organizational citizenship behavior (OCB). OCBs are behaviors and actions that go beyond the formal role and what is laid down in the teacher's job description, and contribute to the achievement of the school's objectives (Oplatka, 2006). OCB in schools is commonly taken to be an individual phenomenon, namely a personality trait or a social response to the behavior or attitude of superiors or to other motivation-based mechanisms in the workplace (Zeinabadi, 2010). This individual approach seems wanting: teachers perform or fail to perform OCBs not in a vacuum but in an organizational context, which most probably serves to encourage or discourage them (Somech and Drach-Zahavy, 2004). Only in the last two decades have some scholars studied OCB as a team or organizational-level concept (e.g. Ehrhart, 2004). Put otherwise, the individual approach probes the possibility that teacher OCB can be better understood as a team or an organizational feature that thrives in a context. This development is important because the aggregate OCB level, not sporadic actions by some individuals, influences organizational effectiveness (Organ, 1988). Moreover, the several studies that have captured OCB as a collective phenomenon examined either its antecedents (e.g. Hu and Liden, 2011) or its consequences (e.g. Yen et al., 2008). Very few researchers, if any, have explored the mediating role of team OCB in relation to specific antecedents and outcomes. The present research seeks to address this deficiency by examining OCB as a team phenomenon. The first premise of the research model is that it is important to measure OCB as a collective structure. Next, the mediating role of team OCB is investigated. Specifically, the research model posits that the contextual variable of a team's justice climate, and the team's collective psychological state (psychological capital), will be positively related to team OCB. Furthermore, team OCB will be positively related to the outcome of team innovation. The model also suggests that team OCB will mediate the relation between the antecedents and the outcome. Teachers' OCB is defined as "[...] teachers voluntarily going out of their way to help their students, colleagues, and others as they engage in the work of teaching and learning" (DiPaola and Hoy, 2005, p. 390). OCBs are essential because formal in-role job descriptions cannot cover the entire array of behaviors needed for achieving school goals (Zeinabadi, 2010). They operate indirectly: they influence schools' social and psychological environment, enhance school effectiveness by freeing up resources for more productive purposes, help coordinate activities within the organization, and enable teachers to adapt more effectively to environmental changes (Sesen and Basim, 2012). Teachers manifest OCB, for instance, by staying after school hours to help a student with learning materials, aid a colleague with a heavy workload, volunteer for unpaid tasks, or make innovative suggestions to improve the school (Somech and Oplatka, 2014). Scholars have identified two broad dimensions of OCB (Sesen and Basim, 2012; Williams and Anderson, 1991). The first is OCBI (OCBs toward individuals in school), namely behaviors that immediately benefit particular individuals, such as colleagues and/or students, hence indirectly contribute to the organization. This dimension includes such behaviors as taking a personal interest in a novice teacher or helping a student after working hours. The second dimension is OCBO (OCB toward the organization), namely behaviors that benefit the school as a whole, for example, organizing special social activities for the school or making innovative suggestions to improve it. According to Podsakoff et al. (2009), these two broad categories capture most previous taxonomies effectively. Although OCBs are acts by individuals, our main argument is that these extra-role behaviors grow in a context, therefore can be extended to the team level. Put simply, because OCBs are interactive or "social" in nature, teams differ in their teachers' tendency to engage in them (Ehrhart, 2004). Social psychology research has shown that teams are powerful instruments of social influence and significantly affect individuals' perceptions, attitudes and behaviors (Mathieu et al., 2000). Two theoretical models may provide the rationale for considering OCB a team phenomenon: the social learning theory and the social information processing theory. The first posits that people learn by observing the behavior of others (Bandura, 1986). The more OCB is modeled by the members of one's team, the more likely one is to behave in a manner consistent with those models, in particular when that behavior is associated with positive social consequences (Ehrhart, 2004). The second the social information processing theory (Salancik and Pfeffer, 1978) underlines the importance of social cues in shaping attitudes. When OCB levels are high among team members, an individual member will be more likely to view such behavior as acceptable and expected. Ultimately this recognition results in higher levels of individual OCB. George (1991), for example, found high similarities among team members in the level of positive and/or negative mood, as well as in the extent they exhibited citizenship behaviors. Conceptualizing OCB as a team phenomenon implies that teachers who do or do not perform OCB do not do so in a vacuum; the team context in which these behaviors are performed undoubtedly serves to encourage or discourage them (Vigoda-Gadot et al., 2007). In other words, this perspective suggests that contextual variables determine when, and sometimes why, teachers do or do not engage in OCBs (Ehrhart, 2004). In this vein, the present model focused on two team-level variables as antecedents of team OCB: justice climate and team psychological capital, and team innovation as a consequence. Furthermore, the model suggested that team OCB serves as a mediator in the relation of justice climate to team psychological capital and team innovation. The relation of justice climate to team psychological capital and team OCB We focused on two elements of team characteristics: team justice climate and team psychological capital. Our choice of these two variables was grounded in recent OCB research emphasizing the critical role of positive team atmosphere in enhancing positive attitudes and behaviors, as expressed by team citizenship behaviors (Lau and Lam, 2008). Team justice climate and team OCB Team climate derives from the array of the work-contextual properties that the team emphasizes and evaluates; these are perceived as critical in shaping team members' attitudes and behaviors (Schneider et al., 2013). The main claim of climate research is the idea of shared perceptions constructed through social interactions among individual employees engaged in teamwork (Spell and Arnold, 2007). Although there are several kinds of climate in every workplace, justice climate is crucial for promoting a positive atmosphere prevailing over team members, as well as between them and their supervisor and the organization as a whole (Colquitt et al., 2002). Building on the perception of a "social" creation of justice perceptions, justice climate is defined as "'shared perceptions' of team fairness in outcomes, treatment, and supervisory interaction" (Moliner et al., 2005, p. 101). Scholars have identified three distinct dimensions of justice climate: distributive justice, procedural justice and interpersonal justice (Li and Cropanzano, 2008). Distributive justice climate addresses the organizational reward system and reflects fairness of outcomes, namely fairness in distribution of tangible or intangible rewards and/or resources (Li and Cropanzano, 2008); procedural justice climate is conceptualized as fairness of procedures and of the decision-making process (i.e. the degree to which decisions on reward allocation are made according to fair methods and guidelines) (Niehoff and Moorman, 1993); interpersonal justice climate refers to the treatment accorded to team members when decisions are made. It marks the extent to which team members are treated with respect and dignity during the decision-making process (Li and Cropanzano, 2008). The link between justice climate and team OCB can be perceived through social exchange theory (Blau, 1964) and through the norm of reciprocity (Gouldner, 1960). These two models suggest that team members seek to balance their inputs and outcomes in relation to others (Elstad et al., 2012). Generally, with high levels of justice climate, team members share the perception that their organization and/or its representatives can be trusted to protect their interests; this engenders an obligation to reciprocate in positive, beneficial ways, namely OCBs (Lambert, 2000). More specifically, when the collective experience of the team is that they are rewarded well (distributive justice), the team as the whole is treated with respect and dignity (interpersonal justice), and the process leading to outcomes is fairly applied (procedural justice); team members will seek to reciprocate by investing in citizenship behaviors (Parzefall and Salin, 2010). OCB is especially suitable for reciprocity because whereas traditional in-role behaviors can be enforced on the basis of formal role obligations (e.g. Konovsky and Pugh, 1994), OCB consists of informal contributions that team members can choose to exhibit or withhold (Somech and Oplatka, 2014). For example, Ehrhart (2004), who found a positive association between procedural justice climate and team OCB, suggested: "If group members perceive they are treated fairly and this perspective is shared among them [...] their social interactions will also result in similar responses to that assigned meaning, such as the performance of higher levels of OCB" (p. 67). Accordingly, we propose a positive link between justice climate and team OCB: H1. A positive relationship will be found between team justice climate and team OCB. Team psychological capital and team OCB Team psychological capital (team PsyCap) is a team's positive psychological state of development (Luthans et al., 2008), and is characterized by four psychological components: hope, optimism, collective efficacy and resilience (Avey et al., 2009; Heled et al., 2015). Hope is defined as a positive motivational state of the team that is based on a feeling that is fueled by successful agency (goal-oriented energy) and pathways (planning to meet goals) (Snyder et al., 1996). Optimism is the team's ability to create a positive attribution with respect to success in both the present and the future. Collective efficacy means that "a group shares a common belief in regards to the standardization of the capacities of the group members" (Bandura, 1997, p. 477). And resilience is a positive psychological ability of the team "to respond productively to significant change that disrupts the expected pattern of events without engaging in an extended period of regressive behavior" (Horne and Orr, 1997, p. 31). Although each of these four components has the potential to influence teamwork, studies support the notion that together, they have a much greater impact than each variable individually (Luthans et al., 2007). Team PsyCap is a construct of shared psychological state of team members, and can be considered an emergent state (Heled et al., 2015). According to Marks et al. (2001), emergent states are "constructs that characterize properties of the team that are typically dynamic in nature and vary as a function of team context, inputs, processes, and outcomes. Emergent states describe cognitive, motivational, and affective states of teams [...]" (p. 357). Put simply, team PsyCap is a team resource that can be developed and managed, and may thereby influence team processes and outcomes (Mathe-Soulek et al., 2014). The present study posits a positive link between team PsyCap and team OCB. Generally, teams with high levels of PsyCap are confident that good things happen at work; they believe they create their success and can overcome failure (Avey et al., 2011). Accordingly, we argue that team PsyCap plays an essential role in promoting positive attitudes and behaviors among team members, as expressed by team OCBs. The affective approach, which derives from the social-psychological literature (e.g. Williams and Shiaw, 1999), may provide the theoretical foundation for the relationship (George, 1991). This approach suggests that when people are in a positive mood, and hold positive attitudes; they are more likely to be helpful, namely to exhibit high levels of citizenship behaviors. Put simply, team members who experience a good mood, a sense of optimism and hope tend to perceive other people and situations in a more positive light (George, 1996), which may predispose individuals to exhibit behaviors that benefit team members, enhance organizational outcomes, or both (Heled et al., 2015). In other words, teams characterized by optimism, hope and resilience will tend to demonstrate pro-social behaviors that preserve their positive emotional state (Srivastava et al., 2006). Furthermore, this approach suggests that teams in a good mood are more helpful because being so enables them to maintain their positive mood (George, 1991). For example, Heled et al. (2015), who examined 82 school management teams, found positive relations between team PsyCap and the individual's job satisfaction and team OCB. Accordingly, we propose a positive link between Team PsyCap and team OCB: H2. A positive relation will be found between team PsyCap and team OCB. The relation between team OCB and team innovation Educational innovation is perceived as a key factor for promoting school effectiveness (Gilad-Hai and Somech, 2016). Innovation enables schools to cope with a complex and changing environment (Chen, 2010; Tubin, 2009). Team innovation is defined as "the intentional introduction and application within a team, of ideas, processes, products or procedures new to the team, designed to significantly benefit the individual, the team, the organization, or wider society" (West and Wallace, 1991, p. 303). Similarly, educational innovation is defined as "an idea or practice that reflects an emergent trend and offers an optimistic alternative for a current situation of dissatisfaction" (Tubin, 2009, p. 406). Teams may offer an alternative approach to learning, teaching and/or administrative routines (Gilad-Hai and Somech, 2016). Overall, innovation undergoes two phases: the creativity phase, namely generation or production of ideas that are both novel and useful (George, 2007), and the implementation phase, namely "the process of gaining targeted employees' appropriate and committed use of an innovation" (Klein and Sorra, 1996, p. 1055). According to the present model, team OCB can be linked to team innovation. There are logical reasons to expect a positive relation between team OCB and team innovation. According to Organ (1997), OCBs support the social and psychological environment within which the technical core functions. Accordingly, teachers who are willing to collaborate with their team members, to actively participate in meetings, to share knowledge, to explore new alternatives, or to learn new skills, contribute to the team's ability to be innovative (Jimmieson et al., 2010). Moreover, at the stage of innovation implementation, when job definitions are ambiguous, schools depend more on teams that are willing to contribute to successful change, regardless of formal job requirements (Oplatka, 2006). Thus, team OCB is assumed to help schools adopt changes and innovations that are necessary in the consistently changing environment (Savelyeva and Lee, 2012): H3. A positive relationship will be found between team OCB and team innovation. The mediating role of team psychological capital Finally, we propose that team OCB will serve as a mediator between the antecedents of team justice climate and team PsyCap and the outcomes: team innovation (Figure 1). This argument can be supported by the classic team effectiveness models of input-process-outcomes (Hackman and Morris, 1975) wherein team effectiveness involves a three-stage process. In the first stage, team and job characteristics (inputs) that structure teamwork influence the work processes in stage 2 (processes), which in turn lead to positive results in stage 3 (outcomes). Accordingly, we suggest that justice climate and team PsyCap set the stage for encouraging team members to invest in the team above a beyond the call of duty, namely to exhibit high levels of citizenship behaviors, which in turn enhance team innovation: H4a. Team OCB will mediate the relationship between justice climate and team innovation. H4b. Team OCB will mediate the relationship between team PsyCap and team innovation. Sample and procedure This study covered 78 disciplinary teams sampled from 13 junior-high schools (grades 7-9), located in the Northern and Central parts of Israel. In each school, we sampled six disciplinary teams (linguistics, civics, history, English, mathematics, science), totaling 214 team members, 78 coordinators and 13 school principals. We conducted pre-assessment interviews with the principal of each school, and learned that all team members interacted regularly to achieve shared goals regarding teaching issues and other school matters. They also depended on one another for knowledge and effort by means of several permanent structures such as scheduled staff meetings, and joint refresher workshops. Team size ranged from three to six members, with an average of 3.26 (SD=1.73). Of the team members, 181 were women (84.5 percent) and 33 were men. Average age was 45.66 years (SD=7.85), with average job tenure 6.73 (SD=4.42). In education, 80 percent of participants had a bachelor's degree, 20 percent had a master's degree. Of the coordinators, 56 were women (71.8 percent), average age 45 years (SD=7.35). They had served an average of 5.42 years (SD=4.39) in their coordinating role. In education, 62 percent had a Bachelor's degree and 38 percent a Master's degree. Of the school principals, 69 percent were women. Their average age was 48.82 years (SD=6.53), and they had an average of 10.42 years' experience directing their current school (SD=9.57). In education, 69 percent had a Bachelor's degree, and 31 percent had a master's degree. Data were obtained through a survey, and gathered from three different sources to avoid one-source bias. Team members were asked to fill out the scales of team justice climate and team psychological capital; team coordinators were asked to fill out the scale of team OCB; and school principals evaluated team innovation. All participants completed a demographic questionnaire. Data were collected on site at each school and all respondents were assured anonymity of responses. Measures All scales were measured on a five-point Likert scale from 1 (strongly disagree) to 5 (strongly agree). Team justice climate Team justice climate was measured by the 14-item scale of Li, Cropanzano and Benson (2007). The scale consists of three dimensions: first, team distributive justice climate (five items: e.g. "Some of my teammates have received more praise for the team projects than they deserved" (r)), a=0.71. Second, team procedural justice climate (five items: e.g. "The way my teammates make decisions is free from personal bias"), a=0.75. Third, team interpersonal justice climate (four items: e.g. "My teammates treat each other with respect"), a=0.34. Due to the low internal reliability of the interpersonal justice sub-scale (a=0.34), the model's analyses covered only the two sub-scales of team justice climate: distributive justice and procedural justice. Team PsyCap Team psychological capital was measured by the collective PsyCap scale developed by Mathe-Soulek et al. (2014), and adapted to the school context by Heled et al. (2015). The scale has eight items that measure the four dimensions of PsyCap: self-efficacy (two items: e.g. "Members of this team confidently contribute to discussions about the team's strategy"); hope (two items: e.g. "Members of this team think of many ways to attain work goals"); resilience (two items: e.g. "Members of this team usually take stressful things at work in their stride"); and optimism (two items: e.g. "Members of this team are optimistic about what will happen to them in the future as it pertains to work"). The questionnaire's reliability was 0.78. Team OCB Team OCB was measured on Vigoda-Gadot et al.'s (2007) scale, which was developed in the educational context. The scale has 18 items that measure two dimensions: OCBI, namely OCB directed at and contributing to a certain person at school (nine items: e.g. "Members of this team help teachers who have heavy workloads") (a=0.72), and OCBO, namely OCB directed to the school as a whole (nine items: e.g. "Our team makes innovative suggestions to improve school life") (a=0.81). The questionnaire's reliability was 0.80. Team innovation Team innovation was measured on a four-item scale adapted from West and Wallace (1991). The items reflected the extent to which in the previous six months the team had initiated changes in each of four job areas: work objectives, working methods, teaching methods and development of skills (e.g. "The team initiated new teaching methods) (a=0.93). Control variable Team size was chosen as a control variable. Previous studies showed that team size affects the team's process and its outcomes (Ancona and Caldwell, 1992; Heled et al., 2015; Keller, 2001). Level of analysis The unit of theory in the present study was the team. Accordingly, the hypotheses were posited at the team level, and the study variables (justice climate and team PsyCap) were aggregates of individual responses to the team level of analysis. The variables: team OCB, and team innovation were measured at the team level, and were evaluated by the team coordinator and the school principal, respectively. To justify aggregation of the variables team PsyCap and team OCB, it was critical to demonstrate high within-team agreement (rwg: James et al., 1993). A value of 0.70 or above is suggested as a "good" amount of within-group interrater agreement (James et al., 1993). All scales exceeded this criterion: average scores were 0.96 for team PsyCap, 0.90 for distributive justice and 0.90 for procedural justice. We also obtained the following ICC(1) and ICC(2) values: team PsyCap 0.42 and 0.71; distributive justice 0.64 and 0.83; procedural justice 0.45 and 0.70. As indicated by Bliese (2000), ICC(1) generally ranges from 0 to 0.50 with a median of 0.12. There are, however, no definite guidelines for determining acceptable values. All the above values were comparable to the median or to recommended ICC values so we concluded that aggregation was justified for these variables. Table I shows the means, standard deviations and correlations for the study variables. To test our model we used PROCESS, a computational tool that applies a bootstrapping technique (5,000 boots) to estimate the confidence interval (CI) of the mediation effect (indirect relationships) (Hayes, 2013). Bootstrapping is a non-parametric method based on re-sampling with replacement which is done many times, e.g. 5,000 times. From each of these samples, the indirect effect is computed and a sampling distribution can be empirically generated. Two mediation models were tested. The first examined the mediating role of team OCB (OCBI, OCBO) in the relation of justice climate (distributive justice and procedural justice) to team innovation. The second examined the mediating role of team OCB in the team PsyCap - team innovation relation. To analyze the two models we conducted hierarchical linear modeling analyses because teams (level 1) were nested within schools (level 2). The results are presented in Tables II and III. H1 proposed a positive relation between justice climate (distributive justice and procedural justice) and team OCB (OCBI, OCBO). A positive and significant relationship was indicated between procedural justice and team OCBI (b=0.78, t=2.97, p<0.01) and between procedural justice and team OCBO (b=0.79, t=2.73, p<0.01) (see Table II). However, no significant relation was found between distributive justice and OCBI and OCBO (p>0.05). H2 proposed a positive relationship between team PsyCap and team OCB (OCBI, OCBO): a positive and significant relation was indicated between PsyCap and team OCBI (b=0.55, t=2.69, p<0.01) and between PsyCap and team OCBO (b=0.99, t=3.21, p<0.01) (see Table II). Accordingly, H1 and H2 were supported. H3 proposed a positive relation between OCB (OCBI, OCBO) and team innovation. The results supported the hypothesis, showing a positive and significant relation between team OCBI and team innovation (b=0.25, t=5.95, p<0.001) and between team OCBI and team innovation (b=0.39, t=8.62, p<0.01) (see Table II). H4a proposed that team OCB (OCBI, OCBO) mediated the relation between justice climate and team innovation. The mediation analyses, based on bootstrapping, supported the hypothesis for the mediating role of OCBI (ab=0.21, CI 95% (0.04, 0.47)) and of OCBO (ab=0.34, CI 95% (0.13, 0.61)) in the relation of procedural justice to team innovation (see Table III). Finally, H4b proposed that team OCB (OCBI, OCBO) mediated the relation between team PsyCap and team innovation. The results supported the hypothesis for OCBI (ab=0.13, CI 95% (0.03, 0.45)), and for OCBO (ab=0.31, CI 95% (0.12, 0.60)). To summarize, regarding the role of justice climate's two dimensions - distributive justice and procedural justice - in promoting OCB, and in turn team innovation, only the latter was positively and significantly associated with team OCBO and OCBI. Furthermore, team OCBI and team OCBO alike fully mediated the relationship between procedural justice and team innovation. Regarding the role of team PsyCap in advancing team OCB and team innovation, the results showed positively and significantly associations between team PsyCap and team OCBO and OCBI. Also, team OCBI as well as team OCBO fully mediated the relationship between team PsyCap and team innovation. The present study explored team OCB from a context perspective. This course is important because to date most studies on OCB in schools have focused on teacher OCB as an individual feature, without regard for its contextual nature (Somech and Oplatka, 2014). Put simply, although OCB is performed by individuals, willingness to engage in OCB is not generated in a vacuum, and the team context most likely serves to encourage or discourage people in respect of exerting these extra-role behaviors (Vigoda-Gadot et al., 2007). In this regard, our findings highlighted the importance of team-level antecedents and outcomes for better understanding how between team differences in OCB are developed and what their consequences are. The results contribute to the educational administration literature in several respects. In general, the results support the premise that citizenship behaviors grow in a context. The OCB construct is "social" in nature, so the extent teams exhibit citizenship behaviors depends chiefly on teams' characteristics and context (Jackson, 2009). Specifically, the two contextual variables of team justice climate and team PsyCap proved positively associated with team OCB. This is important because it emphasizes the idea that leaders exercise crucial influence on teachers' tendency to contribute above and beyond their formal obligations (Somech and Oplatka, 2014). Leaders can create a team environment that determines when, and why, team members do or do not engage in OCBs (e.g. Ehrhart, 2004). Regarding the link between justice climate and OCB, the relationships between procedural justice and OCBI and OCBO were found positive. This suggests that when team members feel they are treated fairly they tend to exhibit higher levels of helping behaviors toward their colleagues (OCBI), as well as investing extra effort in promoting team and organization goals (OCBO). These findings may further support the exchange approach (Blau, 1964), as the cognitive mechanism sustaining these relationships. Team members who experience the decision-making processes and procedures in their team as fair will want to reciprocate by engaging in citizenship behaviors that may contribute the team and the school as a whole. However, unlike the procedural justice climate, we did not find significant relationships between distributive justice and the two dimensions of team OCB (OCBI and OCBO). These contrasting results accord with the pattern of relations found in previous studies on the links between perceived organizational justice and OCB at the individual level. For example, Moorman (1991) found a positive relation between perceptions of procedural justice and four out of five dimensions of OCB. Conversely, perceptions of distributive justice failed to influence any dimension of citizenship behavior. One possible explanation is that the distributive justice component refers not to social norms and procedures but to allocation of rewards and resources (Masterson et al., 2000). It does not possess the "social aspect" that team procedural justice climate does (Ambrose et al., 2002). Accordingly, distributive justice is less vital for establishing social exchange relationships, making it especially relevant in the context of parties that do not share social bonds. Thus, it does not contribute to the social exchange outcomes, as the procedural justice climate does (Ben-Sasson and Somech, 2015). Regarding the antecedent of team PsyCap, the results indicated, as expected, positive associations between team PsyCap and team OCBI and OCBO. In team PsyCap, high levels of optimism, hope, collective efficacy and resilience are positive psychological resources available to team members to share. They enable them to develop their team's strengths, as well as positive coping methods in daily life. All these create a positive atmosphere in the team and promote positive interactions and dynamics among its members, enhancing their willingness to exhibit citizenship behaviors to advance team and organizational effectiveness (Heled et al., 2015). This finding may provide additional support for the position that positive emotions reinforce individuals' and groups' tendency to engage in pro-social behaviors (George, 1996). Teams that perceive themselves as competent, in an upbeat mood, seek to maintain their positive emotional state by exhibiting positive social behaviors such as OCBs, which contribute to their social and psychological environment (Srivastava et al., 2006). This perspective may also suggest that the relationship between team PsyCap and team OCB is spiral: that is, teams high in PsyCap tend to exhibit high OCB levels, which in turn may further strengthen their team PsyCap levels. Luthans et al. (2011) proposed: "[...] overall PsyCap and its facets can be developed through intentional activities [...]. These intentional activities can increase levels of positivity and well-being in general (Lyubomirsky, 2007) [...]. In other words, positivity can beget more positivity" (p. 340). Furthermore, with respect to outcomes at the team level, the study findings support the H3 of a positive relation between team OCB (OCBI and OCBO) and team innovation. These findings promote the idea that teams that exhibit behaviors of helping, collaboration and support (OCBI), as well as exerting extra effort for the team and the school as a whole (OCBO), are more willing to seek challenges, to exhibit "out-of-the-box" thinking and exploration, namely to be innovative (Oldham and Cummings, 1996). This positive link between team OCB and team innovation may also be supported by Fredrickson's (1998, 2009) broaden-and-build theory, holding that a positive environment broadens thought-action tendencies. According to the theory, positivity broadens psychological, cognitive and physical resources that may motivate team members to exhibit high levels of innovation. The present results also support the mediating role of team OCB in the relationship between the two antecedents of procedural justice climate and team PsyCap and team innovation. Specifically, the findings demonstrated that procedural justice climate was positively related to team OCB, which in turn was positively related to team innovation. The results also demonstrated a positive link between team PsyCap and team OCB, which in turn was positively related to team innovation. These findings support the classic model of input-process-outcomes (Hackman and Morris, 1975), which emphasizes that team OCB serves as a vehicle whereby the contextual variables of team procedural justice climate and team PsyCap enhance team innovation. Accordingly, the present results demonstrate that team procedural justice climate and team PsyCap "set the stage" for encouraging the team to invest above and beyond the call of duty, namely to engage in citizenship behaviors, which create a positive and supporting psychological and social environment, in order to maximize their innovation (Yi, 2005). Beyond this study's theoretical contribution, its findings also have practical implications. By taking a team perspective, the findings offer evidence that despite the usual approach treating OCB as a personal or individual phenomenon; it may be regarded as shared norms of behaviors at the team level. In other words, interactions among team members over time, in a certain context, determine the extent team members exhibit citizenship behaviors, which are important for promoting team effectiveness (Podsakoff et al., 2009). This contextual approach may urge principals and other educational leaders to realize that teachers' willingness to invest extra effort in school is mainly the result of an appropriate team context, which can be shaped and developed. Specifically, by creating an atmosphere of procedural justice and by enhancing a sense of team's competence, resilience, optimism and hope (team PsyCap), leaders may motivate educational team members to exhibit high levels of OCBs, which in turn may improve those outcomes that promote the success of the individual, the team, and the entire school. Finally, as in any research, this study has some limitations as well. First, data were gathered using questionnaires so certain potential biases may be present, such as one-source bias or subjectivity of the responses. Nevertheless, recall that in this study data were gathered from three different sources (team members, team coordinators, principals) to avoid one-source bias (Avolio et al., 1991). It is, however, recommended that in future studies researchers employ more objective measures to further validate the present results. Second, the nature of the data collecting limits the extent our logical explanations can be conclusively supported by the data. The causal direction may be the reverse, and teams that are more innovative may exhibit higher levels of OCBs; alternately, engaging in citizenship behaviors may affect the levels of justice climate and/or team PsyCap. Future longitudinal research should be conducted before causal explanations. Finally, the present research focused on two variables as antecedents: team justice climate and team PsyCap. We recommend examining team OCB with respect to additional context variables, such as team composition as promoting or inhibiting team OCB. Similarly, although in the present study, we focused on a crucial team-level outcome of innovation, we encourage researchers to expand future research and examine additional variables as outcomes at the individual, team and school levels.
|
The present study explored team OCB by from a context perspective. This is important because to date most studies on OCB in schools have focused on teacher OCB as an individual phenomenon, disregarding its contextual nature.
|
[SECTION: Purpose] Recent trends in the evolution of marketing have delivered return on investment (ROI)-driven brand managers to an important crossroads. Should they choose event marketing or consumer sampling? Can it ever be both? If so, under what circumstances? Just a few years ago in 2004, the leading brands spent close to a million dollars a year on event marketing. That volume of expenditure declined sharply in the subsequent two years, however, slipping 13 percent in 2005 and another 16 percent in 2006. Conversely, spending on product sampling rose 9 percent in 2006, according to the 2007 PROMO Industry Trends Report. The increase was fueled by that invaluable marketer's tool - measurability. As brand developers become more aware of the bang they are getting (or not getting) from their investment buck, spending patterns are bound to change. This paper is meant to help marketers understand the dynamics of the different types of sampling, and to identify the best marketing program investment opportunities. A decade ago, event marketing rarely was a major component of a brand's promotional plan. But event marketing, or live marketing as it is sometimes known, has grown in popularity. In fact, event marketing now claims the largest portion of many promotional budgets and accounts for half of all promotional spending (PROMO Industry Trends Report, 2006). George P. Johnson Co.'s annual survey, EventView, reported that 96 percent of marketing executives use events in their mix. Unfortunately, half of all marketers admit to not measuring events (PROMO Magazine, 2006), while the other half have barely enough data to know what is working and what is not. This uncertainty accounts, perhaps, for the decline in event spending. Other marketing tactics, such as coupons or product samples, have been around for decades and are measured as a matter of standard practice by most brands that use them. While brand marketers know these tactics will provide a positive ROI, they are choosing events based on the hope of achieving a better consumer connection, and as a result, a more dramatic ROI. Event marketing choices, nevertheless, ought to be based on more than wishful thinking. Events do have their place in promotion plans and do not need to be eliminated completely. But smart marketers will conduct their "due-diligence" to determine where events will deliver a good trial experience - and where they will not. Experience has shown that most often, events are the best choice when there is no other way to provide a trial opportunity to a consumer. Of all the tactics that brands use as tie-ins to events, sampling leads the list. More than 50 percent of marketers say they distribute samples at events (PROMO Industry Trends Report, 2006). Is this the best use of a limited resource? While sampling at events is not a bad idea in every situation, successful managers are careful to use a pre-event ROI analysis and other measurement tools to make an informed decision. Brands often distribute samples at events simply because they want to make the most of the overall strategic investment that motivated the event. That strategy, usually aimed broadly at creating favorable brand impression, can include goals like generating awareness, banking good will and other public relations objectives. Product sampling goals are very different. Product sampling is all about trial; it is about making the best use of a limited resource to gain trial - to get as many different consumers as possible to sample the product. If an event does not present the best opportunity for trial, then the brand has not maximized its resources properly. Risk is another key reason not to use samples at an event. Event sampling drastically increases complexity. Some of the risk factors that are most threatening to event sampling ROI are as follows: poor attendance, unappealing brand representatives, weather-related issues and re-sampling. Event sampling works better for some product categories than others. If a product category needs some type of preparation, involves special storage or requires a demonstration, an event may be the only way to truly sample the product correctly. If a brand does not have an adequate sample (a one-use product experience), then event marketing can be an effective way of making a connection with targeted consumers. Which product categories are better served with targeted, direct-to-consumer sampling programs and which are better served with events or event sampling (see Table I)? Most marketers believe they need to do something more to win over today's young adults - the target of most brand samples. Many marketers developing promotion plans hope to make a mark on the business by creating "cool" newsworthy promotions. Some brands choose to sample at events in the hope that interest in the event will deliver additional brand loyalty. Some marketers believe that integrating all elements of the marketing plan will deliver optimal results. This leads them to believe that their sampling results will be strongest in a program that has tied various promotional activities together. More importantly, does it provide a better ROI? Here are some instructive case studies.Situation no. 1 A new shampoo brand could reach 500,000 consumers with samples via a health-club sampling program for $235,000 ($.22 distribution/$.25 sample cost). If this same brand chose to do a 20-city campaign, distributing 25,000 samples per event via a mobile tour, the cost would be more than $335,000. That is because sample costs are the same but distribution costs are almost double (see Table II).Analysis The brand may be expecting higher purchase numbers from the event, since it anticipates a stronger consumer connection. But the reality is that trial rates are likely to be lower, resulting in lower purchase numbers. Sample waste is likely to be much higher with an event sampling program.In this example, the brand team needs to ask the difficult questions: By participating in the mobile tour, were some consumers who did not receive a sample converted? Or, how much did the brand interaction at the event favorably influence other purchase results? In this case, any other activity would have to more than double the results, which is very unlikely.Situation no. 2 An OTC/health care brand has the opportunity to sample up to 150,000 "competitive users" via a direct-mail program. The sampling vendor has a database of households that purchased a competitive brand in the past 12-months. The cost of this program would be $195,000. The other option is to sponsor ten events in different cities, sampling up to 15,000 consumers at each event. The cost of this program would be almost $165,000 (see Table III).Analysis The brand believes it might receive higher trial and purchase numbers from the event, based on an assumption that direct mail results would be low. The reality is, however, that the event would have lower trial and purchase numbers and higher sample waste.In this case, the costs associated with buying the competitive-user list and mailing the sample made this program more expensive than the event. By targeting better and maintaining better sample controls, the brand is able to get a much better ROI from the first program. It would be highly unlikely that the brand could cover the loss from the event-sampling program through any other added benefits the event might provide.Case study methodology The costs in these case studies were quoted in actual programs presented to brands. The trial and purchase numbers were actual averages from sampling effectiveness studies for these types of programs (an industry-accepted method of measurement). Average number of samples received in traditional programs was taken from actual sampling effectiveness studies conducted by SEA over the past three years. Average number of samples received at events was determined by an independent consumer survey, conducted by National Research Inc., June 2001. Even when a targeted, direct-to-consumer sampling program has higher distribution costs, due to using a highly targeted list or having to mail the sample, it is unlikely that an event-sampling program could return a better ROI. That's because sample controls are not as good and trial numbers are likely to be lower, resulting in lower purchase numbers. It is also highly unlikely that any other benefit derived from an event could increase ROI enough to cover the substantial trial and purchase differences.What steps should marketers take to achieve the highest ROI behind product sampling? First and foremost, brand marketers must know their consumers and apply the best sampling strategy to achieve the brand's goals. Events may play a role in that strategy, but that is the exception rather than the rule. Events are more appropriate for brands that cannot be tried any other way. Marketers will benefit from doing a pre-event ROI analysis to see which type of program will deliver the highest ROI. Those assumptions should be confirmed via market research.If marketers will follow these important STEPS, their programs will return a higher ROI:* Situation must be right for trial.* Target needs to be receptive to the brands message.* Effective sample controls must be in place.* Program distribution costs should be reasonable.* Sampling program must be measured.By following these STEPS, brand marketers will have the data needed to make the best program choices and will significantly improve the ROI of future sampling programs. Opens in a new window.Table I Sampling matrix Opens in a new window.Table II Situation no. 1 Opens in a new window.Table III Situation no. 2
|
- The purpose of this paper is to examine commonly relied upon product sampling strategies, direct-to-consumer sampling and event sampling, to determine which method can deliver the greatest return on investment in a variety of situations.
|
[SECTION: Method] Recent trends in the evolution of marketing have delivered return on investment (ROI)-driven brand managers to an important crossroads. Should they choose event marketing or consumer sampling? Can it ever be both? If so, under what circumstances? Just a few years ago in 2004, the leading brands spent close to a million dollars a year on event marketing. That volume of expenditure declined sharply in the subsequent two years, however, slipping 13 percent in 2005 and another 16 percent in 2006. Conversely, spending on product sampling rose 9 percent in 2006, according to the 2007 PROMO Industry Trends Report. The increase was fueled by that invaluable marketer's tool - measurability. As brand developers become more aware of the bang they are getting (or not getting) from their investment buck, spending patterns are bound to change. This paper is meant to help marketers understand the dynamics of the different types of sampling, and to identify the best marketing program investment opportunities. A decade ago, event marketing rarely was a major component of a brand's promotional plan. But event marketing, or live marketing as it is sometimes known, has grown in popularity. In fact, event marketing now claims the largest portion of many promotional budgets and accounts for half of all promotional spending (PROMO Industry Trends Report, 2006). George P. Johnson Co.'s annual survey, EventView, reported that 96 percent of marketing executives use events in their mix. Unfortunately, half of all marketers admit to not measuring events (PROMO Magazine, 2006), while the other half have barely enough data to know what is working and what is not. This uncertainty accounts, perhaps, for the decline in event spending. Other marketing tactics, such as coupons or product samples, have been around for decades and are measured as a matter of standard practice by most brands that use them. While brand marketers know these tactics will provide a positive ROI, they are choosing events based on the hope of achieving a better consumer connection, and as a result, a more dramatic ROI. Event marketing choices, nevertheless, ought to be based on more than wishful thinking. Events do have their place in promotion plans and do not need to be eliminated completely. But smart marketers will conduct their "due-diligence" to determine where events will deliver a good trial experience - and where they will not. Experience has shown that most often, events are the best choice when there is no other way to provide a trial opportunity to a consumer. Of all the tactics that brands use as tie-ins to events, sampling leads the list. More than 50 percent of marketers say they distribute samples at events (PROMO Industry Trends Report, 2006). Is this the best use of a limited resource? While sampling at events is not a bad idea in every situation, successful managers are careful to use a pre-event ROI analysis and other measurement tools to make an informed decision. Brands often distribute samples at events simply because they want to make the most of the overall strategic investment that motivated the event. That strategy, usually aimed broadly at creating favorable brand impression, can include goals like generating awareness, banking good will and other public relations objectives. Product sampling goals are very different. Product sampling is all about trial; it is about making the best use of a limited resource to gain trial - to get as many different consumers as possible to sample the product. If an event does not present the best opportunity for trial, then the brand has not maximized its resources properly. Risk is another key reason not to use samples at an event. Event sampling drastically increases complexity. Some of the risk factors that are most threatening to event sampling ROI are as follows: poor attendance, unappealing brand representatives, weather-related issues and re-sampling. Event sampling works better for some product categories than others. If a product category needs some type of preparation, involves special storage or requires a demonstration, an event may be the only way to truly sample the product correctly. If a brand does not have an adequate sample (a one-use product experience), then event marketing can be an effective way of making a connection with targeted consumers. Which product categories are better served with targeted, direct-to-consumer sampling programs and which are better served with events or event sampling (see Table I)? Most marketers believe they need to do something more to win over today's young adults - the target of most brand samples. Many marketers developing promotion plans hope to make a mark on the business by creating "cool" newsworthy promotions. Some brands choose to sample at events in the hope that interest in the event will deliver additional brand loyalty. Some marketers believe that integrating all elements of the marketing plan will deliver optimal results. This leads them to believe that their sampling results will be strongest in a program that has tied various promotional activities together. More importantly, does it provide a better ROI? Here are some instructive case studies.Situation no. 1 A new shampoo brand could reach 500,000 consumers with samples via a health-club sampling program for $235,000 ($.22 distribution/$.25 sample cost). If this same brand chose to do a 20-city campaign, distributing 25,000 samples per event via a mobile tour, the cost would be more than $335,000. That is because sample costs are the same but distribution costs are almost double (see Table II).Analysis The brand may be expecting higher purchase numbers from the event, since it anticipates a stronger consumer connection. But the reality is that trial rates are likely to be lower, resulting in lower purchase numbers. Sample waste is likely to be much higher with an event sampling program.In this example, the brand team needs to ask the difficult questions: By participating in the mobile tour, were some consumers who did not receive a sample converted? Or, how much did the brand interaction at the event favorably influence other purchase results? In this case, any other activity would have to more than double the results, which is very unlikely.Situation no. 2 An OTC/health care brand has the opportunity to sample up to 150,000 "competitive users" via a direct-mail program. The sampling vendor has a database of households that purchased a competitive brand in the past 12-months. The cost of this program would be $195,000. The other option is to sponsor ten events in different cities, sampling up to 15,000 consumers at each event. The cost of this program would be almost $165,000 (see Table III).Analysis The brand believes it might receive higher trial and purchase numbers from the event, based on an assumption that direct mail results would be low. The reality is, however, that the event would have lower trial and purchase numbers and higher sample waste.In this case, the costs associated with buying the competitive-user list and mailing the sample made this program more expensive than the event. By targeting better and maintaining better sample controls, the brand is able to get a much better ROI from the first program. It would be highly unlikely that the brand could cover the loss from the event-sampling program through any other added benefits the event might provide.Case study methodology The costs in these case studies were quoted in actual programs presented to brands. The trial and purchase numbers were actual averages from sampling effectiveness studies for these types of programs (an industry-accepted method of measurement). Average number of samples received in traditional programs was taken from actual sampling effectiveness studies conducted by SEA over the past three years. Average number of samples received at events was determined by an independent consumer survey, conducted by National Research Inc., June 2001. Even when a targeted, direct-to-consumer sampling program has higher distribution costs, due to using a highly targeted list or having to mail the sample, it is unlikely that an event-sampling program could return a better ROI. That's because sample controls are not as good and trial numbers are likely to be lower, resulting in lower purchase numbers. It is also highly unlikely that any other benefit derived from an event could increase ROI enough to cover the substantial trial and purchase differences.What steps should marketers take to achieve the highest ROI behind product sampling? First and foremost, brand marketers must know their consumers and apply the best sampling strategy to achieve the brand's goals. Events may play a role in that strategy, but that is the exception rather than the rule. Events are more appropriate for brands that cannot be tried any other way. Marketers will benefit from doing a pre-event ROI analysis to see which type of program will deliver the highest ROI. Those assumptions should be confirmed via market research.If marketers will follow these important STEPS, their programs will return a higher ROI:* Situation must be right for trial.* Target needs to be receptive to the brands message.* Effective sample controls must be in place.* Program distribution costs should be reasonable.* Sampling program must be measured.By following these STEPS, brand marketers will have the data needed to make the best program choices and will significantly improve the ROI of future sampling programs. Opens in a new window.Table I Sampling matrix Opens in a new window.Table II Situation no. 1 Opens in a new window.Table III Situation no. 2
|
- The paper explores information and data collected via case studies in actual programs presented to brands. The trial, sample, and purchase numbers were actual averages from sampling effectiveness studies for these types of programs.
|
[SECTION: Findings] Recent trends in the evolution of marketing have delivered return on investment (ROI)-driven brand managers to an important crossroads. Should they choose event marketing or consumer sampling? Can it ever be both? If so, under what circumstances? Just a few years ago in 2004, the leading brands spent close to a million dollars a year on event marketing. That volume of expenditure declined sharply in the subsequent two years, however, slipping 13 percent in 2005 and another 16 percent in 2006. Conversely, spending on product sampling rose 9 percent in 2006, according to the 2007 PROMO Industry Trends Report. The increase was fueled by that invaluable marketer's tool - measurability. As brand developers become more aware of the bang they are getting (or not getting) from their investment buck, spending patterns are bound to change. This paper is meant to help marketers understand the dynamics of the different types of sampling, and to identify the best marketing program investment opportunities. A decade ago, event marketing rarely was a major component of a brand's promotional plan. But event marketing, or live marketing as it is sometimes known, has grown in popularity. In fact, event marketing now claims the largest portion of many promotional budgets and accounts for half of all promotional spending (PROMO Industry Trends Report, 2006). George P. Johnson Co.'s annual survey, EventView, reported that 96 percent of marketing executives use events in their mix. Unfortunately, half of all marketers admit to not measuring events (PROMO Magazine, 2006), while the other half have barely enough data to know what is working and what is not. This uncertainty accounts, perhaps, for the decline in event spending. Other marketing tactics, such as coupons or product samples, have been around for decades and are measured as a matter of standard practice by most brands that use them. While brand marketers know these tactics will provide a positive ROI, they are choosing events based on the hope of achieving a better consumer connection, and as a result, a more dramatic ROI. Event marketing choices, nevertheless, ought to be based on more than wishful thinking. Events do have their place in promotion plans and do not need to be eliminated completely. But smart marketers will conduct their "due-diligence" to determine where events will deliver a good trial experience - and where they will not. Experience has shown that most often, events are the best choice when there is no other way to provide a trial opportunity to a consumer. Of all the tactics that brands use as tie-ins to events, sampling leads the list. More than 50 percent of marketers say they distribute samples at events (PROMO Industry Trends Report, 2006). Is this the best use of a limited resource? While sampling at events is not a bad idea in every situation, successful managers are careful to use a pre-event ROI analysis and other measurement tools to make an informed decision. Brands often distribute samples at events simply because they want to make the most of the overall strategic investment that motivated the event. That strategy, usually aimed broadly at creating favorable brand impression, can include goals like generating awareness, banking good will and other public relations objectives. Product sampling goals are very different. Product sampling is all about trial; it is about making the best use of a limited resource to gain trial - to get as many different consumers as possible to sample the product. If an event does not present the best opportunity for trial, then the brand has not maximized its resources properly. Risk is another key reason not to use samples at an event. Event sampling drastically increases complexity. Some of the risk factors that are most threatening to event sampling ROI are as follows: poor attendance, unappealing brand representatives, weather-related issues and re-sampling. Event sampling works better for some product categories than others. If a product category needs some type of preparation, involves special storage or requires a demonstration, an event may be the only way to truly sample the product correctly. If a brand does not have an adequate sample (a one-use product experience), then event marketing can be an effective way of making a connection with targeted consumers. Which product categories are better served with targeted, direct-to-consumer sampling programs and which are better served with events or event sampling (see Table I)? Most marketers believe they need to do something more to win over today's young adults - the target of most brand samples. Many marketers developing promotion plans hope to make a mark on the business by creating "cool" newsworthy promotions. Some brands choose to sample at events in the hope that interest in the event will deliver additional brand loyalty. Some marketers believe that integrating all elements of the marketing plan will deliver optimal results. This leads them to believe that their sampling results will be strongest in a program that has tied various promotional activities together. More importantly, does it provide a better ROI? Here are some instructive case studies.Situation no. 1 A new shampoo brand could reach 500,000 consumers with samples via a health-club sampling program for $235,000 ($.22 distribution/$.25 sample cost). If this same brand chose to do a 20-city campaign, distributing 25,000 samples per event via a mobile tour, the cost would be more than $335,000. That is because sample costs are the same but distribution costs are almost double (see Table II).Analysis The brand may be expecting higher purchase numbers from the event, since it anticipates a stronger consumer connection. But the reality is that trial rates are likely to be lower, resulting in lower purchase numbers. Sample waste is likely to be much higher with an event sampling program.In this example, the brand team needs to ask the difficult questions: By participating in the mobile tour, were some consumers who did not receive a sample converted? Or, how much did the brand interaction at the event favorably influence other purchase results? In this case, any other activity would have to more than double the results, which is very unlikely.Situation no. 2 An OTC/health care brand has the opportunity to sample up to 150,000 "competitive users" via a direct-mail program. The sampling vendor has a database of households that purchased a competitive brand in the past 12-months. The cost of this program would be $195,000. The other option is to sponsor ten events in different cities, sampling up to 15,000 consumers at each event. The cost of this program would be almost $165,000 (see Table III).Analysis The brand believes it might receive higher trial and purchase numbers from the event, based on an assumption that direct mail results would be low. The reality is, however, that the event would have lower trial and purchase numbers and higher sample waste.In this case, the costs associated with buying the competitive-user list and mailing the sample made this program more expensive than the event. By targeting better and maintaining better sample controls, the brand is able to get a much better ROI from the first program. It would be highly unlikely that the brand could cover the loss from the event-sampling program through any other added benefits the event might provide.Case study methodology The costs in these case studies were quoted in actual programs presented to brands. The trial and purchase numbers were actual averages from sampling effectiveness studies for these types of programs (an industry-accepted method of measurement). Average number of samples received in traditional programs was taken from actual sampling effectiveness studies conducted by SEA over the past three years. Average number of samples received at events was determined by an independent consumer survey, conducted by National Research Inc., June 2001. Even when a targeted, direct-to-consumer sampling program has higher distribution costs, due to using a highly targeted list or having to mail the sample, it is unlikely that an event-sampling program could return a better ROI. That's because sample controls are not as good and trial numbers are likely to be lower, resulting in lower purchase numbers. It is also highly unlikely that any other benefit derived from an event could increase ROI enough to cover the substantial trial and purchase differences.What steps should marketers take to achieve the highest ROI behind product sampling? First and foremost, brand marketers must know their consumers and apply the best sampling strategy to achieve the brand's goals. Events may play a role in that strategy, but that is the exception rather than the rule. Events are more appropriate for brands that cannot be tried any other way. Marketers will benefit from doing a pre-event ROI analysis to see which type of program will deliver the highest ROI. Those assumptions should be confirmed via market research.If marketers will follow these important STEPS, their programs will return a higher ROI:* Situation must be right for trial.* Target needs to be receptive to the brands message.* Effective sample controls must be in place.* Program distribution costs should be reasonable.* Sampling program must be measured.By following these STEPS, brand marketers will have the data needed to make the best program choices and will significantly improve the ROI of future sampling programs. Opens in a new window.Table I Sampling matrix Opens in a new window.Table II Situation no. 1 Opens in a new window.Table III Situation no. 2
|
- The paper identifies and segments different types of products and the method by which the products are most effectively implemented into trial and sampling programs. ICOM reveals hard statistics on the return on investment of programs utilizing multiple methods of sampling including point-of-use, direct mail and event sampling.
|
[SECTION: Value] Recent trends in the evolution of marketing have delivered return on investment (ROI)-driven brand managers to an important crossroads. Should they choose event marketing or consumer sampling? Can it ever be both? If so, under what circumstances? Just a few years ago in 2004, the leading brands spent close to a million dollars a year on event marketing. That volume of expenditure declined sharply in the subsequent two years, however, slipping 13 percent in 2005 and another 16 percent in 2006. Conversely, spending on product sampling rose 9 percent in 2006, according to the 2007 PROMO Industry Trends Report. The increase was fueled by that invaluable marketer's tool - measurability. As brand developers become more aware of the bang they are getting (or not getting) from their investment buck, spending patterns are bound to change. This paper is meant to help marketers understand the dynamics of the different types of sampling, and to identify the best marketing program investment opportunities. A decade ago, event marketing rarely was a major component of a brand's promotional plan. But event marketing, or live marketing as it is sometimes known, has grown in popularity. In fact, event marketing now claims the largest portion of many promotional budgets and accounts for half of all promotional spending (PROMO Industry Trends Report, 2006). George P. Johnson Co.'s annual survey, EventView, reported that 96 percent of marketing executives use events in their mix. Unfortunately, half of all marketers admit to not measuring events (PROMO Magazine, 2006), while the other half have barely enough data to know what is working and what is not. This uncertainty accounts, perhaps, for the decline in event spending. Other marketing tactics, such as coupons or product samples, have been around for decades and are measured as a matter of standard practice by most brands that use them. While brand marketers know these tactics will provide a positive ROI, they are choosing events based on the hope of achieving a better consumer connection, and as a result, a more dramatic ROI. Event marketing choices, nevertheless, ought to be based on more than wishful thinking. Events do have their place in promotion plans and do not need to be eliminated completely. But smart marketers will conduct their "due-diligence" to determine where events will deliver a good trial experience - and where they will not. Experience has shown that most often, events are the best choice when there is no other way to provide a trial opportunity to a consumer. Of all the tactics that brands use as tie-ins to events, sampling leads the list. More than 50 percent of marketers say they distribute samples at events (PROMO Industry Trends Report, 2006). Is this the best use of a limited resource? While sampling at events is not a bad idea in every situation, successful managers are careful to use a pre-event ROI analysis and other measurement tools to make an informed decision. Brands often distribute samples at events simply because they want to make the most of the overall strategic investment that motivated the event. That strategy, usually aimed broadly at creating favorable brand impression, can include goals like generating awareness, banking good will and other public relations objectives. Product sampling goals are very different. Product sampling is all about trial; it is about making the best use of a limited resource to gain trial - to get as many different consumers as possible to sample the product. If an event does not present the best opportunity for trial, then the brand has not maximized its resources properly. Risk is another key reason not to use samples at an event. Event sampling drastically increases complexity. Some of the risk factors that are most threatening to event sampling ROI are as follows: poor attendance, unappealing brand representatives, weather-related issues and re-sampling. Event sampling works better for some product categories than others. If a product category needs some type of preparation, involves special storage or requires a demonstration, an event may be the only way to truly sample the product correctly. If a brand does not have an adequate sample (a one-use product experience), then event marketing can be an effective way of making a connection with targeted consumers. Which product categories are better served with targeted, direct-to-consumer sampling programs and which are better served with events or event sampling (see Table I)? Most marketers believe they need to do something more to win over today's young adults - the target of most brand samples. Many marketers developing promotion plans hope to make a mark on the business by creating "cool" newsworthy promotions. Some brands choose to sample at events in the hope that interest in the event will deliver additional brand loyalty. Some marketers believe that integrating all elements of the marketing plan will deliver optimal results. This leads them to believe that their sampling results will be strongest in a program that has tied various promotional activities together. More importantly, does it provide a better ROI? Here are some instructive case studies.Situation no. 1 A new shampoo brand could reach 500,000 consumers with samples via a health-club sampling program for $235,000 ($.22 distribution/$.25 sample cost). If this same brand chose to do a 20-city campaign, distributing 25,000 samples per event via a mobile tour, the cost would be more than $335,000. That is because sample costs are the same but distribution costs are almost double (see Table II).Analysis The brand may be expecting higher purchase numbers from the event, since it anticipates a stronger consumer connection. But the reality is that trial rates are likely to be lower, resulting in lower purchase numbers. Sample waste is likely to be much higher with an event sampling program.In this example, the brand team needs to ask the difficult questions: By participating in the mobile tour, were some consumers who did not receive a sample converted? Or, how much did the brand interaction at the event favorably influence other purchase results? In this case, any other activity would have to more than double the results, which is very unlikely.Situation no. 2 An OTC/health care brand has the opportunity to sample up to 150,000 "competitive users" via a direct-mail program. The sampling vendor has a database of households that purchased a competitive brand in the past 12-months. The cost of this program would be $195,000. The other option is to sponsor ten events in different cities, sampling up to 15,000 consumers at each event. The cost of this program would be almost $165,000 (see Table III).Analysis The brand believes it might receive higher trial and purchase numbers from the event, based on an assumption that direct mail results would be low. The reality is, however, that the event would have lower trial and purchase numbers and higher sample waste.In this case, the costs associated with buying the competitive-user list and mailing the sample made this program more expensive than the event. By targeting better and maintaining better sample controls, the brand is able to get a much better ROI from the first program. It would be highly unlikely that the brand could cover the loss from the event-sampling program through any other added benefits the event might provide.Case study methodology The costs in these case studies were quoted in actual programs presented to brands. The trial and purchase numbers were actual averages from sampling effectiveness studies for these types of programs (an industry-accepted method of measurement). Average number of samples received in traditional programs was taken from actual sampling effectiveness studies conducted by SEA over the past three years. Average number of samples received at events was determined by an independent consumer survey, conducted by National Research Inc., June 2001. Even when a targeted, direct-to-consumer sampling program has higher distribution costs, due to using a highly targeted list or having to mail the sample, it is unlikely that an event-sampling program could return a better ROI. That's because sample controls are not as good and trial numbers are likely to be lower, resulting in lower purchase numbers. It is also highly unlikely that any other benefit derived from an event could increase ROI enough to cover the substantial trial and purchase differences.What steps should marketers take to achieve the highest ROI behind product sampling? First and foremost, brand marketers must know their consumers and apply the best sampling strategy to achieve the brand's goals. Events may play a role in that strategy, but that is the exception rather than the rule. Events are more appropriate for brands that cannot be tried any other way. Marketers will benefit from doing a pre-event ROI analysis to see which type of program will deliver the highest ROI. Those assumptions should be confirmed via market research.If marketers will follow these important STEPS, their programs will return a higher ROI:* Situation must be right for trial.* Target needs to be receptive to the brands message.* Effective sample controls must be in place.* Program distribution costs should be reasonable.* Sampling program must be measured.By following these STEPS, brand marketers will have the data needed to make the best program choices and will significantly improve the ROI of future sampling programs. Opens in a new window.Table I Sampling matrix Opens in a new window.Table II Situation no. 1 Opens in a new window.Table III Situation no. 2
|
- The paper reveals reasons for a growing shift in corporate budgeted marketing dollars from event marketing to direct consumer product sampling. While event sampling is not always an ineffective or inferior marketing method,the reader can discover methods for a pre-event return on investment analysis that will reveal the sampling strategy sure to deliver the most "bang" for the marketing "buck".
|
[SECTION: Purpose] An executive summary for managers and executive readers can be found at the end of this article. The objective of this research article is to shed light on the evolution of brand management into a crucial strategic tool for international business operations. On basis of the literature available in this field, we analyze the largest 100 brands (hereafter categorized as megabrands) in terms of ranking and value modifications in the 2001 to 2005 period, a mature globalization period, with the first ranking referring to pre-09/11 findings. The sample and its analysis provide us with significant findings that open crucial questions about US/non-US brand strategy and perceptions, and the future application of global megabrand policies. We then shed light on the causal factors that global terrorism may contain and tentatively propose brand strategy solutions, but do not exclude other causal factors or co-factors that will need further inquiry. Overall, our hypothesis is that brands serve to bring security[1]. Accordingly, if the source of that brand is less secure, then it will be less effective as a brand. This hypothesis needs to be qualified: In particular, would one not expect that the short run reaction to insecurity is to be more, rather than less brand loyalty? The findings of this study have a strong indication that this assumption can be reversed, and we indicate that indirect impacts of global terrorism might be the reason. Further, it is important to show that negative security shifts in the US have been greater than a general increase in malaise in the global markets where the brands are sold. Again, the data indicates that movements in brand value and ranking appear to respond to more than such a malaise.Being a simple "identification tool" at its very start, brand names have become a critical part of a company's strategy. Academic research has shown that one major historic reason for brand success is the diminished risk perceived by the consumer (Roselius, 1971; Kapferer, 1991; Keller, 1998; Riezebos, 2003). McCarthy (1971) highlights the three primary roles of brand:* identification and purchase simplification function;* brand has a projective, symbolic and imaginary function and provides the consumer with a status; and* brand guarantees quality, protection and risk reduction for the consumer by pointing out to its source.For these reasons, companies are willing to consider brands as an important asset of their balance sheets or to invest huge amounts of capital to buy them (Laforet and Saunders, 1994).The power of brands is founded on consumers' aversion to uncertainty. For a long time, consumers made their food buying decisions based only on a product's visual aspect, ignoring its brand name, accepting instead the grocery store owner's opinion as selection criteria (Boyer, 2002). Later on producers introduced clearly visible signals that identified their products and consumers then got used to preferring the signal as opposed to the product visual characteristics (Keller, 1998); that is, brand became more important than the product itself (Riezebos, 2003).Even at the present, perceived risk reduction is the first reason consumers have for choosing a brand and this guides brand management evolution (Kapferer, 2003). When consumers perceive a risk in making a buying decision, they will deploy different strategies for reducing it. Five major risks are considered by consumers:1. Financial risk ("making a bad deal", which increases the importance of the brand compared with the unit price of the product).2. Physical risk (being harmed by the product, especially food products).3. Technological risk (being disappointed by the product performance, it is the risk of functionality).4. Psychological risk (feeling guilty or irresponsible for temptation, especially in impulsive decisions - or associating harm or risk to the brand, either associated to fear or sadness).5. Social risk (what pairs will say or think about choices. Therefore brand is a sign of possession for a community, but also a sign of adherence, of patriotism or of association to or away from particular social issues.)Risk reduction function directly related to the brand has been increased by the macro-economic context, especially after 09/11 because a fragile and complex environment is expected to increase the role the brand has to play in reassuring the consumers' buying decisions. Even though, we later argue that the capacity of brands to link producers and consumers has been rudely challenged. There have been drastic changes of consumption habits in some markets, such as the acceleration of the coming on-stream of the hard-discounters in Europe, with a new approach to the quality - price relationship and the weakening of the brand, of low-cost airlines, and of non-brand textiles from low-cost production. Companies have reacted to these new challenges. This new environment has notably changed the way in which big international companies conceive of their brands. Brand guarantee and its image are shelter points for consumers: normally, the higher the risk the more helpful the brand. Consequently, brands have learned a different way of communication (e.g. emphasizing safety themes, as carmakers do already), to change their relationship with the environment or towards the Third World (e.g., Nike reconsidering its production policy in order to improve its brand image) but also with globalization (e.g. being more respectful of local brands, as Nestle is). Brands also start working on ethical matters (The Body Shop's cosmetics products), fair trade (Malongo coffee) or social responsibility. But one of the major facets of this adaptation of brands and firms to the new situation is the coming on-stream and acceleration of megabrands within companies. Traditionally, choosing brand strategies is the focal point for companies, whether they are multinational groups or local companies (Schuiling and Kapferer, 2004). Supposing that a firm has different sources of competition, one of the strategic issues is whether it uses one or several brands. Strebinger (2002) states that one of the most critical problems in branding relates to the management of a mono or multi-brand system while Riezebos (2003) questions whether it is feasible having just a single-brand strategy in the company, with a prime focus on one brand and then developing additional brands from it.The historical development of branding includes some deeply contradictory factors, as shown in Figure 1.This figure visualizes and conceptualizes a company's willingness and need to have numerous products able to meet the different customers' demands as appropriate as possible, to assure their expansion and international development, that is, to counteract all risk of being a single-brand company. Likewise, there is a need for limiting the number of brands because of a second risk: that of a brand overexposure or over usage, including the financial risk of dispersing the investment.The first risk leads companies wishing to develop to buy or launch more brands in order to enter markets, segments or customers inaccessible with only one brand. This may be an "inflationist" process in terms of markets as it leads to create many brands.The second risk takes the same companies in the opposite direction, trying to limit the number of brands in order to maximize investment per brand, thereby making the brands stronger and covering more territory.But this process is intrinsically schizophrenic and raises the question of the strategic equilibrium of branding (Riezebos, 2003). Strategic choices may become brand choices, choices of brand organization or choices about the kind of relationship between brands that a company wants to maintain. One of the purposes of these choices is to maximize the equity of its different brands.As a way to escape from this process, many companies turn to megabranding.At its origins, the evolution of the brand universe towards megabrands comes from big corporations that discovered, in the early 80s, that they could create value by capitalizing on the transnational concepts carried in supranational brands so as to attain maximum return on investment (Kapferer, 2000). This new strategy reduced internal brand management costs and the costs of launching new innovative products. This simple idea has allowed many companies to focus on the strongest brands, or on brands with high growth potential or on highly internationalized brands, and to abandon or minimize all others. Indeed, at the beginning, economic reasons were the main inspiration for this rationalization process: first of all trying to concentrate all human and economic resources on a few brands and, especially, cutting advertising costs related to the launching and maintenance process of multiple brands.The megabrand concept, thus, is a core concern for most leading transnational firms because, as the competitive environment becomes more and more complex, and with a high level of risks of every nature, companies focus on megabrand strategy and attempt to assure their expansion and international development.In the early 1990s many companies did inform the market of their intentions to reduce their brand numbers: The most extreme case being that of Unilever, which planned to reduce from 1,600 down to 400 brands in the 2000-2004 period. Anthony Simon[2], President of Unilever-BestFoods marketing, underlined that "Unilever's objective is to reduce the number of brands in order to make them stronger. Four strategies support this decision: category, segment, channel and geography".In a megabrand strategy, a brand name may be used for horizontal extensions (inside the same price layer, common for mass consumption products) or vertical extensions (in different price layers, common for durable goods). This strategy can be very successful; a well-developed brand can provide a sustainable competitive advantage. To ensure continuous success, the operation of a megabrand strategy demands permanent innovation, strong R&D investment, a communicational style hard to imitate and a brand image not based on the product but on associations and perceptions.Megabrand management changes the focus of marketing to a superior, strategic decision-making level (Baldinger, 1990; Trinquecoste, 1999), as it implicitly involves focusing on the whole company instead of on individual brands (Riezebos, 2003). Both, Juga (1999) and Reynaud (2001) show that by displacing competition to this superior level, competitive advantages become harder to understand (less tangible) and to imitate.The increasing recognition of brands as a source of sustainable competitive advantage stresses the importance of conceptual models about organizational brand strategies (Louro and Cunha, 2001). Therefore, our research goal is to explore the megabranding field and to evaluate its strategic dimension as a new and more complex and durable source of competitive advantage in times of international adversity and the challenges of 09/11-type terrorism. We have chosen to analyze the evolution of the value of megabrands over a five year period. The sample consists of those brands ranked in "The 100 best global brands" annually by Interbrand corporation for Business Week magazine.Interbrand defined seven criteria (see Appendix) which evaluate brands much in the way analysts value other assets, i.e. on the basis of how much they are likely to earn in the future.To qualify for the list each brand must:* have a value greater than $1 billion;* derive about a third of its earnings outside its home country; and* have publicly available marketing and financial data.For these reasons Interbrand specifies that such heavyweights as Visa, Wal-Mart, Mars or CNN are eliminated from the rankings. Only brands are taken in account (and not parent companies such as Procter and Gamble), and airlines are not ranked because it is too hard to separate their brand impact on sales from factors such as routes and schedules.Despite its limits, this ranking provides a global vision of the value of the main megabrands. This ranking has gained importance over the past years as a main reference for brand strategy. In addition, the assessment and evaluation method has not changed over the past five years. The rankings we refer to were published at the following dates: 6 August 2001/5 August 2002/4 August 2003/22 July 2004/21 July 2005.We present these five rankings in the Appendix. The first ranking refers to the period prior to the 09/11 events. We have, at the same time, conducted in-depth research into the question whether other factors may be responsible for the results we have found. Charts that summarize these findings are also presented in the Appendix, and - while it is certainly impossible to be exhaustive- exclude any major movements, evolutions, malaises or crises that could have the effects found (covering empirical research into size-trend and relative to industry, profitability-trend and relative to others in the industry, industry stage of life cycle, leverage-how vulnerable they are they to taking risks, country of origin to observe movements including characteristics such as access to capital, human resources, competition, and an index of insecurity, movements in scopes of megabrands (global reach, horizontal and vertical branding), or change in type of customers (e.g., services, package goods, durables, business); data chosen for illustration only covers main developments). By including such variables in the analysis, we strive to make it possible to determine that the risk hypothesis can be supported after controlling for other factors that lead to success, identifying other general factors that would lead to shift in megabrand positioning over time. No unexpected development in such has been found. The top 100 brands We first analyze the evolution of the one hundred top brands worldwide, in absolute value and in relative value.The results are summarized in Tables I and II:* The value of the 100 top brands increased by 5.7 percent, from 988.21 billion dollars to 1044.58 billion dollars in five years.* Amongst the 100 top brands worldwide, the total value of US brands declined from 737.55 billion dollars in August 2001 to 701.13 billion in July 2005, i.e. decreased by 4.9 percent. At the same time the value of non-US brands in this ranking increased from 250,66 billion to 343,45 billion (plus 37 percent). This evolution is far too important to be explained solely by currency and exchange rate conditions.* The number of US brands in this ranking declined from 63 brands to 52; while the number of non-US brands increased from 37 to 48 brands over the five -year period.Figure 2 illustrates the ratio of non-US compared to US brands in the top one hundred. This relative weight of the one to the other evolves dramatically from in the post - 09/11 era. It is important to note that the initial measure is dated July 2001, i.e. two months before 09/11.The top 20 brands We then focus on the top 20 brands of the ranking and their evolution.This focus is considered as especially important because these 20 top brands represent more than 50 percent of the global value of the whole 100 top worldwide brands. The data extracted here appear in Tables III and IV:* over the period examined, the number of non-US brands increase from four to seven brands and their value from 90.36 billions to 135.2 billions (plus 56.2 percent); and* the number of US brands decreases at the same time from 16 to 13 brands and their value decreases by 10.2 percent.Following this comparative analysis in the period studied, we also analyze the data of the US/non-US brand value ratio through a separation into zones, i.e. the top 20 US and the top 20 non-US brands: Now one is to pay special attention to the top ten, the top 20 and also the second/lower half of the top 20 brands.Table V shows that the lower half of the top twenty brands have, on the US side, suffered more than those in the top half. To the contrary, the bottom part of the top twenty also shows that non-US brands have very strongly increased their value.Top and bottom halves of the top 10 US and non-US brands We then analyze the value of the top ten US brands versus the value of the top ten non-US brands, as shown in Tables VI and VII: the value of the ten top US brands decreased by 5.25 percent while the value of the ten top non-US brands increases by 8.1 percent.The significance of the gap (i.e. the difference in value) between the top ten US and the ten top - non-US brands is illustrated in Figure 3. This figure is calculated via the total value as it evolves over time, i.e. dividing the total of 2002 by that of 2001, and so on for each year. The figure clearly demonstrates the increasing gap between the US and non-US megabrands over time, and the decrease of the curve because the gap strongly increases between 2002 and 2005 (by calculus of linear regression). Our initial assumption for this research was that international corporations adapted their brand marketing to globalization. We began by reviewing megabrand strategies that were put into effect over three decades, an option chosen by a wide range of companies to secure global, relatively easy and cost-efficient management of brands. We then raised the question of how megabrands evolved over the five years from 2000, with an objective to study the validity of this strategy through the analysis of the value evolution in the ensemble of megabrands worldwideThe data analysis provides strong empirical findings and raises an important set of questions: The value of US top brands worldwide declined significantly after 2001, and over the past rankings of world megabrands, while non-US brands experienced significant expansion over the same period. This evolution is confirmed on all three levels of analysis that we developed: total of 100 leading brands, total of the 20 leading brands, and comparison between the leading ten US and non-US brands.Why is the value gap more significant in the top 20 brands than in the top ten ones? Are second tier brands of this sample more vulnerable, and if so, why? The further we decrease the ranks of top brands in the top 100, the bigger the gap becomes between US and non-US brands, and this to the benefit of non-US ones. Are business cycle trends responsible for this trend? Are these brands are particularly symbolic in terms of nationality and risk perception since 2001 and global terrorism? Will consumers feel uncomfortable with certain brands since 9/11, and if so, what indications could allow us to understand this phenomenon? Is the dot.com burst responsible for this?With these questions in mind, further analysis provides the following indications: The following tables, one with the top 20 evolutions and one with the bottom 20 worst evolutions of brands, only refer to brands for which data is available for all five years.The findings indicate that in the top 20 "best evolution" international mega-brands only eight brands are US American and 12 are non-US, that the two best performers by value are non US (Samsung and Louis Vuitton), that Pepsi Co is the top US brand (interestingly it is the leading one in the US and in terms of brand name the competitor Coca-Cola is part of the bottom 20 brands, though still best known); that 2nd and 3rd best American mega-brands are Dell and Apple brands, and those who one could consider best known as Microsoft and Oracle are in the bottom 20. Does this mean that demand remained constant but strong US image made them fall? Due to the diversity of products and sectors represented, we believe that the dot.com bubble, highly sector-dependent, cannot be causal or solely causal to the megabrand evolutions that we note.Also, currency fluctuations in that time period would rather imply opposite effects.If thus the evolution of brands value over this five year period, of megabrands, is linked to brand nationality, and in this case that of US or non-US origins, this would imply that corporations need to invest in megabrands emanating from different regions. If one considers that US brands may be more sensitive to risk perceptions from global terrorism, by the consumer, than non-US brands, and that this terrorism could be a causal factor, because the data modifies after 2001, then the managerial objective is immunity to the consequences of such events. Given the crucial significance of such cause to strategy, we provide some basis for understanding and potentially resilience. Alexander et al. (1979, p. 4) define terrorism as "the systematic threat or use of violence to attain a political goal or communicate a political message through fear, coercion, or intimidation of particular persons or the general public". We can assume that the citizen and consumer in this general public is, therefore, exposed to stress scenarios that differs from typical scenarios, and therefore alter his or her purchasing behavior.It is widely admitted that with 9/11/2001, terrorism has become more global (Schneckener, 2002). 9/11-type terrorism is characterized by a proximity to western civilization and its psychological impact is reinforced through wide spread media coverage. Contemporary terrorist activities share a number of common features which are inter-related and of a recently resurrected nature: These features include the increasing link of terrorist activity to a quasi-legitimization on basis of allegedly religious motivation, modern business-like leadership structures, asymmetric warfare, and the use of the victim mostly as part of a communication strategy. The objectives of terrorists are to convey a triple message:1. Government is not capable of guaranteeing security of a society or citizen, nor of service or product safety.2. Corporations, investors and travelers are safe nowhere, and that symbols of a country, culture and society that they convey are potential targets for any type of attack.3. Any measure taken against terrorism is insufficient by nature.These messages have a powerful impact on many. Psychological effects (defined above as any of the extremes, from feeling guilt or irresponsible for temptation, especially in impulsive decisions to associating harm or risk to the brand, either associated to fear or sadness) instill uncertainty into the economy, and have been found elsewhere to significantly affect the economic, organizational and governmental environment (Suder, 2004). Given this, we adduce that consumer behavior and corporate strategy may be affected. For instance, just as in times of war, the consumer may adopt a "stocking/storing" behavior for particular types of food and medicines if he/she perceives a terror-based threat.Therefore, we hypothesized elsewhere that a firm's performance under uncertainty and risk of terrorism will be a function of its ability to reduce its vulnerability to terrorist acts through risk analysis and assessment, through shortened supply lines, and a decreased need for economic redundancy (Suder, 2006). This is even more so in the case of 9/11-type terrorism; a terrorism that has globalized and that hits the global activities of firms in addition to those at the location of a strike. In this section, we therefore focus on the question whether top management of megabrands should take into account a corporation's vulnerability to terrorist threat felt by consumers. If a brand has national symbolism - like Coca Cola -then its goods or services are exposed to threat or acts of terrorism. Will the consumer turn away from the brand, or in fact rather increase its faith in it? Our study could be interpreted to show possible link on a quantitative basis by comparative approach. In this case, is a megabrand strategy still a reasonable option?To be deemed reliable, enterprises must be able to keep their brands resilient in the event of a catastrophe. The US airlines that were victims to the hijacking of its planes crashed into the WTC in New York are the first illustration of the psychological impacts of brands that are related to terrorism threat. The symbolic relation to the events though entirely involuntary had dramatic consequences for bith American and United airlines. Also, a tendency of clients to rather fly shorter distances, on separate flights and with non-nationally related aircraft such as low-cost airlines emerged since 9/11 (MacBain, 2003, Tourism Queensland, 2006 et.al.).Markets melt down or freeze with great speed in case of threat or terrorism acts, other markets can rise because consider unrelated to the threat (Suder and Czinkota, 2007). Another example is the reluctance of Londoners to use public transport after the double-attacks of summer 2005; the bicycle-market however boomed almost immediately. The terrible human costs of terrorism are clearly unacceptable to any logic or ethics. Given that terrorism has existed in various forms over history, people, companies and industry now need be knowledgeable about 9/11 type threat and its impacts, and to adapt. International terrorism adds an important determinant to the definition of a firm's brand strategy. As an uncontrollable force in its external environment, terrorism events may lead to direct (mainly physical) or indirect (for instance consumer behavior and brand perception) disruptions. In the preliminary phase of threatened violence, or the following phase of the attack's aftermath (for details of this classification, see Suder, 2004), consumer demand for the firm's goods and services may alter but does not always decline (e.g. the demand for security equipment and services increases); any related disruption to the value chain perceivable by the consumer such as supply difficulties of needed inputs, resources, and services; or government policies and laws enacted to deal with terrorism alter the conduct of brand strategy. Macroeconomic phenomena, and shifts in international relations also modify behaviors. Media plays an important role in the intensification of the related psychological effects. For instance, the political differences between some European states and the USA in terms of the conduct of a war against terrorism, in particular concerning the invasion of Iraq, significantly modified consumer behaviors in the USA towards French and German brands (such as Roquefort cheese, Perrier water, ...and even French fries, solely based on their denomination). In those different dimensions, terrorism threat, act and aftermath affect:* ways of life;* perceptions;* consumption habits of millions of people all around the world; and* the company-client relationship.The responsiveness of consumers to a global threat is particularly high because it is intangible, close- by, and may strike anyone anywhere, in an expression of the "flatness" of the world. The incalculable uncertainty becomes a certainty that terror events happen and society and business adapt. The only certainty is that events will always be symbolic, whether that applies to locations, victims or the relation to the "hated" society. In this society anyone and anything can potentially identify with victims to attacks, whether human or object, whether a site, a product or a group. We therefore assume that this is so for brands in their dependence on perceptions and image.For a corporation, brand strategy and the administration of price shifts, communications, distribution strategies, buyers and suppliers, logistics, import and export are directly exposed to cultural issues, image responsibilities, and consequences of actions. For a consumer, brands have the particular capacity to link producers and consumers who trust in a specific set of quality, service and security "guarantees" linked psychologically to a particular brand. Brand marketing is symbolic and related on confidence, quite at the opposite of fear or panic. The consumer will hence turn to (or turn away) from brands in proportion to the strength that the brand relates to the threat, and expose brand strategy to risks non-related to their good performance. A brand is by definition the symbol of an object or a service, as well as a model of the consumption society (Keller, 1998). One major weakness of the megabrand approach is to expose the company to a major risk: A single brand, a single image. Needless to say, if a problem occurs with this brand the whole company's stability is at stake. But consumers are also citizens and so the brand may be a broader social and economic battleground amongst companies with respect to consumers. For example, brands also represent an important political space were virulent political battles can be fought (Semprini, 1992). Some movements embody or oppose lifestyles symbolised by brands and their influence, sometimes in a very radical way, the consumer's society becomes represented by companies and their brands (Klein, 2002). This contesting opposition must be taken into consideration when developing brands and their territories in order to avoid vulnerability of a single-brand strategy and extreme exposure. Various authors already tackled this notion under the theme of brand capital or brand equity (Farquhar, 1989; Baldinger, 1990; Kapferer, 1991; Aaker, 1992; Keller, 1998). For Aaker (1992), brand capital is a unit consisting of the name and symbolic meaning of a brand that can add or decrease the value of a product or service, and that delivers value to the client and to the firm. An appropriate strategy this reinforces the value of brands while an inappropriate strategy diminishes the value. On basis of our findings and the exposed nexus that one may establish with 9/11, we suggest that megabrand strategy allows corporations to obtain critical size (specially facing the distribution channels), face the growth limits of existing brands, share, soften, and pool certain costs (research, industrialization, marketing) although the megabrand building process is time related and based on a variety of experiences. One can hence suggest that, if here lies the causal link, in post-9/11 megabrands allow for better control of risks by the company and increase the value of brands better if locally or regionally embedded. If megabrand strategy overexposes brands as symbolic for a mode of consumption rejected by or associated to terrorism, then megabrand overexposure diminishes the value of brands, by overexposing firms to risks, band devaluation and increasing company vulnerability. The findings of this research imply that brand strategy is highly dependent on exterior factors and need to be adapted to those if competitive advantages shall not erode and shift considerably. While the causal link to 9/11 terrorism can not be clearly established, it does appear as one of the sensible explanations or co-factors for the dramatic evolution that were found. These findings in themselves hope to make a contribution to the understanding of megabrand strategy in mature globalization. It appears from our research that brand nationality, and thus brand associations to the various effects of terrorism (victimization or identification), may define the behavior of consumers and have an impact on brands value and ranking. For a future that may have to cope with 9/11-type terrorism, megabrands (except for the very strongest ones perhaps) may therefore not qualify as the best option for companies that wish to reduce risk and immunize brands and performances. If this is confirmed, the firms are well-advised to invest in mega-brands anchored into regions, through a transnational rather than a global strategy.Clearly further research into the potential causalities is needed: Whether terrorism, business cycles, currency issues, the bubble effect, all of these events united or none, international business scholars and practitioners are advised to study these links together, in each sector and market so as improve understandings and capabilities to respond appropriately to the evolution of megabrands in ranking and value since 2001. Opens in a new window.Figure 1 Contradictory tendencies in brand development Opens in a new window.Figure 2 Ratio of non-US compared to US brands in the top 100 Opens in a new window.Figure 3 Gap (i.e. the difference in value) between the top ten US and the ten top non-US brands; base 100% in 2001 Opens in a new window.Table I Evolution of the 100 biggest worldwide brands and split between US and non-US brands Opens in a new window.Table II Evolution of the 100 biggest worldwide brands and split between US and non-US brands Opens in a new window.Table IV Evolution of the 20 first worldwide brands and split between US and non-US brands Opens in a new window.Table III Evolution of the 20 first worldwide brands and split between US and non-US brands Opens in a new window.Table V Detailed analysis of gaps Opens in a new window.Table VII Evolution of the ten biggest US brands and of the ten biggest non-US brands Opens in a new window.Table VI The top ten US brands Opens in a new window.Table AI Inderbrand seven-criteria ranking Opens in a new window.Table AII The five relevant megabrand rankings. 2001-2005 Opens in a new window.Table AIII Exclusion of other potentially relevant causal factors. Analysis of the brands' industry in and out Opens in a new window.Table AIV Exclusion of other potentially relevant causal factors. Brands in and out linked to industry and sector Opens in a new window.Table AV Exclusion of other potentially relevant causal factors. Analysis of significant changes (technology sector + financial), and performance Opens in a new window.Table AVI Country ranking by risk category (Coface, 2006)
|
- The primary purpose of this study is to identify if and how international terrorism has altered the rank and value of brands, and whether the increasing uncertainty of globalizing risks need an adaptation of international brand management.
|
[SECTION: Method] An executive summary for managers and executive readers can be found at the end of this article. The objective of this research article is to shed light on the evolution of brand management into a crucial strategic tool for international business operations. On basis of the literature available in this field, we analyze the largest 100 brands (hereafter categorized as megabrands) in terms of ranking and value modifications in the 2001 to 2005 period, a mature globalization period, with the first ranking referring to pre-09/11 findings. The sample and its analysis provide us with significant findings that open crucial questions about US/non-US brand strategy and perceptions, and the future application of global megabrand policies. We then shed light on the causal factors that global terrorism may contain and tentatively propose brand strategy solutions, but do not exclude other causal factors or co-factors that will need further inquiry. Overall, our hypothesis is that brands serve to bring security[1]. Accordingly, if the source of that brand is less secure, then it will be less effective as a brand. This hypothesis needs to be qualified: In particular, would one not expect that the short run reaction to insecurity is to be more, rather than less brand loyalty? The findings of this study have a strong indication that this assumption can be reversed, and we indicate that indirect impacts of global terrorism might be the reason. Further, it is important to show that negative security shifts in the US have been greater than a general increase in malaise in the global markets where the brands are sold. Again, the data indicates that movements in brand value and ranking appear to respond to more than such a malaise.Being a simple "identification tool" at its very start, brand names have become a critical part of a company's strategy. Academic research has shown that one major historic reason for brand success is the diminished risk perceived by the consumer (Roselius, 1971; Kapferer, 1991; Keller, 1998; Riezebos, 2003). McCarthy (1971) highlights the three primary roles of brand:* identification and purchase simplification function;* brand has a projective, symbolic and imaginary function and provides the consumer with a status; and* brand guarantees quality, protection and risk reduction for the consumer by pointing out to its source.For these reasons, companies are willing to consider brands as an important asset of their balance sheets or to invest huge amounts of capital to buy them (Laforet and Saunders, 1994).The power of brands is founded on consumers' aversion to uncertainty. For a long time, consumers made their food buying decisions based only on a product's visual aspect, ignoring its brand name, accepting instead the grocery store owner's opinion as selection criteria (Boyer, 2002). Later on producers introduced clearly visible signals that identified their products and consumers then got used to preferring the signal as opposed to the product visual characteristics (Keller, 1998); that is, brand became more important than the product itself (Riezebos, 2003).Even at the present, perceived risk reduction is the first reason consumers have for choosing a brand and this guides brand management evolution (Kapferer, 2003). When consumers perceive a risk in making a buying decision, they will deploy different strategies for reducing it. Five major risks are considered by consumers:1. Financial risk ("making a bad deal", which increases the importance of the brand compared with the unit price of the product).2. Physical risk (being harmed by the product, especially food products).3. Technological risk (being disappointed by the product performance, it is the risk of functionality).4. Psychological risk (feeling guilty or irresponsible for temptation, especially in impulsive decisions - or associating harm or risk to the brand, either associated to fear or sadness).5. Social risk (what pairs will say or think about choices. Therefore brand is a sign of possession for a community, but also a sign of adherence, of patriotism or of association to or away from particular social issues.)Risk reduction function directly related to the brand has been increased by the macro-economic context, especially after 09/11 because a fragile and complex environment is expected to increase the role the brand has to play in reassuring the consumers' buying decisions. Even though, we later argue that the capacity of brands to link producers and consumers has been rudely challenged. There have been drastic changes of consumption habits in some markets, such as the acceleration of the coming on-stream of the hard-discounters in Europe, with a new approach to the quality - price relationship and the weakening of the brand, of low-cost airlines, and of non-brand textiles from low-cost production. Companies have reacted to these new challenges. This new environment has notably changed the way in which big international companies conceive of their brands. Brand guarantee and its image are shelter points for consumers: normally, the higher the risk the more helpful the brand. Consequently, brands have learned a different way of communication (e.g. emphasizing safety themes, as carmakers do already), to change their relationship with the environment or towards the Third World (e.g., Nike reconsidering its production policy in order to improve its brand image) but also with globalization (e.g. being more respectful of local brands, as Nestle is). Brands also start working on ethical matters (The Body Shop's cosmetics products), fair trade (Malongo coffee) or social responsibility. But one of the major facets of this adaptation of brands and firms to the new situation is the coming on-stream and acceleration of megabrands within companies. Traditionally, choosing brand strategies is the focal point for companies, whether they are multinational groups or local companies (Schuiling and Kapferer, 2004). Supposing that a firm has different sources of competition, one of the strategic issues is whether it uses one or several brands. Strebinger (2002) states that one of the most critical problems in branding relates to the management of a mono or multi-brand system while Riezebos (2003) questions whether it is feasible having just a single-brand strategy in the company, with a prime focus on one brand and then developing additional brands from it.The historical development of branding includes some deeply contradictory factors, as shown in Figure 1.This figure visualizes and conceptualizes a company's willingness and need to have numerous products able to meet the different customers' demands as appropriate as possible, to assure their expansion and international development, that is, to counteract all risk of being a single-brand company. Likewise, there is a need for limiting the number of brands because of a second risk: that of a brand overexposure or over usage, including the financial risk of dispersing the investment.The first risk leads companies wishing to develop to buy or launch more brands in order to enter markets, segments or customers inaccessible with only one brand. This may be an "inflationist" process in terms of markets as it leads to create many brands.The second risk takes the same companies in the opposite direction, trying to limit the number of brands in order to maximize investment per brand, thereby making the brands stronger and covering more territory.But this process is intrinsically schizophrenic and raises the question of the strategic equilibrium of branding (Riezebos, 2003). Strategic choices may become brand choices, choices of brand organization or choices about the kind of relationship between brands that a company wants to maintain. One of the purposes of these choices is to maximize the equity of its different brands.As a way to escape from this process, many companies turn to megabranding.At its origins, the evolution of the brand universe towards megabrands comes from big corporations that discovered, in the early 80s, that they could create value by capitalizing on the transnational concepts carried in supranational brands so as to attain maximum return on investment (Kapferer, 2000). This new strategy reduced internal brand management costs and the costs of launching new innovative products. This simple idea has allowed many companies to focus on the strongest brands, or on brands with high growth potential or on highly internationalized brands, and to abandon or minimize all others. Indeed, at the beginning, economic reasons were the main inspiration for this rationalization process: first of all trying to concentrate all human and economic resources on a few brands and, especially, cutting advertising costs related to the launching and maintenance process of multiple brands.The megabrand concept, thus, is a core concern for most leading transnational firms because, as the competitive environment becomes more and more complex, and with a high level of risks of every nature, companies focus on megabrand strategy and attempt to assure their expansion and international development.In the early 1990s many companies did inform the market of their intentions to reduce their brand numbers: The most extreme case being that of Unilever, which planned to reduce from 1,600 down to 400 brands in the 2000-2004 period. Anthony Simon[2], President of Unilever-BestFoods marketing, underlined that "Unilever's objective is to reduce the number of brands in order to make them stronger. Four strategies support this decision: category, segment, channel and geography".In a megabrand strategy, a brand name may be used for horizontal extensions (inside the same price layer, common for mass consumption products) or vertical extensions (in different price layers, common for durable goods). This strategy can be very successful; a well-developed brand can provide a sustainable competitive advantage. To ensure continuous success, the operation of a megabrand strategy demands permanent innovation, strong R&D investment, a communicational style hard to imitate and a brand image not based on the product but on associations and perceptions.Megabrand management changes the focus of marketing to a superior, strategic decision-making level (Baldinger, 1990; Trinquecoste, 1999), as it implicitly involves focusing on the whole company instead of on individual brands (Riezebos, 2003). Both, Juga (1999) and Reynaud (2001) show that by displacing competition to this superior level, competitive advantages become harder to understand (less tangible) and to imitate.The increasing recognition of brands as a source of sustainable competitive advantage stresses the importance of conceptual models about organizational brand strategies (Louro and Cunha, 2001). Therefore, our research goal is to explore the megabranding field and to evaluate its strategic dimension as a new and more complex and durable source of competitive advantage in times of international adversity and the challenges of 09/11-type terrorism. We have chosen to analyze the evolution of the value of megabrands over a five year period. The sample consists of those brands ranked in "The 100 best global brands" annually by Interbrand corporation for Business Week magazine.Interbrand defined seven criteria (see Appendix) which evaluate brands much in the way analysts value other assets, i.e. on the basis of how much they are likely to earn in the future.To qualify for the list each brand must:* have a value greater than $1 billion;* derive about a third of its earnings outside its home country; and* have publicly available marketing and financial data.For these reasons Interbrand specifies that such heavyweights as Visa, Wal-Mart, Mars or CNN are eliminated from the rankings. Only brands are taken in account (and not parent companies such as Procter and Gamble), and airlines are not ranked because it is too hard to separate their brand impact on sales from factors such as routes and schedules.Despite its limits, this ranking provides a global vision of the value of the main megabrands. This ranking has gained importance over the past years as a main reference for brand strategy. In addition, the assessment and evaluation method has not changed over the past five years. The rankings we refer to were published at the following dates: 6 August 2001/5 August 2002/4 August 2003/22 July 2004/21 July 2005.We present these five rankings in the Appendix. The first ranking refers to the period prior to the 09/11 events. We have, at the same time, conducted in-depth research into the question whether other factors may be responsible for the results we have found. Charts that summarize these findings are also presented in the Appendix, and - while it is certainly impossible to be exhaustive- exclude any major movements, evolutions, malaises or crises that could have the effects found (covering empirical research into size-trend and relative to industry, profitability-trend and relative to others in the industry, industry stage of life cycle, leverage-how vulnerable they are they to taking risks, country of origin to observe movements including characteristics such as access to capital, human resources, competition, and an index of insecurity, movements in scopes of megabrands (global reach, horizontal and vertical branding), or change in type of customers (e.g., services, package goods, durables, business); data chosen for illustration only covers main developments). By including such variables in the analysis, we strive to make it possible to determine that the risk hypothesis can be supported after controlling for other factors that lead to success, identifying other general factors that would lead to shift in megabrand positioning over time. No unexpected development in such has been found. The top 100 brands We first analyze the evolution of the one hundred top brands worldwide, in absolute value and in relative value.The results are summarized in Tables I and II:* The value of the 100 top brands increased by 5.7 percent, from 988.21 billion dollars to 1044.58 billion dollars in five years.* Amongst the 100 top brands worldwide, the total value of US brands declined from 737.55 billion dollars in August 2001 to 701.13 billion in July 2005, i.e. decreased by 4.9 percent. At the same time the value of non-US brands in this ranking increased from 250,66 billion to 343,45 billion (plus 37 percent). This evolution is far too important to be explained solely by currency and exchange rate conditions.* The number of US brands in this ranking declined from 63 brands to 52; while the number of non-US brands increased from 37 to 48 brands over the five -year period.Figure 2 illustrates the ratio of non-US compared to US brands in the top one hundred. This relative weight of the one to the other evolves dramatically from in the post - 09/11 era. It is important to note that the initial measure is dated July 2001, i.e. two months before 09/11.The top 20 brands We then focus on the top 20 brands of the ranking and their evolution.This focus is considered as especially important because these 20 top brands represent more than 50 percent of the global value of the whole 100 top worldwide brands. The data extracted here appear in Tables III and IV:* over the period examined, the number of non-US brands increase from four to seven brands and their value from 90.36 billions to 135.2 billions (plus 56.2 percent); and* the number of US brands decreases at the same time from 16 to 13 brands and their value decreases by 10.2 percent.Following this comparative analysis in the period studied, we also analyze the data of the US/non-US brand value ratio through a separation into zones, i.e. the top 20 US and the top 20 non-US brands: Now one is to pay special attention to the top ten, the top 20 and also the second/lower half of the top 20 brands.Table V shows that the lower half of the top twenty brands have, on the US side, suffered more than those in the top half. To the contrary, the bottom part of the top twenty also shows that non-US brands have very strongly increased their value.Top and bottom halves of the top 10 US and non-US brands We then analyze the value of the top ten US brands versus the value of the top ten non-US brands, as shown in Tables VI and VII: the value of the ten top US brands decreased by 5.25 percent while the value of the ten top non-US brands increases by 8.1 percent.The significance of the gap (i.e. the difference in value) between the top ten US and the ten top - non-US brands is illustrated in Figure 3. This figure is calculated via the total value as it evolves over time, i.e. dividing the total of 2002 by that of 2001, and so on for each year. The figure clearly demonstrates the increasing gap between the US and non-US megabrands over time, and the decrease of the curve because the gap strongly increases between 2002 and 2005 (by calculus of linear regression). Our initial assumption for this research was that international corporations adapted their brand marketing to globalization. We began by reviewing megabrand strategies that were put into effect over three decades, an option chosen by a wide range of companies to secure global, relatively easy and cost-efficient management of brands. We then raised the question of how megabrands evolved over the five years from 2000, with an objective to study the validity of this strategy through the analysis of the value evolution in the ensemble of megabrands worldwideThe data analysis provides strong empirical findings and raises an important set of questions: The value of US top brands worldwide declined significantly after 2001, and over the past rankings of world megabrands, while non-US brands experienced significant expansion over the same period. This evolution is confirmed on all three levels of analysis that we developed: total of 100 leading brands, total of the 20 leading brands, and comparison between the leading ten US and non-US brands.Why is the value gap more significant in the top 20 brands than in the top ten ones? Are second tier brands of this sample more vulnerable, and if so, why? The further we decrease the ranks of top brands in the top 100, the bigger the gap becomes between US and non-US brands, and this to the benefit of non-US ones. Are business cycle trends responsible for this trend? Are these brands are particularly symbolic in terms of nationality and risk perception since 2001 and global terrorism? Will consumers feel uncomfortable with certain brands since 9/11, and if so, what indications could allow us to understand this phenomenon? Is the dot.com burst responsible for this?With these questions in mind, further analysis provides the following indications: The following tables, one with the top 20 evolutions and one with the bottom 20 worst evolutions of brands, only refer to brands for which data is available for all five years.The findings indicate that in the top 20 "best evolution" international mega-brands only eight brands are US American and 12 are non-US, that the two best performers by value are non US (Samsung and Louis Vuitton), that Pepsi Co is the top US brand (interestingly it is the leading one in the US and in terms of brand name the competitor Coca-Cola is part of the bottom 20 brands, though still best known); that 2nd and 3rd best American mega-brands are Dell and Apple brands, and those who one could consider best known as Microsoft and Oracle are in the bottom 20. Does this mean that demand remained constant but strong US image made them fall? Due to the diversity of products and sectors represented, we believe that the dot.com bubble, highly sector-dependent, cannot be causal or solely causal to the megabrand evolutions that we note.Also, currency fluctuations in that time period would rather imply opposite effects.If thus the evolution of brands value over this five year period, of megabrands, is linked to brand nationality, and in this case that of US or non-US origins, this would imply that corporations need to invest in megabrands emanating from different regions. If one considers that US brands may be more sensitive to risk perceptions from global terrorism, by the consumer, than non-US brands, and that this terrorism could be a causal factor, because the data modifies after 2001, then the managerial objective is immunity to the consequences of such events. Given the crucial significance of such cause to strategy, we provide some basis for understanding and potentially resilience. Alexander et al. (1979, p. 4) define terrorism as "the systematic threat or use of violence to attain a political goal or communicate a political message through fear, coercion, or intimidation of particular persons or the general public". We can assume that the citizen and consumer in this general public is, therefore, exposed to stress scenarios that differs from typical scenarios, and therefore alter his or her purchasing behavior.It is widely admitted that with 9/11/2001, terrorism has become more global (Schneckener, 2002). 9/11-type terrorism is characterized by a proximity to western civilization and its psychological impact is reinforced through wide spread media coverage. Contemporary terrorist activities share a number of common features which are inter-related and of a recently resurrected nature: These features include the increasing link of terrorist activity to a quasi-legitimization on basis of allegedly religious motivation, modern business-like leadership structures, asymmetric warfare, and the use of the victim mostly as part of a communication strategy. The objectives of terrorists are to convey a triple message:1. Government is not capable of guaranteeing security of a society or citizen, nor of service or product safety.2. Corporations, investors and travelers are safe nowhere, and that symbols of a country, culture and society that they convey are potential targets for any type of attack.3. Any measure taken against terrorism is insufficient by nature.These messages have a powerful impact on many. Psychological effects (defined above as any of the extremes, from feeling guilt or irresponsible for temptation, especially in impulsive decisions to associating harm or risk to the brand, either associated to fear or sadness) instill uncertainty into the economy, and have been found elsewhere to significantly affect the economic, organizational and governmental environment (Suder, 2004). Given this, we adduce that consumer behavior and corporate strategy may be affected. For instance, just as in times of war, the consumer may adopt a "stocking/storing" behavior for particular types of food and medicines if he/she perceives a terror-based threat.Therefore, we hypothesized elsewhere that a firm's performance under uncertainty and risk of terrorism will be a function of its ability to reduce its vulnerability to terrorist acts through risk analysis and assessment, through shortened supply lines, and a decreased need for economic redundancy (Suder, 2006). This is even more so in the case of 9/11-type terrorism; a terrorism that has globalized and that hits the global activities of firms in addition to those at the location of a strike. In this section, we therefore focus on the question whether top management of megabrands should take into account a corporation's vulnerability to terrorist threat felt by consumers. If a brand has national symbolism - like Coca Cola -then its goods or services are exposed to threat or acts of terrorism. Will the consumer turn away from the brand, or in fact rather increase its faith in it? Our study could be interpreted to show possible link on a quantitative basis by comparative approach. In this case, is a megabrand strategy still a reasonable option?To be deemed reliable, enterprises must be able to keep their brands resilient in the event of a catastrophe. The US airlines that were victims to the hijacking of its planes crashed into the WTC in New York are the first illustration of the psychological impacts of brands that are related to terrorism threat. The symbolic relation to the events though entirely involuntary had dramatic consequences for bith American and United airlines. Also, a tendency of clients to rather fly shorter distances, on separate flights and with non-nationally related aircraft such as low-cost airlines emerged since 9/11 (MacBain, 2003, Tourism Queensland, 2006 et.al.).Markets melt down or freeze with great speed in case of threat or terrorism acts, other markets can rise because consider unrelated to the threat (Suder and Czinkota, 2007). Another example is the reluctance of Londoners to use public transport after the double-attacks of summer 2005; the bicycle-market however boomed almost immediately. The terrible human costs of terrorism are clearly unacceptable to any logic or ethics. Given that terrorism has existed in various forms over history, people, companies and industry now need be knowledgeable about 9/11 type threat and its impacts, and to adapt. International terrorism adds an important determinant to the definition of a firm's brand strategy. As an uncontrollable force in its external environment, terrorism events may lead to direct (mainly physical) or indirect (for instance consumer behavior and brand perception) disruptions. In the preliminary phase of threatened violence, or the following phase of the attack's aftermath (for details of this classification, see Suder, 2004), consumer demand for the firm's goods and services may alter but does not always decline (e.g. the demand for security equipment and services increases); any related disruption to the value chain perceivable by the consumer such as supply difficulties of needed inputs, resources, and services; or government policies and laws enacted to deal with terrorism alter the conduct of brand strategy. Macroeconomic phenomena, and shifts in international relations also modify behaviors. Media plays an important role in the intensification of the related psychological effects. For instance, the political differences between some European states and the USA in terms of the conduct of a war against terrorism, in particular concerning the invasion of Iraq, significantly modified consumer behaviors in the USA towards French and German brands (such as Roquefort cheese, Perrier water, ...and even French fries, solely based on their denomination). In those different dimensions, terrorism threat, act and aftermath affect:* ways of life;* perceptions;* consumption habits of millions of people all around the world; and* the company-client relationship.The responsiveness of consumers to a global threat is particularly high because it is intangible, close- by, and may strike anyone anywhere, in an expression of the "flatness" of the world. The incalculable uncertainty becomes a certainty that terror events happen and society and business adapt. The only certainty is that events will always be symbolic, whether that applies to locations, victims or the relation to the "hated" society. In this society anyone and anything can potentially identify with victims to attacks, whether human or object, whether a site, a product or a group. We therefore assume that this is so for brands in their dependence on perceptions and image.For a corporation, brand strategy and the administration of price shifts, communications, distribution strategies, buyers and suppliers, logistics, import and export are directly exposed to cultural issues, image responsibilities, and consequences of actions. For a consumer, brands have the particular capacity to link producers and consumers who trust in a specific set of quality, service and security "guarantees" linked psychologically to a particular brand. Brand marketing is symbolic and related on confidence, quite at the opposite of fear or panic. The consumer will hence turn to (or turn away) from brands in proportion to the strength that the brand relates to the threat, and expose brand strategy to risks non-related to their good performance. A brand is by definition the symbol of an object or a service, as well as a model of the consumption society (Keller, 1998). One major weakness of the megabrand approach is to expose the company to a major risk: A single brand, a single image. Needless to say, if a problem occurs with this brand the whole company's stability is at stake. But consumers are also citizens and so the brand may be a broader social and economic battleground amongst companies with respect to consumers. For example, brands also represent an important political space were virulent political battles can be fought (Semprini, 1992). Some movements embody or oppose lifestyles symbolised by brands and their influence, sometimes in a very radical way, the consumer's society becomes represented by companies and their brands (Klein, 2002). This contesting opposition must be taken into consideration when developing brands and their territories in order to avoid vulnerability of a single-brand strategy and extreme exposure. Various authors already tackled this notion under the theme of brand capital or brand equity (Farquhar, 1989; Baldinger, 1990; Kapferer, 1991; Aaker, 1992; Keller, 1998). For Aaker (1992), brand capital is a unit consisting of the name and symbolic meaning of a brand that can add or decrease the value of a product or service, and that delivers value to the client and to the firm. An appropriate strategy this reinforces the value of brands while an inappropriate strategy diminishes the value. On basis of our findings and the exposed nexus that one may establish with 9/11, we suggest that megabrand strategy allows corporations to obtain critical size (specially facing the distribution channels), face the growth limits of existing brands, share, soften, and pool certain costs (research, industrialization, marketing) although the megabrand building process is time related and based on a variety of experiences. One can hence suggest that, if here lies the causal link, in post-9/11 megabrands allow for better control of risks by the company and increase the value of brands better if locally or regionally embedded. If megabrand strategy overexposes brands as symbolic for a mode of consumption rejected by or associated to terrorism, then megabrand overexposure diminishes the value of brands, by overexposing firms to risks, band devaluation and increasing company vulnerability. The findings of this research imply that brand strategy is highly dependent on exterior factors and need to be adapted to those if competitive advantages shall not erode and shift considerably. While the causal link to 9/11 terrorism can not be clearly established, it does appear as one of the sensible explanations or co-factors for the dramatic evolution that were found. These findings in themselves hope to make a contribution to the understanding of megabrand strategy in mature globalization. It appears from our research that brand nationality, and thus brand associations to the various effects of terrorism (victimization or identification), may define the behavior of consumers and have an impact on brands value and ranking. For a future that may have to cope with 9/11-type terrorism, megabrands (except for the very strongest ones perhaps) may therefore not qualify as the best option for companies that wish to reduce risk and immunize brands and performances. If this is confirmed, the firms are well-advised to invest in mega-brands anchored into regions, through a transnational rather than a global strategy.Clearly further research into the potential causalities is needed: Whether terrorism, business cycles, currency issues, the bubble effect, all of these events united or none, international business scholars and practitioners are advised to study these links together, in each sector and market so as improve understandings and capabilities to respond appropriately to the evolution of megabrands in ranking and value since 2001. Opens in a new window.Figure 1 Contradictory tendencies in brand development Opens in a new window.Figure 2 Ratio of non-US compared to US brands in the top 100 Opens in a new window.Figure 3 Gap (i.e. the difference in value) between the top ten US and the ten top non-US brands; base 100% in 2001 Opens in a new window.Table I Evolution of the 100 biggest worldwide brands and split between US and non-US brands Opens in a new window.Table II Evolution of the 100 biggest worldwide brands and split between US and non-US brands Opens in a new window.Table IV Evolution of the 20 first worldwide brands and split between US and non-US brands Opens in a new window.Table III Evolution of the 20 first worldwide brands and split between US and non-US brands Opens in a new window.Table V Detailed analysis of gaps Opens in a new window.Table VII Evolution of the ten biggest US brands and of the ten biggest non-US brands Opens in a new window.Table VI The top ten US brands Opens in a new window.Table AI Inderbrand seven-criteria ranking Opens in a new window.Table AII The five relevant megabrand rankings. 2001-2005 Opens in a new window.Table AIII Exclusion of other potentially relevant causal factors. Analysis of the brands' industry in and out Opens in a new window.Table AIV Exclusion of other potentially relevant causal factors. Brands in and out linked to industry and sector Opens in a new window.Table AV Exclusion of other potentially relevant causal factors. Analysis of significant changes (technology sector + financial), and performance Opens in a new window.Table AVI Country ranking by risk category (Coface, 2006)
|
- The methodology for this study was exploratory and quantitative at the same time, and utilized longitudinal brand ranking and a cross-sector and cross-industry data in a comparative research design. Both descriptive and relational statistics are used to analyze the data.
|
[SECTION: Findings] An executive summary for managers and executive readers can be found at the end of this article. The objective of this research article is to shed light on the evolution of brand management into a crucial strategic tool for international business operations. On basis of the literature available in this field, we analyze the largest 100 brands (hereafter categorized as megabrands) in terms of ranking and value modifications in the 2001 to 2005 period, a mature globalization period, with the first ranking referring to pre-09/11 findings. The sample and its analysis provide us with significant findings that open crucial questions about US/non-US brand strategy and perceptions, and the future application of global megabrand policies. We then shed light on the causal factors that global terrorism may contain and tentatively propose brand strategy solutions, but do not exclude other causal factors or co-factors that will need further inquiry. Overall, our hypothesis is that brands serve to bring security[1]. Accordingly, if the source of that brand is less secure, then it will be less effective as a brand. This hypothesis needs to be qualified: In particular, would one not expect that the short run reaction to insecurity is to be more, rather than less brand loyalty? The findings of this study have a strong indication that this assumption can be reversed, and we indicate that indirect impacts of global terrorism might be the reason. Further, it is important to show that negative security shifts in the US have been greater than a general increase in malaise in the global markets where the brands are sold. Again, the data indicates that movements in brand value and ranking appear to respond to more than such a malaise.Being a simple "identification tool" at its very start, brand names have become a critical part of a company's strategy. Academic research has shown that one major historic reason for brand success is the diminished risk perceived by the consumer (Roselius, 1971; Kapferer, 1991; Keller, 1998; Riezebos, 2003). McCarthy (1971) highlights the three primary roles of brand:* identification and purchase simplification function;* brand has a projective, symbolic and imaginary function and provides the consumer with a status; and* brand guarantees quality, protection and risk reduction for the consumer by pointing out to its source.For these reasons, companies are willing to consider brands as an important asset of their balance sheets or to invest huge amounts of capital to buy them (Laforet and Saunders, 1994).The power of brands is founded on consumers' aversion to uncertainty. For a long time, consumers made their food buying decisions based only on a product's visual aspect, ignoring its brand name, accepting instead the grocery store owner's opinion as selection criteria (Boyer, 2002). Later on producers introduced clearly visible signals that identified their products and consumers then got used to preferring the signal as opposed to the product visual characteristics (Keller, 1998); that is, brand became more important than the product itself (Riezebos, 2003).Even at the present, perceived risk reduction is the first reason consumers have for choosing a brand and this guides brand management evolution (Kapferer, 2003). When consumers perceive a risk in making a buying decision, they will deploy different strategies for reducing it. Five major risks are considered by consumers:1. Financial risk ("making a bad deal", which increases the importance of the brand compared with the unit price of the product).2. Physical risk (being harmed by the product, especially food products).3. Technological risk (being disappointed by the product performance, it is the risk of functionality).4. Psychological risk (feeling guilty or irresponsible for temptation, especially in impulsive decisions - or associating harm or risk to the brand, either associated to fear or sadness).5. Social risk (what pairs will say or think about choices. Therefore brand is a sign of possession for a community, but also a sign of adherence, of patriotism or of association to or away from particular social issues.)Risk reduction function directly related to the brand has been increased by the macro-economic context, especially after 09/11 because a fragile and complex environment is expected to increase the role the brand has to play in reassuring the consumers' buying decisions. Even though, we later argue that the capacity of brands to link producers and consumers has been rudely challenged. There have been drastic changes of consumption habits in some markets, such as the acceleration of the coming on-stream of the hard-discounters in Europe, with a new approach to the quality - price relationship and the weakening of the brand, of low-cost airlines, and of non-brand textiles from low-cost production. Companies have reacted to these new challenges. This new environment has notably changed the way in which big international companies conceive of their brands. Brand guarantee and its image are shelter points for consumers: normally, the higher the risk the more helpful the brand. Consequently, brands have learned a different way of communication (e.g. emphasizing safety themes, as carmakers do already), to change their relationship with the environment or towards the Third World (e.g., Nike reconsidering its production policy in order to improve its brand image) but also with globalization (e.g. being more respectful of local brands, as Nestle is). Brands also start working on ethical matters (The Body Shop's cosmetics products), fair trade (Malongo coffee) or social responsibility. But one of the major facets of this adaptation of brands and firms to the new situation is the coming on-stream and acceleration of megabrands within companies. Traditionally, choosing brand strategies is the focal point for companies, whether they are multinational groups or local companies (Schuiling and Kapferer, 2004). Supposing that a firm has different sources of competition, one of the strategic issues is whether it uses one or several brands. Strebinger (2002) states that one of the most critical problems in branding relates to the management of a mono or multi-brand system while Riezebos (2003) questions whether it is feasible having just a single-brand strategy in the company, with a prime focus on one brand and then developing additional brands from it.The historical development of branding includes some deeply contradictory factors, as shown in Figure 1.This figure visualizes and conceptualizes a company's willingness and need to have numerous products able to meet the different customers' demands as appropriate as possible, to assure their expansion and international development, that is, to counteract all risk of being a single-brand company. Likewise, there is a need for limiting the number of brands because of a second risk: that of a brand overexposure or over usage, including the financial risk of dispersing the investment.The first risk leads companies wishing to develop to buy or launch more brands in order to enter markets, segments or customers inaccessible with only one brand. This may be an "inflationist" process in terms of markets as it leads to create many brands.The second risk takes the same companies in the opposite direction, trying to limit the number of brands in order to maximize investment per brand, thereby making the brands stronger and covering more territory.But this process is intrinsically schizophrenic and raises the question of the strategic equilibrium of branding (Riezebos, 2003). Strategic choices may become brand choices, choices of brand organization or choices about the kind of relationship between brands that a company wants to maintain. One of the purposes of these choices is to maximize the equity of its different brands.As a way to escape from this process, many companies turn to megabranding.At its origins, the evolution of the brand universe towards megabrands comes from big corporations that discovered, in the early 80s, that they could create value by capitalizing on the transnational concepts carried in supranational brands so as to attain maximum return on investment (Kapferer, 2000). This new strategy reduced internal brand management costs and the costs of launching new innovative products. This simple idea has allowed many companies to focus on the strongest brands, or on brands with high growth potential or on highly internationalized brands, and to abandon or minimize all others. Indeed, at the beginning, economic reasons were the main inspiration for this rationalization process: first of all trying to concentrate all human and economic resources on a few brands and, especially, cutting advertising costs related to the launching and maintenance process of multiple brands.The megabrand concept, thus, is a core concern for most leading transnational firms because, as the competitive environment becomes more and more complex, and with a high level of risks of every nature, companies focus on megabrand strategy and attempt to assure their expansion and international development.In the early 1990s many companies did inform the market of their intentions to reduce their brand numbers: The most extreme case being that of Unilever, which planned to reduce from 1,600 down to 400 brands in the 2000-2004 period. Anthony Simon[2], President of Unilever-BestFoods marketing, underlined that "Unilever's objective is to reduce the number of brands in order to make them stronger. Four strategies support this decision: category, segment, channel and geography".In a megabrand strategy, a brand name may be used for horizontal extensions (inside the same price layer, common for mass consumption products) or vertical extensions (in different price layers, common for durable goods). This strategy can be very successful; a well-developed brand can provide a sustainable competitive advantage. To ensure continuous success, the operation of a megabrand strategy demands permanent innovation, strong R&D investment, a communicational style hard to imitate and a brand image not based on the product but on associations and perceptions.Megabrand management changes the focus of marketing to a superior, strategic decision-making level (Baldinger, 1990; Trinquecoste, 1999), as it implicitly involves focusing on the whole company instead of on individual brands (Riezebos, 2003). Both, Juga (1999) and Reynaud (2001) show that by displacing competition to this superior level, competitive advantages become harder to understand (less tangible) and to imitate.The increasing recognition of brands as a source of sustainable competitive advantage stresses the importance of conceptual models about organizational brand strategies (Louro and Cunha, 2001). Therefore, our research goal is to explore the megabranding field and to evaluate its strategic dimension as a new and more complex and durable source of competitive advantage in times of international adversity and the challenges of 09/11-type terrorism. We have chosen to analyze the evolution of the value of megabrands over a five year period. The sample consists of those brands ranked in "The 100 best global brands" annually by Interbrand corporation for Business Week magazine.Interbrand defined seven criteria (see Appendix) which evaluate brands much in the way analysts value other assets, i.e. on the basis of how much they are likely to earn in the future.To qualify for the list each brand must:* have a value greater than $1 billion;* derive about a third of its earnings outside its home country; and* have publicly available marketing and financial data.For these reasons Interbrand specifies that such heavyweights as Visa, Wal-Mart, Mars or CNN are eliminated from the rankings. Only brands are taken in account (and not parent companies such as Procter and Gamble), and airlines are not ranked because it is too hard to separate their brand impact on sales from factors such as routes and schedules.Despite its limits, this ranking provides a global vision of the value of the main megabrands. This ranking has gained importance over the past years as a main reference for brand strategy. In addition, the assessment and evaluation method has not changed over the past five years. The rankings we refer to were published at the following dates: 6 August 2001/5 August 2002/4 August 2003/22 July 2004/21 July 2005.We present these five rankings in the Appendix. The first ranking refers to the period prior to the 09/11 events. We have, at the same time, conducted in-depth research into the question whether other factors may be responsible for the results we have found. Charts that summarize these findings are also presented in the Appendix, and - while it is certainly impossible to be exhaustive- exclude any major movements, evolutions, malaises or crises that could have the effects found (covering empirical research into size-trend and relative to industry, profitability-trend and relative to others in the industry, industry stage of life cycle, leverage-how vulnerable they are they to taking risks, country of origin to observe movements including characteristics such as access to capital, human resources, competition, and an index of insecurity, movements in scopes of megabrands (global reach, horizontal and vertical branding), or change in type of customers (e.g., services, package goods, durables, business); data chosen for illustration only covers main developments). By including such variables in the analysis, we strive to make it possible to determine that the risk hypothesis can be supported after controlling for other factors that lead to success, identifying other general factors that would lead to shift in megabrand positioning over time. No unexpected development in such has been found. The top 100 brands We first analyze the evolution of the one hundred top brands worldwide, in absolute value and in relative value.The results are summarized in Tables I and II:* The value of the 100 top brands increased by 5.7 percent, from 988.21 billion dollars to 1044.58 billion dollars in five years.* Amongst the 100 top brands worldwide, the total value of US brands declined from 737.55 billion dollars in August 2001 to 701.13 billion in July 2005, i.e. decreased by 4.9 percent. At the same time the value of non-US brands in this ranking increased from 250,66 billion to 343,45 billion (plus 37 percent). This evolution is far too important to be explained solely by currency and exchange rate conditions.* The number of US brands in this ranking declined from 63 brands to 52; while the number of non-US brands increased from 37 to 48 brands over the five -year period.Figure 2 illustrates the ratio of non-US compared to US brands in the top one hundred. This relative weight of the one to the other evolves dramatically from in the post - 09/11 era. It is important to note that the initial measure is dated July 2001, i.e. two months before 09/11.The top 20 brands We then focus on the top 20 brands of the ranking and their evolution.This focus is considered as especially important because these 20 top brands represent more than 50 percent of the global value of the whole 100 top worldwide brands. The data extracted here appear in Tables III and IV:* over the period examined, the number of non-US brands increase from four to seven brands and their value from 90.36 billions to 135.2 billions (plus 56.2 percent); and* the number of US brands decreases at the same time from 16 to 13 brands and their value decreases by 10.2 percent.Following this comparative analysis in the period studied, we also analyze the data of the US/non-US brand value ratio through a separation into zones, i.e. the top 20 US and the top 20 non-US brands: Now one is to pay special attention to the top ten, the top 20 and also the second/lower half of the top 20 brands.Table V shows that the lower half of the top twenty brands have, on the US side, suffered more than those in the top half. To the contrary, the bottom part of the top twenty also shows that non-US brands have very strongly increased their value.Top and bottom halves of the top 10 US and non-US brands We then analyze the value of the top ten US brands versus the value of the top ten non-US brands, as shown in Tables VI and VII: the value of the ten top US brands decreased by 5.25 percent while the value of the ten top non-US brands increases by 8.1 percent.The significance of the gap (i.e. the difference in value) between the top ten US and the ten top - non-US brands is illustrated in Figure 3. This figure is calculated via the total value as it evolves over time, i.e. dividing the total of 2002 by that of 2001, and so on for each year. The figure clearly demonstrates the increasing gap between the US and non-US megabrands over time, and the decrease of the curve because the gap strongly increases between 2002 and 2005 (by calculus of linear regression). Our initial assumption for this research was that international corporations adapted their brand marketing to globalization. We began by reviewing megabrand strategies that were put into effect over three decades, an option chosen by a wide range of companies to secure global, relatively easy and cost-efficient management of brands. We then raised the question of how megabrands evolved over the five years from 2000, with an objective to study the validity of this strategy through the analysis of the value evolution in the ensemble of megabrands worldwideThe data analysis provides strong empirical findings and raises an important set of questions: The value of US top brands worldwide declined significantly after 2001, and over the past rankings of world megabrands, while non-US brands experienced significant expansion over the same period. This evolution is confirmed on all three levels of analysis that we developed: total of 100 leading brands, total of the 20 leading brands, and comparison between the leading ten US and non-US brands.Why is the value gap more significant in the top 20 brands than in the top ten ones? Are second tier brands of this sample more vulnerable, and if so, why? The further we decrease the ranks of top brands in the top 100, the bigger the gap becomes between US and non-US brands, and this to the benefit of non-US ones. Are business cycle trends responsible for this trend? Are these brands are particularly symbolic in terms of nationality and risk perception since 2001 and global terrorism? Will consumers feel uncomfortable with certain brands since 9/11, and if so, what indications could allow us to understand this phenomenon? Is the dot.com burst responsible for this?With these questions in mind, further analysis provides the following indications: The following tables, one with the top 20 evolutions and one with the bottom 20 worst evolutions of brands, only refer to brands for which data is available for all five years.The findings indicate that in the top 20 "best evolution" international mega-brands only eight brands are US American and 12 are non-US, that the two best performers by value are non US (Samsung and Louis Vuitton), that Pepsi Co is the top US brand (interestingly it is the leading one in the US and in terms of brand name the competitor Coca-Cola is part of the bottom 20 brands, though still best known); that 2nd and 3rd best American mega-brands are Dell and Apple brands, and those who one could consider best known as Microsoft and Oracle are in the bottom 20. Does this mean that demand remained constant but strong US image made them fall? Due to the diversity of products and sectors represented, we believe that the dot.com bubble, highly sector-dependent, cannot be causal or solely causal to the megabrand evolutions that we note.Also, currency fluctuations in that time period would rather imply opposite effects.If thus the evolution of brands value over this five year period, of megabrands, is linked to brand nationality, and in this case that of US or non-US origins, this would imply that corporations need to invest in megabrands emanating from different regions. If one considers that US brands may be more sensitive to risk perceptions from global terrorism, by the consumer, than non-US brands, and that this terrorism could be a causal factor, because the data modifies after 2001, then the managerial objective is immunity to the consequences of such events. Given the crucial significance of such cause to strategy, we provide some basis for understanding and potentially resilience. Alexander et al. (1979, p. 4) define terrorism as "the systematic threat or use of violence to attain a political goal or communicate a political message through fear, coercion, or intimidation of particular persons or the general public". We can assume that the citizen and consumer in this general public is, therefore, exposed to stress scenarios that differs from typical scenarios, and therefore alter his or her purchasing behavior.It is widely admitted that with 9/11/2001, terrorism has become more global (Schneckener, 2002). 9/11-type terrorism is characterized by a proximity to western civilization and its psychological impact is reinforced through wide spread media coverage. Contemporary terrorist activities share a number of common features which are inter-related and of a recently resurrected nature: These features include the increasing link of terrorist activity to a quasi-legitimization on basis of allegedly religious motivation, modern business-like leadership structures, asymmetric warfare, and the use of the victim mostly as part of a communication strategy. The objectives of terrorists are to convey a triple message:1. Government is not capable of guaranteeing security of a society or citizen, nor of service or product safety.2. Corporations, investors and travelers are safe nowhere, and that symbols of a country, culture and society that they convey are potential targets for any type of attack.3. Any measure taken against terrorism is insufficient by nature.These messages have a powerful impact on many. Psychological effects (defined above as any of the extremes, from feeling guilt or irresponsible for temptation, especially in impulsive decisions to associating harm or risk to the brand, either associated to fear or sadness) instill uncertainty into the economy, and have been found elsewhere to significantly affect the economic, organizational and governmental environment (Suder, 2004). Given this, we adduce that consumer behavior and corporate strategy may be affected. For instance, just as in times of war, the consumer may adopt a "stocking/storing" behavior for particular types of food and medicines if he/she perceives a terror-based threat.Therefore, we hypothesized elsewhere that a firm's performance under uncertainty and risk of terrorism will be a function of its ability to reduce its vulnerability to terrorist acts through risk analysis and assessment, through shortened supply lines, and a decreased need for economic redundancy (Suder, 2006). This is even more so in the case of 9/11-type terrorism; a terrorism that has globalized and that hits the global activities of firms in addition to those at the location of a strike. In this section, we therefore focus on the question whether top management of megabrands should take into account a corporation's vulnerability to terrorist threat felt by consumers. If a brand has national symbolism - like Coca Cola -then its goods or services are exposed to threat or acts of terrorism. Will the consumer turn away from the brand, or in fact rather increase its faith in it? Our study could be interpreted to show possible link on a quantitative basis by comparative approach. In this case, is a megabrand strategy still a reasonable option?To be deemed reliable, enterprises must be able to keep their brands resilient in the event of a catastrophe. The US airlines that were victims to the hijacking of its planes crashed into the WTC in New York are the first illustration of the psychological impacts of brands that are related to terrorism threat. The symbolic relation to the events though entirely involuntary had dramatic consequences for bith American and United airlines. Also, a tendency of clients to rather fly shorter distances, on separate flights and with non-nationally related aircraft such as low-cost airlines emerged since 9/11 (MacBain, 2003, Tourism Queensland, 2006 et.al.).Markets melt down or freeze with great speed in case of threat or terrorism acts, other markets can rise because consider unrelated to the threat (Suder and Czinkota, 2007). Another example is the reluctance of Londoners to use public transport after the double-attacks of summer 2005; the bicycle-market however boomed almost immediately. The terrible human costs of terrorism are clearly unacceptable to any logic or ethics. Given that terrorism has existed in various forms over history, people, companies and industry now need be knowledgeable about 9/11 type threat and its impacts, and to adapt. International terrorism adds an important determinant to the definition of a firm's brand strategy. As an uncontrollable force in its external environment, terrorism events may lead to direct (mainly physical) or indirect (for instance consumer behavior and brand perception) disruptions. In the preliminary phase of threatened violence, or the following phase of the attack's aftermath (for details of this classification, see Suder, 2004), consumer demand for the firm's goods and services may alter but does not always decline (e.g. the demand for security equipment and services increases); any related disruption to the value chain perceivable by the consumer such as supply difficulties of needed inputs, resources, and services; or government policies and laws enacted to deal with terrorism alter the conduct of brand strategy. Macroeconomic phenomena, and shifts in international relations also modify behaviors. Media plays an important role in the intensification of the related psychological effects. For instance, the political differences between some European states and the USA in terms of the conduct of a war against terrorism, in particular concerning the invasion of Iraq, significantly modified consumer behaviors in the USA towards French and German brands (such as Roquefort cheese, Perrier water, ...and even French fries, solely based on their denomination). In those different dimensions, terrorism threat, act and aftermath affect:* ways of life;* perceptions;* consumption habits of millions of people all around the world; and* the company-client relationship.The responsiveness of consumers to a global threat is particularly high because it is intangible, close- by, and may strike anyone anywhere, in an expression of the "flatness" of the world. The incalculable uncertainty becomes a certainty that terror events happen and society and business adapt. The only certainty is that events will always be symbolic, whether that applies to locations, victims or the relation to the "hated" society. In this society anyone and anything can potentially identify with victims to attacks, whether human or object, whether a site, a product or a group. We therefore assume that this is so for brands in their dependence on perceptions and image.For a corporation, brand strategy and the administration of price shifts, communications, distribution strategies, buyers and suppliers, logistics, import and export are directly exposed to cultural issues, image responsibilities, and consequences of actions. For a consumer, brands have the particular capacity to link producers and consumers who trust in a specific set of quality, service and security "guarantees" linked psychologically to a particular brand. Brand marketing is symbolic and related on confidence, quite at the opposite of fear or panic. The consumer will hence turn to (or turn away) from brands in proportion to the strength that the brand relates to the threat, and expose brand strategy to risks non-related to their good performance. A brand is by definition the symbol of an object or a service, as well as a model of the consumption society (Keller, 1998). One major weakness of the megabrand approach is to expose the company to a major risk: A single brand, a single image. Needless to say, if a problem occurs with this brand the whole company's stability is at stake. But consumers are also citizens and so the brand may be a broader social and economic battleground amongst companies with respect to consumers. For example, brands also represent an important political space were virulent political battles can be fought (Semprini, 1992). Some movements embody or oppose lifestyles symbolised by brands and their influence, sometimes in a very radical way, the consumer's society becomes represented by companies and their brands (Klein, 2002). This contesting opposition must be taken into consideration when developing brands and their territories in order to avoid vulnerability of a single-brand strategy and extreme exposure. Various authors already tackled this notion under the theme of brand capital or brand equity (Farquhar, 1989; Baldinger, 1990; Kapferer, 1991; Aaker, 1992; Keller, 1998). For Aaker (1992), brand capital is a unit consisting of the name and symbolic meaning of a brand that can add or decrease the value of a product or service, and that delivers value to the client and to the firm. An appropriate strategy this reinforces the value of brands while an inappropriate strategy diminishes the value. On basis of our findings and the exposed nexus that one may establish with 9/11, we suggest that megabrand strategy allows corporations to obtain critical size (specially facing the distribution channels), face the growth limits of existing brands, share, soften, and pool certain costs (research, industrialization, marketing) although the megabrand building process is time related and based on a variety of experiences. One can hence suggest that, if here lies the causal link, in post-9/11 megabrands allow for better control of risks by the company and increase the value of brands better if locally or regionally embedded. If megabrand strategy overexposes brands as symbolic for a mode of consumption rejected by or associated to terrorism, then megabrand overexposure diminishes the value of brands, by overexposing firms to risks, band devaluation and increasing company vulnerability. The findings of this research imply that brand strategy is highly dependent on exterior factors and need to be adapted to those if competitive advantages shall not erode and shift considerably. While the causal link to 9/11 terrorism can not be clearly established, it does appear as one of the sensible explanations or co-factors for the dramatic evolution that were found. These findings in themselves hope to make a contribution to the understanding of megabrand strategy in mature globalization. It appears from our research that brand nationality, and thus brand associations to the various effects of terrorism (victimization or identification), may define the behavior of consumers and have an impact on brands value and ranking. For a future that may have to cope with 9/11-type terrorism, megabrands (except for the very strongest ones perhaps) may therefore not qualify as the best option for companies that wish to reduce risk and immunize brands and performances. If this is confirmed, the firms are well-advised to invest in mega-brands anchored into regions, through a transnational rather than a global strategy.Clearly further research into the potential causalities is needed: Whether terrorism, business cycles, currency issues, the bubble effect, all of these events united or none, international business scholars and practitioners are advised to study these links together, in each sector and market so as improve understandings and capabilities to respond appropriately to the evolution of megabrands in ranking and value since 2001. Opens in a new window.Figure 1 Contradictory tendencies in brand development Opens in a new window.Figure 2 Ratio of non-US compared to US brands in the top 100 Opens in a new window.Figure 3 Gap (i.e. the difference in value) between the top ten US and the ten top non-US brands; base 100% in 2001 Opens in a new window.Table I Evolution of the 100 biggest worldwide brands and split between US and non-US brands Opens in a new window.Table II Evolution of the 100 biggest worldwide brands and split between US and non-US brands Opens in a new window.Table IV Evolution of the 20 first worldwide brands and split between US and non-US brands Opens in a new window.Table III Evolution of the 20 first worldwide brands and split between US and non-US brands Opens in a new window.Table V Detailed analysis of gaps Opens in a new window.Table VII Evolution of the ten biggest US brands and of the ten biggest non-US brands Opens in a new window.Table VI The top ten US brands Opens in a new window.Table AI Inderbrand seven-criteria ranking Opens in a new window.Table AII The five relevant megabrand rankings. 2001-2005 Opens in a new window.Table AIII Exclusion of other potentially relevant causal factors. Analysis of the brands' industry in and out Opens in a new window.Table AIV Exclusion of other potentially relevant causal factors. Brands in and out linked to industry and sector Opens in a new window.Table AV Exclusion of other potentially relevant causal factors. Analysis of significant changes (technology sector + financial), and performance Opens in a new window.Table AVI Country ranking by risk category (Coface, 2006)
|
- The key findings reveal that, in the five consecutive years after 9/11/2001, brands have experienced significant moderation in rank and value. A significant gap in the evolution of US and non-US brands was found in this period of time. The evidence calls for brand management that reflects the risks that globalized at the same pace as brand reach.
|
[SECTION: Value] An executive summary for managers and executive readers can be found at the end of this article. The objective of this research article is to shed light on the evolution of brand management into a crucial strategic tool for international business operations. On basis of the literature available in this field, we analyze the largest 100 brands (hereafter categorized as megabrands) in terms of ranking and value modifications in the 2001 to 2005 period, a mature globalization period, with the first ranking referring to pre-09/11 findings. The sample and its analysis provide us with significant findings that open crucial questions about US/non-US brand strategy and perceptions, and the future application of global megabrand policies. We then shed light on the causal factors that global terrorism may contain and tentatively propose brand strategy solutions, but do not exclude other causal factors or co-factors that will need further inquiry. Overall, our hypothesis is that brands serve to bring security[1]. Accordingly, if the source of that brand is less secure, then it will be less effective as a brand. This hypothesis needs to be qualified: In particular, would one not expect that the short run reaction to insecurity is to be more, rather than less brand loyalty? The findings of this study have a strong indication that this assumption can be reversed, and we indicate that indirect impacts of global terrorism might be the reason. Further, it is important to show that negative security shifts in the US have been greater than a general increase in malaise in the global markets where the brands are sold. Again, the data indicates that movements in brand value and ranking appear to respond to more than such a malaise.Being a simple "identification tool" at its very start, brand names have become a critical part of a company's strategy. Academic research has shown that one major historic reason for brand success is the diminished risk perceived by the consumer (Roselius, 1971; Kapferer, 1991; Keller, 1998; Riezebos, 2003). McCarthy (1971) highlights the three primary roles of brand:* identification and purchase simplification function;* brand has a projective, symbolic and imaginary function and provides the consumer with a status; and* brand guarantees quality, protection and risk reduction for the consumer by pointing out to its source.For these reasons, companies are willing to consider brands as an important asset of their balance sheets or to invest huge amounts of capital to buy them (Laforet and Saunders, 1994).The power of brands is founded on consumers' aversion to uncertainty. For a long time, consumers made their food buying decisions based only on a product's visual aspect, ignoring its brand name, accepting instead the grocery store owner's opinion as selection criteria (Boyer, 2002). Later on producers introduced clearly visible signals that identified their products and consumers then got used to preferring the signal as opposed to the product visual characteristics (Keller, 1998); that is, brand became more important than the product itself (Riezebos, 2003).Even at the present, perceived risk reduction is the first reason consumers have for choosing a brand and this guides brand management evolution (Kapferer, 2003). When consumers perceive a risk in making a buying decision, they will deploy different strategies for reducing it. Five major risks are considered by consumers:1. Financial risk ("making a bad deal", which increases the importance of the brand compared with the unit price of the product).2. Physical risk (being harmed by the product, especially food products).3. Technological risk (being disappointed by the product performance, it is the risk of functionality).4. Psychological risk (feeling guilty or irresponsible for temptation, especially in impulsive decisions - or associating harm or risk to the brand, either associated to fear or sadness).5. Social risk (what pairs will say or think about choices. Therefore brand is a sign of possession for a community, but also a sign of adherence, of patriotism or of association to or away from particular social issues.)Risk reduction function directly related to the brand has been increased by the macro-economic context, especially after 09/11 because a fragile and complex environment is expected to increase the role the brand has to play in reassuring the consumers' buying decisions. Even though, we later argue that the capacity of brands to link producers and consumers has been rudely challenged. There have been drastic changes of consumption habits in some markets, such as the acceleration of the coming on-stream of the hard-discounters in Europe, with a new approach to the quality - price relationship and the weakening of the brand, of low-cost airlines, and of non-brand textiles from low-cost production. Companies have reacted to these new challenges. This new environment has notably changed the way in which big international companies conceive of their brands. Brand guarantee and its image are shelter points for consumers: normally, the higher the risk the more helpful the brand. Consequently, brands have learned a different way of communication (e.g. emphasizing safety themes, as carmakers do already), to change their relationship with the environment or towards the Third World (e.g., Nike reconsidering its production policy in order to improve its brand image) but also with globalization (e.g. being more respectful of local brands, as Nestle is). Brands also start working on ethical matters (The Body Shop's cosmetics products), fair trade (Malongo coffee) or social responsibility. But one of the major facets of this adaptation of brands and firms to the new situation is the coming on-stream and acceleration of megabrands within companies. Traditionally, choosing brand strategies is the focal point for companies, whether they are multinational groups or local companies (Schuiling and Kapferer, 2004). Supposing that a firm has different sources of competition, one of the strategic issues is whether it uses one or several brands. Strebinger (2002) states that one of the most critical problems in branding relates to the management of a mono or multi-brand system while Riezebos (2003) questions whether it is feasible having just a single-brand strategy in the company, with a prime focus on one brand and then developing additional brands from it.The historical development of branding includes some deeply contradictory factors, as shown in Figure 1.This figure visualizes and conceptualizes a company's willingness and need to have numerous products able to meet the different customers' demands as appropriate as possible, to assure their expansion and international development, that is, to counteract all risk of being a single-brand company. Likewise, there is a need for limiting the number of brands because of a second risk: that of a brand overexposure or over usage, including the financial risk of dispersing the investment.The first risk leads companies wishing to develop to buy or launch more brands in order to enter markets, segments or customers inaccessible with only one brand. This may be an "inflationist" process in terms of markets as it leads to create many brands.The second risk takes the same companies in the opposite direction, trying to limit the number of brands in order to maximize investment per brand, thereby making the brands stronger and covering more territory.But this process is intrinsically schizophrenic and raises the question of the strategic equilibrium of branding (Riezebos, 2003). Strategic choices may become brand choices, choices of brand organization or choices about the kind of relationship between brands that a company wants to maintain. One of the purposes of these choices is to maximize the equity of its different brands.As a way to escape from this process, many companies turn to megabranding.At its origins, the evolution of the brand universe towards megabrands comes from big corporations that discovered, in the early 80s, that they could create value by capitalizing on the transnational concepts carried in supranational brands so as to attain maximum return on investment (Kapferer, 2000). This new strategy reduced internal brand management costs and the costs of launching new innovative products. This simple idea has allowed many companies to focus on the strongest brands, or on brands with high growth potential or on highly internationalized brands, and to abandon or minimize all others. Indeed, at the beginning, economic reasons were the main inspiration for this rationalization process: first of all trying to concentrate all human and economic resources on a few brands and, especially, cutting advertising costs related to the launching and maintenance process of multiple brands.The megabrand concept, thus, is a core concern for most leading transnational firms because, as the competitive environment becomes more and more complex, and with a high level of risks of every nature, companies focus on megabrand strategy and attempt to assure their expansion and international development.In the early 1990s many companies did inform the market of their intentions to reduce their brand numbers: The most extreme case being that of Unilever, which planned to reduce from 1,600 down to 400 brands in the 2000-2004 period. Anthony Simon[2], President of Unilever-BestFoods marketing, underlined that "Unilever's objective is to reduce the number of brands in order to make them stronger. Four strategies support this decision: category, segment, channel and geography".In a megabrand strategy, a brand name may be used for horizontal extensions (inside the same price layer, common for mass consumption products) or vertical extensions (in different price layers, common for durable goods). This strategy can be very successful; a well-developed brand can provide a sustainable competitive advantage. To ensure continuous success, the operation of a megabrand strategy demands permanent innovation, strong R&D investment, a communicational style hard to imitate and a brand image not based on the product but on associations and perceptions.Megabrand management changes the focus of marketing to a superior, strategic decision-making level (Baldinger, 1990; Trinquecoste, 1999), as it implicitly involves focusing on the whole company instead of on individual brands (Riezebos, 2003). Both, Juga (1999) and Reynaud (2001) show that by displacing competition to this superior level, competitive advantages become harder to understand (less tangible) and to imitate.The increasing recognition of brands as a source of sustainable competitive advantage stresses the importance of conceptual models about organizational brand strategies (Louro and Cunha, 2001). Therefore, our research goal is to explore the megabranding field and to evaluate its strategic dimension as a new and more complex and durable source of competitive advantage in times of international adversity and the challenges of 09/11-type terrorism. We have chosen to analyze the evolution of the value of megabrands over a five year period. The sample consists of those brands ranked in "The 100 best global brands" annually by Interbrand corporation for Business Week magazine.Interbrand defined seven criteria (see Appendix) which evaluate brands much in the way analysts value other assets, i.e. on the basis of how much they are likely to earn in the future.To qualify for the list each brand must:* have a value greater than $1 billion;* derive about a third of its earnings outside its home country; and* have publicly available marketing and financial data.For these reasons Interbrand specifies that such heavyweights as Visa, Wal-Mart, Mars or CNN are eliminated from the rankings. Only brands are taken in account (and not parent companies such as Procter and Gamble), and airlines are not ranked because it is too hard to separate their brand impact on sales from factors such as routes and schedules.Despite its limits, this ranking provides a global vision of the value of the main megabrands. This ranking has gained importance over the past years as a main reference for brand strategy. In addition, the assessment and evaluation method has not changed over the past five years. The rankings we refer to were published at the following dates: 6 August 2001/5 August 2002/4 August 2003/22 July 2004/21 July 2005.We present these five rankings in the Appendix. The first ranking refers to the period prior to the 09/11 events. We have, at the same time, conducted in-depth research into the question whether other factors may be responsible for the results we have found. Charts that summarize these findings are also presented in the Appendix, and - while it is certainly impossible to be exhaustive- exclude any major movements, evolutions, malaises or crises that could have the effects found (covering empirical research into size-trend and relative to industry, profitability-trend and relative to others in the industry, industry stage of life cycle, leverage-how vulnerable they are they to taking risks, country of origin to observe movements including characteristics such as access to capital, human resources, competition, and an index of insecurity, movements in scopes of megabrands (global reach, horizontal and vertical branding), or change in type of customers (e.g., services, package goods, durables, business); data chosen for illustration only covers main developments). By including such variables in the analysis, we strive to make it possible to determine that the risk hypothesis can be supported after controlling for other factors that lead to success, identifying other general factors that would lead to shift in megabrand positioning over time. No unexpected development in such has been found. The top 100 brands We first analyze the evolution of the one hundred top brands worldwide, in absolute value and in relative value.The results are summarized in Tables I and II:* The value of the 100 top brands increased by 5.7 percent, from 988.21 billion dollars to 1044.58 billion dollars in five years.* Amongst the 100 top brands worldwide, the total value of US brands declined from 737.55 billion dollars in August 2001 to 701.13 billion in July 2005, i.e. decreased by 4.9 percent. At the same time the value of non-US brands in this ranking increased from 250,66 billion to 343,45 billion (plus 37 percent). This evolution is far too important to be explained solely by currency and exchange rate conditions.* The number of US brands in this ranking declined from 63 brands to 52; while the number of non-US brands increased from 37 to 48 brands over the five -year period.Figure 2 illustrates the ratio of non-US compared to US brands in the top one hundred. This relative weight of the one to the other evolves dramatically from in the post - 09/11 era. It is important to note that the initial measure is dated July 2001, i.e. two months before 09/11.The top 20 brands We then focus on the top 20 brands of the ranking and their evolution.This focus is considered as especially important because these 20 top brands represent more than 50 percent of the global value of the whole 100 top worldwide brands. The data extracted here appear in Tables III and IV:* over the period examined, the number of non-US brands increase from four to seven brands and their value from 90.36 billions to 135.2 billions (plus 56.2 percent); and* the number of US brands decreases at the same time from 16 to 13 brands and their value decreases by 10.2 percent.Following this comparative analysis in the period studied, we also analyze the data of the US/non-US brand value ratio through a separation into zones, i.e. the top 20 US and the top 20 non-US brands: Now one is to pay special attention to the top ten, the top 20 and also the second/lower half of the top 20 brands.Table V shows that the lower half of the top twenty brands have, on the US side, suffered more than those in the top half. To the contrary, the bottom part of the top twenty also shows that non-US brands have very strongly increased their value.Top and bottom halves of the top 10 US and non-US brands We then analyze the value of the top ten US brands versus the value of the top ten non-US brands, as shown in Tables VI and VII: the value of the ten top US brands decreased by 5.25 percent while the value of the ten top non-US brands increases by 8.1 percent.The significance of the gap (i.e. the difference in value) between the top ten US and the ten top - non-US brands is illustrated in Figure 3. This figure is calculated via the total value as it evolves over time, i.e. dividing the total of 2002 by that of 2001, and so on for each year. The figure clearly demonstrates the increasing gap between the US and non-US megabrands over time, and the decrease of the curve because the gap strongly increases between 2002 and 2005 (by calculus of linear regression). Our initial assumption for this research was that international corporations adapted their brand marketing to globalization. We began by reviewing megabrand strategies that were put into effect over three decades, an option chosen by a wide range of companies to secure global, relatively easy and cost-efficient management of brands. We then raised the question of how megabrands evolved over the five years from 2000, with an objective to study the validity of this strategy through the analysis of the value evolution in the ensemble of megabrands worldwideThe data analysis provides strong empirical findings and raises an important set of questions: The value of US top brands worldwide declined significantly after 2001, and over the past rankings of world megabrands, while non-US brands experienced significant expansion over the same period. This evolution is confirmed on all three levels of analysis that we developed: total of 100 leading brands, total of the 20 leading brands, and comparison between the leading ten US and non-US brands.Why is the value gap more significant in the top 20 brands than in the top ten ones? Are second tier brands of this sample more vulnerable, and if so, why? The further we decrease the ranks of top brands in the top 100, the bigger the gap becomes between US and non-US brands, and this to the benefit of non-US ones. Are business cycle trends responsible for this trend? Are these brands are particularly symbolic in terms of nationality and risk perception since 2001 and global terrorism? Will consumers feel uncomfortable with certain brands since 9/11, and if so, what indications could allow us to understand this phenomenon? Is the dot.com burst responsible for this?With these questions in mind, further analysis provides the following indications: The following tables, one with the top 20 evolutions and one with the bottom 20 worst evolutions of brands, only refer to brands for which data is available for all five years.The findings indicate that in the top 20 "best evolution" international mega-brands only eight brands are US American and 12 are non-US, that the two best performers by value are non US (Samsung and Louis Vuitton), that Pepsi Co is the top US brand (interestingly it is the leading one in the US and in terms of brand name the competitor Coca-Cola is part of the bottom 20 brands, though still best known); that 2nd and 3rd best American mega-brands are Dell and Apple brands, and those who one could consider best known as Microsoft and Oracle are in the bottom 20. Does this mean that demand remained constant but strong US image made them fall? Due to the diversity of products and sectors represented, we believe that the dot.com bubble, highly sector-dependent, cannot be causal or solely causal to the megabrand evolutions that we note.Also, currency fluctuations in that time period would rather imply opposite effects.If thus the evolution of brands value over this five year period, of megabrands, is linked to brand nationality, and in this case that of US or non-US origins, this would imply that corporations need to invest in megabrands emanating from different regions. If one considers that US brands may be more sensitive to risk perceptions from global terrorism, by the consumer, than non-US brands, and that this terrorism could be a causal factor, because the data modifies after 2001, then the managerial objective is immunity to the consequences of such events. Given the crucial significance of such cause to strategy, we provide some basis for understanding and potentially resilience. Alexander et al. (1979, p. 4) define terrorism as "the systematic threat or use of violence to attain a political goal or communicate a political message through fear, coercion, or intimidation of particular persons or the general public". We can assume that the citizen and consumer in this general public is, therefore, exposed to stress scenarios that differs from typical scenarios, and therefore alter his or her purchasing behavior.It is widely admitted that with 9/11/2001, terrorism has become more global (Schneckener, 2002). 9/11-type terrorism is characterized by a proximity to western civilization and its psychological impact is reinforced through wide spread media coverage. Contemporary terrorist activities share a number of common features which are inter-related and of a recently resurrected nature: These features include the increasing link of terrorist activity to a quasi-legitimization on basis of allegedly religious motivation, modern business-like leadership structures, asymmetric warfare, and the use of the victim mostly as part of a communication strategy. The objectives of terrorists are to convey a triple message:1. Government is not capable of guaranteeing security of a society or citizen, nor of service or product safety.2. Corporations, investors and travelers are safe nowhere, and that symbols of a country, culture and society that they convey are potential targets for any type of attack.3. Any measure taken against terrorism is insufficient by nature.These messages have a powerful impact on many. Psychological effects (defined above as any of the extremes, from feeling guilt or irresponsible for temptation, especially in impulsive decisions to associating harm or risk to the brand, either associated to fear or sadness) instill uncertainty into the economy, and have been found elsewhere to significantly affect the economic, organizational and governmental environment (Suder, 2004). Given this, we adduce that consumer behavior and corporate strategy may be affected. For instance, just as in times of war, the consumer may adopt a "stocking/storing" behavior for particular types of food and medicines if he/she perceives a terror-based threat.Therefore, we hypothesized elsewhere that a firm's performance under uncertainty and risk of terrorism will be a function of its ability to reduce its vulnerability to terrorist acts through risk analysis and assessment, through shortened supply lines, and a decreased need for economic redundancy (Suder, 2006). This is even more so in the case of 9/11-type terrorism; a terrorism that has globalized and that hits the global activities of firms in addition to those at the location of a strike. In this section, we therefore focus on the question whether top management of megabrands should take into account a corporation's vulnerability to terrorist threat felt by consumers. If a brand has national symbolism - like Coca Cola -then its goods or services are exposed to threat or acts of terrorism. Will the consumer turn away from the brand, or in fact rather increase its faith in it? Our study could be interpreted to show possible link on a quantitative basis by comparative approach. In this case, is a megabrand strategy still a reasonable option?To be deemed reliable, enterprises must be able to keep their brands resilient in the event of a catastrophe. The US airlines that were victims to the hijacking of its planes crashed into the WTC in New York are the first illustration of the psychological impacts of brands that are related to terrorism threat. The symbolic relation to the events though entirely involuntary had dramatic consequences for bith American and United airlines. Also, a tendency of clients to rather fly shorter distances, on separate flights and with non-nationally related aircraft such as low-cost airlines emerged since 9/11 (MacBain, 2003, Tourism Queensland, 2006 et.al.).Markets melt down or freeze with great speed in case of threat or terrorism acts, other markets can rise because consider unrelated to the threat (Suder and Czinkota, 2007). Another example is the reluctance of Londoners to use public transport after the double-attacks of summer 2005; the bicycle-market however boomed almost immediately. The terrible human costs of terrorism are clearly unacceptable to any logic or ethics. Given that terrorism has existed in various forms over history, people, companies and industry now need be knowledgeable about 9/11 type threat and its impacts, and to adapt. International terrorism adds an important determinant to the definition of a firm's brand strategy. As an uncontrollable force in its external environment, terrorism events may lead to direct (mainly physical) or indirect (for instance consumer behavior and brand perception) disruptions. In the preliminary phase of threatened violence, or the following phase of the attack's aftermath (for details of this classification, see Suder, 2004), consumer demand for the firm's goods and services may alter but does not always decline (e.g. the demand for security equipment and services increases); any related disruption to the value chain perceivable by the consumer such as supply difficulties of needed inputs, resources, and services; or government policies and laws enacted to deal with terrorism alter the conduct of brand strategy. Macroeconomic phenomena, and shifts in international relations also modify behaviors. Media plays an important role in the intensification of the related psychological effects. For instance, the political differences between some European states and the USA in terms of the conduct of a war against terrorism, in particular concerning the invasion of Iraq, significantly modified consumer behaviors in the USA towards French and German brands (such as Roquefort cheese, Perrier water, ...and even French fries, solely based on their denomination). In those different dimensions, terrorism threat, act and aftermath affect:* ways of life;* perceptions;* consumption habits of millions of people all around the world; and* the company-client relationship.The responsiveness of consumers to a global threat is particularly high because it is intangible, close- by, and may strike anyone anywhere, in an expression of the "flatness" of the world. The incalculable uncertainty becomes a certainty that terror events happen and society and business adapt. The only certainty is that events will always be symbolic, whether that applies to locations, victims or the relation to the "hated" society. In this society anyone and anything can potentially identify with victims to attacks, whether human or object, whether a site, a product or a group. We therefore assume that this is so for brands in their dependence on perceptions and image.For a corporation, brand strategy and the administration of price shifts, communications, distribution strategies, buyers and suppliers, logistics, import and export are directly exposed to cultural issues, image responsibilities, and consequences of actions. For a consumer, brands have the particular capacity to link producers and consumers who trust in a specific set of quality, service and security "guarantees" linked psychologically to a particular brand. Brand marketing is symbolic and related on confidence, quite at the opposite of fear or panic. The consumer will hence turn to (or turn away) from brands in proportion to the strength that the brand relates to the threat, and expose brand strategy to risks non-related to their good performance. A brand is by definition the symbol of an object or a service, as well as a model of the consumption society (Keller, 1998). One major weakness of the megabrand approach is to expose the company to a major risk: A single brand, a single image. Needless to say, if a problem occurs with this brand the whole company's stability is at stake. But consumers are also citizens and so the brand may be a broader social and economic battleground amongst companies with respect to consumers. For example, brands also represent an important political space were virulent political battles can be fought (Semprini, 1992). Some movements embody or oppose lifestyles symbolised by brands and their influence, sometimes in a very radical way, the consumer's society becomes represented by companies and their brands (Klein, 2002). This contesting opposition must be taken into consideration when developing brands and their territories in order to avoid vulnerability of a single-brand strategy and extreme exposure. Various authors already tackled this notion under the theme of brand capital or brand equity (Farquhar, 1989; Baldinger, 1990; Kapferer, 1991; Aaker, 1992; Keller, 1998). For Aaker (1992), brand capital is a unit consisting of the name and symbolic meaning of a brand that can add or decrease the value of a product or service, and that delivers value to the client and to the firm. An appropriate strategy this reinforces the value of brands while an inappropriate strategy diminishes the value. On basis of our findings and the exposed nexus that one may establish with 9/11, we suggest that megabrand strategy allows corporations to obtain critical size (specially facing the distribution channels), face the growth limits of existing brands, share, soften, and pool certain costs (research, industrialization, marketing) although the megabrand building process is time related and based on a variety of experiences. One can hence suggest that, if here lies the causal link, in post-9/11 megabrands allow for better control of risks by the company and increase the value of brands better if locally or regionally embedded. If megabrand strategy overexposes brands as symbolic for a mode of consumption rejected by or associated to terrorism, then megabrand overexposure diminishes the value of brands, by overexposing firms to risks, band devaluation and increasing company vulnerability. The findings of this research imply that brand strategy is highly dependent on exterior factors and need to be adapted to those if competitive advantages shall not erode and shift considerably. While the causal link to 9/11 terrorism can not be clearly established, it does appear as one of the sensible explanations or co-factors for the dramatic evolution that were found. These findings in themselves hope to make a contribution to the understanding of megabrand strategy in mature globalization. It appears from our research that brand nationality, and thus brand associations to the various effects of terrorism (victimization or identification), may define the behavior of consumers and have an impact on brands value and ranking. For a future that may have to cope with 9/11-type terrorism, megabrands (except for the very strongest ones perhaps) may therefore not qualify as the best option for companies that wish to reduce risk and immunize brands and performances. If this is confirmed, the firms are well-advised to invest in mega-brands anchored into regions, through a transnational rather than a global strategy.Clearly further research into the potential causalities is needed: Whether terrorism, business cycles, currency issues, the bubble effect, all of these events united or none, international business scholars and practitioners are advised to study these links together, in each sector and market so as improve understandings and capabilities to respond appropriately to the evolution of megabrands in ranking and value since 2001. Opens in a new window.Figure 1 Contradictory tendencies in brand development Opens in a new window.Figure 2 Ratio of non-US compared to US brands in the top 100 Opens in a new window.Figure 3 Gap (i.e. the difference in value) between the top ten US and the ten top non-US brands; base 100% in 2001 Opens in a new window.Table I Evolution of the 100 biggest worldwide brands and split between US and non-US brands Opens in a new window.Table II Evolution of the 100 biggest worldwide brands and split between US and non-US brands Opens in a new window.Table IV Evolution of the 20 first worldwide brands and split between US and non-US brands Opens in a new window.Table III Evolution of the 20 first worldwide brands and split between US and non-US brands Opens in a new window.Table V Detailed analysis of gaps Opens in a new window.Table VII Evolution of the ten biggest US brands and of the ten biggest non-US brands Opens in a new window.Table VI The top ten US brands Opens in a new window.Table AI Inderbrand seven-criteria ranking Opens in a new window.Table AII The five relevant megabrand rankings. 2001-2005 Opens in a new window.Table AIII Exclusion of other potentially relevant causal factors. Analysis of the brands' industry in and out Opens in a new window.Table AIV Exclusion of other potentially relevant causal factors. Brands in and out linked to industry and sector Opens in a new window.Table AV Exclusion of other potentially relevant causal factors. Analysis of significant changes (technology sector + financial), and performance Opens in a new window.Table AVI Country ranking by risk category (Coface, 2006)
|
- The primary purpose of this study is to identify if and how international terrorism has altered the rank and value of brands, and whether the increasing uncertainty of globalizing risks need an adaptation of international brand management.
|
[SECTION: Purpose] Within the past few years, the phenomenon of social networking web sites (SNWs) on the internet has exploded into the mainstream. Further, this online information has begun to be used for purposes beyond its intended use. Owing to the vast amount of personal information on these web sites, employers have begun to tap into this information as a source of applicant data in an effort to improve hiring decisions. This study evaluates the use of the SNWs in employment selection. Specifically, can trained judges consistently and accurately assess important organizational characteristics such as personality, intelligence, and performance using only a target's SNW information? In addition, the use of this information may lead to discrimination against applicants, given the wide range of available personal information such as gender, race, age, religion, and disability status otherwise illegal to use when making employment decisions.Social networking web sites SNWs focus on building online communities of people who share interests and activities, or who are interested in exploring the interests and activities of others. Most provide a variety of ways for users to interact, such as e-mail and instant messaging services. SNWs are designed to connect users to each other and to visually display each individual's network of friends. The number of users for these web sites and the daily traffic created by these web sites are staggering. According to the "About Us" section of the various sites, MySpace is the largest with over 248 million registered users. Other SNWs also have millions of users registered, such as Facebook (110 million), Friendster (85 million), Hi5 (80 million), Orkut/Google (37 million), and LinkedIn (25 million).While these sites differ in the features that are available, most have a mechanism for posting pictures, music and videos, keeping blogs, sharing links, and displaying interests. These sites vary in user demographics. For example, although Facebook is currently open to anyone, it started as a high school and college web site exclusively, with about 90 percent of students registered for the site (van der Werf, 2006).Social networking web sites in selection Owing to the increasing prevalence of SNWs in conjunction with the large volume of information available to the viewer, employers have begun using SNWs to assist in the selection process for new employees. About 50 percent of the employers attending college career fairs use online technology, including both search engines and SNWs to screen candidates (Shea and Wesley, 2006). Currently, between 20 (Framingham, 2008) and 25 percent (Taylor, 2007; NACE, 2006) of managers are using SNWs to screen candidates, with 40 percent indicating that they are likely to use them within the next year (Zeidner, 2007). These employers apparently feel justified in electronic screening using SNWs. The president of a small Chicago consulting firm, when asked about an applicant that applied for an internship and had questionable content on his page, replied, "A lot of it makes me think, what kind of judgment does this person have? Why are you allowing this to be viewed publicly?" (Finder, 2006).While it may be common practice to monitor web site content it may not be legal. Lance Chou, Director of Career Development at Stanford University, noted that:[...] some employers might try and learn something about the student's personality and whether it would be appropriate for the job. However, there is information on Facebook that is not relevant to the job, but may be used inappropriately by employers to assess a candidate (Fuller, 2006).According to George Lenard, an employment attorney, employers can inadvertently learn about matters such as candidates' age, marital status, and other topics typically are off limits in job interviews, and organizations can be sued for discrimination if these candidates are not hired (Frauenheim, 2006). The question of whether employer monitoring of SNWs is illegal may relate to equal employment opportunity (EEO) law. EEO severely limits the type of information an interviewer may ascertain and use. For example, during interviews employers may not ask questions regarding race, religion, sexual preference, or marital status. However, all of this information can be easily located using SNWs (Kowske and Southwell, 2006). The ethics and legality behind electronic screening using SNWs has become an important issue for human resource (HR) practitioners (Zeidner, 2007).Since no academic studies, to our knowledge, have assessed whether using SNWs in employment selection are a reliable and valid predictor of important organizational outcomes, their value in employment screening is unknown. Until the reliability and validity of information on SNWs is established, employers should use caution when using SNWs to make hiring decisions. Just as with structured interview methods (Campion et al., 1997), focusing on job-related information in SNWs should minimize the use of less job-relevant information that might bias the hiring decision. This study focuses on showing the validity and reliability of three such categories of job-relevant characteristics which have potential to be rated consistently and accurately and to serve as valid predictors of job performance. The assessment of personality, intelligence, and a global measure of performance through SNWs provide multiple possible avenues regarding validity generalization of job-relevant characteristics, and thereby justifying the use of SNWs in the screening process.Assessment of personality The 1990s have seen a huge growth in the use of personality assessment within personnel selection practice and research (Barrick and Mount, 1991; Frei and McDaniel, 1997; Ones et al., 1993; Salgado, 1998; Tett et al., 1991). These studies provide positive evidence for the criterion-related validity of personality. When it comes to the prediction of overall job performance, conscientiousness was found to be the best predictor, showing consistent predictions across all occupational groups. In addition, extroversion, agreeableness, and neuroticism were shown to predict job performance in certain jobs. Finally, although typically unrelated to job performance, openness to experience has been found to predict training performance.Beyond the predominant focus on self-reported personality assessment, personality can also be measured using other-ratings. It is well established that people can assess the personality of others, even after relatively short exposure, but that the accuracy of these assessments depends on the information available to the observer (Barrick et al., 2000). Observer ratings may be more valuable than self-ratings in employment selection, particularly when targets are not available for self-reports, self-reports are untrustworthy, and researchers wish to improve accuracy by aggregating multiple raters (Hofstee, 1994; McCrae and Weiss, 2007). Other-rated personality via SNWs seems quite promising in the selection context, since other-ratings of personality have been found to predict job performance. Motowidlo and colleagues (1996) found that interviewer-rated extroversion and conscientiousness (r=0.27 and 0.20, respectively) significantly predict supervisor rated job performance. In addition, Mount et al. (1994) show that observer ratings of extroversion and conscientiousness from supervisors, coworkers, and customers significantly predicts sales performance, even beyond self-rated personality.The validity of other-rated personality, however, can depend on the relationship the subject has with the rater and the quality of the information available to the rater. A meta analysis conducted by Connolly and Viswesvaran (1998) show low accuracy for strangers in predicting personality and moderate prediction by peers for each of the big-five traits. In addition, Barrick et al. (2000) developed a personality-based job interview for the purpose of assessing the personality of the applicant. They found that personality-based job interviews could be used to accurately predict three of the big-five. Considering the average interview is approximately 40 minutes in length (Campion et al., 1997), it appears that an interviewer can assess some aspects of personality as effectively as a close acquaintance of the applicant. We propose that the information available in SNWs provides a similar means to assess personality.Furthermore, since SNW ratings are obtained from a wide range of personal information that is a reflection of ongoing behaviors and interactions with other users of the networks, web sites may actually provide unique information not found with other selection methods. Recent issues of Personnel Psychology (Morgeson et al., 2007a, b; Ones et al., 2007; Tett and Christiansen, 2007) and Industrial and Organizational Psychology (Hough and Oswald, 2008; Griffith and Peterson, 2008) have focused on the limited criterion validity of self-reported personality measures as well as the complex issue of social desirability and faking of self-reported personality measures. To increase validity and address issues of socially desirable responding "we should look at other ways of assessing personality. There are a variety of ways of finding out about people's stable pattern of behavior" (Morgeson et al., 2007a, b, p. 719). SNWs may prove to be an appropriate means of assessing personality in this way. Beyond personality assessment, similar issues emerge in relation to a host of employment selection methods. Resumes, interviews, job applications, and many other forms of employee selection include a certain element of self-presentation, reflecting "maximal" instead of "typical" work performance (Sackett, 2007; Sackett et al., 1988). Employment selection methods using social networking are likely to be based on "typical" behaviors, and therefore may be more accurate than other selection methods. At a minimum, this method should provide information that is distinct from "maximal" selection methods, thereby allowing for a stronger likelihood that using SNWs will yield incremental validity beyond established methods of employment selection[1].This is not to say that SNWs are not susceptible to manipulation and faking, similar to that of self-report personality measures and job interviews. In fact, as users of SNWs become more aware that their profiles are being evaluated by potential employers, information provided on profiles is likely to be skewed in an effort to be viewed more favorably. However, there are aspects of SNWs which would make the process of skewing information difficult. Much of the information present in a given social networking profile is submitted by other members of the network, such as tagged photos and writing on another's wall[2]. Although some negative information can be deleted by the user, the user has more limited control over this aspect of their profile. In addition, some of the information controlled by the users themselves would be difficult to fake. For example, extroversion may be tied to the number of friends a user has in the social network. Artificially inflating a substantial number of friends in the network would pose great difficulty, as the user cannot control who accepts their friend request. As another example, rater assessment of personality traits might be drawn in part from photos, which are similarly difficult to fake. Thus, while faking would appear to be an issue, it is likely that the impact of faking is less than with other selection methods. Future research should assess the impact of faking in the context of SNWs.Despite the potential for faking, Vazire and Gosling (2004) used personal web sites to accurately assess personality. Although personal web sites are similar in some ways to SNWs, they are used by such a small percentage of potential applicants that they are impractical for purposes of employment selection. In addition, SNWs provide additional information not included in personal web sites, such as a list of the user's friends and a list of the interest groups a user has joined. However, Vazire and Gosling provide initial evidence for the accurate prediction of personality using personal information available on the internet.The types of information available on SNWs may be particularly effective in predicting the big-five personality traits. SNWs contain various sources of information which could be used to assess behaviors related to personality. For example, the types and number of interest groups the user has joined, comments that have been left for the user, comments made by the user on other people's "walls," "tagging" photos, updating "status messages"[3], and listing books and intellectual interests in the "personal information" section[4]. These examples provide only a very preliminary introduction to the various sources of personal information available in SNWs which might indicate an individual's personality. It should also be noted that the use of SNWs may vary based on characteristics beyond personality. For example, an individual who is more adept with this form of technology may be more likely to participate in social networking and may post more information more frequently than those individuals lacking in these technological skills. Since age has been shown to relate to technology acceptance and use (Morris and Venkatesh, 2000), this also brings the issue of age into the possible groups which could be adversely impacted by the use of this type of technology by HR departments in hiring. Future research should explore this issue further.Assessment of intelligence Since the very earliest research on personnel selection, cognitive ability has been one of the major methods used to attempt to discriminate between candidates and to predict subsequent job performance (Robertson and Smith, 2001). Intelligence is the single most effective predictor known of individual performance at school and on the job (Gottfredson, 1998), accounting for approximately 25 percent of the variance in job performance (Hunter and Hunter, 1984). Cognitive ability provides criterion-related validity that generalizes across more or less all occupational areas (Robertson and Smith, 2001). In addition, judge ratings of intelligence have been shown to predict intelligence test scores (Borkenau et al., 2004). Furthermore, biographical data have been shown to predict intelligence (Schmidt and Hunter, 1998). These results provide some evidence that assessment of intelligence should be viable within the context of SNWs.Assessment of global performance Beyond personality and intelligence, SNWs may also contain additional information which may be useful in employment selection. Owing to the large volume of information contained in SNWs, information may also be obtained which relate to the user's writing skills, job experiences, or a variety of knowledge, skills, abilities, or other criteria which might relate to job or organizational fit in a given employment selection context. A more global assessment of performance would include a variety of information. In fact, due to the broad range of information available on SNWs and the lack of consistency in this information across individuals, the approach of assessing broad characteristics is likely to be more practical than assessing more narrow aspects of social networking profiles that may be unavailable and/or inconsistent for a large segment of the profiles.Rater consensus/interrater agreement Furr and Funder (2007) identify rater consensus as the first psychometric concern when assessing characteristics through judge-rated behavioral observation. Thus, the primary focus of this paper, as the first to address the potential use of SNW information for the purpose of employment selection, is to focus on the issue of rater consensus.To the degree to which SNWs provide a distinguishable and consistent basis of evaluation, judges will reach a certain level of consensus in their impressions of these SNWs. Thus, consensus is expected across judges:H1. Ratings of the big-five dimensions of personality, intelligence, and performance based on SNW information are consistent across raters.Accuracy of ratings A second psychometric concern is that of validity (Furr and Funder, 2007). If rater consensus is established, the next step will be to establish various forms of validity. This study evaluates validity by assessing whether judges are accurate in their assessment of personality intelligence, and performance.Personality, intelligence, and performance impressions based on judge assessments have been shown to be quite accurate, even when the amount of information is limited (Borkenau et al., 2004). This growing body of research suggests that people have a natural talent for judging one another accurately (Vazire and Gosling, 2004). Given the high volume of information available to assess behavioral cues on SNWs, a suitable level of accuracy is expected:H2. Raters assessing an individual's personality, intelligence, and performance through SNWs are able to distinguish between those individuals who are high on each characteristic from those who are low on that characteristic. Participants and procedures This study was conducted at a large public university in the southern USA. A total of 63 students enrolled in an employment selection course participated in this project for course credit. Participants were 49 percent male, 90 percent Caucasian, averaged 24 years of age, and worked an average of 26 hours per week. Participants had prerequisites in HRs and statistics. As part of the employment selection course, these participants were trained in both personality/intelligence testing and effective utilization of rating scales, participated in a one hour training session for this project (reviewing the definitions of the big-five personality traits, general mental ability, and academic performance; viewing Facebook profiles to identify specific information which could be used to assess the focal characteristics of the study; and familiarizing the participants with the rating form to be used when conducting the assessments), and participated in a series of practice assessments prior to conducting the ratings for this study (an assignment to evaluate SNWs and identify specific information that could be used to assess each of the focal characteristics of the study, follow-up class discussion of these observations, and a practice session in which each participant conducted two assessments of current Facebook users by using the researcher designed rating form). All participants had personal involvement with SNWs. Participants were asked to spend ten minutes evaluating each of the six social networking profiles, consider multiple aspects of the profiles which could relate to a specific trait, then complete the rating form based on their overall impression of the social networking profile.The choice of the three male and three female social networking profiles was randomly generated from a list of volunteers in an introductory management course. Along with volunteering to have their SNWs evaluated[5], the volunteers also completed demographics and personality questionnaires, an intelligence test, and consented to allow the researchers to obtain their grade point average (GPA) from the university registrar.Measures Judge ratings of big-five personality traits - measured with 25 items from the bipolar adjective checklist (Goldberg, 1992) on a nine-point scale.Judge ratings of general mental ability - a single item measure was used to measure intelligence quotient (IQ) based on Reilly and Mulhern (1995). Judges were asked to "Estimate the user's IQ. Remember that the average IQ is 100, and one-sixth of the population have IQs less than 85, with one-sixth scoring over 115."Judge ratings of performance - a single item measure was used to measure academic performance based on the format used to measure IQ. Judges were asked to "Estimate the user's GPA. Remember that an average GPA is 3.0 and the maximum is 4.0."Ratee self-reported big-five personality traits (referred to as true scores) - measured with 150 items from the international personality item pool - IPIP (Goldberg et al., 2006) on a five-point scale ranging from strongly disagree to strongly agree. The as ranged from 0.92 for conscientiousness to 0.81 for openness.Ratee general mental ability (IQ true score) - measured with the Wonderlic personnel test (Wonderlic, 2000), a 12 minute/50 question timed test of intelligence.Ratee performance (performance true score) - academic performance was obtained via GPA from the University Registrar. Although academic performance is less ideal than job performance in the context of employment selection, it represents an objective measure to test the hypotheses presented. The as for the judge ratings of personality were calculated for each of the six ratees. These six as were then averaged for each of the big-five to estimate the overall internal consistency of the scales. In order to assess H1, interrater agreement in the form of average measures intraclass correlation coefficients (ICCs) for the judge ratings are included in Table I. The scaled scores for the big-five personality traits and the single item scores for IQ and performance were evaluated for interrater agreement. The 378 total ratings (63 ratersx6 ratings each) were used to calculate the ICCs. The ICC values were all adequate, ranging from 0.93 for extroversion to 0.99 for conscientiousness and performance. Since ICCs are expected to be higher with a larger number of raters, Table II also includes the number of raters for each characteristic which would be necessary to achieve a 0.50 ICC value. Although there are no guidelines for level of agreement, 0.50 was used in the analyses as it should provide a minimum level of acceptable agreement across judges. The Spearman Brown prophecy formula was used to determine how many raters would be required to obtain an adequate (0.50) ICC value. Based on the 63 raters from this study, it was determined that between two (for conscientiousness and performance) and six (for emotional stability and extroversion) raters would be required to obtain a satisfactory level of interrater agreement.H2 was evaluated by conducting t-tests on score means in order to determine whether or not the means are statistically different from one another. In order to determine which means to test, the true scores (self-reported big-five personality scores, intelligence scores, and GPA) of the six rated subjects were evaluated. For each of the seven characteristics, the individual with the highest true score and the individual with the lowest true score were selected for analysis. Judge mean ratings for these subjects were then compared to determine whether or not raters are able to distinguish individuals high on a characteristic from those low on the same characteristic. This method also allows for evaluation of the direction of the relationship, such that (in addition to evaluating mean differences) the judge rating of the subject with the higher true score should be higher than the judge rating of the lower true score. Results demonstrate that the mean judge ratings for the subject highest on the seven characteristics were statistically different from the subjects lowest on those characteristics. In addition, with the exception of openness to experience, the judges mean ratings were higher for those with the highest true score, indicating the ability of judges to distinguish the traits of conscientiousness, emotional stability, agreeableness, extroversion, intelligence, and performance by evaluating SNWs[6].Post hoc analyses were conducted to determine the impact of intelligence and personality on judge consistency and accuracy. Prior research has demonstrated mixed findings related to the impact of personality traits of the rater on rating accuracy. Ambady et al. (1995) found that less sociable (extroverted) raters were more accurate, while Lippa and Dietz (2000) found that only openness indicated more accurate raters. In addition, narcissistic raters have been found to be less accurate (John and Robbins, 1994), which may be relevant to the big-five since narcissism relates strongly to neuroticism. Finally, intelligence has also been reported to positively relate to rater accuracy (Lippa and Dietz, 2000). In the current study, the 63 judges were asked to take the same intelligence and personality tests as the SNW subjects. The analyses conducted above were then re-evaluated based on high versus low groups based on intelligence and the big-five. Results show no difference in interrater agreement based on these characteristics. However, judges who are more intelligent and more emotionally stable were shown to be more accurate in their judgments. More specifically, when the raters were split into high and low groups based on intelligence scores (the 31 highest scores versus the 31 lowest scores), the high intelligence group significantly and more accurately differentiated between high and low characteristics for conscientiousness, emotional stability, openness, and performance. For example, with all 63 raters combined, the difference between rater means for conscientiousness in Table II is 0.38 (8.03 for the high ratee score and 7.65 for the low ratee score). When assessing high and low intelligence raters independently, the mean difference for the 31 high intelligence raters is 0.61, but only 0.14 for the 31 low intelligence raters. Thus, more intelligent raters seem to be more capable of assessing this trait than less intelligent raters. Similarly, raters who are the most emotionally stable also rate more accurately for conscientiousness, emotional stability, openness, and performance. For example, the mean difference across raters for high and low ratee conscientiousness is again 0.38, but is 0.73 for the 31 raters who are the most emotionally stable and 0.03 for the 31 raters who are the least emotionally stable. These results indicate the potential need for researchers to consider intelligence and emotional stability when selecting individuals who will serve as raters of characteristics such as personality. Based on the large volume of personal information available on SNWs, judges' ratings of the big-five dimensions of personality, intelligence, and global performance were consistent across the 63 raters in this study, demonstrating adequate internal consistency reliability and interrater agreement. In addition, the trained raters were able to accurately distinguish between individuals who scored high and individuals who scored low on four of the big-five personality traits, intelligence, and performance, providing initial evidence that raters can accurately determine these organizationally relevant traits by viewing SNW information.As stated earlier, other rated personality has been shown to predict job performance. Considering that other methods of other-reported personality are unlikely to be viable in an employment selection context, SNW ratings of personality may be a practical approach. Owing to the theoretical and methodological differences between self-reported and other-rated personality, it is likely that ratings of personality via SNWs will provide a context for incremental prediction of job performance beyond the predominant self-report approach. In addition, the differences in context between SNWs and a job interview (i.e. socially desirable responding in the job interview as well as the unique nature of information contained in SNWs) should similarly allow for unique prediction of job performance beyond what can be evaluated through personality assessment in the employment interview. This approach may be particularly valuable since these assessments take only a fraction of the time involved with other selection methods.This study is not without limitations. Although the analyses testing the consistency of the relationships of SNW ratings are based on 378 judge ratings from 63 raters, the analyses testing rater accuracy were conducted by testing for significant differences between the high and low performer on the seven characteristics for only six subjects. Future research should assess accuracy over a larger sample of subjects.We hope that the results of this preliminary study will not be used by organizations to support their use of SNWs in employment selection. Without further validation in a variety of studies, with larger samples and in a variety of organizational contexts, caution should be used when interpreting the implications of this study. This is particularly true given the potential for employer legal liability due to the vast amount of personal information available on SNWs. Information regarding gender, race, age, disabilities, and other criteria which should not be used when making hiring decisions will most certainly, consciously or not, influence who gets hired. Even if this information does not bias the hiring decision, disparate impact issues may still exist. Future research should also examine the potential issues of adverse impact and potentially illegal information in hiring decisions using personal information from SNWs. In addition, research should be conducted to compare assessments of SNWs to other employment selection methods, such as personality assessment, intelligence testing, and employment interviews.Based on the relative absence of research evidence in this newly developing area, particularly regarding the potential for adverse impact and the lack of validity evidence, we believe the most important practical implication of this paper is for organizations to use SNWs with these issues in mind. Organizational representatives assessing SNWs should ask themselves two important questions. First, is the organization assessing (or could be perceived as assessing) information which could lead to discrimination against a legally protected group? Second, is the specific social networking information used to help make a hiring decision valid in determining who will perform better on the job? The approach used in this paper of assessing personality traits, intelligence, or general performance begin to provide answers to these questions. Opens in a new window.Table I Judge rating as and ICCs Opens in a new window.Table II Differences in judge ratings means comparing high and low ratee scores
|
- The use of social networking web sites (SNWs), like Facebook and MySpace, has become extremely popular, particularly with today's emerging workforce. Employers, aware of this phenomenon, have begun to use the personal information available on SNWs to make hiring decisions. The purpose of this paper is to examine the feasibility of using applicant personal information currently available on SNWs to improve employment selection decisions.
|
[SECTION: Method] Within the past few years, the phenomenon of social networking web sites (SNWs) on the internet has exploded into the mainstream. Further, this online information has begun to be used for purposes beyond its intended use. Owing to the vast amount of personal information on these web sites, employers have begun to tap into this information as a source of applicant data in an effort to improve hiring decisions. This study evaluates the use of the SNWs in employment selection. Specifically, can trained judges consistently and accurately assess important organizational characteristics such as personality, intelligence, and performance using only a target's SNW information? In addition, the use of this information may lead to discrimination against applicants, given the wide range of available personal information such as gender, race, age, religion, and disability status otherwise illegal to use when making employment decisions.Social networking web sites SNWs focus on building online communities of people who share interests and activities, or who are interested in exploring the interests and activities of others. Most provide a variety of ways for users to interact, such as e-mail and instant messaging services. SNWs are designed to connect users to each other and to visually display each individual's network of friends. The number of users for these web sites and the daily traffic created by these web sites are staggering. According to the "About Us" section of the various sites, MySpace is the largest with over 248 million registered users. Other SNWs also have millions of users registered, such as Facebook (110 million), Friendster (85 million), Hi5 (80 million), Orkut/Google (37 million), and LinkedIn (25 million).While these sites differ in the features that are available, most have a mechanism for posting pictures, music and videos, keeping blogs, sharing links, and displaying interests. These sites vary in user demographics. For example, although Facebook is currently open to anyone, it started as a high school and college web site exclusively, with about 90 percent of students registered for the site (van der Werf, 2006).Social networking web sites in selection Owing to the increasing prevalence of SNWs in conjunction with the large volume of information available to the viewer, employers have begun using SNWs to assist in the selection process for new employees. About 50 percent of the employers attending college career fairs use online technology, including both search engines and SNWs to screen candidates (Shea and Wesley, 2006). Currently, between 20 (Framingham, 2008) and 25 percent (Taylor, 2007; NACE, 2006) of managers are using SNWs to screen candidates, with 40 percent indicating that they are likely to use them within the next year (Zeidner, 2007). These employers apparently feel justified in electronic screening using SNWs. The president of a small Chicago consulting firm, when asked about an applicant that applied for an internship and had questionable content on his page, replied, "A lot of it makes me think, what kind of judgment does this person have? Why are you allowing this to be viewed publicly?" (Finder, 2006).While it may be common practice to monitor web site content it may not be legal. Lance Chou, Director of Career Development at Stanford University, noted that:[...] some employers might try and learn something about the student's personality and whether it would be appropriate for the job. However, there is information on Facebook that is not relevant to the job, but may be used inappropriately by employers to assess a candidate (Fuller, 2006).According to George Lenard, an employment attorney, employers can inadvertently learn about matters such as candidates' age, marital status, and other topics typically are off limits in job interviews, and organizations can be sued for discrimination if these candidates are not hired (Frauenheim, 2006). The question of whether employer monitoring of SNWs is illegal may relate to equal employment opportunity (EEO) law. EEO severely limits the type of information an interviewer may ascertain and use. For example, during interviews employers may not ask questions regarding race, religion, sexual preference, or marital status. However, all of this information can be easily located using SNWs (Kowske and Southwell, 2006). The ethics and legality behind electronic screening using SNWs has become an important issue for human resource (HR) practitioners (Zeidner, 2007).Since no academic studies, to our knowledge, have assessed whether using SNWs in employment selection are a reliable and valid predictor of important organizational outcomes, their value in employment screening is unknown. Until the reliability and validity of information on SNWs is established, employers should use caution when using SNWs to make hiring decisions. Just as with structured interview methods (Campion et al., 1997), focusing on job-related information in SNWs should minimize the use of less job-relevant information that might bias the hiring decision. This study focuses on showing the validity and reliability of three such categories of job-relevant characteristics which have potential to be rated consistently and accurately and to serve as valid predictors of job performance. The assessment of personality, intelligence, and a global measure of performance through SNWs provide multiple possible avenues regarding validity generalization of job-relevant characteristics, and thereby justifying the use of SNWs in the screening process.Assessment of personality The 1990s have seen a huge growth in the use of personality assessment within personnel selection practice and research (Barrick and Mount, 1991; Frei and McDaniel, 1997; Ones et al., 1993; Salgado, 1998; Tett et al., 1991). These studies provide positive evidence for the criterion-related validity of personality. When it comes to the prediction of overall job performance, conscientiousness was found to be the best predictor, showing consistent predictions across all occupational groups. In addition, extroversion, agreeableness, and neuroticism were shown to predict job performance in certain jobs. Finally, although typically unrelated to job performance, openness to experience has been found to predict training performance.Beyond the predominant focus on self-reported personality assessment, personality can also be measured using other-ratings. It is well established that people can assess the personality of others, even after relatively short exposure, but that the accuracy of these assessments depends on the information available to the observer (Barrick et al., 2000). Observer ratings may be more valuable than self-ratings in employment selection, particularly when targets are not available for self-reports, self-reports are untrustworthy, and researchers wish to improve accuracy by aggregating multiple raters (Hofstee, 1994; McCrae and Weiss, 2007). Other-rated personality via SNWs seems quite promising in the selection context, since other-ratings of personality have been found to predict job performance. Motowidlo and colleagues (1996) found that interviewer-rated extroversion and conscientiousness (r=0.27 and 0.20, respectively) significantly predict supervisor rated job performance. In addition, Mount et al. (1994) show that observer ratings of extroversion and conscientiousness from supervisors, coworkers, and customers significantly predicts sales performance, even beyond self-rated personality.The validity of other-rated personality, however, can depend on the relationship the subject has with the rater and the quality of the information available to the rater. A meta analysis conducted by Connolly and Viswesvaran (1998) show low accuracy for strangers in predicting personality and moderate prediction by peers for each of the big-five traits. In addition, Barrick et al. (2000) developed a personality-based job interview for the purpose of assessing the personality of the applicant. They found that personality-based job interviews could be used to accurately predict three of the big-five. Considering the average interview is approximately 40 minutes in length (Campion et al., 1997), it appears that an interviewer can assess some aspects of personality as effectively as a close acquaintance of the applicant. We propose that the information available in SNWs provides a similar means to assess personality.Furthermore, since SNW ratings are obtained from a wide range of personal information that is a reflection of ongoing behaviors and interactions with other users of the networks, web sites may actually provide unique information not found with other selection methods. Recent issues of Personnel Psychology (Morgeson et al., 2007a, b; Ones et al., 2007; Tett and Christiansen, 2007) and Industrial and Organizational Psychology (Hough and Oswald, 2008; Griffith and Peterson, 2008) have focused on the limited criterion validity of self-reported personality measures as well as the complex issue of social desirability and faking of self-reported personality measures. To increase validity and address issues of socially desirable responding "we should look at other ways of assessing personality. There are a variety of ways of finding out about people's stable pattern of behavior" (Morgeson et al., 2007a, b, p. 719). SNWs may prove to be an appropriate means of assessing personality in this way. Beyond personality assessment, similar issues emerge in relation to a host of employment selection methods. Resumes, interviews, job applications, and many other forms of employee selection include a certain element of self-presentation, reflecting "maximal" instead of "typical" work performance (Sackett, 2007; Sackett et al., 1988). Employment selection methods using social networking are likely to be based on "typical" behaviors, and therefore may be more accurate than other selection methods. At a minimum, this method should provide information that is distinct from "maximal" selection methods, thereby allowing for a stronger likelihood that using SNWs will yield incremental validity beyond established methods of employment selection[1].This is not to say that SNWs are not susceptible to manipulation and faking, similar to that of self-report personality measures and job interviews. In fact, as users of SNWs become more aware that their profiles are being evaluated by potential employers, information provided on profiles is likely to be skewed in an effort to be viewed more favorably. However, there are aspects of SNWs which would make the process of skewing information difficult. Much of the information present in a given social networking profile is submitted by other members of the network, such as tagged photos and writing on another's wall[2]. Although some negative information can be deleted by the user, the user has more limited control over this aspect of their profile. In addition, some of the information controlled by the users themselves would be difficult to fake. For example, extroversion may be tied to the number of friends a user has in the social network. Artificially inflating a substantial number of friends in the network would pose great difficulty, as the user cannot control who accepts their friend request. As another example, rater assessment of personality traits might be drawn in part from photos, which are similarly difficult to fake. Thus, while faking would appear to be an issue, it is likely that the impact of faking is less than with other selection methods. Future research should assess the impact of faking in the context of SNWs.Despite the potential for faking, Vazire and Gosling (2004) used personal web sites to accurately assess personality. Although personal web sites are similar in some ways to SNWs, they are used by such a small percentage of potential applicants that they are impractical for purposes of employment selection. In addition, SNWs provide additional information not included in personal web sites, such as a list of the user's friends and a list of the interest groups a user has joined. However, Vazire and Gosling provide initial evidence for the accurate prediction of personality using personal information available on the internet.The types of information available on SNWs may be particularly effective in predicting the big-five personality traits. SNWs contain various sources of information which could be used to assess behaviors related to personality. For example, the types and number of interest groups the user has joined, comments that have been left for the user, comments made by the user on other people's "walls," "tagging" photos, updating "status messages"[3], and listing books and intellectual interests in the "personal information" section[4]. These examples provide only a very preliminary introduction to the various sources of personal information available in SNWs which might indicate an individual's personality. It should also be noted that the use of SNWs may vary based on characteristics beyond personality. For example, an individual who is more adept with this form of technology may be more likely to participate in social networking and may post more information more frequently than those individuals lacking in these technological skills. Since age has been shown to relate to technology acceptance and use (Morris and Venkatesh, 2000), this also brings the issue of age into the possible groups which could be adversely impacted by the use of this type of technology by HR departments in hiring. Future research should explore this issue further.Assessment of intelligence Since the very earliest research on personnel selection, cognitive ability has been one of the major methods used to attempt to discriminate between candidates and to predict subsequent job performance (Robertson and Smith, 2001). Intelligence is the single most effective predictor known of individual performance at school and on the job (Gottfredson, 1998), accounting for approximately 25 percent of the variance in job performance (Hunter and Hunter, 1984). Cognitive ability provides criterion-related validity that generalizes across more or less all occupational areas (Robertson and Smith, 2001). In addition, judge ratings of intelligence have been shown to predict intelligence test scores (Borkenau et al., 2004). Furthermore, biographical data have been shown to predict intelligence (Schmidt and Hunter, 1998). These results provide some evidence that assessment of intelligence should be viable within the context of SNWs.Assessment of global performance Beyond personality and intelligence, SNWs may also contain additional information which may be useful in employment selection. Owing to the large volume of information contained in SNWs, information may also be obtained which relate to the user's writing skills, job experiences, or a variety of knowledge, skills, abilities, or other criteria which might relate to job or organizational fit in a given employment selection context. A more global assessment of performance would include a variety of information. In fact, due to the broad range of information available on SNWs and the lack of consistency in this information across individuals, the approach of assessing broad characteristics is likely to be more practical than assessing more narrow aspects of social networking profiles that may be unavailable and/or inconsistent for a large segment of the profiles.Rater consensus/interrater agreement Furr and Funder (2007) identify rater consensus as the first psychometric concern when assessing characteristics through judge-rated behavioral observation. Thus, the primary focus of this paper, as the first to address the potential use of SNW information for the purpose of employment selection, is to focus on the issue of rater consensus.To the degree to which SNWs provide a distinguishable and consistent basis of evaluation, judges will reach a certain level of consensus in their impressions of these SNWs. Thus, consensus is expected across judges:H1. Ratings of the big-five dimensions of personality, intelligence, and performance based on SNW information are consistent across raters.Accuracy of ratings A second psychometric concern is that of validity (Furr and Funder, 2007). If rater consensus is established, the next step will be to establish various forms of validity. This study evaluates validity by assessing whether judges are accurate in their assessment of personality intelligence, and performance.Personality, intelligence, and performance impressions based on judge assessments have been shown to be quite accurate, even when the amount of information is limited (Borkenau et al., 2004). This growing body of research suggests that people have a natural talent for judging one another accurately (Vazire and Gosling, 2004). Given the high volume of information available to assess behavioral cues on SNWs, a suitable level of accuracy is expected:H2. Raters assessing an individual's personality, intelligence, and performance through SNWs are able to distinguish between those individuals who are high on each characteristic from those who are low on that characteristic. Participants and procedures This study was conducted at a large public university in the southern USA. A total of 63 students enrolled in an employment selection course participated in this project for course credit. Participants were 49 percent male, 90 percent Caucasian, averaged 24 years of age, and worked an average of 26 hours per week. Participants had prerequisites in HRs and statistics. As part of the employment selection course, these participants were trained in both personality/intelligence testing and effective utilization of rating scales, participated in a one hour training session for this project (reviewing the definitions of the big-five personality traits, general mental ability, and academic performance; viewing Facebook profiles to identify specific information which could be used to assess the focal characteristics of the study; and familiarizing the participants with the rating form to be used when conducting the assessments), and participated in a series of practice assessments prior to conducting the ratings for this study (an assignment to evaluate SNWs and identify specific information that could be used to assess each of the focal characteristics of the study, follow-up class discussion of these observations, and a practice session in which each participant conducted two assessments of current Facebook users by using the researcher designed rating form). All participants had personal involvement with SNWs. Participants were asked to spend ten minutes evaluating each of the six social networking profiles, consider multiple aspects of the profiles which could relate to a specific trait, then complete the rating form based on their overall impression of the social networking profile.The choice of the three male and three female social networking profiles was randomly generated from a list of volunteers in an introductory management course. Along with volunteering to have their SNWs evaluated[5], the volunteers also completed demographics and personality questionnaires, an intelligence test, and consented to allow the researchers to obtain their grade point average (GPA) from the university registrar.Measures Judge ratings of big-five personality traits - measured with 25 items from the bipolar adjective checklist (Goldberg, 1992) on a nine-point scale.Judge ratings of general mental ability - a single item measure was used to measure intelligence quotient (IQ) based on Reilly and Mulhern (1995). Judges were asked to "Estimate the user's IQ. Remember that the average IQ is 100, and one-sixth of the population have IQs less than 85, with one-sixth scoring over 115."Judge ratings of performance - a single item measure was used to measure academic performance based on the format used to measure IQ. Judges were asked to "Estimate the user's GPA. Remember that an average GPA is 3.0 and the maximum is 4.0."Ratee self-reported big-five personality traits (referred to as true scores) - measured with 150 items from the international personality item pool - IPIP (Goldberg et al., 2006) on a five-point scale ranging from strongly disagree to strongly agree. The as ranged from 0.92 for conscientiousness to 0.81 for openness.Ratee general mental ability (IQ true score) - measured with the Wonderlic personnel test (Wonderlic, 2000), a 12 minute/50 question timed test of intelligence.Ratee performance (performance true score) - academic performance was obtained via GPA from the University Registrar. Although academic performance is less ideal than job performance in the context of employment selection, it represents an objective measure to test the hypotheses presented. The as for the judge ratings of personality were calculated for each of the six ratees. These six as were then averaged for each of the big-five to estimate the overall internal consistency of the scales. In order to assess H1, interrater agreement in the form of average measures intraclass correlation coefficients (ICCs) for the judge ratings are included in Table I. The scaled scores for the big-five personality traits and the single item scores for IQ and performance were evaluated for interrater agreement. The 378 total ratings (63 ratersx6 ratings each) were used to calculate the ICCs. The ICC values were all adequate, ranging from 0.93 for extroversion to 0.99 for conscientiousness and performance. Since ICCs are expected to be higher with a larger number of raters, Table II also includes the number of raters for each characteristic which would be necessary to achieve a 0.50 ICC value. Although there are no guidelines for level of agreement, 0.50 was used in the analyses as it should provide a minimum level of acceptable agreement across judges. The Spearman Brown prophecy formula was used to determine how many raters would be required to obtain an adequate (0.50) ICC value. Based on the 63 raters from this study, it was determined that between two (for conscientiousness and performance) and six (for emotional stability and extroversion) raters would be required to obtain a satisfactory level of interrater agreement.H2 was evaluated by conducting t-tests on score means in order to determine whether or not the means are statistically different from one another. In order to determine which means to test, the true scores (self-reported big-five personality scores, intelligence scores, and GPA) of the six rated subjects were evaluated. For each of the seven characteristics, the individual with the highest true score and the individual with the lowest true score were selected for analysis. Judge mean ratings for these subjects were then compared to determine whether or not raters are able to distinguish individuals high on a characteristic from those low on the same characteristic. This method also allows for evaluation of the direction of the relationship, such that (in addition to evaluating mean differences) the judge rating of the subject with the higher true score should be higher than the judge rating of the lower true score. Results demonstrate that the mean judge ratings for the subject highest on the seven characteristics were statistically different from the subjects lowest on those characteristics. In addition, with the exception of openness to experience, the judges mean ratings were higher for those with the highest true score, indicating the ability of judges to distinguish the traits of conscientiousness, emotional stability, agreeableness, extroversion, intelligence, and performance by evaluating SNWs[6].Post hoc analyses were conducted to determine the impact of intelligence and personality on judge consistency and accuracy. Prior research has demonstrated mixed findings related to the impact of personality traits of the rater on rating accuracy. Ambady et al. (1995) found that less sociable (extroverted) raters were more accurate, while Lippa and Dietz (2000) found that only openness indicated more accurate raters. In addition, narcissistic raters have been found to be less accurate (John and Robbins, 1994), which may be relevant to the big-five since narcissism relates strongly to neuroticism. Finally, intelligence has also been reported to positively relate to rater accuracy (Lippa and Dietz, 2000). In the current study, the 63 judges were asked to take the same intelligence and personality tests as the SNW subjects. The analyses conducted above were then re-evaluated based on high versus low groups based on intelligence and the big-five. Results show no difference in interrater agreement based on these characteristics. However, judges who are more intelligent and more emotionally stable were shown to be more accurate in their judgments. More specifically, when the raters were split into high and low groups based on intelligence scores (the 31 highest scores versus the 31 lowest scores), the high intelligence group significantly and more accurately differentiated between high and low characteristics for conscientiousness, emotional stability, openness, and performance. For example, with all 63 raters combined, the difference between rater means for conscientiousness in Table II is 0.38 (8.03 for the high ratee score and 7.65 for the low ratee score). When assessing high and low intelligence raters independently, the mean difference for the 31 high intelligence raters is 0.61, but only 0.14 for the 31 low intelligence raters. Thus, more intelligent raters seem to be more capable of assessing this trait than less intelligent raters. Similarly, raters who are the most emotionally stable also rate more accurately for conscientiousness, emotional stability, openness, and performance. For example, the mean difference across raters for high and low ratee conscientiousness is again 0.38, but is 0.73 for the 31 raters who are the most emotionally stable and 0.03 for the 31 raters who are the least emotionally stable. These results indicate the potential need for researchers to consider intelligence and emotional stability when selecting individuals who will serve as raters of characteristics such as personality. Based on the large volume of personal information available on SNWs, judges' ratings of the big-five dimensions of personality, intelligence, and global performance were consistent across the 63 raters in this study, demonstrating adequate internal consistency reliability and interrater agreement. In addition, the trained raters were able to accurately distinguish between individuals who scored high and individuals who scored low on four of the big-five personality traits, intelligence, and performance, providing initial evidence that raters can accurately determine these organizationally relevant traits by viewing SNW information.As stated earlier, other rated personality has been shown to predict job performance. Considering that other methods of other-reported personality are unlikely to be viable in an employment selection context, SNW ratings of personality may be a practical approach. Owing to the theoretical and methodological differences between self-reported and other-rated personality, it is likely that ratings of personality via SNWs will provide a context for incremental prediction of job performance beyond the predominant self-report approach. In addition, the differences in context between SNWs and a job interview (i.e. socially desirable responding in the job interview as well as the unique nature of information contained in SNWs) should similarly allow for unique prediction of job performance beyond what can be evaluated through personality assessment in the employment interview. This approach may be particularly valuable since these assessments take only a fraction of the time involved with other selection methods.This study is not without limitations. Although the analyses testing the consistency of the relationships of SNW ratings are based on 378 judge ratings from 63 raters, the analyses testing rater accuracy were conducted by testing for significant differences between the high and low performer on the seven characteristics for only six subjects. Future research should assess accuracy over a larger sample of subjects.We hope that the results of this preliminary study will not be used by organizations to support their use of SNWs in employment selection. Without further validation in a variety of studies, with larger samples and in a variety of organizational contexts, caution should be used when interpreting the implications of this study. This is particularly true given the potential for employer legal liability due to the vast amount of personal information available on SNWs. Information regarding gender, race, age, disabilities, and other criteria which should not be used when making hiring decisions will most certainly, consciously or not, influence who gets hired. Even if this information does not bias the hiring decision, disparate impact issues may still exist. Future research should also examine the potential issues of adverse impact and potentially illegal information in hiring decisions using personal information from SNWs. In addition, research should be conducted to compare assessments of SNWs to other employment selection methods, such as personality assessment, intelligence testing, and employment interviews.Based on the relative absence of research evidence in this newly developing area, particularly regarding the potential for adverse impact and the lack of validity evidence, we believe the most important practical implication of this paper is for organizations to use SNWs with these issues in mind. Organizational representatives assessing SNWs should ask themselves two important questions. First, is the organization assessing (or could be perceived as assessing) information which could lead to discrimination against a legally protected group? Second, is the specific social networking information used to help make a hiring decision valid in determining who will perform better on the job? The approach used in this paper of assessing personality traits, intelligence, or general performance begin to provide answers to these questions. Opens in a new window.Table I Judge rating as and ICCs Opens in a new window.Table II Differences in judge ratings means comparing high and low ratee scores
|
- A total of 378 judge ratings (63 ratersx6 subjects) are evaluated to determine if raters can reliably and accurately determine the big-five personality traits, intelligence, and performance based only on information available on SNWs. Interrater reliability is assessed to determine rater consistency, followed by an assessment of rater accuracy.
|
[SECTION: Findings] Within the past few years, the phenomenon of social networking web sites (SNWs) on the internet has exploded into the mainstream. Further, this online information has begun to be used for purposes beyond its intended use. Owing to the vast amount of personal information on these web sites, employers have begun to tap into this information as a source of applicant data in an effort to improve hiring decisions. This study evaluates the use of the SNWs in employment selection. Specifically, can trained judges consistently and accurately assess important organizational characteristics such as personality, intelligence, and performance using only a target's SNW information? In addition, the use of this information may lead to discrimination against applicants, given the wide range of available personal information such as gender, race, age, religion, and disability status otherwise illegal to use when making employment decisions.Social networking web sites SNWs focus on building online communities of people who share interests and activities, or who are interested in exploring the interests and activities of others. Most provide a variety of ways for users to interact, such as e-mail and instant messaging services. SNWs are designed to connect users to each other and to visually display each individual's network of friends. The number of users for these web sites and the daily traffic created by these web sites are staggering. According to the "About Us" section of the various sites, MySpace is the largest with over 248 million registered users. Other SNWs also have millions of users registered, such as Facebook (110 million), Friendster (85 million), Hi5 (80 million), Orkut/Google (37 million), and LinkedIn (25 million).While these sites differ in the features that are available, most have a mechanism for posting pictures, music and videos, keeping blogs, sharing links, and displaying interests. These sites vary in user demographics. For example, although Facebook is currently open to anyone, it started as a high school and college web site exclusively, with about 90 percent of students registered for the site (van der Werf, 2006).Social networking web sites in selection Owing to the increasing prevalence of SNWs in conjunction with the large volume of information available to the viewer, employers have begun using SNWs to assist in the selection process for new employees. About 50 percent of the employers attending college career fairs use online technology, including both search engines and SNWs to screen candidates (Shea and Wesley, 2006). Currently, between 20 (Framingham, 2008) and 25 percent (Taylor, 2007; NACE, 2006) of managers are using SNWs to screen candidates, with 40 percent indicating that they are likely to use them within the next year (Zeidner, 2007). These employers apparently feel justified in electronic screening using SNWs. The president of a small Chicago consulting firm, when asked about an applicant that applied for an internship and had questionable content on his page, replied, "A lot of it makes me think, what kind of judgment does this person have? Why are you allowing this to be viewed publicly?" (Finder, 2006).While it may be common practice to monitor web site content it may not be legal. Lance Chou, Director of Career Development at Stanford University, noted that:[...] some employers might try and learn something about the student's personality and whether it would be appropriate for the job. However, there is information on Facebook that is not relevant to the job, but may be used inappropriately by employers to assess a candidate (Fuller, 2006).According to George Lenard, an employment attorney, employers can inadvertently learn about matters such as candidates' age, marital status, and other topics typically are off limits in job interviews, and organizations can be sued for discrimination if these candidates are not hired (Frauenheim, 2006). The question of whether employer monitoring of SNWs is illegal may relate to equal employment opportunity (EEO) law. EEO severely limits the type of information an interviewer may ascertain and use. For example, during interviews employers may not ask questions regarding race, religion, sexual preference, or marital status. However, all of this information can be easily located using SNWs (Kowske and Southwell, 2006). The ethics and legality behind electronic screening using SNWs has become an important issue for human resource (HR) practitioners (Zeidner, 2007).Since no academic studies, to our knowledge, have assessed whether using SNWs in employment selection are a reliable and valid predictor of important organizational outcomes, their value in employment screening is unknown. Until the reliability and validity of information on SNWs is established, employers should use caution when using SNWs to make hiring decisions. Just as with structured interview methods (Campion et al., 1997), focusing on job-related information in SNWs should minimize the use of less job-relevant information that might bias the hiring decision. This study focuses on showing the validity and reliability of three such categories of job-relevant characteristics which have potential to be rated consistently and accurately and to serve as valid predictors of job performance. The assessment of personality, intelligence, and a global measure of performance through SNWs provide multiple possible avenues regarding validity generalization of job-relevant characteristics, and thereby justifying the use of SNWs in the screening process.Assessment of personality The 1990s have seen a huge growth in the use of personality assessment within personnel selection practice and research (Barrick and Mount, 1991; Frei and McDaniel, 1997; Ones et al., 1993; Salgado, 1998; Tett et al., 1991). These studies provide positive evidence for the criterion-related validity of personality. When it comes to the prediction of overall job performance, conscientiousness was found to be the best predictor, showing consistent predictions across all occupational groups. In addition, extroversion, agreeableness, and neuroticism were shown to predict job performance in certain jobs. Finally, although typically unrelated to job performance, openness to experience has been found to predict training performance.Beyond the predominant focus on self-reported personality assessment, personality can also be measured using other-ratings. It is well established that people can assess the personality of others, even after relatively short exposure, but that the accuracy of these assessments depends on the information available to the observer (Barrick et al., 2000). Observer ratings may be more valuable than self-ratings in employment selection, particularly when targets are not available for self-reports, self-reports are untrustworthy, and researchers wish to improve accuracy by aggregating multiple raters (Hofstee, 1994; McCrae and Weiss, 2007). Other-rated personality via SNWs seems quite promising in the selection context, since other-ratings of personality have been found to predict job performance. Motowidlo and colleagues (1996) found that interviewer-rated extroversion and conscientiousness (r=0.27 and 0.20, respectively) significantly predict supervisor rated job performance. In addition, Mount et al. (1994) show that observer ratings of extroversion and conscientiousness from supervisors, coworkers, and customers significantly predicts sales performance, even beyond self-rated personality.The validity of other-rated personality, however, can depend on the relationship the subject has with the rater and the quality of the information available to the rater. A meta analysis conducted by Connolly and Viswesvaran (1998) show low accuracy for strangers in predicting personality and moderate prediction by peers for each of the big-five traits. In addition, Barrick et al. (2000) developed a personality-based job interview for the purpose of assessing the personality of the applicant. They found that personality-based job interviews could be used to accurately predict three of the big-five. Considering the average interview is approximately 40 minutes in length (Campion et al., 1997), it appears that an interviewer can assess some aspects of personality as effectively as a close acquaintance of the applicant. We propose that the information available in SNWs provides a similar means to assess personality.Furthermore, since SNW ratings are obtained from a wide range of personal information that is a reflection of ongoing behaviors and interactions with other users of the networks, web sites may actually provide unique information not found with other selection methods. Recent issues of Personnel Psychology (Morgeson et al., 2007a, b; Ones et al., 2007; Tett and Christiansen, 2007) and Industrial and Organizational Psychology (Hough and Oswald, 2008; Griffith and Peterson, 2008) have focused on the limited criterion validity of self-reported personality measures as well as the complex issue of social desirability and faking of self-reported personality measures. To increase validity and address issues of socially desirable responding "we should look at other ways of assessing personality. There are a variety of ways of finding out about people's stable pattern of behavior" (Morgeson et al., 2007a, b, p. 719). SNWs may prove to be an appropriate means of assessing personality in this way. Beyond personality assessment, similar issues emerge in relation to a host of employment selection methods. Resumes, interviews, job applications, and many other forms of employee selection include a certain element of self-presentation, reflecting "maximal" instead of "typical" work performance (Sackett, 2007; Sackett et al., 1988). Employment selection methods using social networking are likely to be based on "typical" behaviors, and therefore may be more accurate than other selection methods. At a minimum, this method should provide information that is distinct from "maximal" selection methods, thereby allowing for a stronger likelihood that using SNWs will yield incremental validity beyond established methods of employment selection[1].This is not to say that SNWs are not susceptible to manipulation and faking, similar to that of self-report personality measures and job interviews. In fact, as users of SNWs become more aware that their profiles are being evaluated by potential employers, information provided on profiles is likely to be skewed in an effort to be viewed more favorably. However, there are aspects of SNWs which would make the process of skewing information difficult. Much of the information present in a given social networking profile is submitted by other members of the network, such as tagged photos and writing on another's wall[2]. Although some negative information can be deleted by the user, the user has more limited control over this aspect of their profile. In addition, some of the information controlled by the users themselves would be difficult to fake. For example, extroversion may be tied to the number of friends a user has in the social network. Artificially inflating a substantial number of friends in the network would pose great difficulty, as the user cannot control who accepts their friend request. As another example, rater assessment of personality traits might be drawn in part from photos, which are similarly difficult to fake. Thus, while faking would appear to be an issue, it is likely that the impact of faking is less than with other selection methods. Future research should assess the impact of faking in the context of SNWs.Despite the potential for faking, Vazire and Gosling (2004) used personal web sites to accurately assess personality. Although personal web sites are similar in some ways to SNWs, they are used by such a small percentage of potential applicants that they are impractical for purposes of employment selection. In addition, SNWs provide additional information not included in personal web sites, such as a list of the user's friends and a list of the interest groups a user has joined. However, Vazire and Gosling provide initial evidence for the accurate prediction of personality using personal information available on the internet.The types of information available on SNWs may be particularly effective in predicting the big-five personality traits. SNWs contain various sources of information which could be used to assess behaviors related to personality. For example, the types and number of interest groups the user has joined, comments that have been left for the user, comments made by the user on other people's "walls," "tagging" photos, updating "status messages"[3], and listing books and intellectual interests in the "personal information" section[4]. These examples provide only a very preliminary introduction to the various sources of personal information available in SNWs which might indicate an individual's personality. It should also be noted that the use of SNWs may vary based on characteristics beyond personality. For example, an individual who is more adept with this form of technology may be more likely to participate in social networking and may post more information more frequently than those individuals lacking in these technological skills. Since age has been shown to relate to technology acceptance and use (Morris and Venkatesh, 2000), this also brings the issue of age into the possible groups which could be adversely impacted by the use of this type of technology by HR departments in hiring. Future research should explore this issue further.Assessment of intelligence Since the very earliest research on personnel selection, cognitive ability has been one of the major methods used to attempt to discriminate between candidates and to predict subsequent job performance (Robertson and Smith, 2001). Intelligence is the single most effective predictor known of individual performance at school and on the job (Gottfredson, 1998), accounting for approximately 25 percent of the variance in job performance (Hunter and Hunter, 1984). Cognitive ability provides criterion-related validity that generalizes across more or less all occupational areas (Robertson and Smith, 2001). In addition, judge ratings of intelligence have been shown to predict intelligence test scores (Borkenau et al., 2004). Furthermore, biographical data have been shown to predict intelligence (Schmidt and Hunter, 1998). These results provide some evidence that assessment of intelligence should be viable within the context of SNWs.Assessment of global performance Beyond personality and intelligence, SNWs may also contain additional information which may be useful in employment selection. Owing to the large volume of information contained in SNWs, information may also be obtained which relate to the user's writing skills, job experiences, or a variety of knowledge, skills, abilities, or other criteria which might relate to job or organizational fit in a given employment selection context. A more global assessment of performance would include a variety of information. In fact, due to the broad range of information available on SNWs and the lack of consistency in this information across individuals, the approach of assessing broad characteristics is likely to be more practical than assessing more narrow aspects of social networking profiles that may be unavailable and/or inconsistent for a large segment of the profiles.Rater consensus/interrater agreement Furr and Funder (2007) identify rater consensus as the first psychometric concern when assessing characteristics through judge-rated behavioral observation. Thus, the primary focus of this paper, as the first to address the potential use of SNW information for the purpose of employment selection, is to focus on the issue of rater consensus.To the degree to which SNWs provide a distinguishable and consistent basis of evaluation, judges will reach a certain level of consensus in their impressions of these SNWs. Thus, consensus is expected across judges:H1. Ratings of the big-five dimensions of personality, intelligence, and performance based on SNW information are consistent across raters.Accuracy of ratings A second psychometric concern is that of validity (Furr and Funder, 2007). If rater consensus is established, the next step will be to establish various forms of validity. This study evaluates validity by assessing whether judges are accurate in their assessment of personality intelligence, and performance.Personality, intelligence, and performance impressions based on judge assessments have been shown to be quite accurate, even when the amount of information is limited (Borkenau et al., 2004). This growing body of research suggests that people have a natural talent for judging one another accurately (Vazire and Gosling, 2004). Given the high volume of information available to assess behavioral cues on SNWs, a suitable level of accuracy is expected:H2. Raters assessing an individual's personality, intelligence, and performance through SNWs are able to distinguish between those individuals who are high on each characteristic from those who are low on that characteristic. Participants and procedures This study was conducted at a large public university in the southern USA. A total of 63 students enrolled in an employment selection course participated in this project for course credit. Participants were 49 percent male, 90 percent Caucasian, averaged 24 years of age, and worked an average of 26 hours per week. Participants had prerequisites in HRs and statistics. As part of the employment selection course, these participants were trained in both personality/intelligence testing and effective utilization of rating scales, participated in a one hour training session for this project (reviewing the definitions of the big-five personality traits, general mental ability, and academic performance; viewing Facebook profiles to identify specific information which could be used to assess the focal characteristics of the study; and familiarizing the participants with the rating form to be used when conducting the assessments), and participated in a series of practice assessments prior to conducting the ratings for this study (an assignment to evaluate SNWs and identify specific information that could be used to assess each of the focal characteristics of the study, follow-up class discussion of these observations, and a practice session in which each participant conducted two assessments of current Facebook users by using the researcher designed rating form). All participants had personal involvement with SNWs. Participants were asked to spend ten minutes evaluating each of the six social networking profiles, consider multiple aspects of the profiles which could relate to a specific trait, then complete the rating form based on their overall impression of the social networking profile.The choice of the three male and three female social networking profiles was randomly generated from a list of volunteers in an introductory management course. Along with volunteering to have their SNWs evaluated[5], the volunteers also completed demographics and personality questionnaires, an intelligence test, and consented to allow the researchers to obtain their grade point average (GPA) from the university registrar.Measures Judge ratings of big-five personality traits - measured with 25 items from the bipolar adjective checklist (Goldberg, 1992) on a nine-point scale.Judge ratings of general mental ability - a single item measure was used to measure intelligence quotient (IQ) based on Reilly and Mulhern (1995). Judges were asked to "Estimate the user's IQ. Remember that the average IQ is 100, and one-sixth of the population have IQs less than 85, with one-sixth scoring over 115."Judge ratings of performance - a single item measure was used to measure academic performance based on the format used to measure IQ. Judges were asked to "Estimate the user's GPA. Remember that an average GPA is 3.0 and the maximum is 4.0."Ratee self-reported big-five personality traits (referred to as true scores) - measured with 150 items from the international personality item pool - IPIP (Goldberg et al., 2006) on a five-point scale ranging from strongly disagree to strongly agree. The as ranged from 0.92 for conscientiousness to 0.81 for openness.Ratee general mental ability (IQ true score) - measured with the Wonderlic personnel test (Wonderlic, 2000), a 12 minute/50 question timed test of intelligence.Ratee performance (performance true score) - academic performance was obtained via GPA from the University Registrar. Although academic performance is less ideal than job performance in the context of employment selection, it represents an objective measure to test the hypotheses presented. The as for the judge ratings of personality were calculated for each of the six ratees. These six as were then averaged for each of the big-five to estimate the overall internal consistency of the scales. In order to assess H1, interrater agreement in the form of average measures intraclass correlation coefficients (ICCs) for the judge ratings are included in Table I. The scaled scores for the big-five personality traits and the single item scores for IQ and performance were evaluated for interrater agreement. The 378 total ratings (63 ratersx6 ratings each) were used to calculate the ICCs. The ICC values were all adequate, ranging from 0.93 for extroversion to 0.99 for conscientiousness and performance. Since ICCs are expected to be higher with a larger number of raters, Table II also includes the number of raters for each characteristic which would be necessary to achieve a 0.50 ICC value. Although there are no guidelines for level of agreement, 0.50 was used in the analyses as it should provide a minimum level of acceptable agreement across judges. The Spearman Brown prophecy formula was used to determine how many raters would be required to obtain an adequate (0.50) ICC value. Based on the 63 raters from this study, it was determined that between two (for conscientiousness and performance) and six (for emotional stability and extroversion) raters would be required to obtain a satisfactory level of interrater agreement.H2 was evaluated by conducting t-tests on score means in order to determine whether or not the means are statistically different from one another. In order to determine which means to test, the true scores (self-reported big-five personality scores, intelligence scores, and GPA) of the six rated subjects were evaluated. For each of the seven characteristics, the individual with the highest true score and the individual with the lowest true score were selected for analysis. Judge mean ratings for these subjects were then compared to determine whether or not raters are able to distinguish individuals high on a characteristic from those low on the same characteristic. This method also allows for evaluation of the direction of the relationship, such that (in addition to evaluating mean differences) the judge rating of the subject with the higher true score should be higher than the judge rating of the lower true score. Results demonstrate that the mean judge ratings for the subject highest on the seven characteristics were statistically different from the subjects lowest on those characteristics. In addition, with the exception of openness to experience, the judges mean ratings were higher for those with the highest true score, indicating the ability of judges to distinguish the traits of conscientiousness, emotional stability, agreeableness, extroversion, intelligence, and performance by evaluating SNWs[6].Post hoc analyses were conducted to determine the impact of intelligence and personality on judge consistency and accuracy. Prior research has demonstrated mixed findings related to the impact of personality traits of the rater on rating accuracy. Ambady et al. (1995) found that less sociable (extroverted) raters were more accurate, while Lippa and Dietz (2000) found that only openness indicated more accurate raters. In addition, narcissistic raters have been found to be less accurate (John and Robbins, 1994), which may be relevant to the big-five since narcissism relates strongly to neuroticism. Finally, intelligence has also been reported to positively relate to rater accuracy (Lippa and Dietz, 2000). In the current study, the 63 judges were asked to take the same intelligence and personality tests as the SNW subjects. The analyses conducted above were then re-evaluated based on high versus low groups based on intelligence and the big-five. Results show no difference in interrater agreement based on these characteristics. However, judges who are more intelligent and more emotionally stable were shown to be more accurate in their judgments. More specifically, when the raters were split into high and low groups based on intelligence scores (the 31 highest scores versus the 31 lowest scores), the high intelligence group significantly and more accurately differentiated between high and low characteristics for conscientiousness, emotional stability, openness, and performance. For example, with all 63 raters combined, the difference between rater means for conscientiousness in Table II is 0.38 (8.03 for the high ratee score and 7.65 for the low ratee score). When assessing high and low intelligence raters independently, the mean difference for the 31 high intelligence raters is 0.61, but only 0.14 for the 31 low intelligence raters. Thus, more intelligent raters seem to be more capable of assessing this trait than less intelligent raters. Similarly, raters who are the most emotionally stable also rate more accurately for conscientiousness, emotional stability, openness, and performance. For example, the mean difference across raters for high and low ratee conscientiousness is again 0.38, but is 0.73 for the 31 raters who are the most emotionally stable and 0.03 for the 31 raters who are the least emotionally stable. These results indicate the potential need for researchers to consider intelligence and emotional stability when selecting individuals who will serve as raters of characteristics such as personality. Based on the large volume of personal information available on SNWs, judges' ratings of the big-five dimensions of personality, intelligence, and global performance were consistent across the 63 raters in this study, demonstrating adequate internal consistency reliability and interrater agreement. In addition, the trained raters were able to accurately distinguish between individuals who scored high and individuals who scored low on four of the big-five personality traits, intelligence, and performance, providing initial evidence that raters can accurately determine these organizationally relevant traits by viewing SNW information.As stated earlier, other rated personality has been shown to predict job performance. Considering that other methods of other-reported personality are unlikely to be viable in an employment selection context, SNW ratings of personality may be a practical approach. Owing to the theoretical and methodological differences between self-reported and other-rated personality, it is likely that ratings of personality via SNWs will provide a context for incremental prediction of job performance beyond the predominant self-report approach. In addition, the differences in context between SNWs and a job interview (i.e. socially desirable responding in the job interview as well as the unique nature of information contained in SNWs) should similarly allow for unique prediction of job performance beyond what can be evaluated through personality assessment in the employment interview. This approach may be particularly valuable since these assessments take only a fraction of the time involved with other selection methods.This study is not without limitations. Although the analyses testing the consistency of the relationships of SNW ratings are based on 378 judge ratings from 63 raters, the analyses testing rater accuracy were conducted by testing for significant differences between the high and low performer on the seven characteristics for only six subjects. Future research should assess accuracy over a larger sample of subjects.We hope that the results of this preliminary study will not be used by organizations to support their use of SNWs in employment selection. Without further validation in a variety of studies, with larger samples and in a variety of organizational contexts, caution should be used when interpreting the implications of this study. This is particularly true given the potential for employer legal liability due to the vast amount of personal information available on SNWs. Information regarding gender, race, age, disabilities, and other criteria which should not be used when making hiring decisions will most certainly, consciously or not, influence who gets hired. Even if this information does not bias the hiring decision, disparate impact issues may still exist. Future research should also examine the potential issues of adverse impact and potentially illegal information in hiring decisions using personal information from SNWs. In addition, research should be conducted to compare assessments of SNWs to other employment selection methods, such as personality assessment, intelligence testing, and employment interviews.Based on the relative absence of research evidence in this newly developing area, particularly regarding the potential for adverse impact and the lack of validity evidence, we believe the most important practical implication of this paper is for organizations to use SNWs with these issues in mind. Organizational representatives assessing SNWs should ask themselves two important questions. First, is the organization assessing (or could be perceived as assessing) information which could lead to discrimination against a legally protected group? Second, is the specific social networking information used to help make a hiring decision valid in determining who will perform better on the job? The approach used in this paper of assessing personality traits, intelligence, or general performance begin to provide answers to these questions. Opens in a new window.Table I Judge rating as and ICCs Opens in a new window.Table II Differences in judge ratings means comparing high and low ratee scores
|
- Based solely on viewing social networking profiles, judges are consistent in their ratings across subjects and typically able to accurately distinguish high from low performers. In addition, raters who are more intelligent and emotionally stable outperformed their counterparts.
|
[SECTION: Value] Within the past few years, the phenomenon of social networking web sites (SNWs) on the internet has exploded into the mainstream. Further, this online information has begun to be used for purposes beyond its intended use. Owing to the vast amount of personal information on these web sites, employers have begun to tap into this information as a source of applicant data in an effort to improve hiring decisions. This study evaluates the use of the SNWs in employment selection. Specifically, can trained judges consistently and accurately assess important organizational characteristics such as personality, intelligence, and performance using only a target's SNW information? In addition, the use of this information may lead to discrimination against applicants, given the wide range of available personal information such as gender, race, age, religion, and disability status otherwise illegal to use when making employment decisions.Social networking web sites SNWs focus on building online communities of people who share interests and activities, or who are interested in exploring the interests and activities of others. Most provide a variety of ways for users to interact, such as e-mail and instant messaging services. SNWs are designed to connect users to each other and to visually display each individual's network of friends. The number of users for these web sites and the daily traffic created by these web sites are staggering. According to the "About Us" section of the various sites, MySpace is the largest with over 248 million registered users. Other SNWs also have millions of users registered, such as Facebook (110 million), Friendster (85 million), Hi5 (80 million), Orkut/Google (37 million), and LinkedIn (25 million).While these sites differ in the features that are available, most have a mechanism for posting pictures, music and videos, keeping blogs, sharing links, and displaying interests. These sites vary in user demographics. For example, although Facebook is currently open to anyone, it started as a high school and college web site exclusively, with about 90 percent of students registered for the site (van der Werf, 2006).Social networking web sites in selection Owing to the increasing prevalence of SNWs in conjunction with the large volume of information available to the viewer, employers have begun using SNWs to assist in the selection process for new employees. About 50 percent of the employers attending college career fairs use online technology, including both search engines and SNWs to screen candidates (Shea and Wesley, 2006). Currently, between 20 (Framingham, 2008) and 25 percent (Taylor, 2007; NACE, 2006) of managers are using SNWs to screen candidates, with 40 percent indicating that they are likely to use them within the next year (Zeidner, 2007). These employers apparently feel justified in electronic screening using SNWs. The president of a small Chicago consulting firm, when asked about an applicant that applied for an internship and had questionable content on his page, replied, "A lot of it makes me think, what kind of judgment does this person have? Why are you allowing this to be viewed publicly?" (Finder, 2006).While it may be common practice to monitor web site content it may not be legal. Lance Chou, Director of Career Development at Stanford University, noted that:[...] some employers might try and learn something about the student's personality and whether it would be appropriate for the job. However, there is information on Facebook that is not relevant to the job, but may be used inappropriately by employers to assess a candidate (Fuller, 2006).According to George Lenard, an employment attorney, employers can inadvertently learn about matters such as candidates' age, marital status, and other topics typically are off limits in job interviews, and organizations can be sued for discrimination if these candidates are not hired (Frauenheim, 2006). The question of whether employer monitoring of SNWs is illegal may relate to equal employment opportunity (EEO) law. EEO severely limits the type of information an interviewer may ascertain and use. For example, during interviews employers may not ask questions regarding race, religion, sexual preference, or marital status. However, all of this information can be easily located using SNWs (Kowske and Southwell, 2006). The ethics and legality behind electronic screening using SNWs has become an important issue for human resource (HR) practitioners (Zeidner, 2007).Since no academic studies, to our knowledge, have assessed whether using SNWs in employment selection are a reliable and valid predictor of important organizational outcomes, their value in employment screening is unknown. Until the reliability and validity of information on SNWs is established, employers should use caution when using SNWs to make hiring decisions. Just as with structured interview methods (Campion et al., 1997), focusing on job-related information in SNWs should minimize the use of less job-relevant information that might bias the hiring decision. This study focuses on showing the validity and reliability of three such categories of job-relevant characteristics which have potential to be rated consistently and accurately and to serve as valid predictors of job performance. The assessment of personality, intelligence, and a global measure of performance through SNWs provide multiple possible avenues regarding validity generalization of job-relevant characteristics, and thereby justifying the use of SNWs in the screening process.Assessment of personality The 1990s have seen a huge growth in the use of personality assessment within personnel selection practice and research (Barrick and Mount, 1991; Frei and McDaniel, 1997; Ones et al., 1993; Salgado, 1998; Tett et al., 1991). These studies provide positive evidence for the criterion-related validity of personality. When it comes to the prediction of overall job performance, conscientiousness was found to be the best predictor, showing consistent predictions across all occupational groups. In addition, extroversion, agreeableness, and neuroticism were shown to predict job performance in certain jobs. Finally, although typically unrelated to job performance, openness to experience has been found to predict training performance.Beyond the predominant focus on self-reported personality assessment, personality can also be measured using other-ratings. It is well established that people can assess the personality of others, even after relatively short exposure, but that the accuracy of these assessments depends on the information available to the observer (Barrick et al., 2000). Observer ratings may be more valuable than self-ratings in employment selection, particularly when targets are not available for self-reports, self-reports are untrustworthy, and researchers wish to improve accuracy by aggregating multiple raters (Hofstee, 1994; McCrae and Weiss, 2007). Other-rated personality via SNWs seems quite promising in the selection context, since other-ratings of personality have been found to predict job performance. Motowidlo and colleagues (1996) found that interviewer-rated extroversion and conscientiousness (r=0.27 and 0.20, respectively) significantly predict supervisor rated job performance. In addition, Mount et al. (1994) show that observer ratings of extroversion and conscientiousness from supervisors, coworkers, and customers significantly predicts sales performance, even beyond self-rated personality.The validity of other-rated personality, however, can depend on the relationship the subject has with the rater and the quality of the information available to the rater. A meta analysis conducted by Connolly and Viswesvaran (1998) show low accuracy for strangers in predicting personality and moderate prediction by peers for each of the big-five traits. In addition, Barrick et al. (2000) developed a personality-based job interview for the purpose of assessing the personality of the applicant. They found that personality-based job interviews could be used to accurately predict three of the big-five. Considering the average interview is approximately 40 minutes in length (Campion et al., 1997), it appears that an interviewer can assess some aspects of personality as effectively as a close acquaintance of the applicant. We propose that the information available in SNWs provides a similar means to assess personality.Furthermore, since SNW ratings are obtained from a wide range of personal information that is a reflection of ongoing behaviors and interactions with other users of the networks, web sites may actually provide unique information not found with other selection methods. Recent issues of Personnel Psychology (Morgeson et al., 2007a, b; Ones et al., 2007; Tett and Christiansen, 2007) and Industrial and Organizational Psychology (Hough and Oswald, 2008; Griffith and Peterson, 2008) have focused on the limited criterion validity of self-reported personality measures as well as the complex issue of social desirability and faking of self-reported personality measures. To increase validity and address issues of socially desirable responding "we should look at other ways of assessing personality. There are a variety of ways of finding out about people's stable pattern of behavior" (Morgeson et al., 2007a, b, p. 719). SNWs may prove to be an appropriate means of assessing personality in this way. Beyond personality assessment, similar issues emerge in relation to a host of employment selection methods. Resumes, interviews, job applications, and many other forms of employee selection include a certain element of self-presentation, reflecting "maximal" instead of "typical" work performance (Sackett, 2007; Sackett et al., 1988). Employment selection methods using social networking are likely to be based on "typical" behaviors, and therefore may be more accurate than other selection methods. At a minimum, this method should provide information that is distinct from "maximal" selection methods, thereby allowing for a stronger likelihood that using SNWs will yield incremental validity beyond established methods of employment selection[1].This is not to say that SNWs are not susceptible to manipulation and faking, similar to that of self-report personality measures and job interviews. In fact, as users of SNWs become more aware that their profiles are being evaluated by potential employers, information provided on profiles is likely to be skewed in an effort to be viewed more favorably. However, there are aspects of SNWs which would make the process of skewing information difficult. Much of the information present in a given social networking profile is submitted by other members of the network, such as tagged photos and writing on another's wall[2]. Although some negative information can be deleted by the user, the user has more limited control over this aspect of their profile. In addition, some of the information controlled by the users themselves would be difficult to fake. For example, extroversion may be tied to the number of friends a user has in the social network. Artificially inflating a substantial number of friends in the network would pose great difficulty, as the user cannot control who accepts their friend request. As another example, rater assessment of personality traits might be drawn in part from photos, which are similarly difficult to fake. Thus, while faking would appear to be an issue, it is likely that the impact of faking is less than with other selection methods. Future research should assess the impact of faking in the context of SNWs.Despite the potential for faking, Vazire and Gosling (2004) used personal web sites to accurately assess personality. Although personal web sites are similar in some ways to SNWs, they are used by such a small percentage of potential applicants that they are impractical for purposes of employment selection. In addition, SNWs provide additional information not included in personal web sites, such as a list of the user's friends and a list of the interest groups a user has joined. However, Vazire and Gosling provide initial evidence for the accurate prediction of personality using personal information available on the internet.The types of information available on SNWs may be particularly effective in predicting the big-five personality traits. SNWs contain various sources of information which could be used to assess behaviors related to personality. For example, the types and number of interest groups the user has joined, comments that have been left for the user, comments made by the user on other people's "walls," "tagging" photos, updating "status messages"[3], and listing books and intellectual interests in the "personal information" section[4]. These examples provide only a very preliminary introduction to the various sources of personal information available in SNWs which might indicate an individual's personality. It should also be noted that the use of SNWs may vary based on characteristics beyond personality. For example, an individual who is more adept with this form of technology may be more likely to participate in social networking and may post more information more frequently than those individuals lacking in these technological skills. Since age has been shown to relate to technology acceptance and use (Morris and Venkatesh, 2000), this also brings the issue of age into the possible groups which could be adversely impacted by the use of this type of technology by HR departments in hiring. Future research should explore this issue further.Assessment of intelligence Since the very earliest research on personnel selection, cognitive ability has been one of the major methods used to attempt to discriminate between candidates and to predict subsequent job performance (Robertson and Smith, 2001). Intelligence is the single most effective predictor known of individual performance at school and on the job (Gottfredson, 1998), accounting for approximately 25 percent of the variance in job performance (Hunter and Hunter, 1984). Cognitive ability provides criterion-related validity that generalizes across more or less all occupational areas (Robertson and Smith, 2001). In addition, judge ratings of intelligence have been shown to predict intelligence test scores (Borkenau et al., 2004). Furthermore, biographical data have been shown to predict intelligence (Schmidt and Hunter, 1998). These results provide some evidence that assessment of intelligence should be viable within the context of SNWs.Assessment of global performance Beyond personality and intelligence, SNWs may also contain additional information which may be useful in employment selection. Owing to the large volume of information contained in SNWs, information may also be obtained which relate to the user's writing skills, job experiences, or a variety of knowledge, skills, abilities, or other criteria which might relate to job or organizational fit in a given employment selection context. A more global assessment of performance would include a variety of information. In fact, due to the broad range of information available on SNWs and the lack of consistency in this information across individuals, the approach of assessing broad characteristics is likely to be more practical than assessing more narrow aspects of social networking profiles that may be unavailable and/or inconsistent for a large segment of the profiles.Rater consensus/interrater agreement Furr and Funder (2007) identify rater consensus as the first psychometric concern when assessing characteristics through judge-rated behavioral observation. Thus, the primary focus of this paper, as the first to address the potential use of SNW information for the purpose of employment selection, is to focus on the issue of rater consensus.To the degree to which SNWs provide a distinguishable and consistent basis of evaluation, judges will reach a certain level of consensus in their impressions of these SNWs. Thus, consensus is expected across judges:H1. Ratings of the big-five dimensions of personality, intelligence, and performance based on SNW information are consistent across raters.Accuracy of ratings A second psychometric concern is that of validity (Furr and Funder, 2007). If rater consensus is established, the next step will be to establish various forms of validity. This study evaluates validity by assessing whether judges are accurate in their assessment of personality intelligence, and performance.Personality, intelligence, and performance impressions based on judge assessments have been shown to be quite accurate, even when the amount of information is limited (Borkenau et al., 2004). This growing body of research suggests that people have a natural talent for judging one another accurately (Vazire and Gosling, 2004). Given the high volume of information available to assess behavioral cues on SNWs, a suitable level of accuracy is expected:H2. Raters assessing an individual's personality, intelligence, and performance through SNWs are able to distinguish between those individuals who are high on each characteristic from those who are low on that characteristic. Participants and procedures This study was conducted at a large public university in the southern USA. A total of 63 students enrolled in an employment selection course participated in this project for course credit. Participants were 49 percent male, 90 percent Caucasian, averaged 24 years of age, and worked an average of 26 hours per week. Participants had prerequisites in HRs and statistics. As part of the employment selection course, these participants were trained in both personality/intelligence testing and effective utilization of rating scales, participated in a one hour training session for this project (reviewing the definitions of the big-five personality traits, general mental ability, and academic performance; viewing Facebook profiles to identify specific information which could be used to assess the focal characteristics of the study; and familiarizing the participants with the rating form to be used when conducting the assessments), and participated in a series of practice assessments prior to conducting the ratings for this study (an assignment to evaluate SNWs and identify specific information that could be used to assess each of the focal characteristics of the study, follow-up class discussion of these observations, and a practice session in which each participant conducted two assessments of current Facebook users by using the researcher designed rating form). All participants had personal involvement with SNWs. Participants were asked to spend ten minutes evaluating each of the six social networking profiles, consider multiple aspects of the profiles which could relate to a specific trait, then complete the rating form based on their overall impression of the social networking profile.The choice of the three male and three female social networking profiles was randomly generated from a list of volunteers in an introductory management course. Along with volunteering to have their SNWs evaluated[5], the volunteers also completed demographics and personality questionnaires, an intelligence test, and consented to allow the researchers to obtain their grade point average (GPA) from the university registrar.Measures Judge ratings of big-five personality traits - measured with 25 items from the bipolar adjective checklist (Goldberg, 1992) on a nine-point scale.Judge ratings of general mental ability - a single item measure was used to measure intelligence quotient (IQ) based on Reilly and Mulhern (1995). Judges were asked to "Estimate the user's IQ. Remember that the average IQ is 100, and one-sixth of the population have IQs less than 85, with one-sixth scoring over 115."Judge ratings of performance - a single item measure was used to measure academic performance based on the format used to measure IQ. Judges were asked to "Estimate the user's GPA. Remember that an average GPA is 3.0 and the maximum is 4.0."Ratee self-reported big-five personality traits (referred to as true scores) - measured with 150 items from the international personality item pool - IPIP (Goldberg et al., 2006) on a five-point scale ranging from strongly disagree to strongly agree. The as ranged from 0.92 for conscientiousness to 0.81 for openness.Ratee general mental ability (IQ true score) - measured with the Wonderlic personnel test (Wonderlic, 2000), a 12 minute/50 question timed test of intelligence.Ratee performance (performance true score) - academic performance was obtained via GPA from the University Registrar. Although academic performance is less ideal than job performance in the context of employment selection, it represents an objective measure to test the hypotheses presented. The as for the judge ratings of personality were calculated for each of the six ratees. These six as were then averaged for each of the big-five to estimate the overall internal consistency of the scales. In order to assess H1, interrater agreement in the form of average measures intraclass correlation coefficients (ICCs) for the judge ratings are included in Table I. The scaled scores for the big-five personality traits and the single item scores for IQ and performance were evaluated for interrater agreement. The 378 total ratings (63 ratersx6 ratings each) were used to calculate the ICCs. The ICC values were all adequate, ranging from 0.93 for extroversion to 0.99 for conscientiousness and performance. Since ICCs are expected to be higher with a larger number of raters, Table II also includes the number of raters for each characteristic which would be necessary to achieve a 0.50 ICC value. Although there are no guidelines for level of agreement, 0.50 was used in the analyses as it should provide a minimum level of acceptable agreement across judges. The Spearman Brown prophecy formula was used to determine how many raters would be required to obtain an adequate (0.50) ICC value. Based on the 63 raters from this study, it was determined that between two (for conscientiousness and performance) and six (for emotional stability and extroversion) raters would be required to obtain a satisfactory level of interrater agreement.H2 was evaluated by conducting t-tests on score means in order to determine whether or not the means are statistically different from one another. In order to determine which means to test, the true scores (self-reported big-five personality scores, intelligence scores, and GPA) of the six rated subjects were evaluated. For each of the seven characteristics, the individual with the highest true score and the individual with the lowest true score were selected for analysis. Judge mean ratings for these subjects were then compared to determine whether or not raters are able to distinguish individuals high on a characteristic from those low on the same characteristic. This method also allows for evaluation of the direction of the relationship, such that (in addition to evaluating mean differences) the judge rating of the subject with the higher true score should be higher than the judge rating of the lower true score. Results demonstrate that the mean judge ratings for the subject highest on the seven characteristics were statistically different from the subjects lowest on those characteristics. In addition, with the exception of openness to experience, the judges mean ratings were higher for those with the highest true score, indicating the ability of judges to distinguish the traits of conscientiousness, emotional stability, agreeableness, extroversion, intelligence, and performance by evaluating SNWs[6].Post hoc analyses were conducted to determine the impact of intelligence and personality on judge consistency and accuracy. Prior research has demonstrated mixed findings related to the impact of personality traits of the rater on rating accuracy. Ambady et al. (1995) found that less sociable (extroverted) raters were more accurate, while Lippa and Dietz (2000) found that only openness indicated more accurate raters. In addition, narcissistic raters have been found to be less accurate (John and Robbins, 1994), which may be relevant to the big-five since narcissism relates strongly to neuroticism. Finally, intelligence has also been reported to positively relate to rater accuracy (Lippa and Dietz, 2000). In the current study, the 63 judges were asked to take the same intelligence and personality tests as the SNW subjects. The analyses conducted above were then re-evaluated based on high versus low groups based on intelligence and the big-five. Results show no difference in interrater agreement based on these characteristics. However, judges who are more intelligent and more emotionally stable were shown to be more accurate in their judgments. More specifically, when the raters were split into high and low groups based on intelligence scores (the 31 highest scores versus the 31 lowest scores), the high intelligence group significantly and more accurately differentiated between high and low characteristics for conscientiousness, emotional stability, openness, and performance. For example, with all 63 raters combined, the difference between rater means for conscientiousness in Table II is 0.38 (8.03 for the high ratee score and 7.65 for the low ratee score). When assessing high and low intelligence raters independently, the mean difference for the 31 high intelligence raters is 0.61, but only 0.14 for the 31 low intelligence raters. Thus, more intelligent raters seem to be more capable of assessing this trait than less intelligent raters. Similarly, raters who are the most emotionally stable also rate more accurately for conscientiousness, emotional stability, openness, and performance. For example, the mean difference across raters for high and low ratee conscientiousness is again 0.38, but is 0.73 for the 31 raters who are the most emotionally stable and 0.03 for the 31 raters who are the least emotionally stable. These results indicate the potential need for researchers to consider intelligence and emotional stability when selecting individuals who will serve as raters of characteristics such as personality. Based on the large volume of personal information available on SNWs, judges' ratings of the big-five dimensions of personality, intelligence, and global performance were consistent across the 63 raters in this study, demonstrating adequate internal consistency reliability and interrater agreement. In addition, the trained raters were able to accurately distinguish between individuals who scored high and individuals who scored low on four of the big-five personality traits, intelligence, and performance, providing initial evidence that raters can accurately determine these organizationally relevant traits by viewing SNW information.As stated earlier, other rated personality has been shown to predict job performance. Considering that other methods of other-reported personality are unlikely to be viable in an employment selection context, SNW ratings of personality may be a practical approach. Owing to the theoretical and methodological differences between self-reported and other-rated personality, it is likely that ratings of personality via SNWs will provide a context for incremental prediction of job performance beyond the predominant self-report approach. In addition, the differences in context between SNWs and a job interview (i.e. socially desirable responding in the job interview as well as the unique nature of information contained in SNWs) should similarly allow for unique prediction of job performance beyond what can be evaluated through personality assessment in the employment interview. This approach may be particularly valuable since these assessments take only a fraction of the time involved with other selection methods.This study is not without limitations. Although the analyses testing the consistency of the relationships of SNW ratings are based on 378 judge ratings from 63 raters, the analyses testing rater accuracy were conducted by testing for significant differences between the high and low performer on the seven characteristics for only six subjects. Future research should assess accuracy over a larger sample of subjects.We hope that the results of this preliminary study will not be used by organizations to support their use of SNWs in employment selection. Without further validation in a variety of studies, with larger samples and in a variety of organizational contexts, caution should be used when interpreting the implications of this study. This is particularly true given the potential for employer legal liability due to the vast amount of personal information available on SNWs. Information regarding gender, race, age, disabilities, and other criteria which should not be used when making hiring decisions will most certainly, consciously or not, influence who gets hired. Even if this information does not bias the hiring decision, disparate impact issues may still exist. Future research should also examine the potential issues of adverse impact and potentially illegal information in hiring decisions using personal information from SNWs. In addition, research should be conducted to compare assessments of SNWs to other employment selection methods, such as personality assessment, intelligence testing, and employment interviews.Based on the relative absence of research evidence in this newly developing area, particularly regarding the potential for adverse impact and the lack of validity evidence, we believe the most important practical implication of this paper is for organizations to use SNWs with these issues in mind. Organizational representatives assessing SNWs should ask themselves two important questions. First, is the organization assessing (or could be perceived as assessing) information which could lead to discrimination against a legally protected group? Second, is the specific social networking information used to help make a hiring decision valid in determining who will perform better on the job? The approach used in this paper of assessing personality traits, intelligence, or general performance begin to provide answers to these questions. Opens in a new window.Table I Judge rating as and ICCs Opens in a new window.Table II Differences in judge ratings means comparing high and low ratee scores
|
- This paper is the first, as far as the authors are concerned, to address the use of SNWs in employment selection, despite their current utilization by HR practitioners.
|
[SECTION: Purpose] Stringent labour market constraints are expected to pose serious obstacles to firm performance and economic growth. A wide range of literature finds that rigid labour regulations would induce lower labour force participation and higher unemployment rates (e.g. Botero et al., 2004; Besley and Burgess, 2004; Amin, 2009; Djankov and Ramalho, 2009), and would prevent labour markets from being efficient leading to losses in productivity (e.g. Kaplan, 2009). Another strand of literature inspects the problem of labour skill shortages or "skill deficits", which can be defined as the divergence between the educational attainments of workers and the skill requirements of jobs (Kiker et al., 1997). This literature regularly indicates that accentuated labour skill shortages impose significant restrictions on employment creation and economic growth (e.g. Pissarides and Veganzones-Varoudakis, 2007; Bhattacharya and Wolde, 2012), and could eventually inflict severe impacts on economic performance and labour market outcomes (e.g. Allen and van der Velden, 2001). The implications of labour market constraints are deemed to be of particular concerns for the Middle East and North Africa (MENA) region. Although the MENA region has generally realized higher annual employment growth rates compared to other geo-economic regions (International Labour Organization (ILO), 2012), rigid labour regulations and lack of suitable labour skills were often identified as main obstacles to firm operation and development in the MENA region (Drzeniek-Hanouz and Dusek, 2013). MENA countries have one of the least flexible labour markets, which remain too tight compared to labour markets in other developing countries (Angel-Urdinola and Kuddo, 2010; Bhattacharya and Wolde, 2012 [1]). Table I presents the percentages of firms identifying different business obstacles as major or very severe (henceforth, major/severe) constraints across geo-economic regions. The statistics emphasize that the MENA region has the highest percentages of firms that consider labour regulations and labour skill shortages to be major/severe business constraints compared to other geo-economic regions. For example, labour regulations and labour skill shortages are identified as major/severe problems to 25 and 38 per cent of firms located in the MENA region, respectively, compared to only 9 and 16 per cent of firms located in South Asia[2]. Also, in the case of the MENA region, Table I shows that the constraint of labour skill shortages is associated with one of the highest percentages compared to other business obstacles, whereas the constraint of labour regulations occupy an intermediate rank. Understanding the factors influencing employability constraints constitute an essential requirement towards the implementation of policies aiming at improving the regulatory and institutional environment for business (Pierre and Scarpetta, 2004; Kaplan and Pathania, 2010). This is particularly relevant for MENA countries that have recently undergone significant labour reforms, where employment creation is listed among the primary objectives (Angel-Urdinola and Kuddo, 2010). Several studies underline systematic differences in firm perceptions of labour market constraints through firm characteristics, as well as across countries and sectors (e.g. Pierre and Scarpetta, 2004; Kaplan and Pathania, 2010; Lyon et al., 2012; Meyer and Vandenberg, 2013). Therefore, designing policies that alleviate the adverse implications of labour market constraints requires a comprehension of the categories of firms that are more likely to endure the burden of these constraints. This paper examines the implications of firm characteristics for the perceptions of firms located in the MENA region concerning the stringency of labour regulations and labour skill shortages. It also examines the heterogeneity in firm perceptions of labour market constraints across MENA countries and across sectors. We use a data set sourced from the World Bank's Enterprise Surveys database, which includes information on firm perceptions of many business constraints. Firm perception data are collected through surveys, which are generally answered by senior managers and business owners. Hence, following the previous literature (e.g. Pierre and Scarpetta, 2004; Kaplan and Pathania, 2010; Meyer and Vandenberg, 2013), our dependent variables are determined as perception-based indicators. We implement overall examination and comparative cross-country and cross-sector empirical analyses of the factors influencing firm perceptions of labour regulations and labour skill shortages for the MENA region. MENA countries are characterized by a considerable level of diversity in macroeconomic and industrial factors, and by important differences in labour market conditions and institutions (Angel-Urdinola and Kuddo, 2010), suggesting potential cross-country heterogeneity. Also, cross-sector heterogeneity is expected given the varying structures in labour requirements between sectors. Through the basic regressions, we use a bivariate probit model that allows for firm perceptions of labour regulations and labour skill shortages to be jointly formulated. This estimation model takes into account potential correlation between the error terms in the two labour market constraints' equations. 2.1. Labour regulations Labour regulations are normally expressed through laws that regulate and govern the employment relationships between employees, employers, unions, and government in order to protect the basic standards of fair treatment for workers (Kaplan, 2009) and to maximize the social welfare (Botero et al., 2004). Governments commonly intervene in labour markets through regulations because employers may mistreat workers, leading to unfairness and inefficient outcomes for workers such as unfair dismissals, unfair minimum wage, lay-offs for economic reasons, and under-payment (Djankov and Ramalho, 2009 [3]). These regulations and laws may be written through labour codes, current legislations, and norms set by collective agreements (Pierre and Scarpetta, 2004). Several studies report variations in firm perceptions of labour regulations through firm characteristics and across countries and sectors. Gelb et al. (2007) use data covering firms located in 26 Sub-Saharan African countries to examine the implications of various factors for firm perceptions of the business environment. Among the results, they find that large firms have higher propensities to complain about the rigidity of labour regulations. They argue that, in developing countries, small firms are more likely to work within restricted markets and, therefore, they are less visible to regulators, and are less appealing targets for officials. They also find that exporting activities and private foreign ownership do not exhibit statistically significant impacts on firm perceptions of labour regulations. They also report that firm perceptions of labour regulations do not exert considerable variations across sectors. Hallward-Driemeier and Aterido (2009) investigate the role of firm characteristics in determining firm perceptions of various business constraints, using data covering firms located in 105 countries. The results suggest that small firms may face more problems in dealing with the requirements of poor labour regulations, and may become easier targets in corrupted environments in developing countries. Clarke (2010) uses data covering firms located in South Africa to study the factors influencing firm perceptions of business constraints. Among the results, he finds that large firms and exporting firms are more likely to report labour regulations as a significant obstacle to their development. Also, he finds that firm age and type of firm ownership do not exercise statistically significant role in explaining the variations in firm perceptions of labour regulations. Bartelsman et al. (2010) document how business constraints, as perceived by individual firms, differ across countries and across firm characteristics. They show that firm age, ownership type, export orientation, industry binary variables, and country binary variables are important indicators in explaining the differences in firm perceptions of business constraints between Europe and Central Asia (ECA) and Latin America and the Caribbean regions. The results also show that large firms and exporting firms in the ECA region are more likely to be affected by rigid labour regulations, reflecting costs of labour adjustment. Vargas (2012) implements the analysis for firms located in Bolivia, focusing on the implications of firm size for firm perceptions of business constraints. Among the results, he finds that small firms are less likely to identify labour regulations as an important business constraint. Also, he finds that manufacturing firms are more likely to report labour regulations as a significant business constraint compared to firms belonging to other sectors. Meyer and Vandenberg (2013) analyse the relationship between firm characteristics and firm perceptions of labour regulations using data covering firms located in five Asian economies (Bangladesh, Indonesia, Pakistan, the Philippines, and Vietnam). They find systematic differences in firm perceptions of labour regulations across firm characteristics and sectors. In four of the five countries examined, they find that exporting firms are more likely to perceive labour regulations as an important obstacle to their operations compared to non-exporting firms. They also find that, in four of the five countries, large firms have higher propensities to perceive labour regulations as a significant business constraint. They argue that small firms may enjoy de facto or de jure exemption from the enforcement of labour regulations. Also, they show that young firms have higher tendencies to report difficulties in coping with labour regulations. 2.2. Labour skill shortages Mismatches between worker skills and job requirements are often classified among the significant obstacles facing firm productivity and growth, particularly in developing countries (Almeida and Aterido, 2011). For example, skill mismatches form an important constraint on hiring decisions, leading to higher levels of unemployment and important costs for firms (O'Sullivan et al., 2011) and causing higher turnover rates (Hersch, 1991). Gupta et al. (2010) argue that weak performances of labour-intensive industries could be related to labour skill shortages. Piore (1986) suggests that a low-skilled workforce may decrease the internal flexibility, leading to inefficient functioning of firms. Empirical evidence on the factors influencing firm perceptions of labour skill shortages points out to large variations through firm characteristics and across countries and sectors. Gelb et al. (2007) argue that the problem of labour skill shortages tends to be more prevalent for firms that are large, more productive, and use advanced technologies. Additionally, they note that small firms may use less sophisticated production techniques that require fewer skilled workers. Using data set covering firms located in Sub-Saharan Africa, Gelb et al. (2007) find that large firms are more likely to report labour skill shortages as one main business constraint. They also find that exporting activities and type of firm ownership do not exhibit important influences on firm perceptions of labour skill shortages. They note that the availability of skilled labour may become a more binding constraint as economies become more sophisticated and as governments magnify their capacities in enforcing labour regulations. Hallward-Driemeier and Aterido (2009) find that large firms and exporting firms are more likely to report labour skill shortages as a significant obstacle to business operations. They also find that government-owned firms are less likely to perceive labour skill shortages as an important business constraint. In the case of firms located in South Africa, Clarke (2010) finds that large firms and exporting firms have higher propensities to perceive labour skill shortages as a significant business obstacle. He also reports that firm age and type of firm ownership do not exhibit important influences on firm perceptions of labour skill shortages. Kaplan and Pathania (2010) find that large firms are more likely to identify labour skill shortages as one main business constraint. However, Vargas (2012) does not find significant influences of firm size, but he reports important variations across sectors. Using data covering firms located in Canada, Sabourin (2001) shows that firm size and technological intensity are negatively correlated with the probability of firms to identify labour skill shortages as one primary business obstacle. Also, he finds that type of firm ownership and Research and Development activities have no statistically significant influences on these firm perceptions. Baldwin and Lin (2002) find that young firms are more likely to designate labour skill shortages as a considerable problem using data covering Canadian manufacturing firms. However, they find that firm size and ownership type do no exhibit important influences. Green et al. (1998) implement the empirical analysis for firms located in the UK. They find that higher skilled share of employment accentuates the problem of labour skill shortages as perceived by employers. They also report that the share of part-time workers does not exhibit important effects. Lyon et al. (2012) find that medium and large size firms are more likely to designate labour skill shortages as an important constraint on firm operations and growth. They find that exporting firms have lower propensities to report labour skill shortages as a significant business constraint compared to non-exporting firms. Also, they do not find important variations across sectors. The empirical analysis is carried out for the perceived levels of labour market constraints as reported by the respondents (e.g. senior managers, business owners) through the World Bank's Enterprise Surveys database. Pierre and Scarpetta (2004, 2006) examine the relationship between the perceived and actual stringency of labour regulations using national labour protection indices (i.e. de jure labour laws). They find that the reported perceptions are closely related to the actual levels of labour regulations' constraints. Specifically, countries with higher national indices on the stringency of labour regulations are associated with higher proportions of firms perceiving labour regulations as being significant constraints. Kaplan and Pathania (2010) indicate that the perception-based indicators capture the actual institutional environment (e.g. laws, regulations, infrastructure, corruption), particularly when objective measures are inadequate. They argue that the perception-based indicators reflect the de facto severity of business constraints, and serve as a normal check vis-a-vis national indicators. Clarke (2010) notes that the collection of objective measures could be difficult, particularly when it comes to sensitive indicators (e.g. corruption). In such cases, perception-based indicators can serve as a substitute. Hallward-Driemeier and Aterido (2009) find that the perception-based and objective measures of business constraints are closely correlated with each other. They indicate that subjective measures enable the ranking of business constraints within and between countries, particularly when comparing the rigidity of the business environment. Also, they argue that perception-based measures provide information on the most severe obstacles for businesses growth and development as identified by the business entity, and would eventually assist managers and policy makers to recognize reform priorities. Hence, perception-based measures tend to reflect the actual rigidity of business constraints faced by firms. However, there are some concerns regarding perception-based measures. Specifically, these measures could be subjected to: first, potential measurement errors leading to biased results; second, variations in managers willingness to report actual negative or positive responses due to differences in personalities, views, and culture; third, performance bias where managers tend to rank business obstacles according to the performance of their firms or to the business environment conditions through which the firm operates; and finally, defects due to imperfections in question formulations and answer interpretations (Bertrand and Mullainathan, 2001; Senik, 2005; Hallward-Driemeier and Aterido, 2009). Some studies (e.g. Glaeser et al., 2004; Kaplan and Pathania, 2010) argue that caution should be exercised when using perception-based measures as independent variables through empirical analyses. This is because these measures may embody information on growth, which would lead to reverse causality issues. Meanwhile, these studies indicate that using these measures as dependent variables is more intuitive to capture growth in the business environment. These indications complement the aforementioned empirical findings reflecting a close correspondence between perceived and actual measures of business constraints. It is important to note that the empirical analysis in this paper exclusively covers firms that exist in the market (i.e. those that have already entered the market, and that did not exit the market). This is to say that the empirical analysis examines the business obstacles as perceived by the existing firms in the market. Hausmann and Velasco (2005) depict this issue through a camel-hippopotamus analogy. Camels living in the desert do not identify access to water as a primary problem since they have adjusted to the situation by learning how to conserve water. Meanwhile, hippopotami lack the ability to stay in the desert since water scarcity is a critical obstacle for their survival. Hence, by generating and analysing data from interviews with "camels" about water constraints, we may be missing distinct information from "hippopotami". Gelb et al. (2007) argue that such selection is incomplete since many firms choose to enter the market despite the severity of business constraints. They also indicate that the ability to adjust to a business constraint does not mean that firms do not recognize it anymore as a serious problem. They provide an illustration using the World Bank's Enterprises Surveys database: the perceptions of generator-owning firms regarding the electricity as a business constraint are not distinguishable from the perceptions of firms that do not own generators. Furthermore, they note that the intensity of complaints are often highly correlated with corresponding national indices. Dethier et al. (2011) suggest that empirical analyses covering "hippopotami" should be carried out using different frameworks (e.g. entry or exit models[4]). We use a data set sourced from the World Bank's Enterprise Surveys database. This database represents a comprehensive source of firm-level data in emerging and developing economies. It covers firms operating in the manufacturing, service, and other sectors. It contains information on various aspects of the business environment such as, access to finance, corruption, workforce characteristics, innovation and technology, and trade. It should be noted that one of many advantages of using data from these surveys is that the questions are identical through firms across all countries. The basic data set used in this paper covers 5,052 firms located in eight developing Arab countries of the MENA region: Algeria, Egypt, Jordan, Lebanon, Morocco, Oman, Syria, and Yemen[5]. Through these surveys, firms' representatives (e.g. senior managers, business owners) are asked whether labour regulations and labour skill shortages are considered prominent constraints on business operations and development. The responses are used to generate the basic dependent binary variables, which equal one when a firm declares the corresponding labour market constraint as a major/severe business obstacle and zero otherwise. Following the previous literature on business constraints (e.g. Gelb et al., 2007; Kaplan and Pathania, 2010; Clarke, 2010; Meyer and Vandenberg, 2013), the explanatory variables cover firm characteristics' variables which comprise: first, a binary variable that equals one for firms characterized by private foreign ownership and zero otherwise[6]; second, a binary variable that equals one for firms engaged in exporting activities and zero otherwise; third, firm size depicted through the number of total employment and presented in hundreds of workers through the regressions; fourth, firm age measured by the number of years since firm establishment; fifth, firm use of Information and Communication Technology (ICT) depicted through a binary variable that takes the value of one for firms regularly using the internet to communicate with clients and to source information and zero otherwise; finally, two firm labour composition variables: the first is represented through the fraction of skilled production workers in total number of production workers, and the second is depicted through the fraction of non-production workers in total employment. The use of skilled production workers' share and the use of non-production workers' share through the regressions imply that the results are relative to the use of unskilled production workers and to the use of production workers, respectively. The explanatory variables also include country-specific binary variables and sector-specific binary variables to capture cross-country and cross-sector variations. Table II displays summary statistics of the variables used in the empirical analysis. The first panel shows the results for the dependent variables. We find that around 24.6 per cent of firms in the regression data set report that labour regulations represent a major/severe business constraint, while 38.0 per cent of firms report that labour skill shortages exercise a major/severe business constraint[7]. The second panel shows statistics for the explanatory variables. The results indicate that 4.8 per cent of firms are characterized by private foreign ownership and that 22.5 per cent of firms are engaged in exporting activities. The average firm size is 124.0 full-time workers (with a standard deviation of 317.1 full-time workers) and the average firm age is 20.9 years (with a standard deviation of 16.6 years). Also, we find that 50.4 per cent of firms use the internet to communicate with clients and to source information. The percentage of skilled production workers in total number of production workers and the percentage of non-production workers in total employment have averages of 61.6 and 26.1 per cent, respectively. Consider a given firm j (j=1, ..., J) belonging to sector k (k=1, ..., K) and located in country c (c=1, ..., C). Firm perception levels of constraints related to labour regulations and those related to labour skill shortages are depicted through the latent variables (Equation 1) and (Equation 2), respectively. These latent variables are not observed. However, we observe the perceptions of firms through dichotomous responses on whether labour regulations and labour skill shortages do or do not pose major/severe obstacles on firm operations and development. Let R jkc depict a binary variable that takes the value of one when the corresponding firm identifies labour regulations as a major/severe business constraint and zero otherwise. Also, let S jkc represent a binary variable that takes the value of one when the corresponding firm identifies labour skill shortages as a major/severe business constraint and zero otherwise. The benchmark empirical specifications can be represented as: (Equation 3) (Equation 4) where X j is the vector of variables depicting firm characteristics, Y k is the vector of binary variables depicting sectors, Z c is the vector of binary variables depicting countries, and (Equation 5) and (Equation 6) are the stochastic error terms of the corresponding equations. The univariate probit estimator would produce biased estimates when there are some unobserved or omitted characteristics that simultaneously affect firm perceptions of labour regulations and labour skill shortages (Deadman and MacDonald, 2004; Greene, 2008). We allow for the errors in these two labour market constraints' equations to be potentially correlated. Thus, the two equations are jointly modelled using a bivariate probit estimator. The error terms are assumed to be independently and identically distributed as bivariate normal with r depicting the correlation parameter. Specifically, we have: (Equation 7) Through the empirical analysis, we use the Wald test to determine whether the correlation parameter r is statistically significant[8]. The rejection of the null hypothesis indicates that firm perceptions of the labour market constraints are jointly formulated (Greene, 2008 [9]). The bivariate probit specification allows us to estimate unconditional marginal effects, but also conditional and joint marginal effects of variables influencing firm perceptions of the labour market constraints. Table III presents the marginal effects from the benchmark bivariate probit estimation carried out for the pooled data set covering existing firms' perceptions of labour market constraints. The Wald test rejects the null hypothesis of zero correlation between the errors in the two labour market constraints' equations and, hence, it indicates that the model should be estimated through the bivariate probit estimator rather than through the univariate probit estimator. The estimated coefficient of correlation between the errors in the two equations is positive and statistically significant at the 1 per cent level. Table III displays the unconditional marginal effects for Pr(R jkc =1) and Pr(S jkc =1). It also includes the joint marginal effects for Pr(R jkc =1, S jkc =1) and Pr(R jkc =0, S jkc =0), and the conditional marginal effects for Pr(R jkc =1|S jkc =1) and Pr(S jkc =1|R jkc =1[10]). These marginal effects are determined at the mean values of the explanatory variables. For explanatory binary variables, the marginal effects are calculated through discrete changes in probabilities as the binary variable changes from 0 to 1. 6.1. Firm characteristics We find that larger firms have lower propensities to identify labour market constraints as major/severe obstacles facing business operations and development. This is in line with the results reported in some other empirical studies (e.g. Pierre and Scarpetta, 2004; Hallward-Driemeier and Aterido, 2009). The unconditional marginal effects imply that an increase in firm size by one hundred full-time workers reduces the likelihood of firms to perceive labour regulations and labour skill shortages as major/severe constraints on business operations by 0.8 and 0.5 percentage points, respectively. Also, the joint marginal effect of firm size for Pr(R jkc =1, S jkc =1) indicates that an increase in firm size by 100 workers decreases the likelihood of firms to jointly identify labour regulations and labour skill shortages as major/severe business constraints by 0.5 percentage points. The results on the relationship between firm size and firm perceptions of labour regulations are consistent with arguments provided in the literature, which suggest that smaller firms have more limited abilities to realize internal adjustments in response to labour regulations (Meyer and Vandenberg, 2013). Hence, smaller firms have higher tendencies to be more adversely affected by labour regulations. These results are also in line with some other findings showing that labour regulations have disproportional effects on small firms (Aterido et al., 2011). The results on the relationship between firm size and firm perceptions of labour skill shortages imply that smaller firms in MENA countries may encounter more difficulties in realizing future growth, since adequacy of educated workforce is essential for firm competitiveness and performance (Storey, 1994; Jensen and McGuckin, 1997). These results could also suggest that smaller firms have more difficulties in accessing skilled labour markets, identifying required labour skills, and investing in workforce training programmes (Jansen and Lanz, 2013). The implication of exporting activities for firm perceptions of labour regulations is not statistically significant. These results are similar to those found in some previous studies, such as Gelb et al. (2007) in the case of Sub-Saharan African firms. Yet, they deviate from those reported in some studies covering firms located in other developing regions. For example, Meyer and Vandenberg (2013) find that exporting firms located in some South East Asian countries have higher tendencies to perceive labour regulations as a major/severe business constraint compared to non-exporting firms. They attributed these findings to variations in labour demand requirements between exporting and non-exporting firms. Also, Lyon et al. (2012) find that non-exporting firms are more likely to identify labour skill shortages as a major/severe business constraint compared to exporting firms. The effect of firm age is negative and statistically significant in the case of labour skill shortages' equation. The unconditional marginal effect and the conditional marginal effect for Pr(S jkc =1|R jkc =1) indicate that an increase in firm age by ten years reduces the likelihood to perceive labour skill shortages as a major/severe business constraint by 1 percentage point. Although relatively small in magnitude, this impact is consistent with some a priori expectations discussed in the literature, suggesting that older firms have more accumulated experiences to attenuate the effects of business obstacles as perceived by the survey respondents (Meyer and Vandenberg, 2013). We find that an increase in the proportion of non-production workers in total employment by 10 percentage points raises the likelihood of firms to identify labour skill shortages as a major/severe business constraint by 1.3 percentage points. Also, we find that an increase in this proportion by 10 percentage points raises this likelihood by 1.6 percentage points, when assessed conditional on identifying labour regulations as a major/severe business constraint. The effect of the proportion of skilled production workers in total number of production workers on firm perceptions of these labour market constraints is not statistically significant. These findings suggest that the implications of labour skill shortages are primarily associated with the inadequacy of non-production workers rather than with the inadequacy of skilled production workers. The results show that firms with private foreign ownership are less likely to perceive labour skill shortages as a major/severe business constraint by 6.0 percentage points. They suggest that foreign affiliates have higher capabilities in attenuating the implications of labour skill shortages compared to local, domestically owned firms. Almeida (2007) indicates that foreign-owned firms have higher tendencies to pay higher wages and to attract educated workforce compared to domestic firms. These tendencies could eventually mitigate the perceptions of firms with private foreign ownership concerning labour skill shortages compared to firms with domestic ownership. 6.2. Countries' binary variables The results emphasize significant variations in the perceptions of firms located in different countries of the MENA region concerning labour market constraints. Hence, they underscore the implications of national characteristics for these firm perceptions. The binary variable depicting firms located in Algeria is set as the reference for MENA countries through the regressions. The unconditional marginal effects reveal that firms located in Lebanon have statistically higher propensities to perceive labour regulations as a major/severe business constraint by 34.4 percentage points compared to firms located in the Maghreb countries (i.e. Algeria and Morocco) and in Yemen. The corresponding propensities for firms located in Egypt, Oman, and Syria are higher by 17.3, 31.9, and 22.3 percentage points, respectively, compared to firms located in the two Maghreb countries and in Yemen. Also, firms located in Egypt, Lebanon, and Oman have statistically higher propensities to perceive labour skill shortages as a major/severe business constraint by 15.4, 17.5, and 13.5 percentage points, respectively, compared to firms located in the other MENA countries[11]. These results complement other empirical findings in the literature, which document international variations in the perceived magnitude of labour market constraints. For example, Pierre and Scarpetta (2004) show that countries with more stringent de jure labour laws have higher proportions of firms identifying labour regulations as being restrictive. Gelb et al. (2007) find significant variations in firm perceptions of labour market constraints, among other business obstacles, across Sub-Saharan African countries. Also, Lyon et al. (2012) present some important regional differences in firm perceptions of labour skill shortages. Drzeniek-Hanouz and Dusek (2013) indicate that labour market constraints constitute significant obstacles on business development in the Arab world. They also describe considerable variations in the perceptions of firms located in different MENA sub-regions: North Africa (including Egypt), the Levant, and the Gulf sub-regions. 6.3. Sectors' binary variables We find that firm perceptions of labour market constraints vary across sectors. The binary variable for firms belonging to the service sector is set as the reference for sectors through the regressions. The unconditional marginal effects indicate that firms belonging to the manufacturing sector and to the category covering the remaining other (non-manufacturing and non-service) sectors have higher propensities to perceive labour regulations as a major/severe business constraint by 9.9 and 10.1 percentage points, respectively, compared to firms belonging to the service sector. The estimates for the labour skill shortages' equation are 12.6 and 11.8 percentage points, respectively. These results suggest that labour market constraints, as perceived by the survey respondents, disproportionally impact manufacturing firms and firms belonging to the other sectors compared to firms belonging to the service sector. Firm perceptions of labour market constraints can be alternatively represented through ordinal responses in the data set. Specifically, firms identify the stringency of labour regulations and labour skill shortages as: no or minor obstacle (outcome 1), moderate obstacle (outcome 2), and major or very severe obstacle (outcome 3). Table IV displays the marginal effects obtained through the ordered probit model. The results are comparable to the corresponding marginal effects for Pr(R jkc =1) and Pr(S jkc =1) that are reported in Table III. 7.1. Empirical results by sector The results from the empirical analysis implemented for data sets covering individual sectors are displayed in Table V. The estimated coefficients of correlation between the errors in the two labour market constraints' equations are positive and statistically significant at the 1 per cent level for all sectors. To save space, Table V does not display the rows of marginal effects of firm characteristics that do not exhibit statistical significance at all. In general, the results for firms belonging to the manufacturing sectors are qualitatively similar to the benchmark results for pooled data set. We also find few differences. For example, the perceptions of manufacturing firms located in Jordan concerning labour skill shortages appear to be less pronounced compared to manufacturing firms located in the other MENA countries. Firms located in Jordan have statistically lower propensities to perceive labour skill shortages as a major/severe business constraint compared to firms located in Algeria and those located in Egypt by 11.3 and (11.3+14.0=) 25.3 percentage points, respectively. The results derived from the service sector's data set reveal some notable differences compared to those obtained from the manufacturing sector's data set[12]. In the case of the service sector, firm perceptions of labour regulations do not exhibit statistically significant heterogeneity across firm characteristics. We find that firm perceptions of labour skill shortages are accentuated with an increase in the ratio of non-production workers to total employment. The marginal effect shows that an increase in this ratio by 10 percentage points raises the likelihood of perceiving labour skill shortages as a major/severe business constraint by 3.1 percentage points. The magnitude of this effect is considerably higher than the one reported for manufacturing firms. Our data set does not contain service firms located in Algeria that report labour regulations and labour skill shortages as major/severe constraints on business operations. In other words, service firms located in Algeria predict failure perfectly and, therefore, they are dropped from the service sector's data set. Then, the binary variable depicting firms located in Egypt is set as the reference for MENA countries. We find that service firms located in Egypt and Lebanon have statistically higher likelihoods than those located in Jordan to perceive labour regulations as a major/severe business constraint by 6.7 and (6.7+2.1=) 8.8 percentage points, respectively. The corresponding likelihood for service firms located in Oman is considerably higher than those located in Jordan by (18.4+6.7=) 25.1 percentage points. We also find that service firms located in Egypt and Lebanon have lower propensities than those located in Jordan to identify labour skill shortages as a major/severe business constraint by 16.1 and (16.1-1.3=) 14.8 percentage points, respectively. The joint marginal effects for Pr(R jkc =1, S jkc =1) indicate that service firms located in Egypt and Jordan have statistically comparable likelihoods to identify both labour market constraints as major/severe obstacles on business operations as perceived by the survey respondents. These likelihoods are lower than the corresponding likelihood for service firms located in Oman by more than 11 percentage points. Finally, the results for the other sectors' category reveal that firms located in Oman are statistically more likely to perceive labour regulations as a major/severe business constraint by 34.9 percentage points compared to firms located in the other MENA countries (except Lebanon). Also, we find that firms located in Egypt and Lebanon have higher propensities to perceive labour skill shortages as a major/severe business constraint by 44.2 and 31.3 percentage points, respectively, compared to firms located in the other MENA countries. 7.2. Empirical results by country The results from the empirical analysis carried out for data sets covering firms located in individual MENA countries are displayed in Table VI. The estimated coefficients of correlation between the errors in the two labour market constraints' equations are positive and statistically significant at the 1 per cent level for all countries. To save space, Table VI does not display the rows of marginal effects of firm characteristics that do not exhibit statistical significance at all. In the case of Algeria, service firms predict failure perfectly and, hence, they are dropped from the data set. Then, the other sectors' category is set as the reference for sectors. We find that the perceptions of firms located in Algeria concerning labour regulations do not exhibit statistically significant variations between the manufacturing sector and the other sectors' category. Exporting firms are less likely to perceive labour regulations as a major/severe business constraint by 13.5 percentage points compared to non-exporting firms. Also, an increase in the skilled production workers' ratio by 10 percentage points reduces the extent of these perceptions by 0.7 percentage points. We find that manufacturing firms have a statistically higher likelihood to identify labour skill shortages as a major/severe business constraint by 12.1 percentage points compared to other sectors' firms. Manufacturing firms located in Egypt have statistically higher propensities to perceive labour regulations and labour skill shortages as major/severe business constraints by 15.7 and 20.6 percentage points, respectively, compared to service firms. The implications of firm size for firm perceptions of labour regulations are similar to the benchmark results. Also, the effects of firm age, private foreign ownership, and the proportion of non-production workers in total employment are reminiscent of those obtained from the pooled regressions. Service firms located in Jordan have statistically higher propensities to identify labour skill shortages as a major/severe business constraint compared to manufacturing and other sector's firms by 19.2 and 13.6 percentage points, respectively. The non-production workers' ratio exhibits positive implications for firm perceptions of labour skill shortages. An increase in this ratio by 10 percentage points raises the magnitude of these perceptions by 3.2 percentage points. In contrast, this ratio has negative impacts on firm perceptions of labour regulations. An increase in this ratio by 10 percentage points induces a decrease in the magnitude of these perceptions by 2.0 percentage points. In the case of Lebanon, manufacturing and service firms have lower propensities to perceive labour skill shortages as a major/severe business constraint by 34.6 percentage points compared to firms belonging to the other sectors. Firm characteristics do not exhibit statistically significant implications, with the exception of firm use of ICT in the case of labour regulations. The perceptions of firms located in Oman concerning labour market constraints do not exhibit statistically significant variations across sectors. The conditional marginal effects for Pr(R jkc =1|S jkc =1) indicate that exporting activities and an increase in firm size by one hundred workers reduce the corresponding probabilities by 34.6 and 38.0 percentage points, respectively. In our data set, firms located in Morocco, Syria, and Yemen belong exclusively to the manufacturing sector. We find that exporting firms located in Morocco are more likely to perceive labour regulations and labour skill shortages as major/severe business constraints by 15.3 and 22.3 percentage points, respectively, compared to non-exporting firms. Also, an increase in the skilled production workers' ratio by 10 percentage points raises the likelihood of identifying labour skill shortages as a major/severe business constraint by 2.9 percentage points. Finally, exporting firms located in Yemen are less likely to perceive labour regulations as a major/severe business constraint by 11.2 percentage points, and are more likely to perceive labour skill shortages as a major/severe business constraint by 34.8 percentage points, compared to non-exporting firms. Labour market constraints are often identified as main business obstacles facing firm operation and development in the MENA region. Therefore, they are naturally listed through the primary items on the labour policy agenda of the MENA countries. The comprehension of the factors influencing the perceived severity of labour market constraints is essential in the design of policies aiming at improving labour market conditions and enhancing business environments. This paper examines the implications of firm characteristics, national locations, and sectoral associations for the perceptions of firms located in the MENA region concerning two primary labour market constraints: labour regulations and labour skill shortages. The empirical analysis is carried out using firm-level data set sourced from the World Bank's Enterprise Surveys database. A bivariate probit estimator is used to account for potential correlations between the errors in the labour regulations' equation and labour skill shortages' equation. The empirical results are generated through overall estimations and by implementing comparative cross-country and cross-sector analyses. The benchmark empirical results obtained from the overall estimations reveal important implications of firm characteristics for firm perceptions of labour market constraints. Larger firms appear to have lower propensities to identify labour market constraints as major/severe business obstacles. These results imply that small firms have limited flexibility in adjusting to labour regulations and to labour skill shortages. Hence, they suggest that small firms may encounter difficulties in realizing future growth through the MENA region. Also, we find that firms with private foreign ownership are less likely to endure the implications of labour skill shortages compared to domestic firms. This is consistent with higher tendencies of foreign affiliates to pay higher wages and to attract educated workforce compared to domestic firms. Meanwhile, exporting activities do not have statistically significant effects on firm perceptions of labour market constraints. This finding could stem from the generally limited engagement of MENA exporting firms in international trade. We find that firms with higher proportions of non-production workers in total employment have higher propensities to perceive labour skill shortages as a major/severe business constraint. However, the proportion of skilled production workers in total number of production workers does not exercise statistically significant influences on firm perceptions. These findings suggest that the implications of labour skill shortages are primarily associated with inadequacy of non-production workers rather than inadequacy of skilled production workers. The results also emphasize considerable cross-country and cross-sector variations in firm perceptions of labour market constraints. Finally, the empirical analysis reveals significant heterogeneity through the implications of firm characteristics and national locations for the perceptions of firms belonging to different sectors concerning labour market constraints. Also, it underlines important heterogeneity through the implications of firm characteristics and sectoral associations for the perceptions of firms located in different countries concerning labour market constraints. Reforms in labour regulations and investment in human capital are important governmental policy interventions for promoting firm development and economic growth. This paper provides policy makers with information needed to design labour policies that attenuate the impact of labour market constraints and enhance the performance of firms in the MENA region. Specifically, the design of labour policies should encompass the findings that the perceptions of labour market constraints significantly vary through firm characteristics, countries, and sectors. Furthermore, given the heterogeneity in the implications of the factors influencing firm perceptions of labour market constraints, labour policies should be customized by country, and should recognize the varying consequences for different sectors. Opens in a new window. Table I Percentage of firms identifying business obstacles as major/severe constraints Opens in a new window. Table II Descriptive statistics Opens in a new window. Table III Benchmark empirical results from the bivariate probit model (marginal effects) Opens in a new window. Table IV Empirical results from the ordered probit model (marginal effects) Opens in a new window. Table V Empirical results by sector (marginal effects from the bivariate probit model) Opens in a new window. Table VI Empirical results by country (marginal effects from the bivariate probit model) Opens in a new window. Table AI Descriptive statistics for the dependent variables by country and by sector Opens in a new window. Table AII Other results from the bivariate probit model Opens in a new window. Equation 1 Opens in a new window. Equation 2 Opens in a new window. Equation 3 Opens in a new window. Equation 4 Opens in a new window. Equation 5 Opens in a new window. Equation 6 Opens in a new window. Equation 7 Opens in a new window. Equation 8 Opens in a new window. Equation 9 Opens in a new window. Equation 10 Opens in a new window. Equation 11 Opens in a new window. Equation 12 Opens in a new window. Equation 13 Opens in a new window. Equation 14 Opens in a new window. Equation 15 Opens in a new window. Equation 16 Opens in a new window. Equation 17
|
Labour market constraints constitute prominent obstacles to firm development and economic growth of countries located in the Middle East and North Africa (MENA) region. The purpose of this paper is to examine the implications of firm characteristics, national locations, and sectoral associations for the perceptions of firms concerning two basic labour market constraints: labour regulations and labour skill shortages.
|
[SECTION: Method] Stringent labour market constraints are expected to pose serious obstacles to firm performance and economic growth. A wide range of literature finds that rigid labour regulations would induce lower labour force participation and higher unemployment rates (e.g. Botero et al., 2004; Besley and Burgess, 2004; Amin, 2009; Djankov and Ramalho, 2009), and would prevent labour markets from being efficient leading to losses in productivity (e.g. Kaplan, 2009). Another strand of literature inspects the problem of labour skill shortages or "skill deficits", which can be defined as the divergence between the educational attainments of workers and the skill requirements of jobs (Kiker et al., 1997). This literature regularly indicates that accentuated labour skill shortages impose significant restrictions on employment creation and economic growth (e.g. Pissarides and Veganzones-Varoudakis, 2007; Bhattacharya and Wolde, 2012), and could eventually inflict severe impacts on economic performance and labour market outcomes (e.g. Allen and van der Velden, 2001). The implications of labour market constraints are deemed to be of particular concerns for the Middle East and North Africa (MENA) region. Although the MENA region has generally realized higher annual employment growth rates compared to other geo-economic regions (International Labour Organization (ILO), 2012), rigid labour regulations and lack of suitable labour skills were often identified as main obstacles to firm operation and development in the MENA region (Drzeniek-Hanouz and Dusek, 2013). MENA countries have one of the least flexible labour markets, which remain too tight compared to labour markets in other developing countries (Angel-Urdinola and Kuddo, 2010; Bhattacharya and Wolde, 2012 [1]). Table I presents the percentages of firms identifying different business obstacles as major or very severe (henceforth, major/severe) constraints across geo-economic regions. The statistics emphasize that the MENA region has the highest percentages of firms that consider labour regulations and labour skill shortages to be major/severe business constraints compared to other geo-economic regions. For example, labour regulations and labour skill shortages are identified as major/severe problems to 25 and 38 per cent of firms located in the MENA region, respectively, compared to only 9 and 16 per cent of firms located in South Asia[2]. Also, in the case of the MENA region, Table I shows that the constraint of labour skill shortages is associated with one of the highest percentages compared to other business obstacles, whereas the constraint of labour regulations occupy an intermediate rank. Understanding the factors influencing employability constraints constitute an essential requirement towards the implementation of policies aiming at improving the regulatory and institutional environment for business (Pierre and Scarpetta, 2004; Kaplan and Pathania, 2010). This is particularly relevant for MENA countries that have recently undergone significant labour reforms, where employment creation is listed among the primary objectives (Angel-Urdinola and Kuddo, 2010). Several studies underline systematic differences in firm perceptions of labour market constraints through firm characteristics, as well as across countries and sectors (e.g. Pierre and Scarpetta, 2004; Kaplan and Pathania, 2010; Lyon et al., 2012; Meyer and Vandenberg, 2013). Therefore, designing policies that alleviate the adverse implications of labour market constraints requires a comprehension of the categories of firms that are more likely to endure the burden of these constraints. This paper examines the implications of firm characteristics for the perceptions of firms located in the MENA region concerning the stringency of labour regulations and labour skill shortages. It also examines the heterogeneity in firm perceptions of labour market constraints across MENA countries and across sectors. We use a data set sourced from the World Bank's Enterprise Surveys database, which includes information on firm perceptions of many business constraints. Firm perception data are collected through surveys, which are generally answered by senior managers and business owners. Hence, following the previous literature (e.g. Pierre and Scarpetta, 2004; Kaplan and Pathania, 2010; Meyer and Vandenberg, 2013), our dependent variables are determined as perception-based indicators. We implement overall examination and comparative cross-country and cross-sector empirical analyses of the factors influencing firm perceptions of labour regulations and labour skill shortages for the MENA region. MENA countries are characterized by a considerable level of diversity in macroeconomic and industrial factors, and by important differences in labour market conditions and institutions (Angel-Urdinola and Kuddo, 2010), suggesting potential cross-country heterogeneity. Also, cross-sector heterogeneity is expected given the varying structures in labour requirements between sectors. Through the basic regressions, we use a bivariate probit model that allows for firm perceptions of labour regulations and labour skill shortages to be jointly formulated. This estimation model takes into account potential correlation between the error terms in the two labour market constraints' equations. 2.1. Labour regulations Labour regulations are normally expressed through laws that regulate and govern the employment relationships between employees, employers, unions, and government in order to protect the basic standards of fair treatment for workers (Kaplan, 2009) and to maximize the social welfare (Botero et al., 2004). Governments commonly intervene in labour markets through regulations because employers may mistreat workers, leading to unfairness and inefficient outcomes for workers such as unfair dismissals, unfair minimum wage, lay-offs for economic reasons, and under-payment (Djankov and Ramalho, 2009 [3]). These regulations and laws may be written through labour codes, current legislations, and norms set by collective agreements (Pierre and Scarpetta, 2004). Several studies report variations in firm perceptions of labour regulations through firm characteristics and across countries and sectors. Gelb et al. (2007) use data covering firms located in 26 Sub-Saharan African countries to examine the implications of various factors for firm perceptions of the business environment. Among the results, they find that large firms have higher propensities to complain about the rigidity of labour regulations. They argue that, in developing countries, small firms are more likely to work within restricted markets and, therefore, they are less visible to regulators, and are less appealing targets for officials. They also find that exporting activities and private foreign ownership do not exhibit statistically significant impacts on firm perceptions of labour regulations. They also report that firm perceptions of labour regulations do not exert considerable variations across sectors. Hallward-Driemeier and Aterido (2009) investigate the role of firm characteristics in determining firm perceptions of various business constraints, using data covering firms located in 105 countries. The results suggest that small firms may face more problems in dealing with the requirements of poor labour regulations, and may become easier targets in corrupted environments in developing countries. Clarke (2010) uses data covering firms located in South Africa to study the factors influencing firm perceptions of business constraints. Among the results, he finds that large firms and exporting firms are more likely to report labour regulations as a significant obstacle to their development. Also, he finds that firm age and type of firm ownership do not exercise statistically significant role in explaining the variations in firm perceptions of labour regulations. Bartelsman et al. (2010) document how business constraints, as perceived by individual firms, differ across countries and across firm characteristics. They show that firm age, ownership type, export orientation, industry binary variables, and country binary variables are important indicators in explaining the differences in firm perceptions of business constraints between Europe and Central Asia (ECA) and Latin America and the Caribbean regions. The results also show that large firms and exporting firms in the ECA region are more likely to be affected by rigid labour regulations, reflecting costs of labour adjustment. Vargas (2012) implements the analysis for firms located in Bolivia, focusing on the implications of firm size for firm perceptions of business constraints. Among the results, he finds that small firms are less likely to identify labour regulations as an important business constraint. Also, he finds that manufacturing firms are more likely to report labour regulations as a significant business constraint compared to firms belonging to other sectors. Meyer and Vandenberg (2013) analyse the relationship between firm characteristics and firm perceptions of labour regulations using data covering firms located in five Asian economies (Bangladesh, Indonesia, Pakistan, the Philippines, and Vietnam). They find systematic differences in firm perceptions of labour regulations across firm characteristics and sectors. In four of the five countries examined, they find that exporting firms are more likely to perceive labour regulations as an important obstacle to their operations compared to non-exporting firms. They also find that, in four of the five countries, large firms have higher propensities to perceive labour regulations as a significant business constraint. They argue that small firms may enjoy de facto or de jure exemption from the enforcement of labour regulations. Also, they show that young firms have higher tendencies to report difficulties in coping with labour regulations. 2.2. Labour skill shortages Mismatches between worker skills and job requirements are often classified among the significant obstacles facing firm productivity and growth, particularly in developing countries (Almeida and Aterido, 2011). For example, skill mismatches form an important constraint on hiring decisions, leading to higher levels of unemployment and important costs for firms (O'Sullivan et al., 2011) and causing higher turnover rates (Hersch, 1991). Gupta et al. (2010) argue that weak performances of labour-intensive industries could be related to labour skill shortages. Piore (1986) suggests that a low-skilled workforce may decrease the internal flexibility, leading to inefficient functioning of firms. Empirical evidence on the factors influencing firm perceptions of labour skill shortages points out to large variations through firm characteristics and across countries and sectors. Gelb et al. (2007) argue that the problem of labour skill shortages tends to be more prevalent for firms that are large, more productive, and use advanced technologies. Additionally, they note that small firms may use less sophisticated production techniques that require fewer skilled workers. Using data set covering firms located in Sub-Saharan Africa, Gelb et al. (2007) find that large firms are more likely to report labour skill shortages as one main business constraint. They also find that exporting activities and type of firm ownership do not exhibit important influences on firm perceptions of labour skill shortages. They note that the availability of skilled labour may become a more binding constraint as economies become more sophisticated and as governments magnify their capacities in enforcing labour regulations. Hallward-Driemeier and Aterido (2009) find that large firms and exporting firms are more likely to report labour skill shortages as a significant obstacle to business operations. They also find that government-owned firms are less likely to perceive labour skill shortages as an important business constraint. In the case of firms located in South Africa, Clarke (2010) finds that large firms and exporting firms have higher propensities to perceive labour skill shortages as a significant business obstacle. He also reports that firm age and type of firm ownership do not exhibit important influences on firm perceptions of labour skill shortages. Kaplan and Pathania (2010) find that large firms are more likely to identify labour skill shortages as one main business constraint. However, Vargas (2012) does not find significant influences of firm size, but he reports important variations across sectors. Using data covering firms located in Canada, Sabourin (2001) shows that firm size and technological intensity are negatively correlated with the probability of firms to identify labour skill shortages as one primary business obstacle. Also, he finds that type of firm ownership and Research and Development activities have no statistically significant influences on these firm perceptions. Baldwin and Lin (2002) find that young firms are more likely to designate labour skill shortages as a considerable problem using data covering Canadian manufacturing firms. However, they find that firm size and ownership type do no exhibit important influences. Green et al. (1998) implement the empirical analysis for firms located in the UK. They find that higher skilled share of employment accentuates the problem of labour skill shortages as perceived by employers. They also report that the share of part-time workers does not exhibit important effects. Lyon et al. (2012) find that medium and large size firms are more likely to designate labour skill shortages as an important constraint on firm operations and growth. They find that exporting firms have lower propensities to report labour skill shortages as a significant business constraint compared to non-exporting firms. Also, they do not find important variations across sectors. The empirical analysis is carried out for the perceived levels of labour market constraints as reported by the respondents (e.g. senior managers, business owners) through the World Bank's Enterprise Surveys database. Pierre and Scarpetta (2004, 2006) examine the relationship between the perceived and actual stringency of labour regulations using national labour protection indices (i.e. de jure labour laws). They find that the reported perceptions are closely related to the actual levels of labour regulations' constraints. Specifically, countries with higher national indices on the stringency of labour regulations are associated with higher proportions of firms perceiving labour regulations as being significant constraints. Kaplan and Pathania (2010) indicate that the perception-based indicators capture the actual institutional environment (e.g. laws, regulations, infrastructure, corruption), particularly when objective measures are inadequate. They argue that the perception-based indicators reflect the de facto severity of business constraints, and serve as a normal check vis-a-vis national indicators. Clarke (2010) notes that the collection of objective measures could be difficult, particularly when it comes to sensitive indicators (e.g. corruption). In such cases, perception-based indicators can serve as a substitute. Hallward-Driemeier and Aterido (2009) find that the perception-based and objective measures of business constraints are closely correlated with each other. They indicate that subjective measures enable the ranking of business constraints within and between countries, particularly when comparing the rigidity of the business environment. Also, they argue that perception-based measures provide information on the most severe obstacles for businesses growth and development as identified by the business entity, and would eventually assist managers and policy makers to recognize reform priorities. Hence, perception-based measures tend to reflect the actual rigidity of business constraints faced by firms. However, there are some concerns regarding perception-based measures. Specifically, these measures could be subjected to: first, potential measurement errors leading to biased results; second, variations in managers willingness to report actual negative or positive responses due to differences in personalities, views, and culture; third, performance bias where managers tend to rank business obstacles according to the performance of their firms or to the business environment conditions through which the firm operates; and finally, defects due to imperfections in question formulations and answer interpretations (Bertrand and Mullainathan, 2001; Senik, 2005; Hallward-Driemeier and Aterido, 2009). Some studies (e.g. Glaeser et al., 2004; Kaplan and Pathania, 2010) argue that caution should be exercised when using perception-based measures as independent variables through empirical analyses. This is because these measures may embody information on growth, which would lead to reverse causality issues. Meanwhile, these studies indicate that using these measures as dependent variables is more intuitive to capture growth in the business environment. These indications complement the aforementioned empirical findings reflecting a close correspondence between perceived and actual measures of business constraints. It is important to note that the empirical analysis in this paper exclusively covers firms that exist in the market (i.e. those that have already entered the market, and that did not exit the market). This is to say that the empirical analysis examines the business obstacles as perceived by the existing firms in the market. Hausmann and Velasco (2005) depict this issue through a camel-hippopotamus analogy. Camels living in the desert do not identify access to water as a primary problem since they have adjusted to the situation by learning how to conserve water. Meanwhile, hippopotami lack the ability to stay in the desert since water scarcity is a critical obstacle for their survival. Hence, by generating and analysing data from interviews with "camels" about water constraints, we may be missing distinct information from "hippopotami". Gelb et al. (2007) argue that such selection is incomplete since many firms choose to enter the market despite the severity of business constraints. They also indicate that the ability to adjust to a business constraint does not mean that firms do not recognize it anymore as a serious problem. They provide an illustration using the World Bank's Enterprises Surveys database: the perceptions of generator-owning firms regarding the electricity as a business constraint are not distinguishable from the perceptions of firms that do not own generators. Furthermore, they note that the intensity of complaints are often highly correlated with corresponding national indices. Dethier et al. (2011) suggest that empirical analyses covering "hippopotami" should be carried out using different frameworks (e.g. entry or exit models[4]). We use a data set sourced from the World Bank's Enterprise Surveys database. This database represents a comprehensive source of firm-level data in emerging and developing economies. It covers firms operating in the manufacturing, service, and other sectors. It contains information on various aspects of the business environment such as, access to finance, corruption, workforce characteristics, innovation and technology, and trade. It should be noted that one of many advantages of using data from these surveys is that the questions are identical through firms across all countries. The basic data set used in this paper covers 5,052 firms located in eight developing Arab countries of the MENA region: Algeria, Egypt, Jordan, Lebanon, Morocco, Oman, Syria, and Yemen[5]. Through these surveys, firms' representatives (e.g. senior managers, business owners) are asked whether labour regulations and labour skill shortages are considered prominent constraints on business operations and development. The responses are used to generate the basic dependent binary variables, which equal one when a firm declares the corresponding labour market constraint as a major/severe business obstacle and zero otherwise. Following the previous literature on business constraints (e.g. Gelb et al., 2007; Kaplan and Pathania, 2010; Clarke, 2010; Meyer and Vandenberg, 2013), the explanatory variables cover firm characteristics' variables which comprise: first, a binary variable that equals one for firms characterized by private foreign ownership and zero otherwise[6]; second, a binary variable that equals one for firms engaged in exporting activities and zero otherwise; third, firm size depicted through the number of total employment and presented in hundreds of workers through the regressions; fourth, firm age measured by the number of years since firm establishment; fifth, firm use of Information and Communication Technology (ICT) depicted through a binary variable that takes the value of one for firms regularly using the internet to communicate with clients and to source information and zero otherwise; finally, two firm labour composition variables: the first is represented through the fraction of skilled production workers in total number of production workers, and the second is depicted through the fraction of non-production workers in total employment. The use of skilled production workers' share and the use of non-production workers' share through the regressions imply that the results are relative to the use of unskilled production workers and to the use of production workers, respectively. The explanatory variables also include country-specific binary variables and sector-specific binary variables to capture cross-country and cross-sector variations. Table II displays summary statistics of the variables used in the empirical analysis. The first panel shows the results for the dependent variables. We find that around 24.6 per cent of firms in the regression data set report that labour regulations represent a major/severe business constraint, while 38.0 per cent of firms report that labour skill shortages exercise a major/severe business constraint[7]. The second panel shows statistics for the explanatory variables. The results indicate that 4.8 per cent of firms are characterized by private foreign ownership and that 22.5 per cent of firms are engaged in exporting activities. The average firm size is 124.0 full-time workers (with a standard deviation of 317.1 full-time workers) and the average firm age is 20.9 years (with a standard deviation of 16.6 years). Also, we find that 50.4 per cent of firms use the internet to communicate with clients and to source information. The percentage of skilled production workers in total number of production workers and the percentage of non-production workers in total employment have averages of 61.6 and 26.1 per cent, respectively. Consider a given firm j (j=1, ..., J) belonging to sector k (k=1, ..., K) and located in country c (c=1, ..., C). Firm perception levels of constraints related to labour regulations and those related to labour skill shortages are depicted through the latent variables (Equation 1) and (Equation 2), respectively. These latent variables are not observed. However, we observe the perceptions of firms through dichotomous responses on whether labour regulations and labour skill shortages do or do not pose major/severe obstacles on firm operations and development. Let R jkc depict a binary variable that takes the value of one when the corresponding firm identifies labour regulations as a major/severe business constraint and zero otherwise. Also, let S jkc represent a binary variable that takes the value of one when the corresponding firm identifies labour skill shortages as a major/severe business constraint and zero otherwise. The benchmark empirical specifications can be represented as: (Equation 3) (Equation 4) where X j is the vector of variables depicting firm characteristics, Y k is the vector of binary variables depicting sectors, Z c is the vector of binary variables depicting countries, and (Equation 5) and (Equation 6) are the stochastic error terms of the corresponding equations. The univariate probit estimator would produce biased estimates when there are some unobserved or omitted characteristics that simultaneously affect firm perceptions of labour regulations and labour skill shortages (Deadman and MacDonald, 2004; Greene, 2008). We allow for the errors in these two labour market constraints' equations to be potentially correlated. Thus, the two equations are jointly modelled using a bivariate probit estimator. The error terms are assumed to be independently and identically distributed as bivariate normal with r depicting the correlation parameter. Specifically, we have: (Equation 7) Through the empirical analysis, we use the Wald test to determine whether the correlation parameter r is statistically significant[8]. The rejection of the null hypothesis indicates that firm perceptions of the labour market constraints are jointly formulated (Greene, 2008 [9]). The bivariate probit specification allows us to estimate unconditional marginal effects, but also conditional and joint marginal effects of variables influencing firm perceptions of the labour market constraints. Table III presents the marginal effects from the benchmark bivariate probit estimation carried out for the pooled data set covering existing firms' perceptions of labour market constraints. The Wald test rejects the null hypothesis of zero correlation between the errors in the two labour market constraints' equations and, hence, it indicates that the model should be estimated through the bivariate probit estimator rather than through the univariate probit estimator. The estimated coefficient of correlation between the errors in the two equations is positive and statistically significant at the 1 per cent level. Table III displays the unconditional marginal effects for Pr(R jkc =1) and Pr(S jkc =1). It also includes the joint marginal effects for Pr(R jkc =1, S jkc =1) and Pr(R jkc =0, S jkc =0), and the conditional marginal effects for Pr(R jkc =1|S jkc =1) and Pr(S jkc =1|R jkc =1[10]). These marginal effects are determined at the mean values of the explanatory variables. For explanatory binary variables, the marginal effects are calculated through discrete changes in probabilities as the binary variable changes from 0 to 1. 6.1. Firm characteristics We find that larger firms have lower propensities to identify labour market constraints as major/severe obstacles facing business operations and development. This is in line with the results reported in some other empirical studies (e.g. Pierre and Scarpetta, 2004; Hallward-Driemeier and Aterido, 2009). The unconditional marginal effects imply that an increase in firm size by one hundred full-time workers reduces the likelihood of firms to perceive labour regulations and labour skill shortages as major/severe constraints on business operations by 0.8 and 0.5 percentage points, respectively. Also, the joint marginal effect of firm size for Pr(R jkc =1, S jkc =1) indicates that an increase in firm size by 100 workers decreases the likelihood of firms to jointly identify labour regulations and labour skill shortages as major/severe business constraints by 0.5 percentage points. The results on the relationship between firm size and firm perceptions of labour regulations are consistent with arguments provided in the literature, which suggest that smaller firms have more limited abilities to realize internal adjustments in response to labour regulations (Meyer and Vandenberg, 2013). Hence, smaller firms have higher tendencies to be more adversely affected by labour regulations. These results are also in line with some other findings showing that labour regulations have disproportional effects on small firms (Aterido et al., 2011). The results on the relationship between firm size and firm perceptions of labour skill shortages imply that smaller firms in MENA countries may encounter more difficulties in realizing future growth, since adequacy of educated workforce is essential for firm competitiveness and performance (Storey, 1994; Jensen and McGuckin, 1997). These results could also suggest that smaller firms have more difficulties in accessing skilled labour markets, identifying required labour skills, and investing in workforce training programmes (Jansen and Lanz, 2013). The implication of exporting activities for firm perceptions of labour regulations is not statistically significant. These results are similar to those found in some previous studies, such as Gelb et al. (2007) in the case of Sub-Saharan African firms. Yet, they deviate from those reported in some studies covering firms located in other developing regions. For example, Meyer and Vandenberg (2013) find that exporting firms located in some South East Asian countries have higher tendencies to perceive labour regulations as a major/severe business constraint compared to non-exporting firms. They attributed these findings to variations in labour demand requirements between exporting and non-exporting firms. Also, Lyon et al. (2012) find that non-exporting firms are more likely to identify labour skill shortages as a major/severe business constraint compared to exporting firms. The effect of firm age is negative and statistically significant in the case of labour skill shortages' equation. The unconditional marginal effect and the conditional marginal effect for Pr(S jkc =1|R jkc =1) indicate that an increase in firm age by ten years reduces the likelihood to perceive labour skill shortages as a major/severe business constraint by 1 percentage point. Although relatively small in magnitude, this impact is consistent with some a priori expectations discussed in the literature, suggesting that older firms have more accumulated experiences to attenuate the effects of business obstacles as perceived by the survey respondents (Meyer and Vandenberg, 2013). We find that an increase in the proportion of non-production workers in total employment by 10 percentage points raises the likelihood of firms to identify labour skill shortages as a major/severe business constraint by 1.3 percentage points. Also, we find that an increase in this proportion by 10 percentage points raises this likelihood by 1.6 percentage points, when assessed conditional on identifying labour regulations as a major/severe business constraint. The effect of the proportion of skilled production workers in total number of production workers on firm perceptions of these labour market constraints is not statistically significant. These findings suggest that the implications of labour skill shortages are primarily associated with the inadequacy of non-production workers rather than with the inadequacy of skilled production workers. The results show that firms with private foreign ownership are less likely to perceive labour skill shortages as a major/severe business constraint by 6.0 percentage points. They suggest that foreign affiliates have higher capabilities in attenuating the implications of labour skill shortages compared to local, domestically owned firms. Almeida (2007) indicates that foreign-owned firms have higher tendencies to pay higher wages and to attract educated workforce compared to domestic firms. These tendencies could eventually mitigate the perceptions of firms with private foreign ownership concerning labour skill shortages compared to firms with domestic ownership. 6.2. Countries' binary variables The results emphasize significant variations in the perceptions of firms located in different countries of the MENA region concerning labour market constraints. Hence, they underscore the implications of national characteristics for these firm perceptions. The binary variable depicting firms located in Algeria is set as the reference for MENA countries through the regressions. The unconditional marginal effects reveal that firms located in Lebanon have statistically higher propensities to perceive labour regulations as a major/severe business constraint by 34.4 percentage points compared to firms located in the Maghreb countries (i.e. Algeria and Morocco) and in Yemen. The corresponding propensities for firms located in Egypt, Oman, and Syria are higher by 17.3, 31.9, and 22.3 percentage points, respectively, compared to firms located in the two Maghreb countries and in Yemen. Also, firms located in Egypt, Lebanon, and Oman have statistically higher propensities to perceive labour skill shortages as a major/severe business constraint by 15.4, 17.5, and 13.5 percentage points, respectively, compared to firms located in the other MENA countries[11]. These results complement other empirical findings in the literature, which document international variations in the perceived magnitude of labour market constraints. For example, Pierre and Scarpetta (2004) show that countries with more stringent de jure labour laws have higher proportions of firms identifying labour regulations as being restrictive. Gelb et al. (2007) find significant variations in firm perceptions of labour market constraints, among other business obstacles, across Sub-Saharan African countries. Also, Lyon et al. (2012) present some important regional differences in firm perceptions of labour skill shortages. Drzeniek-Hanouz and Dusek (2013) indicate that labour market constraints constitute significant obstacles on business development in the Arab world. They also describe considerable variations in the perceptions of firms located in different MENA sub-regions: North Africa (including Egypt), the Levant, and the Gulf sub-regions. 6.3. Sectors' binary variables We find that firm perceptions of labour market constraints vary across sectors. The binary variable for firms belonging to the service sector is set as the reference for sectors through the regressions. The unconditional marginal effects indicate that firms belonging to the manufacturing sector and to the category covering the remaining other (non-manufacturing and non-service) sectors have higher propensities to perceive labour regulations as a major/severe business constraint by 9.9 and 10.1 percentage points, respectively, compared to firms belonging to the service sector. The estimates for the labour skill shortages' equation are 12.6 and 11.8 percentage points, respectively. These results suggest that labour market constraints, as perceived by the survey respondents, disproportionally impact manufacturing firms and firms belonging to the other sectors compared to firms belonging to the service sector. Firm perceptions of labour market constraints can be alternatively represented through ordinal responses in the data set. Specifically, firms identify the stringency of labour regulations and labour skill shortages as: no or minor obstacle (outcome 1), moderate obstacle (outcome 2), and major or very severe obstacle (outcome 3). Table IV displays the marginal effects obtained through the ordered probit model. The results are comparable to the corresponding marginal effects for Pr(R jkc =1) and Pr(S jkc =1) that are reported in Table III. 7.1. Empirical results by sector The results from the empirical analysis implemented for data sets covering individual sectors are displayed in Table V. The estimated coefficients of correlation between the errors in the two labour market constraints' equations are positive and statistically significant at the 1 per cent level for all sectors. To save space, Table V does not display the rows of marginal effects of firm characteristics that do not exhibit statistical significance at all. In general, the results for firms belonging to the manufacturing sectors are qualitatively similar to the benchmark results for pooled data set. We also find few differences. For example, the perceptions of manufacturing firms located in Jordan concerning labour skill shortages appear to be less pronounced compared to manufacturing firms located in the other MENA countries. Firms located in Jordan have statistically lower propensities to perceive labour skill shortages as a major/severe business constraint compared to firms located in Algeria and those located in Egypt by 11.3 and (11.3+14.0=) 25.3 percentage points, respectively. The results derived from the service sector's data set reveal some notable differences compared to those obtained from the manufacturing sector's data set[12]. In the case of the service sector, firm perceptions of labour regulations do not exhibit statistically significant heterogeneity across firm characteristics. We find that firm perceptions of labour skill shortages are accentuated with an increase in the ratio of non-production workers to total employment. The marginal effect shows that an increase in this ratio by 10 percentage points raises the likelihood of perceiving labour skill shortages as a major/severe business constraint by 3.1 percentage points. The magnitude of this effect is considerably higher than the one reported for manufacturing firms. Our data set does not contain service firms located in Algeria that report labour regulations and labour skill shortages as major/severe constraints on business operations. In other words, service firms located in Algeria predict failure perfectly and, therefore, they are dropped from the service sector's data set. Then, the binary variable depicting firms located in Egypt is set as the reference for MENA countries. We find that service firms located in Egypt and Lebanon have statistically higher likelihoods than those located in Jordan to perceive labour regulations as a major/severe business constraint by 6.7 and (6.7+2.1=) 8.8 percentage points, respectively. The corresponding likelihood for service firms located in Oman is considerably higher than those located in Jordan by (18.4+6.7=) 25.1 percentage points. We also find that service firms located in Egypt and Lebanon have lower propensities than those located in Jordan to identify labour skill shortages as a major/severe business constraint by 16.1 and (16.1-1.3=) 14.8 percentage points, respectively. The joint marginal effects for Pr(R jkc =1, S jkc =1) indicate that service firms located in Egypt and Jordan have statistically comparable likelihoods to identify both labour market constraints as major/severe obstacles on business operations as perceived by the survey respondents. These likelihoods are lower than the corresponding likelihood for service firms located in Oman by more than 11 percentage points. Finally, the results for the other sectors' category reveal that firms located in Oman are statistically more likely to perceive labour regulations as a major/severe business constraint by 34.9 percentage points compared to firms located in the other MENA countries (except Lebanon). Also, we find that firms located in Egypt and Lebanon have higher propensities to perceive labour skill shortages as a major/severe business constraint by 44.2 and 31.3 percentage points, respectively, compared to firms located in the other MENA countries. 7.2. Empirical results by country The results from the empirical analysis carried out for data sets covering firms located in individual MENA countries are displayed in Table VI. The estimated coefficients of correlation between the errors in the two labour market constraints' equations are positive and statistically significant at the 1 per cent level for all countries. To save space, Table VI does not display the rows of marginal effects of firm characteristics that do not exhibit statistical significance at all. In the case of Algeria, service firms predict failure perfectly and, hence, they are dropped from the data set. Then, the other sectors' category is set as the reference for sectors. We find that the perceptions of firms located in Algeria concerning labour regulations do not exhibit statistically significant variations between the manufacturing sector and the other sectors' category. Exporting firms are less likely to perceive labour regulations as a major/severe business constraint by 13.5 percentage points compared to non-exporting firms. Also, an increase in the skilled production workers' ratio by 10 percentage points reduces the extent of these perceptions by 0.7 percentage points. We find that manufacturing firms have a statistically higher likelihood to identify labour skill shortages as a major/severe business constraint by 12.1 percentage points compared to other sectors' firms. Manufacturing firms located in Egypt have statistically higher propensities to perceive labour regulations and labour skill shortages as major/severe business constraints by 15.7 and 20.6 percentage points, respectively, compared to service firms. The implications of firm size for firm perceptions of labour regulations are similar to the benchmark results. Also, the effects of firm age, private foreign ownership, and the proportion of non-production workers in total employment are reminiscent of those obtained from the pooled regressions. Service firms located in Jordan have statistically higher propensities to identify labour skill shortages as a major/severe business constraint compared to manufacturing and other sector's firms by 19.2 and 13.6 percentage points, respectively. The non-production workers' ratio exhibits positive implications for firm perceptions of labour skill shortages. An increase in this ratio by 10 percentage points raises the magnitude of these perceptions by 3.2 percentage points. In contrast, this ratio has negative impacts on firm perceptions of labour regulations. An increase in this ratio by 10 percentage points induces a decrease in the magnitude of these perceptions by 2.0 percentage points. In the case of Lebanon, manufacturing and service firms have lower propensities to perceive labour skill shortages as a major/severe business constraint by 34.6 percentage points compared to firms belonging to the other sectors. Firm characteristics do not exhibit statistically significant implications, with the exception of firm use of ICT in the case of labour regulations. The perceptions of firms located in Oman concerning labour market constraints do not exhibit statistically significant variations across sectors. The conditional marginal effects for Pr(R jkc =1|S jkc =1) indicate that exporting activities and an increase in firm size by one hundred workers reduce the corresponding probabilities by 34.6 and 38.0 percentage points, respectively. In our data set, firms located in Morocco, Syria, and Yemen belong exclusively to the manufacturing sector. We find that exporting firms located in Morocco are more likely to perceive labour regulations and labour skill shortages as major/severe business constraints by 15.3 and 22.3 percentage points, respectively, compared to non-exporting firms. Also, an increase in the skilled production workers' ratio by 10 percentage points raises the likelihood of identifying labour skill shortages as a major/severe business constraint by 2.9 percentage points. Finally, exporting firms located in Yemen are less likely to perceive labour regulations as a major/severe business constraint by 11.2 percentage points, and are more likely to perceive labour skill shortages as a major/severe business constraint by 34.8 percentage points, compared to non-exporting firms. Labour market constraints are often identified as main business obstacles facing firm operation and development in the MENA region. Therefore, they are naturally listed through the primary items on the labour policy agenda of the MENA countries. The comprehension of the factors influencing the perceived severity of labour market constraints is essential in the design of policies aiming at improving labour market conditions and enhancing business environments. This paper examines the implications of firm characteristics, national locations, and sectoral associations for the perceptions of firms located in the MENA region concerning two primary labour market constraints: labour regulations and labour skill shortages. The empirical analysis is carried out using firm-level data set sourced from the World Bank's Enterprise Surveys database. A bivariate probit estimator is used to account for potential correlations between the errors in the labour regulations' equation and labour skill shortages' equation. The empirical results are generated through overall estimations and by implementing comparative cross-country and cross-sector analyses. The benchmark empirical results obtained from the overall estimations reveal important implications of firm characteristics for firm perceptions of labour market constraints. Larger firms appear to have lower propensities to identify labour market constraints as major/severe business obstacles. These results imply that small firms have limited flexibility in adjusting to labour regulations and to labour skill shortages. Hence, they suggest that small firms may encounter difficulties in realizing future growth through the MENA region. Also, we find that firms with private foreign ownership are less likely to endure the implications of labour skill shortages compared to domestic firms. This is consistent with higher tendencies of foreign affiliates to pay higher wages and to attract educated workforce compared to domestic firms. Meanwhile, exporting activities do not have statistically significant effects on firm perceptions of labour market constraints. This finding could stem from the generally limited engagement of MENA exporting firms in international trade. We find that firms with higher proportions of non-production workers in total employment have higher propensities to perceive labour skill shortages as a major/severe business constraint. However, the proportion of skilled production workers in total number of production workers does not exercise statistically significant influences on firm perceptions. These findings suggest that the implications of labour skill shortages are primarily associated with inadequacy of non-production workers rather than inadequacy of skilled production workers. The results also emphasize considerable cross-country and cross-sector variations in firm perceptions of labour market constraints. Finally, the empirical analysis reveals significant heterogeneity through the implications of firm characteristics and national locations for the perceptions of firms belonging to different sectors concerning labour market constraints. Also, it underlines important heterogeneity through the implications of firm characteristics and sectoral associations for the perceptions of firms located in different countries concerning labour market constraints. Reforms in labour regulations and investment in human capital are important governmental policy interventions for promoting firm development and economic growth. This paper provides policy makers with information needed to design labour policies that attenuate the impact of labour market constraints and enhance the performance of firms in the MENA region. Specifically, the design of labour policies should encompass the findings that the perceptions of labour market constraints significantly vary through firm characteristics, countries, and sectors. Furthermore, given the heterogeneity in the implications of the factors influencing firm perceptions of labour market constraints, labour policies should be customized by country, and should recognize the varying consequences for different sectors. Opens in a new window. Table I Percentage of firms identifying business obstacles as major/severe constraints Opens in a new window. Table II Descriptive statistics Opens in a new window. Table III Benchmark empirical results from the bivariate probit model (marginal effects) Opens in a new window. Table IV Empirical results from the ordered probit model (marginal effects) Opens in a new window. Table V Empirical results by sector (marginal effects from the bivariate probit model) Opens in a new window. Table VI Empirical results by country (marginal effects from the bivariate probit model) Opens in a new window. Table AI Descriptive statistics for the dependent variables by country and by sector Opens in a new window. Table AII Other results from the bivariate probit model Opens in a new window. Equation 1 Opens in a new window. Equation 2 Opens in a new window. Equation 3 Opens in a new window. Equation 4 Opens in a new window. Equation 5 Opens in a new window. Equation 6 Opens in a new window. Equation 7 Opens in a new window. Equation 8 Opens in a new window. Equation 9 Opens in a new window. Equation 10 Opens in a new window. Equation 11 Opens in a new window. Equation 12 Opens in a new window. Equation 13 Opens in a new window. Equation 14 Opens in a new window. Equation 15 Opens in a new window. Equation 16 Opens in a new window. Equation 17
|
The empirical analysis is carried out using firm-level data set sourced from the World Bank's Enterprise Surveys database. A bivariate probit estimator is used to account for potential correlations between the errors in the two labour market constraints' equations. The authors implement overall estimations and comparative cross-country and cross-sector analyses, and use alternative estimation models.
|
[SECTION: Findings] Stringent labour market constraints are expected to pose serious obstacles to firm performance and economic growth. A wide range of literature finds that rigid labour regulations would induce lower labour force participation and higher unemployment rates (e.g. Botero et al., 2004; Besley and Burgess, 2004; Amin, 2009; Djankov and Ramalho, 2009), and would prevent labour markets from being efficient leading to losses in productivity (e.g. Kaplan, 2009). Another strand of literature inspects the problem of labour skill shortages or "skill deficits", which can be defined as the divergence between the educational attainments of workers and the skill requirements of jobs (Kiker et al., 1997). This literature regularly indicates that accentuated labour skill shortages impose significant restrictions on employment creation and economic growth (e.g. Pissarides and Veganzones-Varoudakis, 2007; Bhattacharya and Wolde, 2012), and could eventually inflict severe impacts on economic performance and labour market outcomes (e.g. Allen and van der Velden, 2001). The implications of labour market constraints are deemed to be of particular concerns for the Middle East and North Africa (MENA) region. Although the MENA region has generally realized higher annual employment growth rates compared to other geo-economic regions (International Labour Organization (ILO), 2012), rigid labour regulations and lack of suitable labour skills were often identified as main obstacles to firm operation and development in the MENA region (Drzeniek-Hanouz and Dusek, 2013). MENA countries have one of the least flexible labour markets, which remain too tight compared to labour markets in other developing countries (Angel-Urdinola and Kuddo, 2010; Bhattacharya and Wolde, 2012 [1]). Table I presents the percentages of firms identifying different business obstacles as major or very severe (henceforth, major/severe) constraints across geo-economic regions. The statistics emphasize that the MENA region has the highest percentages of firms that consider labour regulations and labour skill shortages to be major/severe business constraints compared to other geo-economic regions. For example, labour regulations and labour skill shortages are identified as major/severe problems to 25 and 38 per cent of firms located in the MENA region, respectively, compared to only 9 and 16 per cent of firms located in South Asia[2]. Also, in the case of the MENA region, Table I shows that the constraint of labour skill shortages is associated with one of the highest percentages compared to other business obstacles, whereas the constraint of labour regulations occupy an intermediate rank. Understanding the factors influencing employability constraints constitute an essential requirement towards the implementation of policies aiming at improving the regulatory and institutional environment for business (Pierre and Scarpetta, 2004; Kaplan and Pathania, 2010). This is particularly relevant for MENA countries that have recently undergone significant labour reforms, where employment creation is listed among the primary objectives (Angel-Urdinola and Kuddo, 2010). Several studies underline systematic differences in firm perceptions of labour market constraints through firm characteristics, as well as across countries and sectors (e.g. Pierre and Scarpetta, 2004; Kaplan and Pathania, 2010; Lyon et al., 2012; Meyer and Vandenberg, 2013). Therefore, designing policies that alleviate the adverse implications of labour market constraints requires a comprehension of the categories of firms that are more likely to endure the burden of these constraints. This paper examines the implications of firm characteristics for the perceptions of firms located in the MENA region concerning the stringency of labour regulations and labour skill shortages. It also examines the heterogeneity in firm perceptions of labour market constraints across MENA countries and across sectors. We use a data set sourced from the World Bank's Enterprise Surveys database, which includes information on firm perceptions of many business constraints. Firm perception data are collected through surveys, which are generally answered by senior managers and business owners. Hence, following the previous literature (e.g. Pierre and Scarpetta, 2004; Kaplan and Pathania, 2010; Meyer and Vandenberg, 2013), our dependent variables are determined as perception-based indicators. We implement overall examination and comparative cross-country and cross-sector empirical analyses of the factors influencing firm perceptions of labour regulations and labour skill shortages for the MENA region. MENA countries are characterized by a considerable level of diversity in macroeconomic and industrial factors, and by important differences in labour market conditions and institutions (Angel-Urdinola and Kuddo, 2010), suggesting potential cross-country heterogeneity. Also, cross-sector heterogeneity is expected given the varying structures in labour requirements between sectors. Through the basic regressions, we use a bivariate probit model that allows for firm perceptions of labour regulations and labour skill shortages to be jointly formulated. This estimation model takes into account potential correlation between the error terms in the two labour market constraints' equations. 2.1. Labour regulations Labour regulations are normally expressed through laws that regulate and govern the employment relationships between employees, employers, unions, and government in order to protect the basic standards of fair treatment for workers (Kaplan, 2009) and to maximize the social welfare (Botero et al., 2004). Governments commonly intervene in labour markets through regulations because employers may mistreat workers, leading to unfairness and inefficient outcomes for workers such as unfair dismissals, unfair minimum wage, lay-offs for economic reasons, and under-payment (Djankov and Ramalho, 2009 [3]). These regulations and laws may be written through labour codes, current legislations, and norms set by collective agreements (Pierre and Scarpetta, 2004). Several studies report variations in firm perceptions of labour regulations through firm characteristics and across countries and sectors. Gelb et al. (2007) use data covering firms located in 26 Sub-Saharan African countries to examine the implications of various factors for firm perceptions of the business environment. Among the results, they find that large firms have higher propensities to complain about the rigidity of labour regulations. They argue that, in developing countries, small firms are more likely to work within restricted markets and, therefore, they are less visible to regulators, and are less appealing targets for officials. They also find that exporting activities and private foreign ownership do not exhibit statistically significant impacts on firm perceptions of labour regulations. They also report that firm perceptions of labour regulations do not exert considerable variations across sectors. Hallward-Driemeier and Aterido (2009) investigate the role of firm characteristics in determining firm perceptions of various business constraints, using data covering firms located in 105 countries. The results suggest that small firms may face more problems in dealing with the requirements of poor labour regulations, and may become easier targets in corrupted environments in developing countries. Clarke (2010) uses data covering firms located in South Africa to study the factors influencing firm perceptions of business constraints. Among the results, he finds that large firms and exporting firms are more likely to report labour regulations as a significant obstacle to their development. Also, he finds that firm age and type of firm ownership do not exercise statistically significant role in explaining the variations in firm perceptions of labour regulations. Bartelsman et al. (2010) document how business constraints, as perceived by individual firms, differ across countries and across firm characteristics. They show that firm age, ownership type, export orientation, industry binary variables, and country binary variables are important indicators in explaining the differences in firm perceptions of business constraints between Europe and Central Asia (ECA) and Latin America and the Caribbean regions. The results also show that large firms and exporting firms in the ECA region are more likely to be affected by rigid labour regulations, reflecting costs of labour adjustment. Vargas (2012) implements the analysis for firms located in Bolivia, focusing on the implications of firm size for firm perceptions of business constraints. Among the results, he finds that small firms are less likely to identify labour regulations as an important business constraint. Also, he finds that manufacturing firms are more likely to report labour regulations as a significant business constraint compared to firms belonging to other sectors. Meyer and Vandenberg (2013) analyse the relationship between firm characteristics and firm perceptions of labour regulations using data covering firms located in five Asian economies (Bangladesh, Indonesia, Pakistan, the Philippines, and Vietnam). They find systematic differences in firm perceptions of labour regulations across firm characteristics and sectors. In four of the five countries examined, they find that exporting firms are more likely to perceive labour regulations as an important obstacle to their operations compared to non-exporting firms. They also find that, in four of the five countries, large firms have higher propensities to perceive labour regulations as a significant business constraint. They argue that small firms may enjoy de facto or de jure exemption from the enforcement of labour regulations. Also, they show that young firms have higher tendencies to report difficulties in coping with labour regulations. 2.2. Labour skill shortages Mismatches between worker skills and job requirements are often classified among the significant obstacles facing firm productivity and growth, particularly in developing countries (Almeida and Aterido, 2011). For example, skill mismatches form an important constraint on hiring decisions, leading to higher levels of unemployment and important costs for firms (O'Sullivan et al., 2011) and causing higher turnover rates (Hersch, 1991). Gupta et al. (2010) argue that weak performances of labour-intensive industries could be related to labour skill shortages. Piore (1986) suggests that a low-skilled workforce may decrease the internal flexibility, leading to inefficient functioning of firms. Empirical evidence on the factors influencing firm perceptions of labour skill shortages points out to large variations through firm characteristics and across countries and sectors. Gelb et al. (2007) argue that the problem of labour skill shortages tends to be more prevalent for firms that are large, more productive, and use advanced technologies. Additionally, they note that small firms may use less sophisticated production techniques that require fewer skilled workers. Using data set covering firms located in Sub-Saharan Africa, Gelb et al. (2007) find that large firms are more likely to report labour skill shortages as one main business constraint. They also find that exporting activities and type of firm ownership do not exhibit important influences on firm perceptions of labour skill shortages. They note that the availability of skilled labour may become a more binding constraint as economies become more sophisticated and as governments magnify their capacities in enforcing labour regulations. Hallward-Driemeier and Aterido (2009) find that large firms and exporting firms are more likely to report labour skill shortages as a significant obstacle to business operations. They also find that government-owned firms are less likely to perceive labour skill shortages as an important business constraint. In the case of firms located in South Africa, Clarke (2010) finds that large firms and exporting firms have higher propensities to perceive labour skill shortages as a significant business obstacle. He also reports that firm age and type of firm ownership do not exhibit important influences on firm perceptions of labour skill shortages. Kaplan and Pathania (2010) find that large firms are more likely to identify labour skill shortages as one main business constraint. However, Vargas (2012) does not find significant influences of firm size, but he reports important variations across sectors. Using data covering firms located in Canada, Sabourin (2001) shows that firm size and technological intensity are negatively correlated with the probability of firms to identify labour skill shortages as one primary business obstacle. Also, he finds that type of firm ownership and Research and Development activities have no statistically significant influences on these firm perceptions. Baldwin and Lin (2002) find that young firms are more likely to designate labour skill shortages as a considerable problem using data covering Canadian manufacturing firms. However, they find that firm size and ownership type do no exhibit important influences. Green et al. (1998) implement the empirical analysis for firms located in the UK. They find that higher skilled share of employment accentuates the problem of labour skill shortages as perceived by employers. They also report that the share of part-time workers does not exhibit important effects. Lyon et al. (2012) find that medium and large size firms are more likely to designate labour skill shortages as an important constraint on firm operations and growth. They find that exporting firms have lower propensities to report labour skill shortages as a significant business constraint compared to non-exporting firms. Also, they do not find important variations across sectors. The empirical analysis is carried out for the perceived levels of labour market constraints as reported by the respondents (e.g. senior managers, business owners) through the World Bank's Enterprise Surveys database. Pierre and Scarpetta (2004, 2006) examine the relationship between the perceived and actual stringency of labour regulations using national labour protection indices (i.e. de jure labour laws). They find that the reported perceptions are closely related to the actual levels of labour regulations' constraints. Specifically, countries with higher national indices on the stringency of labour regulations are associated with higher proportions of firms perceiving labour regulations as being significant constraints. Kaplan and Pathania (2010) indicate that the perception-based indicators capture the actual institutional environment (e.g. laws, regulations, infrastructure, corruption), particularly when objective measures are inadequate. They argue that the perception-based indicators reflect the de facto severity of business constraints, and serve as a normal check vis-a-vis national indicators. Clarke (2010) notes that the collection of objective measures could be difficult, particularly when it comes to sensitive indicators (e.g. corruption). In such cases, perception-based indicators can serve as a substitute. Hallward-Driemeier and Aterido (2009) find that the perception-based and objective measures of business constraints are closely correlated with each other. They indicate that subjective measures enable the ranking of business constraints within and between countries, particularly when comparing the rigidity of the business environment. Also, they argue that perception-based measures provide information on the most severe obstacles for businesses growth and development as identified by the business entity, and would eventually assist managers and policy makers to recognize reform priorities. Hence, perception-based measures tend to reflect the actual rigidity of business constraints faced by firms. However, there are some concerns regarding perception-based measures. Specifically, these measures could be subjected to: first, potential measurement errors leading to biased results; second, variations in managers willingness to report actual negative or positive responses due to differences in personalities, views, and culture; third, performance bias where managers tend to rank business obstacles according to the performance of their firms or to the business environment conditions through which the firm operates; and finally, defects due to imperfections in question formulations and answer interpretations (Bertrand and Mullainathan, 2001; Senik, 2005; Hallward-Driemeier and Aterido, 2009). Some studies (e.g. Glaeser et al., 2004; Kaplan and Pathania, 2010) argue that caution should be exercised when using perception-based measures as independent variables through empirical analyses. This is because these measures may embody information on growth, which would lead to reverse causality issues. Meanwhile, these studies indicate that using these measures as dependent variables is more intuitive to capture growth in the business environment. These indications complement the aforementioned empirical findings reflecting a close correspondence between perceived and actual measures of business constraints. It is important to note that the empirical analysis in this paper exclusively covers firms that exist in the market (i.e. those that have already entered the market, and that did not exit the market). This is to say that the empirical analysis examines the business obstacles as perceived by the existing firms in the market. Hausmann and Velasco (2005) depict this issue through a camel-hippopotamus analogy. Camels living in the desert do not identify access to water as a primary problem since they have adjusted to the situation by learning how to conserve water. Meanwhile, hippopotami lack the ability to stay in the desert since water scarcity is a critical obstacle for their survival. Hence, by generating and analysing data from interviews with "camels" about water constraints, we may be missing distinct information from "hippopotami". Gelb et al. (2007) argue that such selection is incomplete since many firms choose to enter the market despite the severity of business constraints. They also indicate that the ability to adjust to a business constraint does not mean that firms do not recognize it anymore as a serious problem. They provide an illustration using the World Bank's Enterprises Surveys database: the perceptions of generator-owning firms regarding the electricity as a business constraint are not distinguishable from the perceptions of firms that do not own generators. Furthermore, they note that the intensity of complaints are often highly correlated with corresponding national indices. Dethier et al. (2011) suggest that empirical analyses covering "hippopotami" should be carried out using different frameworks (e.g. entry or exit models[4]). We use a data set sourced from the World Bank's Enterprise Surveys database. This database represents a comprehensive source of firm-level data in emerging and developing economies. It covers firms operating in the manufacturing, service, and other sectors. It contains information on various aspects of the business environment such as, access to finance, corruption, workforce characteristics, innovation and technology, and trade. It should be noted that one of many advantages of using data from these surveys is that the questions are identical through firms across all countries. The basic data set used in this paper covers 5,052 firms located in eight developing Arab countries of the MENA region: Algeria, Egypt, Jordan, Lebanon, Morocco, Oman, Syria, and Yemen[5]. Through these surveys, firms' representatives (e.g. senior managers, business owners) are asked whether labour regulations and labour skill shortages are considered prominent constraints on business operations and development. The responses are used to generate the basic dependent binary variables, which equal one when a firm declares the corresponding labour market constraint as a major/severe business obstacle and zero otherwise. Following the previous literature on business constraints (e.g. Gelb et al., 2007; Kaplan and Pathania, 2010; Clarke, 2010; Meyer and Vandenberg, 2013), the explanatory variables cover firm characteristics' variables which comprise: first, a binary variable that equals one for firms characterized by private foreign ownership and zero otherwise[6]; second, a binary variable that equals one for firms engaged in exporting activities and zero otherwise; third, firm size depicted through the number of total employment and presented in hundreds of workers through the regressions; fourth, firm age measured by the number of years since firm establishment; fifth, firm use of Information and Communication Technology (ICT) depicted through a binary variable that takes the value of one for firms regularly using the internet to communicate with clients and to source information and zero otherwise; finally, two firm labour composition variables: the first is represented through the fraction of skilled production workers in total number of production workers, and the second is depicted through the fraction of non-production workers in total employment. The use of skilled production workers' share and the use of non-production workers' share through the regressions imply that the results are relative to the use of unskilled production workers and to the use of production workers, respectively. The explanatory variables also include country-specific binary variables and sector-specific binary variables to capture cross-country and cross-sector variations. Table II displays summary statistics of the variables used in the empirical analysis. The first panel shows the results for the dependent variables. We find that around 24.6 per cent of firms in the regression data set report that labour regulations represent a major/severe business constraint, while 38.0 per cent of firms report that labour skill shortages exercise a major/severe business constraint[7]. The second panel shows statistics for the explanatory variables. The results indicate that 4.8 per cent of firms are characterized by private foreign ownership and that 22.5 per cent of firms are engaged in exporting activities. The average firm size is 124.0 full-time workers (with a standard deviation of 317.1 full-time workers) and the average firm age is 20.9 years (with a standard deviation of 16.6 years). Also, we find that 50.4 per cent of firms use the internet to communicate with clients and to source information. The percentage of skilled production workers in total number of production workers and the percentage of non-production workers in total employment have averages of 61.6 and 26.1 per cent, respectively. Consider a given firm j (j=1, ..., J) belonging to sector k (k=1, ..., K) and located in country c (c=1, ..., C). Firm perception levels of constraints related to labour regulations and those related to labour skill shortages are depicted through the latent variables (Equation 1) and (Equation 2), respectively. These latent variables are not observed. However, we observe the perceptions of firms through dichotomous responses on whether labour regulations and labour skill shortages do or do not pose major/severe obstacles on firm operations and development. Let R jkc depict a binary variable that takes the value of one when the corresponding firm identifies labour regulations as a major/severe business constraint and zero otherwise. Also, let S jkc represent a binary variable that takes the value of one when the corresponding firm identifies labour skill shortages as a major/severe business constraint and zero otherwise. The benchmark empirical specifications can be represented as: (Equation 3) (Equation 4) where X j is the vector of variables depicting firm characteristics, Y k is the vector of binary variables depicting sectors, Z c is the vector of binary variables depicting countries, and (Equation 5) and (Equation 6) are the stochastic error terms of the corresponding equations. The univariate probit estimator would produce biased estimates when there are some unobserved or omitted characteristics that simultaneously affect firm perceptions of labour regulations and labour skill shortages (Deadman and MacDonald, 2004; Greene, 2008). We allow for the errors in these two labour market constraints' equations to be potentially correlated. Thus, the two equations are jointly modelled using a bivariate probit estimator. The error terms are assumed to be independently and identically distributed as bivariate normal with r depicting the correlation parameter. Specifically, we have: (Equation 7) Through the empirical analysis, we use the Wald test to determine whether the correlation parameter r is statistically significant[8]. The rejection of the null hypothesis indicates that firm perceptions of the labour market constraints are jointly formulated (Greene, 2008 [9]). The bivariate probit specification allows us to estimate unconditional marginal effects, but also conditional and joint marginal effects of variables influencing firm perceptions of the labour market constraints. Table III presents the marginal effects from the benchmark bivariate probit estimation carried out for the pooled data set covering existing firms' perceptions of labour market constraints. The Wald test rejects the null hypothesis of zero correlation between the errors in the two labour market constraints' equations and, hence, it indicates that the model should be estimated through the bivariate probit estimator rather than through the univariate probit estimator. The estimated coefficient of correlation between the errors in the two equations is positive and statistically significant at the 1 per cent level. Table III displays the unconditional marginal effects for Pr(R jkc =1) and Pr(S jkc =1). It also includes the joint marginal effects for Pr(R jkc =1, S jkc =1) and Pr(R jkc =0, S jkc =0), and the conditional marginal effects for Pr(R jkc =1|S jkc =1) and Pr(S jkc =1|R jkc =1[10]). These marginal effects are determined at the mean values of the explanatory variables. For explanatory binary variables, the marginal effects are calculated through discrete changes in probabilities as the binary variable changes from 0 to 1. 6.1. Firm characteristics We find that larger firms have lower propensities to identify labour market constraints as major/severe obstacles facing business operations and development. This is in line with the results reported in some other empirical studies (e.g. Pierre and Scarpetta, 2004; Hallward-Driemeier and Aterido, 2009). The unconditional marginal effects imply that an increase in firm size by one hundred full-time workers reduces the likelihood of firms to perceive labour regulations and labour skill shortages as major/severe constraints on business operations by 0.8 and 0.5 percentage points, respectively. Also, the joint marginal effect of firm size for Pr(R jkc =1, S jkc =1) indicates that an increase in firm size by 100 workers decreases the likelihood of firms to jointly identify labour regulations and labour skill shortages as major/severe business constraints by 0.5 percentage points. The results on the relationship between firm size and firm perceptions of labour regulations are consistent with arguments provided in the literature, which suggest that smaller firms have more limited abilities to realize internal adjustments in response to labour regulations (Meyer and Vandenberg, 2013). Hence, smaller firms have higher tendencies to be more adversely affected by labour regulations. These results are also in line with some other findings showing that labour regulations have disproportional effects on small firms (Aterido et al., 2011). The results on the relationship between firm size and firm perceptions of labour skill shortages imply that smaller firms in MENA countries may encounter more difficulties in realizing future growth, since adequacy of educated workforce is essential for firm competitiveness and performance (Storey, 1994; Jensen and McGuckin, 1997). These results could also suggest that smaller firms have more difficulties in accessing skilled labour markets, identifying required labour skills, and investing in workforce training programmes (Jansen and Lanz, 2013). The implication of exporting activities for firm perceptions of labour regulations is not statistically significant. These results are similar to those found in some previous studies, such as Gelb et al. (2007) in the case of Sub-Saharan African firms. Yet, they deviate from those reported in some studies covering firms located in other developing regions. For example, Meyer and Vandenberg (2013) find that exporting firms located in some South East Asian countries have higher tendencies to perceive labour regulations as a major/severe business constraint compared to non-exporting firms. They attributed these findings to variations in labour demand requirements between exporting and non-exporting firms. Also, Lyon et al. (2012) find that non-exporting firms are more likely to identify labour skill shortages as a major/severe business constraint compared to exporting firms. The effect of firm age is negative and statistically significant in the case of labour skill shortages' equation. The unconditional marginal effect and the conditional marginal effect for Pr(S jkc =1|R jkc =1) indicate that an increase in firm age by ten years reduces the likelihood to perceive labour skill shortages as a major/severe business constraint by 1 percentage point. Although relatively small in magnitude, this impact is consistent with some a priori expectations discussed in the literature, suggesting that older firms have more accumulated experiences to attenuate the effects of business obstacles as perceived by the survey respondents (Meyer and Vandenberg, 2013). We find that an increase in the proportion of non-production workers in total employment by 10 percentage points raises the likelihood of firms to identify labour skill shortages as a major/severe business constraint by 1.3 percentage points. Also, we find that an increase in this proportion by 10 percentage points raises this likelihood by 1.6 percentage points, when assessed conditional on identifying labour regulations as a major/severe business constraint. The effect of the proportion of skilled production workers in total number of production workers on firm perceptions of these labour market constraints is not statistically significant. These findings suggest that the implications of labour skill shortages are primarily associated with the inadequacy of non-production workers rather than with the inadequacy of skilled production workers. The results show that firms with private foreign ownership are less likely to perceive labour skill shortages as a major/severe business constraint by 6.0 percentage points. They suggest that foreign affiliates have higher capabilities in attenuating the implications of labour skill shortages compared to local, domestically owned firms. Almeida (2007) indicates that foreign-owned firms have higher tendencies to pay higher wages and to attract educated workforce compared to domestic firms. These tendencies could eventually mitigate the perceptions of firms with private foreign ownership concerning labour skill shortages compared to firms with domestic ownership. 6.2. Countries' binary variables The results emphasize significant variations in the perceptions of firms located in different countries of the MENA region concerning labour market constraints. Hence, they underscore the implications of national characteristics for these firm perceptions. The binary variable depicting firms located in Algeria is set as the reference for MENA countries through the regressions. The unconditional marginal effects reveal that firms located in Lebanon have statistically higher propensities to perceive labour regulations as a major/severe business constraint by 34.4 percentage points compared to firms located in the Maghreb countries (i.e. Algeria and Morocco) and in Yemen. The corresponding propensities for firms located in Egypt, Oman, and Syria are higher by 17.3, 31.9, and 22.3 percentage points, respectively, compared to firms located in the two Maghreb countries and in Yemen. Also, firms located in Egypt, Lebanon, and Oman have statistically higher propensities to perceive labour skill shortages as a major/severe business constraint by 15.4, 17.5, and 13.5 percentage points, respectively, compared to firms located in the other MENA countries[11]. These results complement other empirical findings in the literature, which document international variations in the perceived magnitude of labour market constraints. For example, Pierre and Scarpetta (2004) show that countries with more stringent de jure labour laws have higher proportions of firms identifying labour regulations as being restrictive. Gelb et al. (2007) find significant variations in firm perceptions of labour market constraints, among other business obstacles, across Sub-Saharan African countries. Also, Lyon et al. (2012) present some important regional differences in firm perceptions of labour skill shortages. Drzeniek-Hanouz and Dusek (2013) indicate that labour market constraints constitute significant obstacles on business development in the Arab world. They also describe considerable variations in the perceptions of firms located in different MENA sub-regions: North Africa (including Egypt), the Levant, and the Gulf sub-regions. 6.3. Sectors' binary variables We find that firm perceptions of labour market constraints vary across sectors. The binary variable for firms belonging to the service sector is set as the reference for sectors through the regressions. The unconditional marginal effects indicate that firms belonging to the manufacturing sector and to the category covering the remaining other (non-manufacturing and non-service) sectors have higher propensities to perceive labour regulations as a major/severe business constraint by 9.9 and 10.1 percentage points, respectively, compared to firms belonging to the service sector. The estimates for the labour skill shortages' equation are 12.6 and 11.8 percentage points, respectively. These results suggest that labour market constraints, as perceived by the survey respondents, disproportionally impact manufacturing firms and firms belonging to the other sectors compared to firms belonging to the service sector. Firm perceptions of labour market constraints can be alternatively represented through ordinal responses in the data set. Specifically, firms identify the stringency of labour regulations and labour skill shortages as: no or minor obstacle (outcome 1), moderate obstacle (outcome 2), and major or very severe obstacle (outcome 3). Table IV displays the marginal effects obtained through the ordered probit model. The results are comparable to the corresponding marginal effects for Pr(R jkc =1) and Pr(S jkc =1) that are reported in Table III. 7.1. Empirical results by sector The results from the empirical analysis implemented for data sets covering individual sectors are displayed in Table V. The estimated coefficients of correlation between the errors in the two labour market constraints' equations are positive and statistically significant at the 1 per cent level for all sectors. To save space, Table V does not display the rows of marginal effects of firm characteristics that do not exhibit statistical significance at all. In general, the results for firms belonging to the manufacturing sectors are qualitatively similar to the benchmark results for pooled data set. We also find few differences. For example, the perceptions of manufacturing firms located in Jordan concerning labour skill shortages appear to be less pronounced compared to manufacturing firms located in the other MENA countries. Firms located in Jordan have statistically lower propensities to perceive labour skill shortages as a major/severe business constraint compared to firms located in Algeria and those located in Egypt by 11.3 and (11.3+14.0=) 25.3 percentage points, respectively. The results derived from the service sector's data set reveal some notable differences compared to those obtained from the manufacturing sector's data set[12]. In the case of the service sector, firm perceptions of labour regulations do not exhibit statistically significant heterogeneity across firm characteristics. We find that firm perceptions of labour skill shortages are accentuated with an increase in the ratio of non-production workers to total employment. The marginal effect shows that an increase in this ratio by 10 percentage points raises the likelihood of perceiving labour skill shortages as a major/severe business constraint by 3.1 percentage points. The magnitude of this effect is considerably higher than the one reported for manufacturing firms. Our data set does not contain service firms located in Algeria that report labour regulations and labour skill shortages as major/severe constraints on business operations. In other words, service firms located in Algeria predict failure perfectly and, therefore, they are dropped from the service sector's data set. Then, the binary variable depicting firms located in Egypt is set as the reference for MENA countries. We find that service firms located in Egypt and Lebanon have statistically higher likelihoods than those located in Jordan to perceive labour regulations as a major/severe business constraint by 6.7 and (6.7+2.1=) 8.8 percentage points, respectively. The corresponding likelihood for service firms located in Oman is considerably higher than those located in Jordan by (18.4+6.7=) 25.1 percentage points. We also find that service firms located in Egypt and Lebanon have lower propensities than those located in Jordan to identify labour skill shortages as a major/severe business constraint by 16.1 and (16.1-1.3=) 14.8 percentage points, respectively. The joint marginal effects for Pr(R jkc =1, S jkc =1) indicate that service firms located in Egypt and Jordan have statistically comparable likelihoods to identify both labour market constraints as major/severe obstacles on business operations as perceived by the survey respondents. These likelihoods are lower than the corresponding likelihood for service firms located in Oman by more than 11 percentage points. Finally, the results for the other sectors' category reveal that firms located in Oman are statistically more likely to perceive labour regulations as a major/severe business constraint by 34.9 percentage points compared to firms located in the other MENA countries (except Lebanon). Also, we find that firms located in Egypt and Lebanon have higher propensities to perceive labour skill shortages as a major/severe business constraint by 44.2 and 31.3 percentage points, respectively, compared to firms located in the other MENA countries. 7.2. Empirical results by country The results from the empirical analysis carried out for data sets covering firms located in individual MENA countries are displayed in Table VI. The estimated coefficients of correlation between the errors in the two labour market constraints' equations are positive and statistically significant at the 1 per cent level for all countries. To save space, Table VI does not display the rows of marginal effects of firm characteristics that do not exhibit statistical significance at all. In the case of Algeria, service firms predict failure perfectly and, hence, they are dropped from the data set. Then, the other sectors' category is set as the reference for sectors. We find that the perceptions of firms located in Algeria concerning labour regulations do not exhibit statistically significant variations between the manufacturing sector and the other sectors' category. Exporting firms are less likely to perceive labour regulations as a major/severe business constraint by 13.5 percentage points compared to non-exporting firms. Also, an increase in the skilled production workers' ratio by 10 percentage points reduces the extent of these perceptions by 0.7 percentage points. We find that manufacturing firms have a statistically higher likelihood to identify labour skill shortages as a major/severe business constraint by 12.1 percentage points compared to other sectors' firms. Manufacturing firms located in Egypt have statistically higher propensities to perceive labour regulations and labour skill shortages as major/severe business constraints by 15.7 and 20.6 percentage points, respectively, compared to service firms. The implications of firm size for firm perceptions of labour regulations are similar to the benchmark results. Also, the effects of firm age, private foreign ownership, and the proportion of non-production workers in total employment are reminiscent of those obtained from the pooled regressions. Service firms located in Jordan have statistically higher propensities to identify labour skill shortages as a major/severe business constraint compared to manufacturing and other sector's firms by 19.2 and 13.6 percentage points, respectively. The non-production workers' ratio exhibits positive implications for firm perceptions of labour skill shortages. An increase in this ratio by 10 percentage points raises the magnitude of these perceptions by 3.2 percentage points. In contrast, this ratio has negative impacts on firm perceptions of labour regulations. An increase in this ratio by 10 percentage points induces a decrease in the magnitude of these perceptions by 2.0 percentage points. In the case of Lebanon, manufacturing and service firms have lower propensities to perceive labour skill shortages as a major/severe business constraint by 34.6 percentage points compared to firms belonging to the other sectors. Firm characteristics do not exhibit statistically significant implications, with the exception of firm use of ICT in the case of labour regulations. The perceptions of firms located in Oman concerning labour market constraints do not exhibit statistically significant variations across sectors. The conditional marginal effects for Pr(R jkc =1|S jkc =1) indicate that exporting activities and an increase in firm size by one hundred workers reduce the corresponding probabilities by 34.6 and 38.0 percentage points, respectively. In our data set, firms located in Morocco, Syria, and Yemen belong exclusively to the manufacturing sector. We find that exporting firms located in Morocco are more likely to perceive labour regulations and labour skill shortages as major/severe business constraints by 15.3 and 22.3 percentage points, respectively, compared to non-exporting firms. Also, an increase in the skilled production workers' ratio by 10 percentage points raises the likelihood of identifying labour skill shortages as a major/severe business constraint by 2.9 percentage points. Finally, exporting firms located in Yemen are less likely to perceive labour regulations as a major/severe business constraint by 11.2 percentage points, and are more likely to perceive labour skill shortages as a major/severe business constraint by 34.8 percentage points, compared to non-exporting firms. Labour market constraints are often identified as main business obstacles facing firm operation and development in the MENA region. Therefore, they are naturally listed through the primary items on the labour policy agenda of the MENA countries. The comprehension of the factors influencing the perceived severity of labour market constraints is essential in the design of policies aiming at improving labour market conditions and enhancing business environments. This paper examines the implications of firm characteristics, national locations, and sectoral associations for the perceptions of firms located in the MENA region concerning two primary labour market constraints: labour regulations and labour skill shortages. The empirical analysis is carried out using firm-level data set sourced from the World Bank's Enterprise Surveys database. A bivariate probit estimator is used to account for potential correlations between the errors in the labour regulations' equation and labour skill shortages' equation. The empirical results are generated through overall estimations and by implementing comparative cross-country and cross-sector analyses. The benchmark empirical results obtained from the overall estimations reveal important implications of firm characteristics for firm perceptions of labour market constraints. Larger firms appear to have lower propensities to identify labour market constraints as major/severe business obstacles. These results imply that small firms have limited flexibility in adjusting to labour regulations and to labour skill shortages. Hence, they suggest that small firms may encounter difficulties in realizing future growth through the MENA region. Also, we find that firms with private foreign ownership are less likely to endure the implications of labour skill shortages compared to domestic firms. This is consistent with higher tendencies of foreign affiliates to pay higher wages and to attract educated workforce compared to domestic firms. Meanwhile, exporting activities do not have statistically significant effects on firm perceptions of labour market constraints. This finding could stem from the generally limited engagement of MENA exporting firms in international trade. We find that firms with higher proportions of non-production workers in total employment have higher propensities to perceive labour skill shortages as a major/severe business constraint. However, the proportion of skilled production workers in total number of production workers does not exercise statistically significant influences on firm perceptions. These findings suggest that the implications of labour skill shortages are primarily associated with inadequacy of non-production workers rather than inadequacy of skilled production workers. The results also emphasize considerable cross-country and cross-sector variations in firm perceptions of labour market constraints. Finally, the empirical analysis reveals significant heterogeneity through the implications of firm characteristics and national locations for the perceptions of firms belonging to different sectors concerning labour market constraints. Also, it underlines important heterogeneity through the implications of firm characteristics and sectoral associations for the perceptions of firms located in different countries concerning labour market constraints. Reforms in labour regulations and investment in human capital are important governmental policy interventions for promoting firm development and economic growth. This paper provides policy makers with information needed to design labour policies that attenuate the impact of labour market constraints and enhance the performance of firms in the MENA region. Specifically, the design of labour policies should encompass the findings that the perceptions of labour market constraints significantly vary through firm characteristics, countries, and sectors. Furthermore, given the heterogeneity in the implications of the factors influencing firm perceptions of labour market constraints, labour policies should be customized by country, and should recognize the varying consequences for different sectors. Opens in a new window. Table I Percentage of firms identifying business obstacles as major/severe constraints Opens in a new window. Table II Descriptive statistics Opens in a new window. Table III Benchmark empirical results from the bivariate probit model (marginal effects) Opens in a new window. Table IV Empirical results from the ordered probit model (marginal effects) Opens in a new window. Table V Empirical results by sector (marginal effects from the bivariate probit model) Opens in a new window. Table VI Empirical results by country (marginal effects from the bivariate probit model) Opens in a new window. Table AI Descriptive statistics for the dependent variables by country and by sector Opens in a new window. Table AII Other results from the bivariate probit model Opens in a new window. Equation 1 Opens in a new window. Equation 2 Opens in a new window. Equation 3 Opens in a new window. Equation 4 Opens in a new window. Equation 5 Opens in a new window. Equation 6 Opens in a new window. Equation 7 Opens in a new window. Equation 8 Opens in a new window. Equation 9 Opens in a new window. Equation 10 Opens in a new window. Equation 11 Opens in a new window. Equation 12 Opens in a new window. Equation 13 Opens in a new window. Equation 14 Opens in a new window. Equation 15 Opens in a new window. Equation 16 Opens in a new window. Equation 17
|
The empirical results reveal some important implications of firm characteristics (e.g. firm size, labour compositions) for firm perceptions of labour regulations and labour skill shortages. They also delineate important cross-country and cross-sector variations. The authors also find significant heterogeneity in the factors' implications for the perceptions of firms belonging to different sectors and located in different MENA countries.
|
[SECTION: Value] Stringent labour market constraints are expected to pose serious obstacles to firm performance and economic growth. A wide range of literature finds that rigid labour regulations would induce lower labour force participation and higher unemployment rates (e.g. Botero et al., 2004; Besley and Burgess, 2004; Amin, 2009; Djankov and Ramalho, 2009), and would prevent labour markets from being efficient leading to losses in productivity (e.g. Kaplan, 2009). Another strand of literature inspects the problem of labour skill shortages or "skill deficits", which can be defined as the divergence between the educational attainments of workers and the skill requirements of jobs (Kiker et al., 1997). This literature regularly indicates that accentuated labour skill shortages impose significant restrictions on employment creation and economic growth (e.g. Pissarides and Veganzones-Varoudakis, 2007; Bhattacharya and Wolde, 2012), and could eventually inflict severe impacts on economic performance and labour market outcomes (e.g. Allen and van der Velden, 2001). The implications of labour market constraints are deemed to be of particular concerns for the Middle East and North Africa (MENA) region. Although the MENA region has generally realized higher annual employment growth rates compared to other geo-economic regions (International Labour Organization (ILO), 2012), rigid labour regulations and lack of suitable labour skills were often identified as main obstacles to firm operation and development in the MENA region (Drzeniek-Hanouz and Dusek, 2013). MENA countries have one of the least flexible labour markets, which remain too tight compared to labour markets in other developing countries (Angel-Urdinola and Kuddo, 2010; Bhattacharya and Wolde, 2012 [1]). Table I presents the percentages of firms identifying different business obstacles as major or very severe (henceforth, major/severe) constraints across geo-economic regions. The statistics emphasize that the MENA region has the highest percentages of firms that consider labour regulations and labour skill shortages to be major/severe business constraints compared to other geo-economic regions. For example, labour regulations and labour skill shortages are identified as major/severe problems to 25 and 38 per cent of firms located in the MENA region, respectively, compared to only 9 and 16 per cent of firms located in South Asia[2]. Also, in the case of the MENA region, Table I shows that the constraint of labour skill shortages is associated with one of the highest percentages compared to other business obstacles, whereas the constraint of labour regulations occupy an intermediate rank. Understanding the factors influencing employability constraints constitute an essential requirement towards the implementation of policies aiming at improving the regulatory and institutional environment for business (Pierre and Scarpetta, 2004; Kaplan and Pathania, 2010). This is particularly relevant for MENA countries that have recently undergone significant labour reforms, where employment creation is listed among the primary objectives (Angel-Urdinola and Kuddo, 2010). Several studies underline systematic differences in firm perceptions of labour market constraints through firm characteristics, as well as across countries and sectors (e.g. Pierre and Scarpetta, 2004; Kaplan and Pathania, 2010; Lyon et al., 2012; Meyer and Vandenberg, 2013). Therefore, designing policies that alleviate the adverse implications of labour market constraints requires a comprehension of the categories of firms that are more likely to endure the burden of these constraints. This paper examines the implications of firm characteristics for the perceptions of firms located in the MENA region concerning the stringency of labour regulations and labour skill shortages. It also examines the heterogeneity in firm perceptions of labour market constraints across MENA countries and across sectors. We use a data set sourced from the World Bank's Enterprise Surveys database, which includes information on firm perceptions of many business constraints. Firm perception data are collected through surveys, which are generally answered by senior managers and business owners. Hence, following the previous literature (e.g. Pierre and Scarpetta, 2004; Kaplan and Pathania, 2010; Meyer and Vandenberg, 2013), our dependent variables are determined as perception-based indicators. We implement overall examination and comparative cross-country and cross-sector empirical analyses of the factors influencing firm perceptions of labour regulations and labour skill shortages for the MENA region. MENA countries are characterized by a considerable level of diversity in macroeconomic and industrial factors, and by important differences in labour market conditions and institutions (Angel-Urdinola and Kuddo, 2010), suggesting potential cross-country heterogeneity. Also, cross-sector heterogeneity is expected given the varying structures in labour requirements between sectors. Through the basic regressions, we use a bivariate probit model that allows for firm perceptions of labour regulations and labour skill shortages to be jointly formulated. This estimation model takes into account potential correlation between the error terms in the two labour market constraints' equations. 2.1. Labour regulations Labour regulations are normally expressed through laws that regulate and govern the employment relationships between employees, employers, unions, and government in order to protect the basic standards of fair treatment for workers (Kaplan, 2009) and to maximize the social welfare (Botero et al., 2004). Governments commonly intervene in labour markets through regulations because employers may mistreat workers, leading to unfairness and inefficient outcomes for workers such as unfair dismissals, unfair minimum wage, lay-offs for economic reasons, and under-payment (Djankov and Ramalho, 2009 [3]). These regulations and laws may be written through labour codes, current legislations, and norms set by collective agreements (Pierre and Scarpetta, 2004). Several studies report variations in firm perceptions of labour regulations through firm characteristics and across countries and sectors. Gelb et al. (2007) use data covering firms located in 26 Sub-Saharan African countries to examine the implications of various factors for firm perceptions of the business environment. Among the results, they find that large firms have higher propensities to complain about the rigidity of labour regulations. They argue that, in developing countries, small firms are more likely to work within restricted markets and, therefore, they are less visible to regulators, and are less appealing targets for officials. They also find that exporting activities and private foreign ownership do not exhibit statistically significant impacts on firm perceptions of labour regulations. They also report that firm perceptions of labour regulations do not exert considerable variations across sectors. Hallward-Driemeier and Aterido (2009) investigate the role of firm characteristics in determining firm perceptions of various business constraints, using data covering firms located in 105 countries. The results suggest that small firms may face more problems in dealing with the requirements of poor labour regulations, and may become easier targets in corrupted environments in developing countries. Clarke (2010) uses data covering firms located in South Africa to study the factors influencing firm perceptions of business constraints. Among the results, he finds that large firms and exporting firms are more likely to report labour regulations as a significant obstacle to their development. Also, he finds that firm age and type of firm ownership do not exercise statistically significant role in explaining the variations in firm perceptions of labour regulations. Bartelsman et al. (2010) document how business constraints, as perceived by individual firms, differ across countries and across firm characteristics. They show that firm age, ownership type, export orientation, industry binary variables, and country binary variables are important indicators in explaining the differences in firm perceptions of business constraints between Europe and Central Asia (ECA) and Latin America and the Caribbean regions. The results also show that large firms and exporting firms in the ECA region are more likely to be affected by rigid labour regulations, reflecting costs of labour adjustment. Vargas (2012) implements the analysis for firms located in Bolivia, focusing on the implications of firm size for firm perceptions of business constraints. Among the results, he finds that small firms are less likely to identify labour regulations as an important business constraint. Also, he finds that manufacturing firms are more likely to report labour regulations as a significant business constraint compared to firms belonging to other sectors. Meyer and Vandenberg (2013) analyse the relationship between firm characteristics and firm perceptions of labour regulations using data covering firms located in five Asian economies (Bangladesh, Indonesia, Pakistan, the Philippines, and Vietnam). They find systematic differences in firm perceptions of labour regulations across firm characteristics and sectors. In four of the five countries examined, they find that exporting firms are more likely to perceive labour regulations as an important obstacle to their operations compared to non-exporting firms. They also find that, in four of the five countries, large firms have higher propensities to perceive labour regulations as a significant business constraint. They argue that small firms may enjoy de facto or de jure exemption from the enforcement of labour regulations. Also, they show that young firms have higher tendencies to report difficulties in coping with labour regulations. 2.2. Labour skill shortages Mismatches between worker skills and job requirements are often classified among the significant obstacles facing firm productivity and growth, particularly in developing countries (Almeida and Aterido, 2011). For example, skill mismatches form an important constraint on hiring decisions, leading to higher levels of unemployment and important costs for firms (O'Sullivan et al., 2011) and causing higher turnover rates (Hersch, 1991). Gupta et al. (2010) argue that weak performances of labour-intensive industries could be related to labour skill shortages. Piore (1986) suggests that a low-skilled workforce may decrease the internal flexibility, leading to inefficient functioning of firms. Empirical evidence on the factors influencing firm perceptions of labour skill shortages points out to large variations through firm characteristics and across countries and sectors. Gelb et al. (2007) argue that the problem of labour skill shortages tends to be more prevalent for firms that are large, more productive, and use advanced technologies. Additionally, they note that small firms may use less sophisticated production techniques that require fewer skilled workers. Using data set covering firms located in Sub-Saharan Africa, Gelb et al. (2007) find that large firms are more likely to report labour skill shortages as one main business constraint. They also find that exporting activities and type of firm ownership do not exhibit important influences on firm perceptions of labour skill shortages. They note that the availability of skilled labour may become a more binding constraint as economies become more sophisticated and as governments magnify their capacities in enforcing labour regulations. Hallward-Driemeier and Aterido (2009) find that large firms and exporting firms are more likely to report labour skill shortages as a significant obstacle to business operations. They also find that government-owned firms are less likely to perceive labour skill shortages as an important business constraint. In the case of firms located in South Africa, Clarke (2010) finds that large firms and exporting firms have higher propensities to perceive labour skill shortages as a significant business obstacle. He also reports that firm age and type of firm ownership do not exhibit important influences on firm perceptions of labour skill shortages. Kaplan and Pathania (2010) find that large firms are more likely to identify labour skill shortages as one main business constraint. However, Vargas (2012) does not find significant influences of firm size, but he reports important variations across sectors. Using data covering firms located in Canada, Sabourin (2001) shows that firm size and technological intensity are negatively correlated with the probability of firms to identify labour skill shortages as one primary business obstacle. Also, he finds that type of firm ownership and Research and Development activities have no statistically significant influences on these firm perceptions. Baldwin and Lin (2002) find that young firms are more likely to designate labour skill shortages as a considerable problem using data covering Canadian manufacturing firms. However, they find that firm size and ownership type do no exhibit important influences. Green et al. (1998) implement the empirical analysis for firms located in the UK. They find that higher skilled share of employment accentuates the problem of labour skill shortages as perceived by employers. They also report that the share of part-time workers does not exhibit important effects. Lyon et al. (2012) find that medium and large size firms are more likely to designate labour skill shortages as an important constraint on firm operations and growth. They find that exporting firms have lower propensities to report labour skill shortages as a significant business constraint compared to non-exporting firms. Also, they do not find important variations across sectors. The empirical analysis is carried out for the perceived levels of labour market constraints as reported by the respondents (e.g. senior managers, business owners) through the World Bank's Enterprise Surveys database. Pierre and Scarpetta (2004, 2006) examine the relationship between the perceived and actual stringency of labour regulations using national labour protection indices (i.e. de jure labour laws). They find that the reported perceptions are closely related to the actual levels of labour regulations' constraints. Specifically, countries with higher national indices on the stringency of labour regulations are associated with higher proportions of firms perceiving labour regulations as being significant constraints. Kaplan and Pathania (2010) indicate that the perception-based indicators capture the actual institutional environment (e.g. laws, regulations, infrastructure, corruption), particularly when objective measures are inadequate. They argue that the perception-based indicators reflect the de facto severity of business constraints, and serve as a normal check vis-a-vis national indicators. Clarke (2010) notes that the collection of objective measures could be difficult, particularly when it comes to sensitive indicators (e.g. corruption). In such cases, perception-based indicators can serve as a substitute. Hallward-Driemeier and Aterido (2009) find that the perception-based and objective measures of business constraints are closely correlated with each other. They indicate that subjective measures enable the ranking of business constraints within and between countries, particularly when comparing the rigidity of the business environment. Also, they argue that perception-based measures provide information on the most severe obstacles for businesses growth and development as identified by the business entity, and would eventually assist managers and policy makers to recognize reform priorities. Hence, perception-based measures tend to reflect the actual rigidity of business constraints faced by firms. However, there are some concerns regarding perception-based measures. Specifically, these measures could be subjected to: first, potential measurement errors leading to biased results; second, variations in managers willingness to report actual negative or positive responses due to differences in personalities, views, and culture; third, performance bias where managers tend to rank business obstacles according to the performance of their firms or to the business environment conditions through which the firm operates; and finally, defects due to imperfections in question formulations and answer interpretations (Bertrand and Mullainathan, 2001; Senik, 2005; Hallward-Driemeier and Aterido, 2009). Some studies (e.g. Glaeser et al., 2004; Kaplan and Pathania, 2010) argue that caution should be exercised when using perception-based measures as independent variables through empirical analyses. This is because these measures may embody information on growth, which would lead to reverse causality issues. Meanwhile, these studies indicate that using these measures as dependent variables is more intuitive to capture growth in the business environment. These indications complement the aforementioned empirical findings reflecting a close correspondence between perceived and actual measures of business constraints. It is important to note that the empirical analysis in this paper exclusively covers firms that exist in the market (i.e. those that have already entered the market, and that did not exit the market). This is to say that the empirical analysis examines the business obstacles as perceived by the existing firms in the market. Hausmann and Velasco (2005) depict this issue through a camel-hippopotamus analogy. Camels living in the desert do not identify access to water as a primary problem since they have adjusted to the situation by learning how to conserve water. Meanwhile, hippopotami lack the ability to stay in the desert since water scarcity is a critical obstacle for their survival. Hence, by generating and analysing data from interviews with "camels" about water constraints, we may be missing distinct information from "hippopotami". Gelb et al. (2007) argue that such selection is incomplete since many firms choose to enter the market despite the severity of business constraints. They also indicate that the ability to adjust to a business constraint does not mean that firms do not recognize it anymore as a serious problem. They provide an illustration using the World Bank's Enterprises Surveys database: the perceptions of generator-owning firms regarding the electricity as a business constraint are not distinguishable from the perceptions of firms that do not own generators. Furthermore, they note that the intensity of complaints are often highly correlated with corresponding national indices. Dethier et al. (2011) suggest that empirical analyses covering "hippopotami" should be carried out using different frameworks (e.g. entry or exit models[4]). We use a data set sourced from the World Bank's Enterprise Surveys database. This database represents a comprehensive source of firm-level data in emerging and developing economies. It covers firms operating in the manufacturing, service, and other sectors. It contains information on various aspects of the business environment such as, access to finance, corruption, workforce characteristics, innovation and technology, and trade. It should be noted that one of many advantages of using data from these surveys is that the questions are identical through firms across all countries. The basic data set used in this paper covers 5,052 firms located in eight developing Arab countries of the MENA region: Algeria, Egypt, Jordan, Lebanon, Morocco, Oman, Syria, and Yemen[5]. Through these surveys, firms' representatives (e.g. senior managers, business owners) are asked whether labour regulations and labour skill shortages are considered prominent constraints on business operations and development. The responses are used to generate the basic dependent binary variables, which equal one when a firm declares the corresponding labour market constraint as a major/severe business obstacle and zero otherwise. Following the previous literature on business constraints (e.g. Gelb et al., 2007; Kaplan and Pathania, 2010; Clarke, 2010; Meyer and Vandenberg, 2013), the explanatory variables cover firm characteristics' variables which comprise: first, a binary variable that equals one for firms characterized by private foreign ownership and zero otherwise[6]; second, a binary variable that equals one for firms engaged in exporting activities and zero otherwise; third, firm size depicted through the number of total employment and presented in hundreds of workers through the regressions; fourth, firm age measured by the number of years since firm establishment; fifth, firm use of Information and Communication Technology (ICT) depicted through a binary variable that takes the value of one for firms regularly using the internet to communicate with clients and to source information and zero otherwise; finally, two firm labour composition variables: the first is represented through the fraction of skilled production workers in total number of production workers, and the second is depicted through the fraction of non-production workers in total employment. The use of skilled production workers' share and the use of non-production workers' share through the regressions imply that the results are relative to the use of unskilled production workers and to the use of production workers, respectively. The explanatory variables also include country-specific binary variables and sector-specific binary variables to capture cross-country and cross-sector variations. Table II displays summary statistics of the variables used in the empirical analysis. The first panel shows the results for the dependent variables. We find that around 24.6 per cent of firms in the regression data set report that labour regulations represent a major/severe business constraint, while 38.0 per cent of firms report that labour skill shortages exercise a major/severe business constraint[7]. The second panel shows statistics for the explanatory variables. The results indicate that 4.8 per cent of firms are characterized by private foreign ownership and that 22.5 per cent of firms are engaged in exporting activities. The average firm size is 124.0 full-time workers (with a standard deviation of 317.1 full-time workers) and the average firm age is 20.9 years (with a standard deviation of 16.6 years). Also, we find that 50.4 per cent of firms use the internet to communicate with clients and to source information. The percentage of skilled production workers in total number of production workers and the percentage of non-production workers in total employment have averages of 61.6 and 26.1 per cent, respectively. Consider a given firm j (j=1, ..., J) belonging to sector k (k=1, ..., K) and located in country c (c=1, ..., C). Firm perception levels of constraints related to labour regulations and those related to labour skill shortages are depicted through the latent variables (Equation 1) and (Equation 2), respectively. These latent variables are not observed. However, we observe the perceptions of firms through dichotomous responses on whether labour regulations and labour skill shortages do or do not pose major/severe obstacles on firm operations and development. Let R jkc depict a binary variable that takes the value of one when the corresponding firm identifies labour regulations as a major/severe business constraint and zero otherwise. Also, let S jkc represent a binary variable that takes the value of one when the corresponding firm identifies labour skill shortages as a major/severe business constraint and zero otherwise. The benchmark empirical specifications can be represented as: (Equation 3) (Equation 4) where X j is the vector of variables depicting firm characteristics, Y k is the vector of binary variables depicting sectors, Z c is the vector of binary variables depicting countries, and (Equation 5) and (Equation 6) are the stochastic error terms of the corresponding equations. The univariate probit estimator would produce biased estimates when there are some unobserved or omitted characteristics that simultaneously affect firm perceptions of labour regulations and labour skill shortages (Deadman and MacDonald, 2004; Greene, 2008). We allow for the errors in these two labour market constraints' equations to be potentially correlated. Thus, the two equations are jointly modelled using a bivariate probit estimator. The error terms are assumed to be independently and identically distributed as bivariate normal with r depicting the correlation parameter. Specifically, we have: (Equation 7) Through the empirical analysis, we use the Wald test to determine whether the correlation parameter r is statistically significant[8]. The rejection of the null hypothesis indicates that firm perceptions of the labour market constraints are jointly formulated (Greene, 2008 [9]). The bivariate probit specification allows us to estimate unconditional marginal effects, but also conditional and joint marginal effects of variables influencing firm perceptions of the labour market constraints. Table III presents the marginal effects from the benchmark bivariate probit estimation carried out for the pooled data set covering existing firms' perceptions of labour market constraints. The Wald test rejects the null hypothesis of zero correlation between the errors in the two labour market constraints' equations and, hence, it indicates that the model should be estimated through the bivariate probit estimator rather than through the univariate probit estimator. The estimated coefficient of correlation between the errors in the two equations is positive and statistically significant at the 1 per cent level. Table III displays the unconditional marginal effects for Pr(R jkc =1) and Pr(S jkc =1). It also includes the joint marginal effects for Pr(R jkc =1, S jkc =1) and Pr(R jkc =0, S jkc =0), and the conditional marginal effects for Pr(R jkc =1|S jkc =1) and Pr(S jkc =1|R jkc =1[10]). These marginal effects are determined at the mean values of the explanatory variables. For explanatory binary variables, the marginal effects are calculated through discrete changes in probabilities as the binary variable changes from 0 to 1. 6.1. Firm characteristics We find that larger firms have lower propensities to identify labour market constraints as major/severe obstacles facing business operations and development. This is in line with the results reported in some other empirical studies (e.g. Pierre and Scarpetta, 2004; Hallward-Driemeier and Aterido, 2009). The unconditional marginal effects imply that an increase in firm size by one hundred full-time workers reduces the likelihood of firms to perceive labour regulations and labour skill shortages as major/severe constraints on business operations by 0.8 and 0.5 percentage points, respectively. Also, the joint marginal effect of firm size for Pr(R jkc =1, S jkc =1) indicates that an increase in firm size by 100 workers decreases the likelihood of firms to jointly identify labour regulations and labour skill shortages as major/severe business constraints by 0.5 percentage points. The results on the relationship between firm size and firm perceptions of labour regulations are consistent with arguments provided in the literature, which suggest that smaller firms have more limited abilities to realize internal adjustments in response to labour regulations (Meyer and Vandenberg, 2013). Hence, smaller firms have higher tendencies to be more adversely affected by labour regulations. These results are also in line with some other findings showing that labour regulations have disproportional effects on small firms (Aterido et al., 2011). The results on the relationship between firm size and firm perceptions of labour skill shortages imply that smaller firms in MENA countries may encounter more difficulties in realizing future growth, since adequacy of educated workforce is essential for firm competitiveness and performance (Storey, 1994; Jensen and McGuckin, 1997). These results could also suggest that smaller firms have more difficulties in accessing skilled labour markets, identifying required labour skills, and investing in workforce training programmes (Jansen and Lanz, 2013). The implication of exporting activities for firm perceptions of labour regulations is not statistically significant. These results are similar to those found in some previous studies, such as Gelb et al. (2007) in the case of Sub-Saharan African firms. Yet, they deviate from those reported in some studies covering firms located in other developing regions. For example, Meyer and Vandenberg (2013) find that exporting firms located in some South East Asian countries have higher tendencies to perceive labour regulations as a major/severe business constraint compared to non-exporting firms. They attributed these findings to variations in labour demand requirements between exporting and non-exporting firms. Also, Lyon et al. (2012) find that non-exporting firms are more likely to identify labour skill shortages as a major/severe business constraint compared to exporting firms. The effect of firm age is negative and statistically significant in the case of labour skill shortages' equation. The unconditional marginal effect and the conditional marginal effect for Pr(S jkc =1|R jkc =1) indicate that an increase in firm age by ten years reduces the likelihood to perceive labour skill shortages as a major/severe business constraint by 1 percentage point. Although relatively small in magnitude, this impact is consistent with some a priori expectations discussed in the literature, suggesting that older firms have more accumulated experiences to attenuate the effects of business obstacles as perceived by the survey respondents (Meyer and Vandenberg, 2013). We find that an increase in the proportion of non-production workers in total employment by 10 percentage points raises the likelihood of firms to identify labour skill shortages as a major/severe business constraint by 1.3 percentage points. Also, we find that an increase in this proportion by 10 percentage points raises this likelihood by 1.6 percentage points, when assessed conditional on identifying labour regulations as a major/severe business constraint. The effect of the proportion of skilled production workers in total number of production workers on firm perceptions of these labour market constraints is not statistically significant. These findings suggest that the implications of labour skill shortages are primarily associated with the inadequacy of non-production workers rather than with the inadequacy of skilled production workers. The results show that firms with private foreign ownership are less likely to perceive labour skill shortages as a major/severe business constraint by 6.0 percentage points. They suggest that foreign affiliates have higher capabilities in attenuating the implications of labour skill shortages compared to local, domestically owned firms. Almeida (2007) indicates that foreign-owned firms have higher tendencies to pay higher wages and to attract educated workforce compared to domestic firms. These tendencies could eventually mitigate the perceptions of firms with private foreign ownership concerning labour skill shortages compared to firms with domestic ownership. 6.2. Countries' binary variables The results emphasize significant variations in the perceptions of firms located in different countries of the MENA region concerning labour market constraints. Hence, they underscore the implications of national characteristics for these firm perceptions. The binary variable depicting firms located in Algeria is set as the reference for MENA countries through the regressions. The unconditional marginal effects reveal that firms located in Lebanon have statistically higher propensities to perceive labour regulations as a major/severe business constraint by 34.4 percentage points compared to firms located in the Maghreb countries (i.e. Algeria and Morocco) and in Yemen. The corresponding propensities for firms located in Egypt, Oman, and Syria are higher by 17.3, 31.9, and 22.3 percentage points, respectively, compared to firms located in the two Maghreb countries and in Yemen. Also, firms located in Egypt, Lebanon, and Oman have statistically higher propensities to perceive labour skill shortages as a major/severe business constraint by 15.4, 17.5, and 13.5 percentage points, respectively, compared to firms located in the other MENA countries[11]. These results complement other empirical findings in the literature, which document international variations in the perceived magnitude of labour market constraints. For example, Pierre and Scarpetta (2004) show that countries with more stringent de jure labour laws have higher proportions of firms identifying labour regulations as being restrictive. Gelb et al. (2007) find significant variations in firm perceptions of labour market constraints, among other business obstacles, across Sub-Saharan African countries. Also, Lyon et al. (2012) present some important regional differences in firm perceptions of labour skill shortages. Drzeniek-Hanouz and Dusek (2013) indicate that labour market constraints constitute significant obstacles on business development in the Arab world. They also describe considerable variations in the perceptions of firms located in different MENA sub-regions: North Africa (including Egypt), the Levant, and the Gulf sub-regions. 6.3. Sectors' binary variables We find that firm perceptions of labour market constraints vary across sectors. The binary variable for firms belonging to the service sector is set as the reference for sectors through the regressions. The unconditional marginal effects indicate that firms belonging to the manufacturing sector and to the category covering the remaining other (non-manufacturing and non-service) sectors have higher propensities to perceive labour regulations as a major/severe business constraint by 9.9 and 10.1 percentage points, respectively, compared to firms belonging to the service sector. The estimates for the labour skill shortages' equation are 12.6 and 11.8 percentage points, respectively. These results suggest that labour market constraints, as perceived by the survey respondents, disproportionally impact manufacturing firms and firms belonging to the other sectors compared to firms belonging to the service sector. Firm perceptions of labour market constraints can be alternatively represented through ordinal responses in the data set. Specifically, firms identify the stringency of labour regulations and labour skill shortages as: no or minor obstacle (outcome 1), moderate obstacle (outcome 2), and major or very severe obstacle (outcome 3). Table IV displays the marginal effects obtained through the ordered probit model. The results are comparable to the corresponding marginal effects for Pr(R jkc =1) and Pr(S jkc =1) that are reported in Table III. 7.1. Empirical results by sector The results from the empirical analysis implemented for data sets covering individual sectors are displayed in Table V. The estimated coefficients of correlation between the errors in the two labour market constraints' equations are positive and statistically significant at the 1 per cent level for all sectors. To save space, Table V does not display the rows of marginal effects of firm characteristics that do not exhibit statistical significance at all. In general, the results for firms belonging to the manufacturing sectors are qualitatively similar to the benchmark results for pooled data set. We also find few differences. For example, the perceptions of manufacturing firms located in Jordan concerning labour skill shortages appear to be less pronounced compared to manufacturing firms located in the other MENA countries. Firms located in Jordan have statistically lower propensities to perceive labour skill shortages as a major/severe business constraint compared to firms located in Algeria and those located in Egypt by 11.3 and (11.3+14.0=) 25.3 percentage points, respectively. The results derived from the service sector's data set reveal some notable differences compared to those obtained from the manufacturing sector's data set[12]. In the case of the service sector, firm perceptions of labour regulations do not exhibit statistically significant heterogeneity across firm characteristics. We find that firm perceptions of labour skill shortages are accentuated with an increase in the ratio of non-production workers to total employment. The marginal effect shows that an increase in this ratio by 10 percentage points raises the likelihood of perceiving labour skill shortages as a major/severe business constraint by 3.1 percentage points. The magnitude of this effect is considerably higher than the one reported for manufacturing firms. Our data set does not contain service firms located in Algeria that report labour regulations and labour skill shortages as major/severe constraints on business operations. In other words, service firms located in Algeria predict failure perfectly and, therefore, they are dropped from the service sector's data set. Then, the binary variable depicting firms located in Egypt is set as the reference for MENA countries. We find that service firms located in Egypt and Lebanon have statistically higher likelihoods than those located in Jordan to perceive labour regulations as a major/severe business constraint by 6.7 and (6.7+2.1=) 8.8 percentage points, respectively. The corresponding likelihood for service firms located in Oman is considerably higher than those located in Jordan by (18.4+6.7=) 25.1 percentage points. We also find that service firms located in Egypt and Lebanon have lower propensities than those located in Jordan to identify labour skill shortages as a major/severe business constraint by 16.1 and (16.1-1.3=) 14.8 percentage points, respectively. The joint marginal effects for Pr(R jkc =1, S jkc =1) indicate that service firms located in Egypt and Jordan have statistically comparable likelihoods to identify both labour market constraints as major/severe obstacles on business operations as perceived by the survey respondents. These likelihoods are lower than the corresponding likelihood for service firms located in Oman by more than 11 percentage points. Finally, the results for the other sectors' category reveal that firms located in Oman are statistically more likely to perceive labour regulations as a major/severe business constraint by 34.9 percentage points compared to firms located in the other MENA countries (except Lebanon). Also, we find that firms located in Egypt and Lebanon have higher propensities to perceive labour skill shortages as a major/severe business constraint by 44.2 and 31.3 percentage points, respectively, compared to firms located in the other MENA countries. 7.2. Empirical results by country The results from the empirical analysis carried out for data sets covering firms located in individual MENA countries are displayed in Table VI. The estimated coefficients of correlation between the errors in the two labour market constraints' equations are positive and statistically significant at the 1 per cent level for all countries. To save space, Table VI does not display the rows of marginal effects of firm characteristics that do not exhibit statistical significance at all. In the case of Algeria, service firms predict failure perfectly and, hence, they are dropped from the data set. Then, the other sectors' category is set as the reference for sectors. We find that the perceptions of firms located in Algeria concerning labour regulations do not exhibit statistically significant variations between the manufacturing sector and the other sectors' category. Exporting firms are less likely to perceive labour regulations as a major/severe business constraint by 13.5 percentage points compared to non-exporting firms. Also, an increase in the skilled production workers' ratio by 10 percentage points reduces the extent of these perceptions by 0.7 percentage points. We find that manufacturing firms have a statistically higher likelihood to identify labour skill shortages as a major/severe business constraint by 12.1 percentage points compared to other sectors' firms. Manufacturing firms located in Egypt have statistically higher propensities to perceive labour regulations and labour skill shortages as major/severe business constraints by 15.7 and 20.6 percentage points, respectively, compared to service firms. The implications of firm size for firm perceptions of labour regulations are similar to the benchmark results. Also, the effects of firm age, private foreign ownership, and the proportion of non-production workers in total employment are reminiscent of those obtained from the pooled regressions. Service firms located in Jordan have statistically higher propensities to identify labour skill shortages as a major/severe business constraint compared to manufacturing and other sector's firms by 19.2 and 13.6 percentage points, respectively. The non-production workers' ratio exhibits positive implications for firm perceptions of labour skill shortages. An increase in this ratio by 10 percentage points raises the magnitude of these perceptions by 3.2 percentage points. In contrast, this ratio has negative impacts on firm perceptions of labour regulations. An increase in this ratio by 10 percentage points induces a decrease in the magnitude of these perceptions by 2.0 percentage points. In the case of Lebanon, manufacturing and service firms have lower propensities to perceive labour skill shortages as a major/severe business constraint by 34.6 percentage points compared to firms belonging to the other sectors. Firm characteristics do not exhibit statistically significant implications, with the exception of firm use of ICT in the case of labour regulations. The perceptions of firms located in Oman concerning labour market constraints do not exhibit statistically significant variations across sectors. The conditional marginal effects for Pr(R jkc =1|S jkc =1) indicate that exporting activities and an increase in firm size by one hundred workers reduce the corresponding probabilities by 34.6 and 38.0 percentage points, respectively. In our data set, firms located in Morocco, Syria, and Yemen belong exclusively to the manufacturing sector. We find that exporting firms located in Morocco are more likely to perceive labour regulations and labour skill shortages as major/severe business constraints by 15.3 and 22.3 percentage points, respectively, compared to non-exporting firms. Also, an increase in the skilled production workers' ratio by 10 percentage points raises the likelihood of identifying labour skill shortages as a major/severe business constraint by 2.9 percentage points. Finally, exporting firms located in Yemen are less likely to perceive labour regulations as a major/severe business constraint by 11.2 percentage points, and are more likely to perceive labour skill shortages as a major/severe business constraint by 34.8 percentage points, compared to non-exporting firms. Labour market constraints are often identified as main business obstacles facing firm operation and development in the MENA region. Therefore, they are naturally listed through the primary items on the labour policy agenda of the MENA countries. The comprehension of the factors influencing the perceived severity of labour market constraints is essential in the design of policies aiming at improving labour market conditions and enhancing business environments. This paper examines the implications of firm characteristics, national locations, and sectoral associations for the perceptions of firms located in the MENA region concerning two primary labour market constraints: labour regulations and labour skill shortages. The empirical analysis is carried out using firm-level data set sourced from the World Bank's Enterprise Surveys database. A bivariate probit estimator is used to account for potential correlations between the errors in the labour regulations' equation and labour skill shortages' equation. The empirical results are generated through overall estimations and by implementing comparative cross-country and cross-sector analyses. The benchmark empirical results obtained from the overall estimations reveal important implications of firm characteristics for firm perceptions of labour market constraints. Larger firms appear to have lower propensities to identify labour market constraints as major/severe business obstacles. These results imply that small firms have limited flexibility in adjusting to labour regulations and to labour skill shortages. Hence, they suggest that small firms may encounter difficulties in realizing future growth through the MENA region. Also, we find that firms with private foreign ownership are less likely to endure the implications of labour skill shortages compared to domestic firms. This is consistent with higher tendencies of foreign affiliates to pay higher wages and to attract educated workforce compared to domestic firms. Meanwhile, exporting activities do not have statistically significant effects on firm perceptions of labour market constraints. This finding could stem from the generally limited engagement of MENA exporting firms in international trade. We find that firms with higher proportions of non-production workers in total employment have higher propensities to perceive labour skill shortages as a major/severe business constraint. However, the proportion of skilled production workers in total number of production workers does not exercise statistically significant influences on firm perceptions. These findings suggest that the implications of labour skill shortages are primarily associated with inadequacy of non-production workers rather than inadequacy of skilled production workers. The results also emphasize considerable cross-country and cross-sector variations in firm perceptions of labour market constraints. Finally, the empirical analysis reveals significant heterogeneity through the implications of firm characteristics and national locations for the perceptions of firms belonging to different sectors concerning labour market constraints. Also, it underlines important heterogeneity through the implications of firm characteristics and sectoral associations for the perceptions of firms located in different countries concerning labour market constraints. Reforms in labour regulations and investment in human capital are important governmental policy interventions for promoting firm development and economic growth. This paper provides policy makers with information needed to design labour policies that attenuate the impact of labour market constraints and enhance the performance of firms in the MENA region. Specifically, the design of labour policies should encompass the findings that the perceptions of labour market constraints significantly vary through firm characteristics, countries, and sectors. Furthermore, given the heterogeneity in the implications of the factors influencing firm perceptions of labour market constraints, labour policies should be customized by country, and should recognize the varying consequences for different sectors. Opens in a new window. Table I Percentage of firms identifying business obstacles as major/severe constraints Opens in a new window. Table II Descriptive statistics Opens in a new window. Table III Benchmark empirical results from the bivariate probit model (marginal effects) Opens in a new window. Table IV Empirical results from the ordered probit model (marginal effects) Opens in a new window. Table V Empirical results by sector (marginal effects from the bivariate probit model) Opens in a new window. Table VI Empirical results by country (marginal effects from the bivariate probit model) Opens in a new window. Table AI Descriptive statistics for the dependent variables by country and by sector Opens in a new window. Table AII Other results from the bivariate probit model Opens in a new window. Equation 1 Opens in a new window. Equation 2 Opens in a new window. Equation 3 Opens in a new window. Equation 4 Opens in a new window. Equation 5 Opens in a new window. Equation 6 Opens in a new window. Equation 7 Opens in a new window. Equation 8 Opens in a new window. Equation 9 Opens in a new window. Equation 10 Opens in a new window. Equation 11 Opens in a new window. Equation 12 Opens in a new window. Equation 13 Opens in a new window. Equation 14 Opens in a new window. Equation 15 Opens in a new window. Equation 16 Opens in a new window. Equation 17
|
Reforms in labour regulations and investment in human capital are important governmental policy interventions for promoting firm development and economic growth in the MENA region. This paper contributes to the empirical literature by analysing the factors influencing the perceptions of firms located in the MENA region concerning labour regulations and labour skill shortages. It provides policy-makers with information needed in the design of labour policies that attenuate the impacts of labour market constraints and enhance the performance of firms and the long-run economic growth.
|
[SECTION: Purpose] A brand can be seen as a strategic asset that helps a company to be more competitive. In the same way that companies invest in brands, countries can also be seen as such (Anholt, 2005; Huang and Tsai, 2013; Kotler et al., 2006). In considering the role that the image of a country can play in a buyer's behavior, constructs such as the country's brand, the country's image and the country of origin may be attributes that offer the potential for companies to achieve a sustainable competitive advantage (SCA), both in the internal and external market (Baker and Ballington, 2002; Hakala et al., 2013). In this context, based on a country brand, the so-called shared brands may be worked. They have been defined by Tregear and Gorton (2009) as having a single brand as an identity for two or more partner companies, which commit to respect the established agreements, and the management of this shared brand must be performed by an independent entity that represents and responds for the brand. In this way, this article deals with different types of shared brands that fit into this conceptual condition presented by Tregear and Gorton (2009), being geographical indications (GIs), collective brand and sector brand. GIs constitute an instrument for the valuation of goods that are territorially distinguished, that is, they are products coming from a certain region that are recognized for their quality. Collective brands are awarded to a group of organizations that market the same products and wish to link quality to associated companies that do not have to be produced in a delimited geographical region and their registration does not necessarily have to have a geographical name (Castro and Giraldi, 2015). The sector brands are projects developed with the purpose of international representation of the sector through export mechanisms and promotions (Castro and Giraldi, 2015; Lourencao and Giraldi, 2017). Defrancesco et al. (2012) demonstrated the importance of GIs, confirming that foreign consumers are willing to pay a premium for geographical names. They concluded that widely known quality signals, such as protected GIs, would increase the market access of foreign wines by enhancing consumers to recognize and discriminate among terroir-related quality attributes. However, it has to be considered that there are wine enthusiasts and novice consumers, each segment with different needs and degrees of knowledge about wines and the related certifications. It is important that wineries and restaurants be able to understand these differences and devise specific marketing strategies for each audience (Barber et al., 2008). Managers of these establishments need to understand the importance of training to assist the consumer in the purchasing decision process (Dodd et al., 2005). There are also environmental aspects to be considered when analyzing the consumer intention to pay more for certain types of wines. For example, Barber (2010) mentions the expansion of the search for organically or biodynamically produced wines and the consequent reorientation of the blend through repackaging and re-labeling, promoting this in the product dissemination strategy. Barber (2010) suggests that wineries have to focus on developing a work of increasing awareness that organic wine has no lower quality than other traditionally wines (Barber et al., 2010). In this regard, the SCA construct can be defined as that unique attribute that is difficult to be acquired or copied by the competition, which arises from the impossibility of competitors implementing value creation strategies simultaneously (Barney, 1991). Within the SCA theme is the resource-based view (RBV) developing valuable resources and capabilities, which are interchangeable and refer to the tangible and intangible assets that the company uses to develop its strategies (Ray et al., 2004). The study of Vlachveia et al. (2012) in Greece shows that producers invest intangible elements with affect consumer choices. They used a wine tourism and wine routes with an important tool for marketing. They used holistic actions with an important tool for marketing rather than a brand just to a name or a label on the bottle. For this, the wineries used a wine tourism and wine routes. Galati et al. (2014) identified the firms' internal factors that most influence the export intensity and the ability to fit into the international market of the Italian wineries producing sparkling wines through a RBV approach. In particular, the main factors affecting export intensity are innovation degree and intensity of investment in advertising and promotion, among firms' characteristics, and abroad experience, proficiency of foreign languages, educational level and perception barriers, among managers' characteristics, are strategic to ensure the survival of firms into markets and their success. In turn, Amadieu and Viviani (2010) appoint that, for French wine companies, spending money on intangibles is not a guarantee of financial success. The lack of organizational innovation, organizational responsiveness or knowledge-related resources could explain the absence of a positive impact for intangible expenses on performance for French companies in the wine sector. To be financially successful, intangible expenses should be accompanied by organizational and managerial changes, and these changes require time to be learned. Giacomarra et al. (2016) identified that internal motivations drive entrepreneurs to adopt voluntary standards and that wineries adopting voluntary standards exhibit better economic performance than non-certified wine firms. These results suggest that although the adoption of voluntary quality certifications does not have a direct effect in terms of a labor productivity increase, it entails a business management improvement. The work of Alonso and Bressan (2016) shows that it is possible to associate RBV with the strenghts, weakness, opportunities and threats analysis, which is a complementary tool. The accumulated knowledge of the team provides the rare attribute and history, reputation and territory related to the imperfectly imitable/replaceable attribute. Maurel (2009) complements this idea regarding the impact of decision-maker characteristics on export performance by indicating a lack of significant impacts on export performance. Suarez-Ortega and A'lamo-Vera (2005) confirmed that factors influencing export involvement are not the same along the process of export development. Three aspects of export development have been analyzed: intention, propensity and intensity. The effect of industry-specific characteristics on internal export factors has been isolated through the selection of one industry in one country for the empirical research. The study of Wilk and Fensterseifer (2003) was conducted having as its central purpose the identification of the strategic resources and capabilities of a cluster located in southern Brazil and their sustainability that history matters and the cumulative development and upgrading of resources and capabilities can fundamentally shape the future strategic options of a cluster. Although these papers have approached the RBV, when dealing with internal resources, they have not explored and classified the foundations for obtaining a SCA, such as the valuable, rare, imitability/replaceability and organization (VRIO) framework. The seminal work of the RBV was that of Barney (1991), which culminated in the VRIO framework (Barney, 1991, 1995). Two decades later, Barney et al. (2011) show that the theory is mature and some researchers prefer to use the term resource-based theory (RBT) instead of RBV, as they believe that it is currently a consolidated theory and not just a vision (Barney et al., 2011; Newbert, 2007). In analyzing the shared brands and the perspective of SCA, the objective of this article is defined as to investigate if shared brands provide SCA according to the VRIO model adapted to the analyzed context (Brazilian winemaking sector), in the opinion of the government agencies, associations and managers of the wineries. 2.1 Sustainable competitive advantage The SCA arises from the development of a set of features and abilities that cannot be negotiated, imitated or replaced, or implemented simultaneously by current and potential competitors, and use of resources within the company is difficult to access or replicate by competitors (Bharadwaj et al., 1993; Dierickx and Cool, 1989; Hoffman, 2000; Iamin, 2007; Porter, 1996; Teece, 2000; Von Krogh and Roos, 1995; Wernerfelt, 1984). The work of Barney (1986) introduced the position of heterogeneous resources, which is an important theoretical contribution to RBV. The immobility of resources prevents other firms from acquiring or imitating resources easily, making it difficult to make resources unique in the market (Bharadwaj et al., 1993; Hoffman, 2000; Penrose, 1979; Von Krogh and Roos, 1995). Based on the principles of heterogeneity and resource immobility, which is the basis for RBV, Barney (1991) created the VRIS framework (valuable, rare, imperfectly imitable, and imperfectly substitutable) or, for some authors, VRIN (valuable, rare, imperfectly imitable and non-replaceable). The acronyms represent the same idea and differ in the way of explicit terminology with S in VRIS and with N in VRIN, representing, in both, the imperfectly replaceable attribute (Shafeey and Trott, 2014). The VRIS/VRIN model is revised by Barney (1995) and becomes the VRIO framework, in which the imperfectly imitable and imperfectly replaceable attribute is added with the letter "I" in the framework, summing them into the issue of imitation difficulty and adding the item organization, which discusses whether the company is organized to exploit its resources and capabilities (Barney, 1995; Barney and Wright, 1998; Shafeey and Trott, 2014). The organizational process such as the action of managers in the form of structuring the resources and managing them in the market provides the fourth necessary condition to obtain SCA, thus forming the VRIO (Balmer and Gray, 2003; Barney, 1991, 1995; Barney and Hesterly, 2006; Cazurra and Un, 2015; Grant, 1991; Ito and Gimenez, 2011; Shafeey and Trott, 2014; Sirmon et al., 2011; Srivastava et al., 2001). Nowadays, RBV studies are more concise and in the process of maturation, which leads some authors to substitute the term "RBV perspective" for RBT, according to Barney et al. (2011), Crook et al. (2008), Maritan and Peteraf (2011), McWilliams and Siegel (2011) and Molloy et al. (2011). Based on these trends and their relationship with the RBV, RBT with the VRIO framework applicable to the country brand and its related concepts, such as shared brands, is considered by analyzing the attributes (valuable, rare, imperfectly imitable and imperfectly replaceable) based on the internal resources of each country (Zhang and Su, 2009). From the VRIO model of Barney (1991, 1995) and analyzing the applicability of this model to the specific case of shared brands, it is important to make here a specific adaptation in this original model, calling it VRIA. Thus, the "O" of organization is replaced by the "A" of association as for the legal existence of shared brands and according to the definition proposed by Tregear and Gorton (2009). 2.2 Shared brands and sustainable competitive advantage In Brazil, the understanding of shared brands considers GIs, collective brands and sector brands, which can occur in isolation or at the same time. GIs are divided into two types - indications of origin (IO) and denominations of origin (DO). For IO, the reputation of the product is analyzed. In DO, the peculiarities of that region must affect the final result of the product, in an identifiable and measurable way. A food or agricultural product recognized as a denomination of origin must be produced, processed, manufactured or industrialized in its territory of origin, which is not necessarily a rule when it comes to indication of origin [Barbosa et al., 2013, 2014; Instituto Nacional da Propriedade Industrial (INPI), 2014; Junger, 2014; Servico Brasileiro de Apoio as Micro e Pequenas Empresas (SEBRAE) and Instituto Nacional da Propriedade Industrial (INPI) 2011, 2014a, 2014b, 2014c; Seminario Internacional de Indicacoes Geograficas e Marcas Coletivas, 2014 (verbal information)]. For wines, there are six Brazilian indications of origin: Vale dos Vinhedos, Pinto Bandeira, Vale da Uva Goethe, Altos Montes, Monte Belo and Farroupilha [Servico Brasileiro de Apoio as Micro e Pequenas Empresas (SEBRAE) and Instituto Nacional da Propriedade Industrial (INPI) 2011, 2014b; Instituto Nacional da Propriedade Industrial (INPI) 2014, 2015]. In addition, two GIs are under development, namely, Campanha and Vale do Submedio do Sao Francisco (Empresa Brasileira de Pesquisa Agropecuaria - Uva e Vinho (EMBRAPA), 2014). Collective brands, according to industrial property law, are used to identify products or services coming from members of a certain entity [Barbosa et al., 2013; Servico Brasileiro de Apoio as Micro e Pequenas Empresas (SEBRAE) and Instituto Nacional da Propriedade Industrial (INPI), 2014c]. As for the sector brand, there is a diversity of projects and work fronts developed by APEX, promoting the Brazilian industry. One type of share brand that can be classified as a GI is the controlled denomination of origin (DOC). Marchinia et al. (2014) show that the higher number of firms and the investments in marketing contributes to the collective reputation of the DOC. Similarly, the older the DOC designation, the greater its reputation because of the accumulation effect of investments in marketing and consumer experiences. Product differentiation within the DOC areas (subzones, variants of the product, the presence of sweet wine, etc.) represents a growth factor for the reputation of the DOC. Despite the great number of criticisms to the system of DOC wines, the presence of shared and mandatory rules resulted in a collective value that can support SMEs in building their own individual reputation. Olmos (2011), in a study of the appellation of origin (DOC) in the Rioja wine sector, showed that process innovations are not a determinant of internationalization. However, the results show a positive and significant link between a firm's marketing activities and its level of internationalization. The positive effect of advertising in determining the degree of internationalization confirms that wineries in this region can benefit from the collective umbrella brand "qualified denomination of origin" in comparison with wineries in other wine-producing regions. Defrancesco et al. (2012) shows that the protected GIs can constitute a strategic tool for wine producers and it's possible to a strategic differentiation of New World wine producing countries. Fensterseifer and Rastoin (2013) analyzed the contribution potential of the resources of a cluster to generate competitive advantage for local companies in the wine sector. The benefits obtained when companies participated in the cluster were classified from an RBV perspective in the following categories: institutional, specialization, social, reputation and natural capital. Thus, a set of potentially strategic resources for the Serra Gaucha wine cluster was identified. This study was based on a qualitative approach that, for Bauer and Gaskell (2000), aims to understand in detail beliefs, attitudes, values and motivations regarding the behavior of people in specific social contexts (Malhotra, 2006). The research was exploratory and the field study was conducted with in-depth interviews (Bardin, 2011; Cooper and Schindler, 2014; Sampieri et al., 2010). This work was delimited to the study of the viniculture sector. This sector was chosen, as it was possible to analyze the shared brands (GI, collective brand and sector brand), according to studies developed in Brazil (Barbosa et al., 2013, Bruch et al., 2014, Castro and Giraldi, 2015). In the international wine market, considering the past five years, the sparkling wine continued to grow with an annual rate of 7 per cent in value and 6 per cent in volume and bulk wine with an annual rate of 6 per cent in value and 3 per cent in volume. Over the same period, bottled wine grew 5 per cent in value but only 1 per cent in volume in terms of annual rate. Forecast of the future evolution of wine market is challenging and risky in a context dominated by discontinuities, and there is not much research on how markets evolve (Pomarici, 2016). In addition, the Brazilian wine sector is important to the national economy, being the first to obtain a GI in the country, a concept similar to DOC - controlled designation of origin used in Italy, for example (Barbosa et al., 2013). The Brazilian wine sector has already six certified regions, namely, Vale dos Vinhedos, Pinto Bandeira, Monte Belo do Sul, Flores da Cunha, Farroupilha and Urussanga and some collective brands such as ACAVITIS, CPEG and Vinhateiros do Vale and also sector brands (Vinhos do Brasil/Wines of Brasil). The work was delimited to the study of the fine wines that have GIs and are located in the Brazilian state of Rio Grande do Sul. In terms of numbers, the Brazilian viniculture sector is the 19th ranked country in terms of area cultivated with grapes, 12th in grape production and 13th in wine production. From 2004 to 2014, there was a 100 per cent increase in the national production of fine wines, from 25 million to 50 million L (Copello, 2015). The following standard terminology was used for codification: "G" for government agencies, "A" for associations and "W" to represent winery managers followed by numbers in the order in which the interviews were conducted. Interviews were conducted with government agencies such as SEBRAE, MAPA (Ministry of Agriculture, Livestock and Supply), INPI and APEX (Brazilian agency for promotion of exports and investments) represented by IBRAVIN (G1, G2 e G3) and EMBRAPA UVA E VINHO (Embrapa Grape and Wine) (G4) to understand in detail the processes of creation and registration of shared brands. The interviews were conducted in May 2015 and each had an average duration of 40 min. Subsequently, primary data were also collected through interviews with the associations that manage GIs and the collective brands of fine wines, such as APROVALE (Association of Fine Wine Producers of the Vinhedos Valley - A1 e A2); ASPROVINHO (Association of Wine Producers of Pinto Bandeira - A3); APROMONTES (Association of Wine Producers of Altos Montes - A4); APROBELO (Association of Viticulture Farmers of Monte Belo do Sul - A5); and AFAVIN (Association of Wine, Sparkling Wine, Juices and Derivative Producers of Farroupilha - A6); CPEG (A7); AVIGA (Association of Wine Producers of Garibaldi - A8); SINDIVINHORS (Syndicate of the Industry of wine, grape must, vinegar and beverages derived from the grape, and wine of the state of Rio Grande do Sul - A9); and the City Hall of Garibaldi (A10) together with the Secretary of Tourism. Ten interviews were conducted with an average duration of 40 min each. For data collection, six wine-growing municipalities of the State of Rio Grande do Sul were visited: Bento Goncalves, Monte Belo do Sul, Flores da Cunha, Farroupilha, Pinto Bandeira and Garibaldi. Finally, the managers of the wineries were interviewed, with 18 interviews, namely, APROVALE - Don Laurindo (W1), Lidio Carraro (W2, W3), Miolo (W4), Peculiare (W5) and Pizzato (W6); ASPROVINHO - Cooperativa Aurora (W7, W8) and Valmarino (W9); APROBELO - Fae (W10) and Fantin (W11); APROMONTES - Mioranza (W12); CPEG/AVIGA - Batistello (W13), Cooperativa Garibaldi (W14, W15), Milantino (W16), Pedrucci (W17) and Vaccaro (W18). The interviews had an average duration of approximately 40 min each, with about 12 h of testimonials. With this number of interviews, a saturation point was reached, in which the researcher, according to Bauer and Gaskell (2000), evaluated that no new surprises, perceptions or revelations will appear in additional interviews. The wineries interviewed belong to five producing regions in the Rio Grande do Sul (Brazil). The process began with the planting of grapes with the arrival of Italian immigrants in 1847 and only in the 1990s they underwent a process of remodeling, culminating in the current Brazilian winemaking format (W6). Besides the cultivation, producers have started to vinify and to commercialize the products in the internal and external market. The interviewees are large, medium and small, possessing 900, 30 and 8.5 hectares of vineyards of their own, on average. They produce an average of 20 million L/year for the largest, 197,600 L/year for the average wineries and 27,500 L/year for the small wineries interviewed. The following proposition was proposed for this study: shared brands provide SCA, according to the VRIA, in the opinion of the government agencies, associations and managers of the wineries (Barney, 1986, 1991, 1995; Dierickx and Cool, 1989). This proposition is based on the work of Barney (1986, 1991, 1995) and Dierickx and Cool (1989) who postulate that SCA arises when competitors are unable to obtain the same results, even though they use similar marketing strategies (Bharadwaj et al., 1993; Von Krogh and Roos, 1995). For data analysis, content analysis was used, which is a set of communication analysis techniques, in which the analyst delimits the codification or registration units (Bardin, 2011). Categorization consists in organizing the data so that the researcher can make decisions and draw conclusions from them (Bardin, 2011; Cooper and Schindler, 2014). The proposal of this article is composed of four categories denominated C1, C2, C3 and C4, representing the VRIA framework. 4.1 C1 - valuable The valuable category (C1) refers to the benefits, advantages and factors that lead wineries to differentiate themselves from their competitors. The shared brand, being valuable, adds value, becomes distinct and differentiates itself in the minds of consumers (Balmer and Gray, 2003; Barney, 2001). In this way, the interviewees who have GI understand that it adds value and brings visibility to both large- and small-sized wineries. It ends up generating a greater interest of the consumer in knowing that region, fomenting tourism and adding value for the whole chain [O4, Servico Brasileiro de Apoio as Micro e Pequenas Empresas (SEBRAE) and Instituto Nacional da Propriedade Industrial (INPI), 2014c]. The producers understand that with the certificate, they have gained publicity in the media, and Brazilian wine has become more recognized, as can be seen in the following excerpt: It adds a lot of value, brings greater competitiveness in the internal and external market. The issue of the reputation of Vale dos Vinhedos, this intrinsic value of quality, of the region, the culture from here, its own identity, distinction from similar wines. This perception of the brand leads to "this product is more expensive, it has DO, so I think it is a good product" (W4). The IO of Vale dos Vinhedos was the first Brazilian IO and emerged soon after the opening of Brazil to the international market. At that moment, there were changes in the competitive scenario and Brazilian fine wine began to gain space in the international market, presenting itself as a valuable product. The emergence of Vale dos Vinhedos DO - the first DO of Brazilian wines - occurred in 2012 and did not have as much impact as the launch of IO in 2002. This is because of the fact that IO was a novelty and also because DO is more restrictive, covering fewer wineries and fewer certified products (W6): Actually, this is it, we are a smaller winery here, I do not have, let's say, market visibility as a bigger company, I do not have the marketing power to invest in it. When I have a designation of origin, so to speak, the consumer sees my product with different eyes, he does not know the individual brand, but he knows what a designation of origin is, so he can position the product in an easier way. This is a strength to position yourself in the market much greater than if you were individual (W5). They believe that DO is already more traditional overseas such as in Europe, for example, and this facilitates the understanding that it is a wine with production and quality mainly controlled for the consumer who is already accustomed to imported wines of a high standard. They recognize how having DO is a valuable resource for the producing company and for Brazil as the country of origin (W4, W5, W12, W16). The smaller wineries, whose production is practically handcrafted and family operated, denominated as boutique wineries or micro champagne enterprises do not produce enough volume to distribute their products in supermarket chains (W17). It is noteworthy that Vale dos Vinhedos GI was recognized by the European Union in 2007 and represented a step toward visibility in the external market. Before, the wines that went to Europe went as ordinary wines: From the recognition by the European Union, the wines of Vale dos Vinhedos had opened the doors to place their labels. The GI seal gives both region and the products a good reputation (A1, A2, A3, A6). In this sense, the interviewee A10 affirms that CPEG guarantees the quality of the product and this generates recognition by the consumer himself and respectability for the producing region and adds value to the product. This idea is complemented by A9 who believes that this recognition will come through the dissemination and promotion of products. With regard to the sector brand, one interviewee reports: Everyone who works with wines, e.g. journalists who are directly linked to wine, already knows that Brazil produces wine. Now, another step is to make the final consumer know what Brazil produces. The end consumer still does not know that Brazil produces wines, and good wines (G1). Another factor that adds value is the issue of the distribution of products in specialized stores abroad, which counts on a trained employee to perform a personalized service, generating competitive differential (G1, W6). Respondents believe that being part of the Wines of Brazil project and exporting the products adds value, meaning that the product becomes more valuable in the face of international competition (G3, G4, W3). This result relates to Defrancesco et al.'s (2012) study, which demonstrated the importance of GIs as a quality signal that makes consumers willing to pay a premium for geographical names. It is important to mention that this added value may be better perceived by wine enthusiasts than by novices. As Barber et al. (2008) and Dodd et al. (2005) indicate that because of the different needs and degrees of knowledge about wines and related certifications, it may be important for Brazilian wineries to better understand their consumers and provide proper information about the certifications to increase their wines perceived value. In this way, the initiatives of shared brands, whether GI, collective brand or sector brand, add value to the products and differentiate the wineries nationally and internationally. This result found among the interviewees is in agreement with the literature that, when developing valuable resources, the company gains a competitive advantage, as it adds value to the products and manages to keep up with the environmental changes of the competitive market (Barney, 1995; Ray et al., 2004). To illustrate the results schematically, the conceptual map is presented in Figure 1. 4.2 C2 - rare The Rare category (C2), according to Barney (1991, 1995), refers to a differential for the company, in addition to having valuable resources. When it comes to GIs, this rare factor is found in Brazilian legislation and its regulations, which, being detailed and with more requirements, resembles the European model. This differentiates Brazilian wine and brings rarity when compared to the international competitors of the so-called New World (Argentina, Chile, the USA, Australia, South Africa and New Zealand). In general, for the New World, only one delimitation with a geographical name is made and it must own about 65 to 80 per cent of the grape produced in this territory, as well as its winemaking. While for DO in Brazil, 100 per cent of the grape must be produced and vinification performed in the defined region (G4). In this way, the European consumer understands what a GI is, recognizes that the product has differentiation criteria and this, besides adding value, brings rarity to the products. The GI is like a "calling card", informing the consumer of its characteristics: region, origin, history and culture of the product (G4, A6). This result relates to Alonso and Bressan's (2016) study, which suggested that the accumulated knowledge of the team provides the rare attribute and history, and the reputation and territory relate to the imperfectly imitable/replaceable attribute. It also finds support in Wilk and Fensterseifer's (2003), who demonstrated that the wine cluster located in southern Brazil has achieved a high level of resources and capabilities that can give a competitive advantage to this cluster. Respondents say that the consumer generally does not remember the individual brand of wine he has taken, but rather the country of origin of that product: You do not remember a brand of wine you took, the last foreign wine you took will remind you that you took a Chilean wine, you will remember you took a French wine, the brand that the consumer remembers is often the origin, it is not the brand of the product itself (G3). For the global competition, Vale dos Vinhedos is a block that works together for the dissemination of Brazilian wine, showing product differentials. They report that the consumer already has a perception that Vale dos Vinhedos represents certified quality products. However, within the Vale itself there are 22 wineries that also compete with each other, as the tourist chooses among the existing options, and then there is the work to obtain a competitive differential for the individual brand, as well as the collective actions for the competitiveness of the region internationally (W4, W6). The region of Pinto Bandeira, for example, has as main focus the traditional sparkling or methode champenoise (champenoise method). This product is rare because it differs from other producing regions and has its own flavor because of altitude and soil conditions for the production of grapes. And through the traditional method, there is a more careful and time-consuming process in the elaboration that also adds value and provides rarity to the sparkling wine (A3). Similarly, the reality lived in Garibaldi also shows that the rarity of the product lies in the fact that the sparkling wine produced is differentiated and Brazil is already being recognized, especially the region of Serra Gaucha, as a region favorable for the production of sparkling wines (A7, A8, A9). The proposal of the sparkling wine, for example, is more refreshing and seeks a different style from the French, which provides a more structured and fuller-bodied wine (W3, W4). In this way, shared brands, whether GI, collective brand or sector brand, have aspects of the rarity in relation to current and potential competition. This makes it difficult for other companies that wish to implement equivalent strategies (Barney, 1991, 1995). Therefore, for the agencies, associations and managers of the wineries, the shared brand resource is rare because in addition to being valuable, it has conditions that are not widely disseminated among the competitors in the market. The results presented are illustrated in a conceptual map in Figure 2. 4.3 C3 - imperfectly imitable/replaceable The category imperfectly imitable/replaceable (C3) refers to the resource that is difficult to be copied by the competition and has no substitute equivalents. Because they are unique resources, it is difficult for another company to be able to construct the same types of resources (Balmer and Gray, 2003; Barney, 1995; Hoffman, 2000; Iamin, 2007; Porter, 1996; Teece, 2000). Grape production and wine production are difficult to copy, being imperfectly imitable/replaceable because they depend on specific characteristics such as grape acidity, altitude and terroir that are termed as Brazilian terroirs. The quality of this grape cannot be copied and is unique in each region (W4, W6, W11, W13, W14, W15, A3). This issue of terroir as a factor that hinders imitability was also shown in the work of Sharp and Smith (2007), who studied the Champagne region in France and visualized that the territory, climate and subsoil are unique factors of each region. This was also corroborated by Rubini et al. (2013), who pointed out a strong relationship between territory and quality influenced by the geographical origin of the products. In addition, the winemaking process is also unique because each winemaker has a style and adds a personal touch to the process. Each winemaker will make a combination and bring their knowledge and style to the final quality of the wine (W4, W6, W11, A3, W16, W18). There is also the vocation of each region, that is, there are always some grape varieties that fit better in each region. Therefore, the local vocation factor is a competitive differential and also difficult to be imitated or replaced (W3, W6). This idea was described as follows: Vale dos Vinhedos, the main focus is the Merlot wine, here (Pinto Bandeira) the bond is more the sparkling wine, because this region is more prone to the sparkling, here it would be Champenoise sparkling wine from Chardonnay, Riesling and Pinot Noir grapes. In turn, if you take Farroupilha it is a propitious region for Moscatel - sparkling wine from Moscatel. So each region has a context, it has a focus (A3, A6). In addition, American grapes that are not suitable for the production of fine wines are used for the manufacture of grape juice, which has had a great prominence in the international market mainly in the USA (A3, A5, A6, A7, W11). Thus, when analyzing the value delivered to consumers in terms of attributes, benefits and attitudes of shared brands, whether GI, collective brand or sector brand, it is understood that they have imperfectly imitable/replaceable aspects with respect to the competition (Balmer and Gray, 2003; Barney, 1991, 1995; Srivastava et al., 2001). Therefore, for the agencies, associations and managers of the wineries, the shared brand resource is imperfectly imitable/replaceable because in addition to being valuable and rare, it has conditions that make it difficult for competitors to imitate and substitute equivalent resources. In this way, for the VRIA model, a valuable, rare and imperfectly imitable/replaceable resource generates a SCA and performance is above normal in relation to the market average (Barney, 1991, 1995; Barney and Wright, 1998). The results described here are presented in the conceptual map of Figure 3. 4.4 C4 - association The category Association (C4) refers to the equivalent of the letter O in the VRIO framework proposed by Barney (1991, 1995), which represents the organization. In the original form, the model was elaborated for the analysis of an individual organization. However, it is known that shared brands are composed by collectivity, and the trademark registration is granted to a collective entity (Barbosa and Regalado, 2013; Tregear and Gorton, 2009). Thus, for this thesis, the "O" of the organization is replaced by "A" for association, and VRIO will be called VRIA. In the case of the Vale dos Vinhedos region, the wineries are associated with APROVALE who represents a great concentration that generates greater visibility and tourist attraction. These benefits are linked to the growth of the region, the knowledge acquired by the wineries that are technology leaders and the tourist visits that have increased each year (A1, A2, W3, W6). In Pinto Bandeira, the producers are linked to ASPROVINHO. The association has undertaken work to strengthen the region's tourism issue to make it more known and visited. They recognize that the Vale dos Vinhedos was the pioneer and is already consolidated, and that for other regions the disclosure will be slow and gradual, looking for ways that the consumer/tourist is interested in knowing other regions beyond Vale dos Vinhedos. The region is certified with IO. They understand that IO is a stimulus that is making a difference, but it is a tool that is used in conjunction with grape quality and the vinification process among other details (A3, W7, W8, W9). Producers in the Flores da Cunha region are associated with APROMONTES. The work of the association is still incipient and needs to be improved. As reported, there is a lack of further integration by the members and also a team with training in the area of management (A4, W12). In the case of Monte Belo do Sul, there is the APROBELO association. The work developed is still recent, the association does not have headquarters and has associates who participate, but do not have the interest in the indication of origin. These producers intend to continue to produce larger sized bottles of wine and grape juice and will not invest in the production of fine wines (A5, W10). In the Farroupilha region, the producers are linked to AFAVIN. The association carries out publicity actions, event organizations and institutional partnerships. It also works in the promotion of local festivals, as well as participation in international competitions. In Garibaldi, the consortium of CPEG as an entity works toward the objectives of the collectivity, mainly so that the seal of the products is divulged and recognized by consumers as a high-quality product. At present, the initiatives are still recent and there is no specific measurable financial result of CPEG (A7). In relation to the sector brand project, management is carried out by IBRAVIN with the support of APEX. For the international market, respondents believe that Brazilian wine is still seen as an exotic product, but with the work of dissemination and participation in fairs, there is an improvement in the perception of the product (A2). The interviewees understand that participation in fairs should be undertaken collectively, as the wineries are constituents of the so-called "Brazil Block". Producers see themselves as competitors, but to achieve international visibility, they need to work together and in partnership, as they know that individually they would not have resources for such investment (W1, W2, W3, W6). Regarding the domestic market, it was emphasized that there is a need for more public policies and other efforts for the Brazilian consumer to recognize national wine. There are already some actions in this direction carried out by SEBRAE (G4). Thus, for the agencies, associations and managers of the wineries, the shared brand resource has the support of the Association. In this way, the shared brands resource can be considered valuable, rare, imperfectly imitable/replaceable and has the organizational process structured for the management of these brands (Barney, 1991, 1995; Barney and Hesterly, 2006; Barney and Wright, 1998; Grant, 1991; Ito and Gimenez, 2011; Shafeey and Trott, 2014; Sirmon et al., 2011) in the market in question, thus forming the VRIA model. Considering the results presented, it can be concluded that the GIs, the collective brand and the sector brand meet all the VRIA model requirements, in the opinion of the interviewees, creating barriers for imitation by competitors and difficulties in implementing similar strategies in the market (Bharadwaj et al., 1993; Hoffman, 2000; Penrose, 1979; Von Krogh and Roos, 1995). It can be concluded that the proposition that shared brands - GI, collective brands and sector brands - provide SCA, according to the VRIA, can be confirmed, thus fulfilling the four conditions that form the acronym VRIA - valuable, rare, imperfectly imitable/replaceable and association. The value added to the product through the information contained in the shared brands facilitates the establishment of a relationship of trust between producer and consumer, being a source of competitive advantage. When studying the theory about shared brands, this work contributes to the understanding of the concept that the GI, the collective brand and the sector brand can be denominated as types of shared brands from the existent literature. Barney's VRIO framework (Barney, 1991, 1995) hitherto thought for individual companies, has the letter "O" of Organization replaced by the letter "A" of Association, becoming VRIA. In this way, the existence of a shared brand implies in the creation of an association that becomes the owner of the brand. Another theoretical contribution of this research is the deepening of studies on shared brands in Brazil and in the world, as it is still a subject little studied because it is of recent implementation in the market (Barbosa, 2011; Bruch et al., 2014; Castro and Giraldi, 2015; Regalado et al., 2012). 5.1 Practical implications and future research recommendations The practical contributions come from the fact that government agencies, associations and wineries can improve the production process and seek certified products for commercialization in the domestic and foreign markets. These contributions may also, in practice, be used by other sectors and countries. Specifically, it contributes to the fact that government agencies understand the practical result obtained from the management effort in the creation of shared brands. For the associations, the contribution is in improving internal processes and providing advice to members to obtain best practices and results. And finally, for wineries, it points out how much the collective effort in search of shared brands tends to make each individual company more competitive with a source to obtain SCA. Future research studies could be carried on with other wine-growing regions in regions that are beginning to develop the shared brand as in the state of Santa Catarina and Northeast Brazil. A comparative study can also be carried out with other countries of the New World of Wine, such as Argentina, Chile and Uruguay. We also suggest that forthcoming studies consider analyzing the consumers' perspective of the perceived value of wine certifications and the wineries strategies to increase awareness of such added value. 5.2 Limitations The results of qualitative researchers are not reproducible, as the researchers' interpretation is susceptible to changes, as well as the fact that they are specific and unique situations in time for that region or research unit studied, which hinders the generalizations of the study.
|
The purpose of this paper is to investigate if shared brands provide sustainable competitive advantage according to an adapted valuable, rare, imitability/replaceability and organization (VRIO) model to the Brazilian wine sector in the opinion of the government agencies, associations and managers of the wineries.
|
[SECTION: Method] A brand can be seen as a strategic asset that helps a company to be more competitive. In the same way that companies invest in brands, countries can also be seen as such (Anholt, 2005; Huang and Tsai, 2013; Kotler et al., 2006). In considering the role that the image of a country can play in a buyer's behavior, constructs such as the country's brand, the country's image and the country of origin may be attributes that offer the potential for companies to achieve a sustainable competitive advantage (SCA), both in the internal and external market (Baker and Ballington, 2002; Hakala et al., 2013). In this context, based on a country brand, the so-called shared brands may be worked. They have been defined by Tregear and Gorton (2009) as having a single brand as an identity for two or more partner companies, which commit to respect the established agreements, and the management of this shared brand must be performed by an independent entity that represents and responds for the brand. In this way, this article deals with different types of shared brands that fit into this conceptual condition presented by Tregear and Gorton (2009), being geographical indications (GIs), collective brand and sector brand. GIs constitute an instrument for the valuation of goods that are territorially distinguished, that is, they are products coming from a certain region that are recognized for their quality. Collective brands are awarded to a group of organizations that market the same products and wish to link quality to associated companies that do not have to be produced in a delimited geographical region and their registration does not necessarily have to have a geographical name (Castro and Giraldi, 2015). The sector brands are projects developed with the purpose of international representation of the sector through export mechanisms and promotions (Castro and Giraldi, 2015; Lourencao and Giraldi, 2017). Defrancesco et al. (2012) demonstrated the importance of GIs, confirming that foreign consumers are willing to pay a premium for geographical names. They concluded that widely known quality signals, such as protected GIs, would increase the market access of foreign wines by enhancing consumers to recognize and discriminate among terroir-related quality attributes. However, it has to be considered that there are wine enthusiasts and novice consumers, each segment with different needs and degrees of knowledge about wines and the related certifications. It is important that wineries and restaurants be able to understand these differences and devise specific marketing strategies for each audience (Barber et al., 2008). Managers of these establishments need to understand the importance of training to assist the consumer in the purchasing decision process (Dodd et al., 2005). There are also environmental aspects to be considered when analyzing the consumer intention to pay more for certain types of wines. For example, Barber (2010) mentions the expansion of the search for organically or biodynamically produced wines and the consequent reorientation of the blend through repackaging and re-labeling, promoting this in the product dissemination strategy. Barber (2010) suggests that wineries have to focus on developing a work of increasing awareness that organic wine has no lower quality than other traditionally wines (Barber et al., 2010). In this regard, the SCA construct can be defined as that unique attribute that is difficult to be acquired or copied by the competition, which arises from the impossibility of competitors implementing value creation strategies simultaneously (Barney, 1991). Within the SCA theme is the resource-based view (RBV) developing valuable resources and capabilities, which are interchangeable and refer to the tangible and intangible assets that the company uses to develop its strategies (Ray et al., 2004). The study of Vlachveia et al. (2012) in Greece shows that producers invest intangible elements with affect consumer choices. They used a wine tourism and wine routes with an important tool for marketing. They used holistic actions with an important tool for marketing rather than a brand just to a name or a label on the bottle. For this, the wineries used a wine tourism and wine routes. Galati et al. (2014) identified the firms' internal factors that most influence the export intensity and the ability to fit into the international market of the Italian wineries producing sparkling wines through a RBV approach. In particular, the main factors affecting export intensity are innovation degree and intensity of investment in advertising and promotion, among firms' characteristics, and abroad experience, proficiency of foreign languages, educational level and perception barriers, among managers' characteristics, are strategic to ensure the survival of firms into markets and their success. In turn, Amadieu and Viviani (2010) appoint that, for French wine companies, spending money on intangibles is not a guarantee of financial success. The lack of organizational innovation, organizational responsiveness or knowledge-related resources could explain the absence of a positive impact for intangible expenses on performance for French companies in the wine sector. To be financially successful, intangible expenses should be accompanied by organizational and managerial changes, and these changes require time to be learned. Giacomarra et al. (2016) identified that internal motivations drive entrepreneurs to adopt voluntary standards and that wineries adopting voluntary standards exhibit better economic performance than non-certified wine firms. These results suggest that although the adoption of voluntary quality certifications does not have a direct effect in terms of a labor productivity increase, it entails a business management improvement. The work of Alonso and Bressan (2016) shows that it is possible to associate RBV with the strenghts, weakness, opportunities and threats analysis, which is a complementary tool. The accumulated knowledge of the team provides the rare attribute and history, reputation and territory related to the imperfectly imitable/replaceable attribute. Maurel (2009) complements this idea regarding the impact of decision-maker characteristics on export performance by indicating a lack of significant impacts on export performance. Suarez-Ortega and A'lamo-Vera (2005) confirmed that factors influencing export involvement are not the same along the process of export development. Three aspects of export development have been analyzed: intention, propensity and intensity. The effect of industry-specific characteristics on internal export factors has been isolated through the selection of one industry in one country for the empirical research. The study of Wilk and Fensterseifer (2003) was conducted having as its central purpose the identification of the strategic resources and capabilities of a cluster located in southern Brazil and their sustainability that history matters and the cumulative development and upgrading of resources and capabilities can fundamentally shape the future strategic options of a cluster. Although these papers have approached the RBV, when dealing with internal resources, they have not explored and classified the foundations for obtaining a SCA, such as the valuable, rare, imitability/replaceability and organization (VRIO) framework. The seminal work of the RBV was that of Barney (1991), which culminated in the VRIO framework (Barney, 1991, 1995). Two decades later, Barney et al. (2011) show that the theory is mature and some researchers prefer to use the term resource-based theory (RBT) instead of RBV, as they believe that it is currently a consolidated theory and not just a vision (Barney et al., 2011; Newbert, 2007). In analyzing the shared brands and the perspective of SCA, the objective of this article is defined as to investigate if shared brands provide SCA according to the VRIO model adapted to the analyzed context (Brazilian winemaking sector), in the opinion of the government agencies, associations and managers of the wineries. 2.1 Sustainable competitive advantage The SCA arises from the development of a set of features and abilities that cannot be negotiated, imitated or replaced, or implemented simultaneously by current and potential competitors, and use of resources within the company is difficult to access or replicate by competitors (Bharadwaj et al., 1993; Dierickx and Cool, 1989; Hoffman, 2000; Iamin, 2007; Porter, 1996; Teece, 2000; Von Krogh and Roos, 1995; Wernerfelt, 1984). The work of Barney (1986) introduced the position of heterogeneous resources, which is an important theoretical contribution to RBV. The immobility of resources prevents other firms from acquiring or imitating resources easily, making it difficult to make resources unique in the market (Bharadwaj et al., 1993; Hoffman, 2000; Penrose, 1979; Von Krogh and Roos, 1995). Based on the principles of heterogeneity and resource immobility, which is the basis for RBV, Barney (1991) created the VRIS framework (valuable, rare, imperfectly imitable, and imperfectly substitutable) or, for some authors, VRIN (valuable, rare, imperfectly imitable and non-replaceable). The acronyms represent the same idea and differ in the way of explicit terminology with S in VRIS and with N in VRIN, representing, in both, the imperfectly replaceable attribute (Shafeey and Trott, 2014). The VRIS/VRIN model is revised by Barney (1995) and becomes the VRIO framework, in which the imperfectly imitable and imperfectly replaceable attribute is added with the letter "I" in the framework, summing them into the issue of imitation difficulty and adding the item organization, which discusses whether the company is organized to exploit its resources and capabilities (Barney, 1995; Barney and Wright, 1998; Shafeey and Trott, 2014). The organizational process such as the action of managers in the form of structuring the resources and managing them in the market provides the fourth necessary condition to obtain SCA, thus forming the VRIO (Balmer and Gray, 2003; Barney, 1991, 1995; Barney and Hesterly, 2006; Cazurra and Un, 2015; Grant, 1991; Ito and Gimenez, 2011; Shafeey and Trott, 2014; Sirmon et al., 2011; Srivastava et al., 2001). Nowadays, RBV studies are more concise and in the process of maturation, which leads some authors to substitute the term "RBV perspective" for RBT, according to Barney et al. (2011), Crook et al. (2008), Maritan and Peteraf (2011), McWilliams and Siegel (2011) and Molloy et al. (2011). Based on these trends and their relationship with the RBV, RBT with the VRIO framework applicable to the country brand and its related concepts, such as shared brands, is considered by analyzing the attributes (valuable, rare, imperfectly imitable and imperfectly replaceable) based on the internal resources of each country (Zhang and Su, 2009). From the VRIO model of Barney (1991, 1995) and analyzing the applicability of this model to the specific case of shared brands, it is important to make here a specific adaptation in this original model, calling it VRIA. Thus, the "O" of organization is replaced by the "A" of association as for the legal existence of shared brands and according to the definition proposed by Tregear and Gorton (2009). 2.2 Shared brands and sustainable competitive advantage In Brazil, the understanding of shared brands considers GIs, collective brands and sector brands, which can occur in isolation or at the same time. GIs are divided into two types - indications of origin (IO) and denominations of origin (DO). For IO, the reputation of the product is analyzed. In DO, the peculiarities of that region must affect the final result of the product, in an identifiable and measurable way. A food or agricultural product recognized as a denomination of origin must be produced, processed, manufactured or industrialized in its territory of origin, which is not necessarily a rule when it comes to indication of origin [Barbosa et al., 2013, 2014; Instituto Nacional da Propriedade Industrial (INPI), 2014; Junger, 2014; Servico Brasileiro de Apoio as Micro e Pequenas Empresas (SEBRAE) and Instituto Nacional da Propriedade Industrial (INPI) 2011, 2014a, 2014b, 2014c; Seminario Internacional de Indicacoes Geograficas e Marcas Coletivas, 2014 (verbal information)]. For wines, there are six Brazilian indications of origin: Vale dos Vinhedos, Pinto Bandeira, Vale da Uva Goethe, Altos Montes, Monte Belo and Farroupilha [Servico Brasileiro de Apoio as Micro e Pequenas Empresas (SEBRAE) and Instituto Nacional da Propriedade Industrial (INPI) 2011, 2014b; Instituto Nacional da Propriedade Industrial (INPI) 2014, 2015]. In addition, two GIs are under development, namely, Campanha and Vale do Submedio do Sao Francisco (Empresa Brasileira de Pesquisa Agropecuaria - Uva e Vinho (EMBRAPA), 2014). Collective brands, according to industrial property law, are used to identify products or services coming from members of a certain entity [Barbosa et al., 2013; Servico Brasileiro de Apoio as Micro e Pequenas Empresas (SEBRAE) and Instituto Nacional da Propriedade Industrial (INPI), 2014c]. As for the sector brand, there is a diversity of projects and work fronts developed by APEX, promoting the Brazilian industry. One type of share brand that can be classified as a GI is the controlled denomination of origin (DOC). Marchinia et al. (2014) show that the higher number of firms and the investments in marketing contributes to the collective reputation of the DOC. Similarly, the older the DOC designation, the greater its reputation because of the accumulation effect of investments in marketing and consumer experiences. Product differentiation within the DOC areas (subzones, variants of the product, the presence of sweet wine, etc.) represents a growth factor for the reputation of the DOC. Despite the great number of criticisms to the system of DOC wines, the presence of shared and mandatory rules resulted in a collective value that can support SMEs in building their own individual reputation. Olmos (2011), in a study of the appellation of origin (DOC) in the Rioja wine sector, showed that process innovations are not a determinant of internationalization. However, the results show a positive and significant link between a firm's marketing activities and its level of internationalization. The positive effect of advertising in determining the degree of internationalization confirms that wineries in this region can benefit from the collective umbrella brand "qualified denomination of origin" in comparison with wineries in other wine-producing regions. Defrancesco et al. (2012) shows that the protected GIs can constitute a strategic tool for wine producers and it's possible to a strategic differentiation of New World wine producing countries. Fensterseifer and Rastoin (2013) analyzed the contribution potential of the resources of a cluster to generate competitive advantage for local companies in the wine sector. The benefits obtained when companies participated in the cluster were classified from an RBV perspective in the following categories: institutional, specialization, social, reputation and natural capital. Thus, a set of potentially strategic resources for the Serra Gaucha wine cluster was identified. This study was based on a qualitative approach that, for Bauer and Gaskell (2000), aims to understand in detail beliefs, attitudes, values and motivations regarding the behavior of people in specific social contexts (Malhotra, 2006). The research was exploratory and the field study was conducted with in-depth interviews (Bardin, 2011; Cooper and Schindler, 2014; Sampieri et al., 2010). This work was delimited to the study of the viniculture sector. This sector was chosen, as it was possible to analyze the shared brands (GI, collective brand and sector brand), according to studies developed in Brazil (Barbosa et al., 2013, Bruch et al., 2014, Castro and Giraldi, 2015). In the international wine market, considering the past five years, the sparkling wine continued to grow with an annual rate of 7 per cent in value and 6 per cent in volume and bulk wine with an annual rate of 6 per cent in value and 3 per cent in volume. Over the same period, bottled wine grew 5 per cent in value but only 1 per cent in volume in terms of annual rate. Forecast of the future evolution of wine market is challenging and risky in a context dominated by discontinuities, and there is not much research on how markets evolve (Pomarici, 2016). In addition, the Brazilian wine sector is important to the national economy, being the first to obtain a GI in the country, a concept similar to DOC - controlled designation of origin used in Italy, for example (Barbosa et al., 2013). The Brazilian wine sector has already six certified regions, namely, Vale dos Vinhedos, Pinto Bandeira, Monte Belo do Sul, Flores da Cunha, Farroupilha and Urussanga and some collective brands such as ACAVITIS, CPEG and Vinhateiros do Vale and also sector brands (Vinhos do Brasil/Wines of Brasil). The work was delimited to the study of the fine wines that have GIs and are located in the Brazilian state of Rio Grande do Sul. In terms of numbers, the Brazilian viniculture sector is the 19th ranked country in terms of area cultivated with grapes, 12th in grape production and 13th in wine production. From 2004 to 2014, there was a 100 per cent increase in the national production of fine wines, from 25 million to 50 million L (Copello, 2015). The following standard terminology was used for codification: "G" for government agencies, "A" for associations and "W" to represent winery managers followed by numbers in the order in which the interviews were conducted. Interviews were conducted with government agencies such as SEBRAE, MAPA (Ministry of Agriculture, Livestock and Supply), INPI and APEX (Brazilian agency for promotion of exports and investments) represented by IBRAVIN (G1, G2 e G3) and EMBRAPA UVA E VINHO (Embrapa Grape and Wine) (G4) to understand in detail the processes of creation and registration of shared brands. The interviews were conducted in May 2015 and each had an average duration of 40 min. Subsequently, primary data were also collected through interviews with the associations that manage GIs and the collective brands of fine wines, such as APROVALE (Association of Fine Wine Producers of the Vinhedos Valley - A1 e A2); ASPROVINHO (Association of Wine Producers of Pinto Bandeira - A3); APROMONTES (Association of Wine Producers of Altos Montes - A4); APROBELO (Association of Viticulture Farmers of Monte Belo do Sul - A5); and AFAVIN (Association of Wine, Sparkling Wine, Juices and Derivative Producers of Farroupilha - A6); CPEG (A7); AVIGA (Association of Wine Producers of Garibaldi - A8); SINDIVINHORS (Syndicate of the Industry of wine, grape must, vinegar and beverages derived from the grape, and wine of the state of Rio Grande do Sul - A9); and the City Hall of Garibaldi (A10) together with the Secretary of Tourism. Ten interviews were conducted with an average duration of 40 min each. For data collection, six wine-growing municipalities of the State of Rio Grande do Sul were visited: Bento Goncalves, Monte Belo do Sul, Flores da Cunha, Farroupilha, Pinto Bandeira and Garibaldi. Finally, the managers of the wineries were interviewed, with 18 interviews, namely, APROVALE - Don Laurindo (W1), Lidio Carraro (W2, W3), Miolo (W4), Peculiare (W5) and Pizzato (W6); ASPROVINHO - Cooperativa Aurora (W7, W8) and Valmarino (W9); APROBELO - Fae (W10) and Fantin (W11); APROMONTES - Mioranza (W12); CPEG/AVIGA - Batistello (W13), Cooperativa Garibaldi (W14, W15), Milantino (W16), Pedrucci (W17) and Vaccaro (W18). The interviews had an average duration of approximately 40 min each, with about 12 h of testimonials. With this number of interviews, a saturation point was reached, in which the researcher, according to Bauer and Gaskell (2000), evaluated that no new surprises, perceptions or revelations will appear in additional interviews. The wineries interviewed belong to five producing regions in the Rio Grande do Sul (Brazil). The process began with the planting of grapes with the arrival of Italian immigrants in 1847 and only in the 1990s they underwent a process of remodeling, culminating in the current Brazilian winemaking format (W6). Besides the cultivation, producers have started to vinify and to commercialize the products in the internal and external market. The interviewees are large, medium and small, possessing 900, 30 and 8.5 hectares of vineyards of their own, on average. They produce an average of 20 million L/year for the largest, 197,600 L/year for the average wineries and 27,500 L/year for the small wineries interviewed. The following proposition was proposed for this study: shared brands provide SCA, according to the VRIA, in the opinion of the government agencies, associations and managers of the wineries (Barney, 1986, 1991, 1995; Dierickx and Cool, 1989). This proposition is based on the work of Barney (1986, 1991, 1995) and Dierickx and Cool (1989) who postulate that SCA arises when competitors are unable to obtain the same results, even though they use similar marketing strategies (Bharadwaj et al., 1993; Von Krogh and Roos, 1995). For data analysis, content analysis was used, which is a set of communication analysis techniques, in which the analyst delimits the codification or registration units (Bardin, 2011). Categorization consists in organizing the data so that the researcher can make decisions and draw conclusions from them (Bardin, 2011; Cooper and Schindler, 2014). The proposal of this article is composed of four categories denominated C1, C2, C3 and C4, representing the VRIA framework. 4.1 C1 - valuable The valuable category (C1) refers to the benefits, advantages and factors that lead wineries to differentiate themselves from their competitors. The shared brand, being valuable, adds value, becomes distinct and differentiates itself in the minds of consumers (Balmer and Gray, 2003; Barney, 2001). In this way, the interviewees who have GI understand that it adds value and brings visibility to both large- and small-sized wineries. It ends up generating a greater interest of the consumer in knowing that region, fomenting tourism and adding value for the whole chain [O4, Servico Brasileiro de Apoio as Micro e Pequenas Empresas (SEBRAE) and Instituto Nacional da Propriedade Industrial (INPI), 2014c]. The producers understand that with the certificate, they have gained publicity in the media, and Brazilian wine has become more recognized, as can be seen in the following excerpt: It adds a lot of value, brings greater competitiveness in the internal and external market. The issue of the reputation of Vale dos Vinhedos, this intrinsic value of quality, of the region, the culture from here, its own identity, distinction from similar wines. This perception of the brand leads to "this product is more expensive, it has DO, so I think it is a good product" (W4). The IO of Vale dos Vinhedos was the first Brazilian IO and emerged soon after the opening of Brazil to the international market. At that moment, there were changes in the competitive scenario and Brazilian fine wine began to gain space in the international market, presenting itself as a valuable product. The emergence of Vale dos Vinhedos DO - the first DO of Brazilian wines - occurred in 2012 and did not have as much impact as the launch of IO in 2002. This is because of the fact that IO was a novelty and also because DO is more restrictive, covering fewer wineries and fewer certified products (W6): Actually, this is it, we are a smaller winery here, I do not have, let's say, market visibility as a bigger company, I do not have the marketing power to invest in it. When I have a designation of origin, so to speak, the consumer sees my product with different eyes, he does not know the individual brand, but he knows what a designation of origin is, so he can position the product in an easier way. This is a strength to position yourself in the market much greater than if you were individual (W5). They believe that DO is already more traditional overseas such as in Europe, for example, and this facilitates the understanding that it is a wine with production and quality mainly controlled for the consumer who is already accustomed to imported wines of a high standard. They recognize how having DO is a valuable resource for the producing company and for Brazil as the country of origin (W4, W5, W12, W16). The smaller wineries, whose production is practically handcrafted and family operated, denominated as boutique wineries or micro champagne enterprises do not produce enough volume to distribute their products in supermarket chains (W17). It is noteworthy that Vale dos Vinhedos GI was recognized by the European Union in 2007 and represented a step toward visibility in the external market. Before, the wines that went to Europe went as ordinary wines: From the recognition by the European Union, the wines of Vale dos Vinhedos had opened the doors to place their labels. The GI seal gives both region and the products a good reputation (A1, A2, A3, A6). In this sense, the interviewee A10 affirms that CPEG guarantees the quality of the product and this generates recognition by the consumer himself and respectability for the producing region and adds value to the product. This idea is complemented by A9 who believes that this recognition will come through the dissemination and promotion of products. With regard to the sector brand, one interviewee reports: Everyone who works with wines, e.g. journalists who are directly linked to wine, already knows that Brazil produces wine. Now, another step is to make the final consumer know what Brazil produces. The end consumer still does not know that Brazil produces wines, and good wines (G1). Another factor that adds value is the issue of the distribution of products in specialized stores abroad, which counts on a trained employee to perform a personalized service, generating competitive differential (G1, W6). Respondents believe that being part of the Wines of Brazil project and exporting the products adds value, meaning that the product becomes more valuable in the face of international competition (G3, G4, W3). This result relates to Defrancesco et al.'s (2012) study, which demonstrated the importance of GIs as a quality signal that makes consumers willing to pay a premium for geographical names. It is important to mention that this added value may be better perceived by wine enthusiasts than by novices. As Barber et al. (2008) and Dodd et al. (2005) indicate that because of the different needs and degrees of knowledge about wines and related certifications, it may be important for Brazilian wineries to better understand their consumers and provide proper information about the certifications to increase their wines perceived value. In this way, the initiatives of shared brands, whether GI, collective brand or sector brand, add value to the products and differentiate the wineries nationally and internationally. This result found among the interviewees is in agreement with the literature that, when developing valuable resources, the company gains a competitive advantage, as it adds value to the products and manages to keep up with the environmental changes of the competitive market (Barney, 1995; Ray et al., 2004). To illustrate the results schematically, the conceptual map is presented in Figure 1. 4.2 C2 - rare The Rare category (C2), according to Barney (1991, 1995), refers to a differential for the company, in addition to having valuable resources. When it comes to GIs, this rare factor is found in Brazilian legislation and its regulations, which, being detailed and with more requirements, resembles the European model. This differentiates Brazilian wine and brings rarity when compared to the international competitors of the so-called New World (Argentina, Chile, the USA, Australia, South Africa and New Zealand). In general, for the New World, only one delimitation with a geographical name is made and it must own about 65 to 80 per cent of the grape produced in this territory, as well as its winemaking. While for DO in Brazil, 100 per cent of the grape must be produced and vinification performed in the defined region (G4). In this way, the European consumer understands what a GI is, recognizes that the product has differentiation criteria and this, besides adding value, brings rarity to the products. The GI is like a "calling card", informing the consumer of its characteristics: region, origin, history and culture of the product (G4, A6). This result relates to Alonso and Bressan's (2016) study, which suggested that the accumulated knowledge of the team provides the rare attribute and history, and the reputation and territory relate to the imperfectly imitable/replaceable attribute. It also finds support in Wilk and Fensterseifer's (2003), who demonstrated that the wine cluster located in southern Brazil has achieved a high level of resources and capabilities that can give a competitive advantage to this cluster. Respondents say that the consumer generally does not remember the individual brand of wine he has taken, but rather the country of origin of that product: You do not remember a brand of wine you took, the last foreign wine you took will remind you that you took a Chilean wine, you will remember you took a French wine, the brand that the consumer remembers is often the origin, it is not the brand of the product itself (G3). For the global competition, Vale dos Vinhedos is a block that works together for the dissemination of Brazilian wine, showing product differentials. They report that the consumer already has a perception that Vale dos Vinhedos represents certified quality products. However, within the Vale itself there are 22 wineries that also compete with each other, as the tourist chooses among the existing options, and then there is the work to obtain a competitive differential for the individual brand, as well as the collective actions for the competitiveness of the region internationally (W4, W6). The region of Pinto Bandeira, for example, has as main focus the traditional sparkling or methode champenoise (champenoise method). This product is rare because it differs from other producing regions and has its own flavor because of altitude and soil conditions for the production of grapes. And through the traditional method, there is a more careful and time-consuming process in the elaboration that also adds value and provides rarity to the sparkling wine (A3). Similarly, the reality lived in Garibaldi also shows that the rarity of the product lies in the fact that the sparkling wine produced is differentiated and Brazil is already being recognized, especially the region of Serra Gaucha, as a region favorable for the production of sparkling wines (A7, A8, A9). The proposal of the sparkling wine, for example, is more refreshing and seeks a different style from the French, which provides a more structured and fuller-bodied wine (W3, W4). In this way, shared brands, whether GI, collective brand or sector brand, have aspects of the rarity in relation to current and potential competition. This makes it difficult for other companies that wish to implement equivalent strategies (Barney, 1991, 1995). Therefore, for the agencies, associations and managers of the wineries, the shared brand resource is rare because in addition to being valuable, it has conditions that are not widely disseminated among the competitors in the market. The results presented are illustrated in a conceptual map in Figure 2. 4.3 C3 - imperfectly imitable/replaceable The category imperfectly imitable/replaceable (C3) refers to the resource that is difficult to be copied by the competition and has no substitute equivalents. Because they are unique resources, it is difficult for another company to be able to construct the same types of resources (Balmer and Gray, 2003; Barney, 1995; Hoffman, 2000; Iamin, 2007; Porter, 1996; Teece, 2000). Grape production and wine production are difficult to copy, being imperfectly imitable/replaceable because they depend on specific characteristics such as grape acidity, altitude and terroir that are termed as Brazilian terroirs. The quality of this grape cannot be copied and is unique in each region (W4, W6, W11, W13, W14, W15, A3). This issue of terroir as a factor that hinders imitability was also shown in the work of Sharp and Smith (2007), who studied the Champagne region in France and visualized that the territory, climate and subsoil are unique factors of each region. This was also corroborated by Rubini et al. (2013), who pointed out a strong relationship between territory and quality influenced by the geographical origin of the products. In addition, the winemaking process is also unique because each winemaker has a style and adds a personal touch to the process. Each winemaker will make a combination and bring their knowledge and style to the final quality of the wine (W4, W6, W11, A3, W16, W18). There is also the vocation of each region, that is, there are always some grape varieties that fit better in each region. Therefore, the local vocation factor is a competitive differential and also difficult to be imitated or replaced (W3, W6). This idea was described as follows: Vale dos Vinhedos, the main focus is the Merlot wine, here (Pinto Bandeira) the bond is more the sparkling wine, because this region is more prone to the sparkling, here it would be Champenoise sparkling wine from Chardonnay, Riesling and Pinot Noir grapes. In turn, if you take Farroupilha it is a propitious region for Moscatel - sparkling wine from Moscatel. So each region has a context, it has a focus (A3, A6). In addition, American grapes that are not suitable for the production of fine wines are used for the manufacture of grape juice, which has had a great prominence in the international market mainly in the USA (A3, A5, A6, A7, W11). Thus, when analyzing the value delivered to consumers in terms of attributes, benefits and attitudes of shared brands, whether GI, collective brand or sector brand, it is understood that they have imperfectly imitable/replaceable aspects with respect to the competition (Balmer and Gray, 2003; Barney, 1991, 1995; Srivastava et al., 2001). Therefore, for the agencies, associations and managers of the wineries, the shared brand resource is imperfectly imitable/replaceable because in addition to being valuable and rare, it has conditions that make it difficult for competitors to imitate and substitute equivalent resources. In this way, for the VRIA model, a valuable, rare and imperfectly imitable/replaceable resource generates a SCA and performance is above normal in relation to the market average (Barney, 1991, 1995; Barney and Wright, 1998). The results described here are presented in the conceptual map of Figure 3. 4.4 C4 - association The category Association (C4) refers to the equivalent of the letter O in the VRIO framework proposed by Barney (1991, 1995), which represents the organization. In the original form, the model was elaborated for the analysis of an individual organization. However, it is known that shared brands are composed by collectivity, and the trademark registration is granted to a collective entity (Barbosa and Regalado, 2013; Tregear and Gorton, 2009). Thus, for this thesis, the "O" of the organization is replaced by "A" for association, and VRIO will be called VRIA. In the case of the Vale dos Vinhedos region, the wineries are associated with APROVALE who represents a great concentration that generates greater visibility and tourist attraction. These benefits are linked to the growth of the region, the knowledge acquired by the wineries that are technology leaders and the tourist visits that have increased each year (A1, A2, W3, W6). In Pinto Bandeira, the producers are linked to ASPROVINHO. The association has undertaken work to strengthen the region's tourism issue to make it more known and visited. They recognize that the Vale dos Vinhedos was the pioneer and is already consolidated, and that for other regions the disclosure will be slow and gradual, looking for ways that the consumer/tourist is interested in knowing other regions beyond Vale dos Vinhedos. The region is certified with IO. They understand that IO is a stimulus that is making a difference, but it is a tool that is used in conjunction with grape quality and the vinification process among other details (A3, W7, W8, W9). Producers in the Flores da Cunha region are associated with APROMONTES. The work of the association is still incipient and needs to be improved. As reported, there is a lack of further integration by the members and also a team with training in the area of management (A4, W12). In the case of Monte Belo do Sul, there is the APROBELO association. The work developed is still recent, the association does not have headquarters and has associates who participate, but do not have the interest in the indication of origin. These producers intend to continue to produce larger sized bottles of wine and grape juice and will not invest in the production of fine wines (A5, W10). In the Farroupilha region, the producers are linked to AFAVIN. The association carries out publicity actions, event organizations and institutional partnerships. It also works in the promotion of local festivals, as well as participation in international competitions. In Garibaldi, the consortium of CPEG as an entity works toward the objectives of the collectivity, mainly so that the seal of the products is divulged and recognized by consumers as a high-quality product. At present, the initiatives are still recent and there is no specific measurable financial result of CPEG (A7). In relation to the sector brand project, management is carried out by IBRAVIN with the support of APEX. For the international market, respondents believe that Brazilian wine is still seen as an exotic product, but with the work of dissemination and participation in fairs, there is an improvement in the perception of the product (A2). The interviewees understand that participation in fairs should be undertaken collectively, as the wineries are constituents of the so-called "Brazil Block". Producers see themselves as competitors, but to achieve international visibility, they need to work together and in partnership, as they know that individually they would not have resources for such investment (W1, W2, W3, W6). Regarding the domestic market, it was emphasized that there is a need for more public policies and other efforts for the Brazilian consumer to recognize national wine. There are already some actions in this direction carried out by SEBRAE (G4). Thus, for the agencies, associations and managers of the wineries, the shared brand resource has the support of the Association. In this way, the shared brands resource can be considered valuable, rare, imperfectly imitable/replaceable and has the organizational process structured for the management of these brands (Barney, 1991, 1995; Barney and Hesterly, 2006; Barney and Wright, 1998; Grant, 1991; Ito and Gimenez, 2011; Shafeey and Trott, 2014; Sirmon et al., 2011) in the market in question, thus forming the VRIA model. Considering the results presented, it can be concluded that the GIs, the collective brand and the sector brand meet all the VRIA model requirements, in the opinion of the interviewees, creating barriers for imitation by competitors and difficulties in implementing similar strategies in the market (Bharadwaj et al., 1993; Hoffman, 2000; Penrose, 1979; Von Krogh and Roos, 1995). It can be concluded that the proposition that shared brands - GI, collective brands and sector brands - provide SCA, according to the VRIA, can be confirmed, thus fulfilling the four conditions that form the acronym VRIA - valuable, rare, imperfectly imitable/replaceable and association. The value added to the product through the information contained in the shared brands facilitates the establishment of a relationship of trust between producer and consumer, being a source of competitive advantage. When studying the theory about shared brands, this work contributes to the understanding of the concept that the GI, the collective brand and the sector brand can be denominated as types of shared brands from the existent literature. Barney's VRIO framework (Barney, 1991, 1995) hitherto thought for individual companies, has the letter "O" of Organization replaced by the letter "A" of Association, becoming VRIA. In this way, the existence of a shared brand implies in the creation of an association that becomes the owner of the brand. Another theoretical contribution of this research is the deepening of studies on shared brands in Brazil and in the world, as it is still a subject little studied because it is of recent implementation in the market (Barbosa, 2011; Bruch et al., 2014; Castro and Giraldi, 2015; Regalado et al., 2012). 5.1 Practical implications and future research recommendations The practical contributions come from the fact that government agencies, associations and wineries can improve the production process and seek certified products for commercialization in the domestic and foreign markets. These contributions may also, in practice, be used by other sectors and countries. Specifically, it contributes to the fact that government agencies understand the practical result obtained from the management effort in the creation of shared brands. For the associations, the contribution is in improving internal processes and providing advice to members to obtain best practices and results. And finally, for wineries, it points out how much the collective effort in search of shared brands tends to make each individual company more competitive with a source to obtain SCA. Future research studies could be carried on with other wine-growing regions in regions that are beginning to develop the shared brand as in the state of Santa Catarina and Northeast Brazil. A comparative study can also be carried out with other countries of the New World of Wine, such as Argentina, Chile and Uruguay. We also suggest that forthcoming studies consider analyzing the consumers' perspective of the perceived value of wine certifications and the wineries strategies to increase awareness of such added value. 5.2 Limitations The results of qualitative researchers are not reproducible, as the researchers' interpretation is susceptible to changes, as well as the fact that they are specific and unique situations in time for that region or research unit studied, which hinders the generalizations of the study.
|
This study was based on a qualitative and exploratory research, based on in-depth interviews. Fine wines that have geographical indications and are located in the Brazilian state of Rio Grande do Sul were analyzed and content analysis was used to explore data.
|
[SECTION: Findings] A brand can be seen as a strategic asset that helps a company to be more competitive. In the same way that companies invest in brands, countries can also be seen as such (Anholt, 2005; Huang and Tsai, 2013; Kotler et al., 2006). In considering the role that the image of a country can play in a buyer's behavior, constructs such as the country's brand, the country's image and the country of origin may be attributes that offer the potential for companies to achieve a sustainable competitive advantage (SCA), both in the internal and external market (Baker and Ballington, 2002; Hakala et al., 2013). In this context, based on a country brand, the so-called shared brands may be worked. They have been defined by Tregear and Gorton (2009) as having a single brand as an identity for two or more partner companies, which commit to respect the established agreements, and the management of this shared brand must be performed by an independent entity that represents and responds for the brand. In this way, this article deals with different types of shared brands that fit into this conceptual condition presented by Tregear and Gorton (2009), being geographical indications (GIs), collective brand and sector brand. GIs constitute an instrument for the valuation of goods that are territorially distinguished, that is, they are products coming from a certain region that are recognized for their quality. Collective brands are awarded to a group of organizations that market the same products and wish to link quality to associated companies that do not have to be produced in a delimited geographical region and their registration does not necessarily have to have a geographical name (Castro and Giraldi, 2015). The sector brands are projects developed with the purpose of international representation of the sector through export mechanisms and promotions (Castro and Giraldi, 2015; Lourencao and Giraldi, 2017). Defrancesco et al. (2012) demonstrated the importance of GIs, confirming that foreign consumers are willing to pay a premium for geographical names. They concluded that widely known quality signals, such as protected GIs, would increase the market access of foreign wines by enhancing consumers to recognize and discriminate among terroir-related quality attributes. However, it has to be considered that there are wine enthusiasts and novice consumers, each segment with different needs and degrees of knowledge about wines and the related certifications. It is important that wineries and restaurants be able to understand these differences and devise specific marketing strategies for each audience (Barber et al., 2008). Managers of these establishments need to understand the importance of training to assist the consumer in the purchasing decision process (Dodd et al., 2005). There are also environmental aspects to be considered when analyzing the consumer intention to pay more for certain types of wines. For example, Barber (2010) mentions the expansion of the search for organically or biodynamically produced wines and the consequent reorientation of the blend through repackaging and re-labeling, promoting this in the product dissemination strategy. Barber (2010) suggests that wineries have to focus on developing a work of increasing awareness that organic wine has no lower quality than other traditionally wines (Barber et al., 2010). In this regard, the SCA construct can be defined as that unique attribute that is difficult to be acquired or copied by the competition, which arises from the impossibility of competitors implementing value creation strategies simultaneously (Barney, 1991). Within the SCA theme is the resource-based view (RBV) developing valuable resources and capabilities, which are interchangeable and refer to the tangible and intangible assets that the company uses to develop its strategies (Ray et al., 2004). The study of Vlachveia et al. (2012) in Greece shows that producers invest intangible elements with affect consumer choices. They used a wine tourism and wine routes with an important tool for marketing. They used holistic actions with an important tool for marketing rather than a brand just to a name or a label on the bottle. For this, the wineries used a wine tourism and wine routes. Galati et al. (2014) identified the firms' internal factors that most influence the export intensity and the ability to fit into the international market of the Italian wineries producing sparkling wines through a RBV approach. In particular, the main factors affecting export intensity are innovation degree and intensity of investment in advertising and promotion, among firms' characteristics, and abroad experience, proficiency of foreign languages, educational level and perception barriers, among managers' characteristics, are strategic to ensure the survival of firms into markets and their success. In turn, Amadieu and Viviani (2010) appoint that, for French wine companies, spending money on intangibles is not a guarantee of financial success. The lack of organizational innovation, organizational responsiveness or knowledge-related resources could explain the absence of a positive impact for intangible expenses on performance for French companies in the wine sector. To be financially successful, intangible expenses should be accompanied by organizational and managerial changes, and these changes require time to be learned. Giacomarra et al. (2016) identified that internal motivations drive entrepreneurs to adopt voluntary standards and that wineries adopting voluntary standards exhibit better economic performance than non-certified wine firms. These results suggest that although the adoption of voluntary quality certifications does not have a direct effect in terms of a labor productivity increase, it entails a business management improvement. The work of Alonso and Bressan (2016) shows that it is possible to associate RBV with the strenghts, weakness, opportunities and threats analysis, which is a complementary tool. The accumulated knowledge of the team provides the rare attribute and history, reputation and territory related to the imperfectly imitable/replaceable attribute. Maurel (2009) complements this idea regarding the impact of decision-maker characteristics on export performance by indicating a lack of significant impacts on export performance. Suarez-Ortega and A'lamo-Vera (2005) confirmed that factors influencing export involvement are not the same along the process of export development. Three aspects of export development have been analyzed: intention, propensity and intensity. The effect of industry-specific characteristics on internal export factors has been isolated through the selection of one industry in one country for the empirical research. The study of Wilk and Fensterseifer (2003) was conducted having as its central purpose the identification of the strategic resources and capabilities of a cluster located in southern Brazil and their sustainability that history matters and the cumulative development and upgrading of resources and capabilities can fundamentally shape the future strategic options of a cluster. Although these papers have approached the RBV, when dealing with internal resources, they have not explored and classified the foundations for obtaining a SCA, such as the valuable, rare, imitability/replaceability and organization (VRIO) framework. The seminal work of the RBV was that of Barney (1991), which culminated in the VRIO framework (Barney, 1991, 1995). Two decades later, Barney et al. (2011) show that the theory is mature and some researchers prefer to use the term resource-based theory (RBT) instead of RBV, as they believe that it is currently a consolidated theory and not just a vision (Barney et al., 2011; Newbert, 2007). In analyzing the shared brands and the perspective of SCA, the objective of this article is defined as to investigate if shared brands provide SCA according to the VRIO model adapted to the analyzed context (Brazilian winemaking sector), in the opinion of the government agencies, associations and managers of the wineries. 2.1 Sustainable competitive advantage The SCA arises from the development of a set of features and abilities that cannot be negotiated, imitated or replaced, or implemented simultaneously by current and potential competitors, and use of resources within the company is difficult to access or replicate by competitors (Bharadwaj et al., 1993; Dierickx and Cool, 1989; Hoffman, 2000; Iamin, 2007; Porter, 1996; Teece, 2000; Von Krogh and Roos, 1995; Wernerfelt, 1984). The work of Barney (1986) introduced the position of heterogeneous resources, which is an important theoretical contribution to RBV. The immobility of resources prevents other firms from acquiring or imitating resources easily, making it difficult to make resources unique in the market (Bharadwaj et al., 1993; Hoffman, 2000; Penrose, 1979; Von Krogh and Roos, 1995). Based on the principles of heterogeneity and resource immobility, which is the basis for RBV, Barney (1991) created the VRIS framework (valuable, rare, imperfectly imitable, and imperfectly substitutable) or, for some authors, VRIN (valuable, rare, imperfectly imitable and non-replaceable). The acronyms represent the same idea and differ in the way of explicit terminology with S in VRIS and with N in VRIN, representing, in both, the imperfectly replaceable attribute (Shafeey and Trott, 2014). The VRIS/VRIN model is revised by Barney (1995) and becomes the VRIO framework, in which the imperfectly imitable and imperfectly replaceable attribute is added with the letter "I" in the framework, summing them into the issue of imitation difficulty and adding the item organization, which discusses whether the company is organized to exploit its resources and capabilities (Barney, 1995; Barney and Wright, 1998; Shafeey and Trott, 2014). The organizational process such as the action of managers in the form of structuring the resources and managing them in the market provides the fourth necessary condition to obtain SCA, thus forming the VRIO (Balmer and Gray, 2003; Barney, 1991, 1995; Barney and Hesterly, 2006; Cazurra and Un, 2015; Grant, 1991; Ito and Gimenez, 2011; Shafeey and Trott, 2014; Sirmon et al., 2011; Srivastava et al., 2001). Nowadays, RBV studies are more concise and in the process of maturation, which leads some authors to substitute the term "RBV perspective" for RBT, according to Barney et al. (2011), Crook et al. (2008), Maritan and Peteraf (2011), McWilliams and Siegel (2011) and Molloy et al. (2011). Based on these trends and their relationship with the RBV, RBT with the VRIO framework applicable to the country brand and its related concepts, such as shared brands, is considered by analyzing the attributes (valuable, rare, imperfectly imitable and imperfectly replaceable) based on the internal resources of each country (Zhang and Su, 2009). From the VRIO model of Barney (1991, 1995) and analyzing the applicability of this model to the specific case of shared brands, it is important to make here a specific adaptation in this original model, calling it VRIA. Thus, the "O" of organization is replaced by the "A" of association as for the legal existence of shared brands and according to the definition proposed by Tregear and Gorton (2009). 2.2 Shared brands and sustainable competitive advantage In Brazil, the understanding of shared brands considers GIs, collective brands and sector brands, which can occur in isolation or at the same time. GIs are divided into two types - indications of origin (IO) and denominations of origin (DO). For IO, the reputation of the product is analyzed. In DO, the peculiarities of that region must affect the final result of the product, in an identifiable and measurable way. A food or agricultural product recognized as a denomination of origin must be produced, processed, manufactured or industrialized in its territory of origin, which is not necessarily a rule when it comes to indication of origin [Barbosa et al., 2013, 2014; Instituto Nacional da Propriedade Industrial (INPI), 2014; Junger, 2014; Servico Brasileiro de Apoio as Micro e Pequenas Empresas (SEBRAE) and Instituto Nacional da Propriedade Industrial (INPI) 2011, 2014a, 2014b, 2014c; Seminario Internacional de Indicacoes Geograficas e Marcas Coletivas, 2014 (verbal information)]. For wines, there are six Brazilian indications of origin: Vale dos Vinhedos, Pinto Bandeira, Vale da Uva Goethe, Altos Montes, Monte Belo and Farroupilha [Servico Brasileiro de Apoio as Micro e Pequenas Empresas (SEBRAE) and Instituto Nacional da Propriedade Industrial (INPI) 2011, 2014b; Instituto Nacional da Propriedade Industrial (INPI) 2014, 2015]. In addition, two GIs are under development, namely, Campanha and Vale do Submedio do Sao Francisco (Empresa Brasileira de Pesquisa Agropecuaria - Uva e Vinho (EMBRAPA), 2014). Collective brands, according to industrial property law, are used to identify products or services coming from members of a certain entity [Barbosa et al., 2013; Servico Brasileiro de Apoio as Micro e Pequenas Empresas (SEBRAE) and Instituto Nacional da Propriedade Industrial (INPI), 2014c]. As for the sector brand, there is a diversity of projects and work fronts developed by APEX, promoting the Brazilian industry. One type of share brand that can be classified as a GI is the controlled denomination of origin (DOC). Marchinia et al. (2014) show that the higher number of firms and the investments in marketing contributes to the collective reputation of the DOC. Similarly, the older the DOC designation, the greater its reputation because of the accumulation effect of investments in marketing and consumer experiences. Product differentiation within the DOC areas (subzones, variants of the product, the presence of sweet wine, etc.) represents a growth factor for the reputation of the DOC. Despite the great number of criticisms to the system of DOC wines, the presence of shared and mandatory rules resulted in a collective value that can support SMEs in building their own individual reputation. Olmos (2011), in a study of the appellation of origin (DOC) in the Rioja wine sector, showed that process innovations are not a determinant of internationalization. However, the results show a positive and significant link between a firm's marketing activities and its level of internationalization. The positive effect of advertising in determining the degree of internationalization confirms that wineries in this region can benefit from the collective umbrella brand "qualified denomination of origin" in comparison with wineries in other wine-producing regions. Defrancesco et al. (2012) shows that the protected GIs can constitute a strategic tool for wine producers and it's possible to a strategic differentiation of New World wine producing countries. Fensterseifer and Rastoin (2013) analyzed the contribution potential of the resources of a cluster to generate competitive advantage for local companies in the wine sector. The benefits obtained when companies participated in the cluster were classified from an RBV perspective in the following categories: institutional, specialization, social, reputation and natural capital. Thus, a set of potentially strategic resources for the Serra Gaucha wine cluster was identified. This study was based on a qualitative approach that, for Bauer and Gaskell (2000), aims to understand in detail beliefs, attitudes, values and motivations regarding the behavior of people in specific social contexts (Malhotra, 2006). The research was exploratory and the field study was conducted with in-depth interviews (Bardin, 2011; Cooper and Schindler, 2014; Sampieri et al., 2010). This work was delimited to the study of the viniculture sector. This sector was chosen, as it was possible to analyze the shared brands (GI, collective brand and sector brand), according to studies developed in Brazil (Barbosa et al., 2013, Bruch et al., 2014, Castro and Giraldi, 2015). In the international wine market, considering the past five years, the sparkling wine continued to grow with an annual rate of 7 per cent in value and 6 per cent in volume and bulk wine with an annual rate of 6 per cent in value and 3 per cent in volume. Over the same period, bottled wine grew 5 per cent in value but only 1 per cent in volume in terms of annual rate. Forecast of the future evolution of wine market is challenging and risky in a context dominated by discontinuities, and there is not much research on how markets evolve (Pomarici, 2016). In addition, the Brazilian wine sector is important to the national economy, being the first to obtain a GI in the country, a concept similar to DOC - controlled designation of origin used in Italy, for example (Barbosa et al., 2013). The Brazilian wine sector has already six certified regions, namely, Vale dos Vinhedos, Pinto Bandeira, Monte Belo do Sul, Flores da Cunha, Farroupilha and Urussanga and some collective brands such as ACAVITIS, CPEG and Vinhateiros do Vale and also sector brands (Vinhos do Brasil/Wines of Brasil). The work was delimited to the study of the fine wines that have GIs and are located in the Brazilian state of Rio Grande do Sul. In terms of numbers, the Brazilian viniculture sector is the 19th ranked country in terms of area cultivated with grapes, 12th in grape production and 13th in wine production. From 2004 to 2014, there was a 100 per cent increase in the national production of fine wines, from 25 million to 50 million L (Copello, 2015). The following standard terminology was used for codification: "G" for government agencies, "A" for associations and "W" to represent winery managers followed by numbers in the order in which the interviews were conducted. Interviews were conducted with government agencies such as SEBRAE, MAPA (Ministry of Agriculture, Livestock and Supply), INPI and APEX (Brazilian agency for promotion of exports and investments) represented by IBRAVIN (G1, G2 e G3) and EMBRAPA UVA E VINHO (Embrapa Grape and Wine) (G4) to understand in detail the processes of creation and registration of shared brands. The interviews were conducted in May 2015 and each had an average duration of 40 min. Subsequently, primary data were also collected through interviews with the associations that manage GIs and the collective brands of fine wines, such as APROVALE (Association of Fine Wine Producers of the Vinhedos Valley - A1 e A2); ASPROVINHO (Association of Wine Producers of Pinto Bandeira - A3); APROMONTES (Association of Wine Producers of Altos Montes - A4); APROBELO (Association of Viticulture Farmers of Monte Belo do Sul - A5); and AFAVIN (Association of Wine, Sparkling Wine, Juices and Derivative Producers of Farroupilha - A6); CPEG (A7); AVIGA (Association of Wine Producers of Garibaldi - A8); SINDIVINHORS (Syndicate of the Industry of wine, grape must, vinegar and beverages derived from the grape, and wine of the state of Rio Grande do Sul - A9); and the City Hall of Garibaldi (A10) together with the Secretary of Tourism. Ten interviews were conducted with an average duration of 40 min each. For data collection, six wine-growing municipalities of the State of Rio Grande do Sul were visited: Bento Goncalves, Monte Belo do Sul, Flores da Cunha, Farroupilha, Pinto Bandeira and Garibaldi. Finally, the managers of the wineries were interviewed, with 18 interviews, namely, APROVALE - Don Laurindo (W1), Lidio Carraro (W2, W3), Miolo (W4), Peculiare (W5) and Pizzato (W6); ASPROVINHO - Cooperativa Aurora (W7, W8) and Valmarino (W9); APROBELO - Fae (W10) and Fantin (W11); APROMONTES - Mioranza (W12); CPEG/AVIGA - Batistello (W13), Cooperativa Garibaldi (W14, W15), Milantino (W16), Pedrucci (W17) and Vaccaro (W18). The interviews had an average duration of approximately 40 min each, with about 12 h of testimonials. With this number of interviews, a saturation point was reached, in which the researcher, according to Bauer and Gaskell (2000), evaluated that no new surprises, perceptions or revelations will appear in additional interviews. The wineries interviewed belong to five producing regions in the Rio Grande do Sul (Brazil). The process began with the planting of grapes with the arrival of Italian immigrants in 1847 and only in the 1990s they underwent a process of remodeling, culminating in the current Brazilian winemaking format (W6). Besides the cultivation, producers have started to vinify and to commercialize the products in the internal and external market. The interviewees are large, medium and small, possessing 900, 30 and 8.5 hectares of vineyards of their own, on average. They produce an average of 20 million L/year for the largest, 197,600 L/year for the average wineries and 27,500 L/year for the small wineries interviewed. The following proposition was proposed for this study: shared brands provide SCA, according to the VRIA, in the opinion of the government agencies, associations and managers of the wineries (Barney, 1986, 1991, 1995; Dierickx and Cool, 1989). This proposition is based on the work of Barney (1986, 1991, 1995) and Dierickx and Cool (1989) who postulate that SCA arises when competitors are unable to obtain the same results, even though they use similar marketing strategies (Bharadwaj et al., 1993; Von Krogh and Roos, 1995). For data analysis, content analysis was used, which is a set of communication analysis techniques, in which the analyst delimits the codification or registration units (Bardin, 2011). Categorization consists in organizing the data so that the researcher can make decisions and draw conclusions from them (Bardin, 2011; Cooper and Schindler, 2014). The proposal of this article is composed of four categories denominated C1, C2, C3 and C4, representing the VRIA framework. 4.1 C1 - valuable The valuable category (C1) refers to the benefits, advantages and factors that lead wineries to differentiate themselves from their competitors. The shared brand, being valuable, adds value, becomes distinct and differentiates itself in the minds of consumers (Balmer and Gray, 2003; Barney, 2001). In this way, the interviewees who have GI understand that it adds value and brings visibility to both large- and small-sized wineries. It ends up generating a greater interest of the consumer in knowing that region, fomenting tourism and adding value for the whole chain [O4, Servico Brasileiro de Apoio as Micro e Pequenas Empresas (SEBRAE) and Instituto Nacional da Propriedade Industrial (INPI), 2014c]. The producers understand that with the certificate, they have gained publicity in the media, and Brazilian wine has become more recognized, as can be seen in the following excerpt: It adds a lot of value, brings greater competitiveness in the internal and external market. The issue of the reputation of Vale dos Vinhedos, this intrinsic value of quality, of the region, the culture from here, its own identity, distinction from similar wines. This perception of the brand leads to "this product is more expensive, it has DO, so I think it is a good product" (W4). The IO of Vale dos Vinhedos was the first Brazilian IO and emerged soon after the opening of Brazil to the international market. At that moment, there were changes in the competitive scenario and Brazilian fine wine began to gain space in the international market, presenting itself as a valuable product. The emergence of Vale dos Vinhedos DO - the first DO of Brazilian wines - occurred in 2012 and did not have as much impact as the launch of IO in 2002. This is because of the fact that IO was a novelty and also because DO is more restrictive, covering fewer wineries and fewer certified products (W6): Actually, this is it, we are a smaller winery here, I do not have, let's say, market visibility as a bigger company, I do not have the marketing power to invest in it. When I have a designation of origin, so to speak, the consumer sees my product with different eyes, he does not know the individual brand, but he knows what a designation of origin is, so he can position the product in an easier way. This is a strength to position yourself in the market much greater than if you were individual (W5). They believe that DO is already more traditional overseas such as in Europe, for example, and this facilitates the understanding that it is a wine with production and quality mainly controlled for the consumer who is already accustomed to imported wines of a high standard. They recognize how having DO is a valuable resource for the producing company and for Brazil as the country of origin (W4, W5, W12, W16). The smaller wineries, whose production is practically handcrafted and family operated, denominated as boutique wineries or micro champagne enterprises do not produce enough volume to distribute their products in supermarket chains (W17). It is noteworthy that Vale dos Vinhedos GI was recognized by the European Union in 2007 and represented a step toward visibility in the external market. Before, the wines that went to Europe went as ordinary wines: From the recognition by the European Union, the wines of Vale dos Vinhedos had opened the doors to place their labels. The GI seal gives both region and the products a good reputation (A1, A2, A3, A6). In this sense, the interviewee A10 affirms that CPEG guarantees the quality of the product and this generates recognition by the consumer himself and respectability for the producing region and adds value to the product. This idea is complemented by A9 who believes that this recognition will come through the dissemination and promotion of products. With regard to the sector brand, one interviewee reports: Everyone who works with wines, e.g. journalists who are directly linked to wine, already knows that Brazil produces wine. Now, another step is to make the final consumer know what Brazil produces. The end consumer still does not know that Brazil produces wines, and good wines (G1). Another factor that adds value is the issue of the distribution of products in specialized stores abroad, which counts on a trained employee to perform a personalized service, generating competitive differential (G1, W6). Respondents believe that being part of the Wines of Brazil project and exporting the products adds value, meaning that the product becomes more valuable in the face of international competition (G3, G4, W3). This result relates to Defrancesco et al.'s (2012) study, which demonstrated the importance of GIs as a quality signal that makes consumers willing to pay a premium for geographical names. It is important to mention that this added value may be better perceived by wine enthusiasts than by novices. As Barber et al. (2008) and Dodd et al. (2005) indicate that because of the different needs and degrees of knowledge about wines and related certifications, it may be important for Brazilian wineries to better understand their consumers and provide proper information about the certifications to increase their wines perceived value. In this way, the initiatives of shared brands, whether GI, collective brand or sector brand, add value to the products and differentiate the wineries nationally and internationally. This result found among the interviewees is in agreement with the literature that, when developing valuable resources, the company gains a competitive advantage, as it adds value to the products and manages to keep up with the environmental changes of the competitive market (Barney, 1995; Ray et al., 2004). To illustrate the results schematically, the conceptual map is presented in Figure 1. 4.2 C2 - rare The Rare category (C2), according to Barney (1991, 1995), refers to a differential for the company, in addition to having valuable resources. When it comes to GIs, this rare factor is found in Brazilian legislation and its regulations, which, being detailed and with more requirements, resembles the European model. This differentiates Brazilian wine and brings rarity when compared to the international competitors of the so-called New World (Argentina, Chile, the USA, Australia, South Africa and New Zealand). In general, for the New World, only one delimitation with a geographical name is made and it must own about 65 to 80 per cent of the grape produced in this territory, as well as its winemaking. While for DO in Brazil, 100 per cent of the grape must be produced and vinification performed in the defined region (G4). In this way, the European consumer understands what a GI is, recognizes that the product has differentiation criteria and this, besides adding value, brings rarity to the products. The GI is like a "calling card", informing the consumer of its characteristics: region, origin, history and culture of the product (G4, A6). This result relates to Alonso and Bressan's (2016) study, which suggested that the accumulated knowledge of the team provides the rare attribute and history, and the reputation and territory relate to the imperfectly imitable/replaceable attribute. It also finds support in Wilk and Fensterseifer's (2003), who demonstrated that the wine cluster located in southern Brazil has achieved a high level of resources and capabilities that can give a competitive advantage to this cluster. Respondents say that the consumer generally does not remember the individual brand of wine he has taken, but rather the country of origin of that product: You do not remember a brand of wine you took, the last foreign wine you took will remind you that you took a Chilean wine, you will remember you took a French wine, the brand that the consumer remembers is often the origin, it is not the brand of the product itself (G3). For the global competition, Vale dos Vinhedos is a block that works together for the dissemination of Brazilian wine, showing product differentials. They report that the consumer already has a perception that Vale dos Vinhedos represents certified quality products. However, within the Vale itself there are 22 wineries that also compete with each other, as the tourist chooses among the existing options, and then there is the work to obtain a competitive differential for the individual brand, as well as the collective actions for the competitiveness of the region internationally (W4, W6). The region of Pinto Bandeira, for example, has as main focus the traditional sparkling or methode champenoise (champenoise method). This product is rare because it differs from other producing regions and has its own flavor because of altitude and soil conditions for the production of grapes. And through the traditional method, there is a more careful and time-consuming process in the elaboration that also adds value and provides rarity to the sparkling wine (A3). Similarly, the reality lived in Garibaldi also shows that the rarity of the product lies in the fact that the sparkling wine produced is differentiated and Brazil is already being recognized, especially the region of Serra Gaucha, as a region favorable for the production of sparkling wines (A7, A8, A9). The proposal of the sparkling wine, for example, is more refreshing and seeks a different style from the French, which provides a more structured and fuller-bodied wine (W3, W4). In this way, shared brands, whether GI, collective brand or sector brand, have aspects of the rarity in relation to current and potential competition. This makes it difficult for other companies that wish to implement equivalent strategies (Barney, 1991, 1995). Therefore, for the agencies, associations and managers of the wineries, the shared brand resource is rare because in addition to being valuable, it has conditions that are not widely disseminated among the competitors in the market. The results presented are illustrated in a conceptual map in Figure 2. 4.3 C3 - imperfectly imitable/replaceable The category imperfectly imitable/replaceable (C3) refers to the resource that is difficult to be copied by the competition and has no substitute equivalents. Because they are unique resources, it is difficult for another company to be able to construct the same types of resources (Balmer and Gray, 2003; Barney, 1995; Hoffman, 2000; Iamin, 2007; Porter, 1996; Teece, 2000). Grape production and wine production are difficult to copy, being imperfectly imitable/replaceable because they depend on specific characteristics such as grape acidity, altitude and terroir that are termed as Brazilian terroirs. The quality of this grape cannot be copied and is unique in each region (W4, W6, W11, W13, W14, W15, A3). This issue of terroir as a factor that hinders imitability was also shown in the work of Sharp and Smith (2007), who studied the Champagne region in France and visualized that the territory, climate and subsoil are unique factors of each region. This was also corroborated by Rubini et al. (2013), who pointed out a strong relationship between territory and quality influenced by the geographical origin of the products. In addition, the winemaking process is also unique because each winemaker has a style and adds a personal touch to the process. Each winemaker will make a combination and bring their knowledge and style to the final quality of the wine (W4, W6, W11, A3, W16, W18). There is also the vocation of each region, that is, there are always some grape varieties that fit better in each region. Therefore, the local vocation factor is a competitive differential and also difficult to be imitated or replaced (W3, W6). This idea was described as follows: Vale dos Vinhedos, the main focus is the Merlot wine, here (Pinto Bandeira) the bond is more the sparkling wine, because this region is more prone to the sparkling, here it would be Champenoise sparkling wine from Chardonnay, Riesling and Pinot Noir grapes. In turn, if you take Farroupilha it is a propitious region for Moscatel - sparkling wine from Moscatel. So each region has a context, it has a focus (A3, A6). In addition, American grapes that are not suitable for the production of fine wines are used for the manufacture of grape juice, which has had a great prominence in the international market mainly in the USA (A3, A5, A6, A7, W11). Thus, when analyzing the value delivered to consumers in terms of attributes, benefits and attitudes of shared brands, whether GI, collective brand or sector brand, it is understood that they have imperfectly imitable/replaceable aspects with respect to the competition (Balmer and Gray, 2003; Barney, 1991, 1995; Srivastava et al., 2001). Therefore, for the agencies, associations and managers of the wineries, the shared brand resource is imperfectly imitable/replaceable because in addition to being valuable and rare, it has conditions that make it difficult for competitors to imitate and substitute equivalent resources. In this way, for the VRIA model, a valuable, rare and imperfectly imitable/replaceable resource generates a SCA and performance is above normal in relation to the market average (Barney, 1991, 1995; Barney and Wright, 1998). The results described here are presented in the conceptual map of Figure 3. 4.4 C4 - association The category Association (C4) refers to the equivalent of the letter O in the VRIO framework proposed by Barney (1991, 1995), which represents the organization. In the original form, the model was elaborated for the analysis of an individual organization. However, it is known that shared brands are composed by collectivity, and the trademark registration is granted to a collective entity (Barbosa and Regalado, 2013; Tregear and Gorton, 2009). Thus, for this thesis, the "O" of the organization is replaced by "A" for association, and VRIO will be called VRIA. In the case of the Vale dos Vinhedos region, the wineries are associated with APROVALE who represents a great concentration that generates greater visibility and tourist attraction. These benefits are linked to the growth of the region, the knowledge acquired by the wineries that are technology leaders and the tourist visits that have increased each year (A1, A2, W3, W6). In Pinto Bandeira, the producers are linked to ASPROVINHO. The association has undertaken work to strengthen the region's tourism issue to make it more known and visited. They recognize that the Vale dos Vinhedos was the pioneer and is already consolidated, and that for other regions the disclosure will be slow and gradual, looking for ways that the consumer/tourist is interested in knowing other regions beyond Vale dos Vinhedos. The region is certified with IO. They understand that IO is a stimulus that is making a difference, but it is a tool that is used in conjunction with grape quality and the vinification process among other details (A3, W7, W8, W9). Producers in the Flores da Cunha region are associated with APROMONTES. The work of the association is still incipient and needs to be improved. As reported, there is a lack of further integration by the members and also a team with training in the area of management (A4, W12). In the case of Monte Belo do Sul, there is the APROBELO association. The work developed is still recent, the association does not have headquarters and has associates who participate, but do not have the interest in the indication of origin. These producers intend to continue to produce larger sized bottles of wine and grape juice and will not invest in the production of fine wines (A5, W10). In the Farroupilha region, the producers are linked to AFAVIN. The association carries out publicity actions, event organizations and institutional partnerships. It also works in the promotion of local festivals, as well as participation in international competitions. In Garibaldi, the consortium of CPEG as an entity works toward the objectives of the collectivity, mainly so that the seal of the products is divulged and recognized by consumers as a high-quality product. At present, the initiatives are still recent and there is no specific measurable financial result of CPEG (A7). In relation to the sector brand project, management is carried out by IBRAVIN with the support of APEX. For the international market, respondents believe that Brazilian wine is still seen as an exotic product, but with the work of dissemination and participation in fairs, there is an improvement in the perception of the product (A2). The interviewees understand that participation in fairs should be undertaken collectively, as the wineries are constituents of the so-called "Brazil Block". Producers see themselves as competitors, but to achieve international visibility, they need to work together and in partnership, as they know that individually they would not have resources for such investment (W1, W2, W3, W6). Regarding the domestic market, it was emphasized that there is a need for more public policies and other efforts for the Brazilian consumer to recognize national wine. There are already some actions in this direction carried out by SEBRAE (G4). Thus, for the agencies, associations and managers of the wineries, the shared brand resource has the support of the Association. In this way, the shared brands resource can be considered valuable, rare, imperfectly imitable/replaceable and has the organizational process structured for the management of these brands (Barney, 1991, 1995; Barney and Hesterly, 2006; Barney and Wright, 1998; Grant, 1991; Ito and Gimenez, 2011; Shafeey and Trott, 2014; Sirmon et al., 2011) in the market in question, thus forming the VRIA model. Considering the results presented, it can be concluded that the GIs, the collective brand and the sector brand meet all the VRIA model requirements, in the opinion of the interviewees, creating barriers for imitation by competitors and difficulties in implementing similar strategies in the market (Bharadwaj et al., 1993; Hoffman, 2000; Penrose, 1979; Von Krogh and Roos, 1995). It can be concluded that the proposition that shared brands - GI, collective brands and sector brands - provide SCA, according to the VRIA, can be confirmed, thus fulfilling the four conditions that form the acronym VRIA - valuable, rare, imperfectly imitable/replaceable and association. The value added to the product through the information contained in the shared brands facilitates the establishment of a relationship of trust between producer and consumer, being a source of competitive advantage. When studying the theory about shared brands, this work contributes to the understanding of the concept that the GI, the collective brand and the sector brand can be denominated as types of shared brands from the existent literature. Barney's VRIO framework (Barney, 1991, 1995) hitherto thought for individual companies, has the letter "O" of Organization replaced by the letter "A" of Association, becoming VRIA. In this way, the existence of a shared brand implies in the creation of an association that becomes the owner of the brand. Another theoretical contribution of this research is the deepening of studies on shared brands in Brazil and in the world, as it is still a subject little studied because it is of recent implementation in the market (Barbosa, 2011; Bruch et al., 2014; Castro and Giraldi, 2015; Regalado et al., 2012). 5.1 Practical implications and future research recommendations The practical contributions come from the fact that government agencies, associations and wineries can improve the production process and seek certified products for commercialization in the domestic and foreign markets. These contributions may also, in practice, be used by other sectors and countries. Specifically, it contributes to the fact that government agencies understand the practical result obtained from the management effort in the creation of shared brands. For the associations, the contribution is in improving internal processes and providing advice to members to obtain best practices and results. And finally, for wineries, it points out how much the collective effort in search of shared brands tends to make each individual company more competitive with a source to obtain SCA. Future research studies could be carried on with other wine-growing regions in regions that are beginning to develop the shared brand as in the state of Santa Catarina and Northeast Brazil. A comparative study can also be carried out with other countries of the New World of Wine, such as Argentina, Chile and Uruguay. We also suggest that forthcoming studies consider analyzing the consumers' perspective of the perceived value of wine certifications and the wineries strategies to increase awareness of such added value. 5.2 Limitations The results of qualitative researchers are not reproducible, as the researchers' interpretation is susceptible to changes, as well as the fact that they are specific and unique situations in time for that region or research unit studied, which hinders the generalizations of the study.
|
It was concluded that shared brands in the Brazilian wine sector can be considered a source of sustainable competitive advantage according to the resource-based view.
|
[SECTION: Value] A brand can be seen as a strategic asset that helps a company to be more competitive. In the same way that companies invest in brands, countries can also be seen as such (Anholt, 2005; Huang and Tsai, 2013; Kotler et al., 2006). In considering the role that the image of a country can play in a buyer's behavior, constructs such as the country's brand, the country's image and the country of origin may be attributes that offer the potential for companies to achieve a sustainable competitive advantage (SCA), both in the internal and external market (Baker and Ballington, 2002; Hakala et al., 2013). In this context, based on a country brand, the so-called shared brands may be worked. They have been defined by Tregear and Gorton (2009) as having a single brand as an identity for two or more partner companies, which commit to respect the established agreements, and the management of this shared brand must be performed by an independent entity that represents and responds for the brand. In this way, this article deals with different types of shared brands that fit into this conceptual condition presented by Tregear and Gorton (2009), being geographical indications (GIs), collective brand and sector brand. GIs constitute an instrument for the valuation of goods that are territorially distinguished, that is, they are products coming from a certain region that are recognized for their quality. Collective brands are awarded to a group of organizations that market the same products and wish to link quality to associated companies that do not have to be produced in a delimited geographical region and their registration does not necessarily have to have a geographical name (Castro and Giraldi, 2015). The sector brands are projects developed with the purpose of international representation of the sector through export mechanisms and promotions (Castro and Giraldi, 2015; Lourencao and Giraldi, 2017). Defrancesco et al. (2012) demonstrated the importance of GIs, confirming that foreign consumers are willing to pay a premium for geographical names. They concluded that widely known quality signals, such as protected GIs, would increase the market access of foreign wines by enhancing consumers to recognize and discriminate among terroir-related quality attributes. However, it has to be considered that there are wine enthusiasts and novice consumers, each segment with different needs and degrees of knowledge about wines and the related certifications. It is important that wineries and restaurants be able to understand these differences and devise specific marketing strategies for each audience (Barber et al., 2008). Managers of these establishments need to understand the importance of training to assist the consumer in the purchasing decision process (Dodd et al., 2005). There are also environmental aspects to be considered when analyzing the consumer intention to pay more for certain types of wines. For example, Barber (2010) mentions the expansion of the search for organically or biodynamically produced wines and the consequent reorientation of the blend through repackaging and re-labeling, promoting this in the product dissemination strategy. Barber (2010) suggests that wineries have to focus on developing a work of increasing awareness that organic wine has no lower quality than other traditionally wines (Barber et al., 2010). In this regard, the SCA construct can be defined as that unique attribute that is difficult to be acquired or copied by the competition, which arises from the impossibility of competitors implementing value creation strategies simultaneously (Barney, 1991). Within the SCA theme is the resource-based view (RBV) developing valuable resources and capabilities, which are interchangeable and refer to the tangible and intangible assets that the company uses to develop its strategies (Ray et al., 2004). The study of Vlachveia et al. (2012) in Greece shows that producers invest intangible elements with affect consumer choices. They used a wine tourism and wine routes with an important tool for marketing. They used holistic actions with an important tool for marketing rather than a brand just to a name or a label on the bottle. For this, the wineries used a wine tourism and wine routes. Galati et al. (2014) identified the firms' internal factors that most influence the export intensity and the ability to fit into the international market of the Italian wineries producing sparkling wines through a RBV approach. In particular, the main factors affecting export intensity are innovation degree and intensity of investment in advertising and promotion, among firms' characteristics, and abroad experience, proficiency of foreign languages, educational level and perception barriers, among managers' characteristics, are strategic to ensure the survival of firms into markets and their success. In turn, Amadieu and Viviani (2010) appoint that, for French wine companies, spending money on intangibles is not a guarantee of financial success. The lack of organizational innovation, organizational responsiveness or knowledge-related resources could explain the absence of a positive impact for intangible expenses on performance for French companies in the wine sector. To be financially successful, intangible expenses should be accompanied by organizational and managerial changes, and these changes require time to be learned. Giacomarra et al. (2016) identified that internal motivations drive entrepreneurs to adopt voluntary standards and that wineries adopting voluntary standards exhibit better economic performance than non-certified wine firms. These results suggest that although the adoption of voluntary quality certifications does not have a direct effect in terms of a labor productivity increase, it entails a business management improvement. The work of Alonso and Bressan (2016) shows that it is possible to associate RBV with the strenghts, weakness, opportunities and threats analysis, which is a complementary tool. The accumulated knowledge of the team provides the rare attribute and history, reputation and territory related to the imperfectly imitable/replaceable attribute. Maurel (2009) complements this idea regarding the impact of decision-maker characteristics on export performance by indicating a lack of significant impacts on export performance. Suarez-Ortega and A'lamo-Vera (2005) confirmed that factors influencing export involvement are not the same along the process of export development. Three aspects of export development have been analyzed: intention, propensity and intensity. The effect of industry-specific characteristics on internal export factors has been isolated through the selection of one industry in one country for the empirical research. The study of Wilk and Fensterseifer (2003) was conducted having as its central purpose the identification of the strategic resources and capabilities of a cluster located in southern Brazil and their sustainability that history matters and the cumulative development and upgrading of resources and capabilities can fundamentally shape the future strategic options of a cluster. Although these papers have approached the RBV, when dealing with internal resources, they have not explored and classified the foundations for obtaining a SCA, such as the valuable, rare, imitability/replaceability and organization (VRIO) framework. The seminal work of the RBV was that of Barney (1991), which culminated in the VRIO framework (Barney, 1991, 1995). Two decades later, Barney et al. (2011) show that the theory is mature and some researchers prefer to use the term resource-based theory (RBT) instead of RBV, as they believe that it is currently a consolidated theory and not just a vision (Barney et al., 2011; Newbert, 2007). In analyzing the shared brands and the perspective of SCA, the objective of this article is defined as to investigate if shared brands provide SCA according to the VRIO model adapted to the analyzed context (Brazilian winemaking sector), in the opinion of the government agencies, associations and managers of the wineries. 2.1 Sustainable competitive advantage The SCA arises from the development of a set of features and abilities that cannot be negotiated, imitated or replaced, or implemented simultaneously by current and potential competitors, and use of resources within the company is difficult to access or replicate by competitors (Bharadwaj et al., 1993; Dierickx and Cool, 1989; Hoffman, 2000; Iamin, 2007; Porter, 1996; Teece, 2000; Von Krogh and Roos, 1995; Wernerfelt, 1984). The work of Barney (1986) introduced the position of heterogeneous resources, which is an important theoretical contribution to RBV. The immobility of resources prevents other firms from acquiring or imitating resources easily, making it difficult to make resources unique in the market (Bharadwaj et al., 1993; Hoffman, 2000; Penrose, 1979; Von Krogh and Roos, 1995). Based on the principles of heterogeneity and resource immobility, which is the basis for RBV, Barney (1991) created the VRIS framework (valuable, rare, imperfectly imitable, and imperfectly substitutable) or, for some authors, VRIN (valuable, rare, imperfectly imitable and non-replaceable). The acronyms represent the same idea and differ in the way of explicit terminology with S in VRIS and with N in VRIN, representing, in both, the imperfectly replaceable attribute (Shafeey and Trott, 2014). The VRIS/VRIN model is revised by Barney (1995) and becomes the VRIO framework, in which the imperfectly imitable and imperfectly replaceable attribute is added with the letter "I" in the framework, summing them into the issue of imitation difficulty and adding the item organization, which discusses whether the company is organized to exploit its resources and capabilities (Barney, 1995; Barney and Wright, 1998; Shafeey and Trott, 2014). The organizational process such as the action of managers in the form of structuring the resources and managing them in the market provides the fourth necessary condition to obtain SCA, thus forming the VRIO (Balmer and Gray, 2003; Barney, 1991, 1995; Barney and Hesterly, 2006; Cazurra and Un, 2015; Grant, 1991; Ito and Gimenez, 2011; Shafeey and Trott, 2014; Sirmon et al., 2011; Srivastava et al., 2001). Nowadays, RBV studies are more concise and in the process of maturation, which leads some authors to substitute the term "RBV perspective" for RBT, according to Barney et al. (2011), Crook et al. (2008), Maritan and Peteraf (2011), McWilliams and Siegel (2011) and Molloy et al. (2011). Based on these trends and their relationship with the RBV, RBT with the VRIO framework applicable to the country brand and its related concepts, such as shared brands, is considered by analyzing the attributes (valuable, rare, imperfectly imitable and imperfectly replaceable) based on the internal resources of each country (Zhang and Su, 2009). From the VRIO model of Barney (1991, 1995) and analyzing the applicability of this model to the specific case of shared brands, it is important to make here a specific adaptation in this original model, calling it VRIA. Thus, the "O" of organization is replaced by the "A" of association as for the legal existence of shared brands and according to the definition proposed by Tregear and Gorton (2009). 2.2 Shared brands and sustainable competitive advantage In Brazil, the understanding of shared brands considers GIs, collective brands and sector brands, which can occur in isolation or at the same time. GIs are divided into two types - indications of origin (IO) and denominations of origin (DO). For IO, the reputation of the product is analyzed. In DO, the peculiarities of that region must affect the final result of the product, in an identifiable and measurable way. A food or agricultural product recognized as a denomination of origin must be produced, processed, manufactured or industrialized in its territory of origin, which is not necessarily a rule when it comes to indication of origin [Barbosa et al., 2013, 2014; Instituto Nacional da Propriedade Industrial (INPI), 2014; Junger, 2014; Servico Brasileiro de Apoio as Micro e Pequenas Empresas (SEBRAE) and Instituto Nacional da Propriedade Industrial (INPI) 2011, 2014a, 2014b, 2014c; Seminario Internacional de Indicacoes Geograficas e Marcas Coletivas, 2014 (verbal information)]. For wines, there are six Brazilian indications of origin: Vale dos Vinhedos, Pinto Bandeira, Vale da Uva Goethe, Altos Montes, Monte Belo and Farroupilha [Servico Brasileiro de Apoio as Micro e Pequenas Empresas (SEBRAE) and Instituto Nacional da Propriedade Industrial (INPI) 2011, 2014b; Instituto Nacional da Propriedade Industrial (INPI) 2014, 2015]. In addition, two GIs are under development, namely, Campanha and Vale do Submedio do Sao Francisco (Empresa Brasileira de Pesquisa Agropecuaria - Uva e Vinho (EMBRAPA), 2014). Collective brands, according to industrial property law, are used to identify products or services coming from members of a certain entity [Barbosa et al., 2013; Servico Brasileiro de Apoio as Micro e Pequenas Empresas (SEBRAE) and Instituto Nacional da Propriedade Industrial (INPI), 2014c]. As for the sector brand, there is a diversity of projects and work fronts developed by APEX, promoting the Brazilian industry. One type of share brand that can be classified as a GI is the controlled denomination of origin (DOC). Marchinia et al. (2014) show that the higher number of firms and the investments in marketing contributes to the collective reputation of the DOC. Similarly, the older the DOC designation, the greater its reputation because of the accumulation effect of investments in marketing and consumer experiences. Product differentiation within the DOC areas (subzones, variants of the product, the presence of sweet wine, etc.) represents a growth factor for the reputation of the DOC. Despite the great number of criticisms to the system of DOC wines, the presence of shared and mandatory rules resulted in a collective value that can support SMEs in building their own individual reputation. Olmos (2011), in a study of the appellation of origin (DOC) in the Rioja wine sector, showed that process innovations are not a determinant of internationalization. However, the results show a positive and significant link between a firm's marketing activities and its level of internationalization. The positive effect of advertising in determining the degree of internationalization confirms that wineries in this region can benefit from the collective umbrella brand "qualified denomination of origin" in comparison with wineries in other wine-producing regions. Defrancesco et al. (2012) shows that the protected GIs can constitute a strategic tool for wine producers and it's possible to a strategic differentiation of New World wine producing countries. Fensterseifer and Rastoin (2013) analyzed the contribution potential of the resources of a cluster to generate competitive advantage for local companies in the wine sector. The benefits obtained when companies participated in the cluster were classified from an RBV perspective in the following categories: institutional, specialization, social, reputation and natural capital. Thus, a set of potentially strategic resources for the Serra Gaucha wine cluster was identified. This study was based on a qualitative approach that, for Bauer and Gaskell (2000), aims to understand in detail beliefs, attitudes, values and motivations regarding the behavior of people in specific social contexts (Malhotra, 2006). The research was exploratory and the field study was conducted with in-depth interviews (Bardin, 2011; Cooper and Schindler, 2014; Sampieri et al., 2010). This work was delimited to the study of the viniculture sector. This sector was chosen, as it was possible to analyze the shared brands (GI, collective brand and sector brand), according to studies developed in Brazil (Barbosa et al., 2013, Bruch et al., 2014, Castro and Giraldi, 2015). In the international wine market, considering the past five years, the sparkling wine continued to grow with an annual rate of 7 per cent in value and 6 per cent in volume and bulk wine with an annual rate of 6 per cent in value and 3 per cent in volume. Over the same period, bottled wine grew 5 per cent in value but only 1 per cent in volume in terms of annual rate. Forecast of the future evolution of wine market is challenging and risky in a context dominated by discontinuities, and there is not much research on how markets evolve (Pomarici, 2016). In addition, the Brazilian wine sector is important to the national economy, being the first to obtain a GI in the country, a concept similar to DOC - controlled designation of origin used in Italy, for example (Barbosa et al., 2013). The Brazilian wine sector has already six certified regions, namely, Vale dos Vinhedos, Pinto Bandeira, Monte Belo do Sul, Flores da Cunha, Farroupilha and Urussanga and some collective brands such as ACAVITIS, CPEG and Vinhateiros do Vale and also sector brands (Vinhos do Brasil/Wines of Brasil). The work was delimited to the study of the fine wines that have GIs and are located in the Brazilian state of Rio Grande do Sul. In terms of numbers, the Brazilian viniculture sector is the 19th ranked country in terms of area cultivated with grapes, 12th in grape production and 13th in wine production. From 2004 to 2014, there was a 100 per cent increase in the national production of fine wines, from 25 million to 50 million L (Copello, 2015). The following standard terminology was used for codification: "G" for government agencies, "A" for associations and "W" to represent winery managers followed by numbers in the order in which the interviews were conducted. Interviews were conducted with government agencies such as SEBRAE, MAPA (Ministry of Agriculture, Livestock and Supply), INPI and APEX (Brazilian agency for promotion of exports and investments) represented by IBRAVIN (G1, G2 e G3) and EMBRAPA UVA E VINHO (Embrapa Grape and Wine) (G4) to understand in detail the processes of creation and registration of shared brands. The interviews were conducted in May 2015 and each had an average duration of 40 min. Subsequently, primary data were also collected through interviews with the associations that manage GIs and the collective brands of fine wines, such as APROVALE (Association of Fine Wine Producers of the Vinhedos Valley - A1 e A2); ASPROVINHO (Association of Wine Producers of Pinto Bandeira - A3); APROMONTES (Association of Wine Producers of Altos Montes - A4); APROBELO (Association of Viticulture Farmers of Monte Belo do Sul - A5); and AFAVIN (Association of Wine, Sparkling Wine, Juices and Derivative Producers of Farroupilha - A6); CPEG (A7); AVIGA (Association of Wine Producers of Garibaldi - A8); SINDIVINHORS (Syndicate of the Industry of wine, grape must, vinegar and beverages derived from the grape, and wine of the state of Rio Grande do Sul - A9); and the City Hall of Garibaldi (A10) together with the Secretary of Tourism. Ten interviews were conducted with an average duration of 40 min each. For data collection, six wine-growing municipalities of the State of Rio Grande do Sul were visited: Bento Goncalves, Monte Belo do Sul, Flores da Cunha, Farroupilha, Pinto Bandeira and Garibaldi. Finally, the managers of the wineries were interviewed, with 18 interviews, namely, APROVALE - Don Laurindo (W1), Lidio Carraro (W2, W3), Miolo (W4), Peculiare (W5) and Pizzato (W6); ASPROVINHO - Cooperativa Aurora (W7, W8) and Valmarino (W9); APROBELO - Fae (W10) and Fantin (W11); APROMONTES - Mioranza (W12); CPEG/AVIGA - Batistello (W13), Cooperativa Garibaldi (W14, W15), Milantino (W16), Pedrucci (W17) and Vaccaro (W18). The interviews had an average duration of approximately 40 min each, with about 12 h of testimonials. With this number of interviews, a saturation point was reached, in which the researcher, according to Bauer and Gaskell (2000), evaluated that no new surprises, perceptions or revelations will appear in additional interviews. The wineries interviewed belong to five producing regions in the Rio Grande do Sul (Brazil). The process began with the planting of grapes with the arrival of Italian immigrants in 1847 and only in the 1990s they underwent a process of remodeling, culminating in the current Brazilian winemaking format (W6). Besides the cultivation, producers have started to vinify and to commercialize the products in the internal and external market. The interviewees are large, medium and small, possessing 900, 30 and 8.5 hectares of vineyards of their own, on average. They produce an average of 20 million L/year for the largest, 197,600 L/year for the average wineries and 27,500 L/year for the small wineries interviewed. The following proposition was proposed for this study: shared brands provide SCA, according to the VRIA, in the opinion of the government agencies, associations and managers of the wineries (Barney, 1986, 1991, 1995; Dierickx and Cool, 1989). This proposition is based on the work of Barney (1986, 1991, 1995) and Dierickx and Cool (1989) who postulate that SCA arises when competitors are unable to obtain the same results, even though they use similar marketing strategies (Bharadwaj et al., 1993; Von Krogh and Roos, 1995). For data analysis, content analysis was used, which is a set of communication analysis techniques, in which the analyst delimits the codification or registration units (Bardin, 2011). Categorization consists in organizing the data so that the researcher can make decisions and draw conclusions from them (Bardin, 2011; Cooper and Schindler, 2014). The proposal of this article is composed of four categories denominated C1, C2, C3 and C4, representing the VRIA framework. 4.1 C1 - valuable The valuable category (C1) refers to the benefits, advantages and factors that lead wineries to differentiate themselves from their competitors. The shared brand, being valuable, adds value, becomes distinct and differentiates itself in the minds of consumers (Balmer and Gray, 2003; Barney, 2001). In this way, the interviewees who have GI understand that it adds value and brings visibility to both large- and small-sized wineries. It ends up generating a greater interest of the consumer in knowing that region, fomenting tourism and adding value for the whole chain [O4, Servico Brasileiro de Apoio as Micro e Pequenas Empresas (SEBRAE) and Instituto Nacional da Propriedade Industrial (INPI), 2014c]. The producers understand that with the certificate, they have gained publicity in the media, and Brazilian wine has become more recognized, as can be seen in the following excerpt: It adds a lot of value, brings greater competitiveness in the internal and external market. The issue of the reputation of Vale dos Vinhedos, this intrinsic value of quality, of the region, the culture from here, its own identity, distinction from similar wines. This perception of the brand leads to "this product is more expensive, it has DO, so I think it is a good product" (W4). The IO of Vale dos Vinhedos was the first Brazilian IO and emerged soon after the opening of Brazil to the international market. At that moment, there were changes in the competitive scenario and Brazilian fine wine began to gain space in the international market, presenting itself as a valuable product. The emergence of Vale dos Vinhedos DO - the first DO of Brazilian wines - occurred in 2012 and did not have as much impact as the launch of IO in 2002. This is because of the fact that IO was a novelty and also because DO is more restrictive, covering fewer wineries and fewer certified products (W6): Actually, this is it, we are a smaller winery here, I do not have, let's say, market visibility as a bigger company, I do not have the marketing power to invest in it. When I have a designation of origin, so to speak, the consumer sees my product with different eyes, he does not know the individual brand, but he knows what a designation of origin is, so he can position the product in an easier way. This is a strength to position yourself in the market much greater than if you were individual (W5). They believe that DO is already more traditional overseas such as in Europe, for example, and this facilitates the understanding that it is a wine with production and quality mainly controlled for the consumer who is already accustomed to imported wines of a high standard. They recognize how having DO is a valuable resource for the producing company and for Brazil as the country of origin (W4, W5, W12, W16). The smaller wineries, whose production is practically handcrafted and family operated, denominated as boutique wineries or micro champagne enterprises do not produce enough volume to distribute their products in supermarket chains (W17). It is noteworthy that Vale dos Vinhedos GI was recognized by the European Union in 2007 and represented a step toward visibility in the external market. Before, the wines that went to Europe went as ordinary wines: From the recognition by the European Union, the wines of Vale dos Vinhedos had opened the doors to place their labels. The GI seal gives both region and the products a good reputation (A1, A2, A3, A6). In this sense, the interviewee A10 affirms that CPEG guarantees the quality of the product and this generates recognition by the consumer himself and respectability for the producing region and adds value to the product. This idea is complemented by A9 who believes that this recognition will come through the dissemination and promotion of products. With regard to the sector brand, one interviewee reports: Everyone who works with wines, e.g. journalists who are directly linked to wine, already knows that Brazil produces wine. Now, another step is to make the final consumer know what Brazil produces. The end consumer still does not know that Brazil produces wines, and good wines (G1). Another factor that adds value is the issue of the distribution of products in specialized stores abroad, which counts on a trained employee to perform a personalized service, generating competitive differential (G1, W6). Respondents believe that being part of the Wines of Brazil project and exporting the products adds value, meaning that the product becomes more valuable in the face of international competition (G3, G4, W3). This result relates to Defrancesco et al.'s (2012) study, which demonstrated the importance of GIs as a quality signal that makes consumers willing to pay a premium for geographical names. It is important to mention that this added value may be better perceived by wine enthusiasts than by novices. As Barber et al. (2008) and Dodd et al. (2005) indicate that because of the different needs and degrees of knowledge about wines and related certifications, it may be important for Brazilian wineries to better understand their consumers and provide proper information about the certifications to increase their wines perceived value. In this way, the initiatives of shared brands, whether GI, collective brand or sector brand, add value to the products and differentiate the wineries nationally and internationally. This result found among the interviewees is in agreement with the literature that, when developing valuable resources, the company gains a competitive advantage, as it adds value to the products and manages to keep up with the environmental changes of the competitive market (Barney, 1995; Ray et al., 2004). To illustrate the results schematically, the conceptual map is presented in Figure 1. 4.2 C2 - rare The Rare category (C2), according to Barney (1991, 1995), refers to a differential for the company, in addition to having valuable resources. When it comes to GIs, this rare factor is found in Brazilian legislation and its regulations, which, being detailed and with more requirements, resembles the European model. This differentiates Brazilian wine and brings rarity when compared to the international competitors of the so-called New World (Argentina, Chile, the USA, Australia, South Africa and New Zealand). In general, for the New World, only one delimitation with a geographical name is made and it must own about 65 to 80 per cent of the grape produced in this territory, as well as its winemaking. While for DO in Brazil, 100 per cent of the grape must be produced and vinification performed in the defined region (G4). In this way, the European consumer understands what a GI is, recognizes that the product has differentiation criteria and this, besides adding value, brings rarity to the products. The GI is like a "calling card", informing the consumer of its characteristics: region, origin, history and culture of the product (G4, A6). This result relates to Alonso and Bressan's (2016) study, which suggested that the accumulated knowledge of the team provides the rare attribute and history, and the reputation and territory relate to the imperfectly imitable/replaceable attribute. It also finds support in Wilk and Fensterseifer's (2003), who demonstrated that the wine cluster located in southern Brazil has achieved a high level of resources and capabilities that can give a competitive advantage to this cluster. Respondents say that the consumer generally does not remember the individual brand of wine he has taken, but rather the country of origin of that product: You do not remember a brand of wine you took, the last foreign wine you took will remind you that you took a Chilean wine, you will remember you took a French wine, the brand that the consumer remembers is often the origin, it is not the brand of the product itself (G3). For the global competition, Vale dos Vinhedos is a block that works together for the dissemination of Brazilian wine, showing product differentials. They report that the consumer already has a perception that Vale dos Vinhedos represents certified quality products. However, within the Vale itself there are 22 wineries that also compete with each other, as the tourist chooses among the existing options, and then there is the work to obtain a competitive differential for the individual brand, as well as the collective actions for the competitiveness of the region internationally (W4, W6). The region of Pinto Bandeira, for example, has as main focus the traditional sparkling or methode champenoise (champenoise method). This product is rare because it differs from other producing regions and has its own flavor because of altitude and soil conditions for the production of grapes. And through the traditional method, there is a more careful and time-consuming process in the elaboration that also adds value and provides rarity to the sparkling wine (A3). Similarly, the reality lived in Garibaldi also shows that the rarity of the product lies in the fact that the sparkling wine produced is differentiated and Brazil is already being recognized, especially the region of Serra Gaucha, as a region favorable for the production of sparkling wines (A7, A8, A9). The proposal of the sparkling wine, for example, is more refreshing and seeks a different style from the French, which provides a more structured and fuller-bodied wine (W3, W4). In this way, shared brands, whether GI, collective brand or sector brand, have aspects of the rarity in relation to current and potential competition. This makes it difficult for other companies that wish to implement equivalent strategies (Barney, 1991, 1995). Therefore, for the agencies, associations and managers of the wineries, the shared brand resource is rare because in addition to being valuable, it has conditions that are not widely disseminated among the competitors in the market. The results presented are illustrated in a conceptual map in Figure 2. 4.3 C3 - imperfectly imitable/replaceable The category imperfectly imitable/replaceable (C3) refers to the resource that is difficult to be copied by the competition and has no substitute equivalents. Because they are unique resources, it is difficult for another company to be able to construct the same types of resources (Balmer and Gray, 2003; Barney, 1995; Hoffman, 2000; Iamin, 2007; Porter, 1996; Teece, 2000). Grape production and wine production are difficult to copy, being imperfectly imitable/replaceable because they depend on specific characteristics such as grape acidity, altitude and terroir that are termed as Brazilian terroirs. The quality of this grape cannot be copied and is unique in each region (W4, W6, W11, W13, W14, W15, A3). This issue of terroir as a factor that hinders imitability was also shown in the work of Sharp and Smith (2007), who studied the Champagne region in France and visualized that the territory, climate and subsoil are unique factors of each region. This was also corroborated by Rubini et al. (2013), who pointed out a strong relationship between territory and quality influenced by the geographical origin of the products. In addition, the winemaking process is also unique because each winemaker has a style and adds a personal touch to the process. Each winemaker will make a combination and bring their knowledge and style to the final quality of the wine (W4, W6, W11, A3, W16, W18). There is also the vocation of each region, that is, there are always some grape varieties that fit better in each region. Therefore, the local vocation factor is a competitive differential and also difficult to be imitated or replaced (W3, W6). This idea was described as follows: Vale dos Vinhedos, the main focus is the Merlot wine, here (Pinto Bandeira) the bond is more the sparkling wine, because this region is more prone to the sparkling, here it would be Champenoise sparkling wine from Chardonnay, Riesling and Pinot Noir grapes. In turn, if you take Farroupilha it is a propitious region for Moscatel - sparkling wine from Moscatel. So each region has a context, it has a focus (A3, A6). In addition, American grapes that are not suitable for the production of fine wines are used for the manufacture of grape juice, which has had a great prominence in the international market mainly in the USA (A3, A5, A6, A7, W11). Thus, when analyzing the value delivered to consumers in terms of attributes, benefits and attitudes of shared brands, whether GI, collective brand or sector brand, it is understood that they have imperfectly imitable/replaceable aspects with respect to the competition (Balmer and Gray, 2003; Barney, 1991, 1995; Srivastava et al., 2001). Therefore, for the agencies, associations and managers of the wineries, the shared brand resource is imperfectly imitable/replaceable because in addition to being valuable and rare, it has conditions that make it difficult for competitors to imitate and substitute equivalent resources. In this way, for the VRIA model, a valuable, rare and imperfectly imitable/replaceable resource generates a SCA and performance is above normal in relation to the market average (Barney, 1991, 1995; Barney and Wright, 1998). The results described here are presented in the conceptual map of Figure 3. 4.4 C4 - association The category Association (C4) refers to the equivalent of the letter O in the VRIO framework proposed by Barney (1991, 1995), which represents the organization. In the original form, the model was elaborated for the analysis of an individual organization. However, it is known that shared brands are composed by collectivity, and the trademark registration is granted to a collective entity (Barbosa and Regalado, 2013; Tregear and Gorton, 2009). Thus, for this thesis, the "O" of the organization is replaced by "A" for association, and VRIO will be called VRIA. In the case of the Vale dos Vinhedos region, the wineries are associated with APROVALE who represents a great concentration that generates greater visibility and tourist attraction. These benefits are linked to the growth of the region, the knowledge acquired by the wineries that are technology leaders and the tourist visits that have increased each year (A1, A2, W3, W6). In Pinto Bandeira, the producers are linked to ASPROVINHO. The association has undertaken work to strengthen the region's tourism issue to make it more known and visited. They recognize that the Vale dos Vinhedos was the pioneer and is already consolidated, and that for other regions the disclosure will be slow and gradual, looking for ways that the consumer/tourist is interested in knowing other regions beyond Vale dos Vinhedos. The region is certified with IO. They understand that IO is a stimulus that is making a difference, but it is a tool that is used in conjunction with grape quality and the vinification process among other details (A3, W7, W8, W9). Producers in the Flores da Cunha region are associated with APROMONTES. The work of the association is still incipient and needs to be improved. As reported, there is a lack of further integration by the members and also a team with training in the area of management (A4, W12). In the case of Monte Belo do Sul, there is the APROBELO association. The work developed is still recent, the association does not have headquarters and has associates who participate, but do not have the interest in the indication of origin. These producers intend to continue to produce larger sized bottles of wine and grape juice and will not invest in the production of fine wines (A5, W10). In the Farroupilha region, the producers are linked to AFAVIN. The association carries out publicity actions, event organizations and institutional partnerships. It also works in the promotion of local festivals, as well as participation in international competitions. In Garibaldi, the consortium of CPEG as an entity works toward the objectives of the collectivity, mainly so that the seal of the products is divulged and recognized by consumers as a high-quality product. At present, the initiatives are still recent and there is no specific measurable financial result of CPEG (A7). In relation to the sector brand project, management is carried out by IBRAVIN with the support of APEX. For the international market, respondents believe that Brazilian wine is still seen as an exotic product, but with the work of dissemination and participation in fairs, there is an improvement in the perception of the product (A2). The interviewees understand that participation in fairs should be undertaken collectively, as the wineries are constituents of the so-called "Brazil Block". Producers see themselves as competitors, but to achieve international visibility, they need to work together and in partnership, as they know that individually they would not have resources for such investment (W1, W2, W3, W6). Regarding the domestic market, it was emphasized that there is a need for more public policies and other efforts for the Brazilian consumer to recognize national wine. There are already some actions in this direction carried out by SEBRAE (G4). Thus, for the agencies, associations and managers of the wineries, the shared brand resource has the support of the Association. In this way, the shared brands resource can be considered valuable, rare, imperfectly imitable/replaceable and has the organizational process structured for the management of these brands (Barney, 1991, 1995; Barney and Hesterly, 2006; Barney and Wright, 1998; Grant, 1991; Ito and Gimenez, 2011; Shafeey and Trott, 2014; Sirmon et al., 2011) in the market in question, thus forming the VRIA model. Considering the results presented, it can be concluded that the GIs, the collective brand and the sector brand meet all the VRIA model requirements, in the opinion of the interviewees, creating barriers for imitation by competitors and difficulties in implementing similar strategies in the market (Bharadwaj et al., 1993; Hoffman, 2000; Penrose, 1979; Von Krogh and Roos, 1995). It can be concluded that the proposition that shared brands - GI, collective brands and sector brands - provide SCA, according to the VRIA, can be confirmed, thus fulfilling the four conditions that form the acronym VRIA - valuable, rare, imperfectly imitable/replaceable and association. The value added to the product through the information contained in the shared brands facilitates the establishment of a relationship of trust between producer and consumer, being a source of competitive advantage. When studying the theory about shared brands, this work contributes to the understanding of the concept that the GI, the collective brand and the sector brand can be denominated as types of shared brands from the existent literature. Barney's VRIO framework (Barney, 1991, 1995) hitherto thought for individual companies, has the letter "O" of Organization replaced by the letter "A" of Association, becoming VRIA. In this way, the existence of a shared brand implies in the creation of an association that becomes the owner of the brand. Another theoretical contribution of this research is the deepening of studies on shared brands in Brazil and in the world, as it is still a subject little studied because it is of recent implementation in the market (Barbosa, 2011; Bruch et al., 2014; Castro and Giraldi, 2015; Regalado et al., 2012). 5.1 Practical implications and future research recommendations The practical contributions come from the fact that government agencies, associations and wineries can improve the production process and seek certified products for commercialization in the domestic and foreign markets. These contributions may also, in practice, be used by other sectors and countries. Specifically, it contributes to the fact that government agencies understand the practical result obtained from the management effort in the creation of shared brands. For the associations, the contribution is in improving internal processes and providing advice to members to obtain best practices and results. And finally, for wineries, it points out how much the collective effort in search of shared brands tends to make each individual company more competitive with a source to obtain SCA. Future research studies could be carried on with other wine-growing regions in regions that are beginning to develop the shared brand as in the state of Santa Catarina and Northeast Brazil. A comparative study can also be carried out with other countries of the New World of Wine, such as Argentina, Chile and Uruguay. We also suggest that forthcoming studies consider analyzing the consumers' perspective of the perceived value of wine certifications and the wineries strategies to increase awareness of such added value. 5.2 Limitations The results of qualitative researchers are not reproducible, as the researchers' interpretation is susceptible to changes, as well as the fact that they are specific and unique situations in time for that region or research unit studied, which hinders the generalizations of the study.
|
This work contributes to the understanding of the shared brand's concept, including geographical indications, collective brands and the sector brands. The proposition that shared brands provide sustainable competitive advantage, according to an adapted VRIO model was confirmed. Barney's VRIO framework (Barney, 1991, 1995) hitherto thought for individual companies, has the letter "O" of Organization replaced by the letter "A" of Association, becoming VRIA. The authors found that the four conditions that form the here proposed acronym VRIA are valuable, rare, imperfectly imitable/replaceable and association.
|
[SECTION: Purpose] The People's Action Party (PAP) won the May 30, 1959 general election and assumed office and attained self-government from Britain on 3 June 1959. Singapore was a different place then because it was a poor third world country, afflicted with a serious housing shortage (half the population was living in squatter huts), an unemployment rate of 14 per cent, political instability, labour unrest, corruption, and a high crime rate. However, today, after 53 years under the PAP government, Singapore has been transformed into a first world country, which is no longer afflicted by the problems it faced in 1959.This article contends that Singapore's ability to solve the problems it encountered after attaining self-government can be attributed to the effectiveness of the various policies introduced by the PAP government since 1959. The PAP government created the Housing and Development Board (HDB) in February 1960 to tackle the housing shortage and the Economic Development Board (EDB) in August 1961 to create jobs by attracting foreign investment to Singapore. These two statutory boards were formed to reduce the workload of the Singapore Civil Service (SCS), which was not equipped to solve the housing shortage or create jobs. Apart from being handicapped by rigid regulations and inflexibility, the civil servants also had a "colonial mentality" and were not attuned to the problems facing Singapore.Accordingly, the PAP government initiated these four policies to solve the country's problems:1. Reorganization and attitudinal reform of the SCS.2. Enactment of the Prevention of Corruption Act (POCA) in June 1960 to empower the Corrupt Practices Investigation Bureau (CPIB) (2011) to curb corruption effectively.3. Maintaining the tradition of meritocracy introduced by the British colonial government by retaining and decentralizing the Public Service Commission (PSC) to enhance its effectiveness.4. Paying competitive salaries to ministers and senior civil servants from 1972 and benchmarking these salaries from 1995 to private sector salaries to attract the "best and brightest" citizens to the SCS and government and to minimize the brain drain to the private sector.These four policies will be analyzed in turn below before providing the governance indicators for Singapore and assessing the transferability of Singapore's experience in promoting good governance to other Asian countries. The PAP government realized on assuming office in June 1959 that it had to transform the colonial bureaucracy it inherited in order to ensure the effective implementation of its socio-economic development programmes. During the 137 years of British colonial rule1, the SCS focused on performing the traditional "housekeeping" functions of maintaining law and order, building public works and collecting taxes. Consequently, the SCS did not play an important role in national development and did not introduce administrative reforms until after the Second World War (Quah, 1996a, pp. 62-3).Its victory in the May 1959 general election gave the PAP government the mandate to introduce comprehensive reforms to the SCS and to change the civil servants' "colonial mentality" by making them more sensitive to the needs of the population. Accordingly, it began by reorganising the SCS into nine ministries including the new ministries of culture and national development, which were created to deal with nation-building and economic development respectively (Seah, 1971, pp. 82-3). The workload of the SCS was reduced by making the HDB and EDB responsible for providing low-cost public housing and attracting foreign investment to Singapore. Unlike the SCS, which could not implement the public housing and industrialisation programmes swiftly because of its regulations and red tape, the HDB and EDB as statutory boards were better equipped to implement these programmes expeditiously (Quah, 1985, pp. 124-6).In addition to reorganising the SCS, the PAP government realized the necessity of changing the civil servants' attitudes and convincing them to participate in the process of attaining national development goals. The Political Study Centre was opened by Prime Minister Lee Kuan Yew on August 15, 1959. In his opening speech, he hoped that the civil servants would change their "colonial mentality" once they were made aware of the problems facing Singapore. Their responsibility was to ensure the efficiency of the SCS. The purpose of the two-week part-time and non-residential course conducted by the Political Study Centre for senior civil servants was to change their attitudes and make them more aware of the local contextual constraints (Quah, 2010, p. 134).In addition to the Political Study Centre, the PAP government also relied on four additional methods to change the attitudes and behaviour of the civil servants. First, civil servants were encouraged to participate voluntarily in mass civic projects during the weekends to enable them to get better acquainted with the political leaders, and to provide them with an opportunity to engage in manual work. Second, Nanyang University graduates, who were Chinese-educated, were recruited from 1960 for the education service to reduce the predominance of the English-educated in the SCS. Third, civil servants who were found guilty of misconduct were disciplined and the Central Complaints Bureau was formed in 1961 to enable the non-English-educated public to submit complaints against rude or incompetent civil servants. Finally, expatriate civil servants who were due for retirement were encouraged to remain in the SCS if they were competent while their incompetent colleagues were retired prematurely (Quah, 2010, pp. 134-5).In short, the PAP government relied on the reorganisation of the SCS and attitudinal reform to change the "colonial mentality" of the senior civil servants because it needed their support to implement its policies effectively. Corruption was a way of life in Singapore during the British colonial period because of the British colonial government's lack of political will and the ineffective anti-corruption measures adopted. Corruption was made illegal in Singapore with the enactment of the Penal Code of the Straits Settlements of Malacca, Penang and Singapore in 1871. However, even though police corruption was rampant and confirmed by the 1879 and 1886 Commissions of Inquiry, the British colonial government ignored their findings and delayed the enactment of the Prevention of Corruption Ordinance (POCO) until December 1937 (Quah, 2007, pp. 9-12).The POCO was ineffective because it limited the powers of arrest, search and investigation of police officers as warrants were required before arrests could be made; and the penalty of two years' imprisonment and/or a fine of S$10,000 for corrupt offenders did not deter corrupt behaviour (Quah, 1978, p. 9). Similarly, the Anti-Corruption Branch (ACB) was ineffective because of the prevalence of police corruption. As the ACB was part of the Criminal Investigation Department (CID) of the Singapore Police Force (SPF), it was not surprising that the ACB was ineffective in curbing police corruption. Furthermore, the ACB was inadequately staffed with only 17 personnel and had to compete with other branches in the CID for limited manpower and other resources (Quah, 2007, pp. 14-15).The British colonial government only realized the folly of making the ACB responsible for curbing corruption when it was discovered that three police detectives and some senior police officers were involved in robbing 1,800lbs of opium worth S$400,000 (US$133,333) in October 1951 (Tan, 1999, p. 59). The Opium Hijacking scandal exposed the ACB's weaknesses and its inability to curb police corruption. Consequently, the British colonial government formed the CPIB as an independent agency in October 1952 to replace the ineffective ACB.During their campaign for the May 1959 general election, the PAP leaders demonstrated their commitment to curbing corruption by exposing the acceptance of S$700,000 by the Minister for Education, Chew Swee Kee, from some American donors (Quah, 2010, p. 175). The Labour Front government led by Chief Minister Lim Yew Hock was described as "being corrupted from head to toe" by a retired architect, Lee Kip Lin (Yap et al., 2009, p. 555). The PAP's revelation of the Chew Swee Kee scandal enabled it to win the May 30, 1959 general election by capturing 43 of the 51 seats and obtaining 53.4 per cent of the votes cast.When the PAP leaders assumed office in June 1959, an immediate priority was to ensure a clean government by adopting a zero-tolerance policy toward corruption. Minister Mentor Lee Kuan Yew explained in his memoirs why he and his colleagues were determined to keep Singapore free from corruption from the outset of their administration:We were sickened by the greed, corruption and decadence of many Asian leaders. [...] We had a deep sense of mission to establish a clean and effective government. When we took the oath of office [...] in June 1959, we all wore white shirts and white slacks to symbolize purity and honesty in our personal behaviour and our public life. [...] We made sure from the day we took office in June 1959 that every dollar in revenue would be properly accounted for and would reach the beneficiaries at the grass roots as one dollar, without being siphoned off along the way. So from the very beginning we gave special attention to the areas where discretionary powers had been exploited for personal gain and sharpened the instruments that could prevent, detect or deter such practices (Lee, 2000, pp. 182-4).As corruption was a way of life in Singapore in June 1959, the PAP leaders learnt from the mistakes made by the British colonial government in curbing corruption and demonstrated their commitment by enacting the POCA on June 17, 1960 to replace the ineffective POCO and to strengthen the CPIB by providing it with more legal powers, personnel and funding.The POCA has three important features to rectify the POCO's weaknesses and to enhance the CPIB's legal powers and increase its personnel. First, the penalty for corruption has been increased to imprisonment for five years and/or a fine of S$10,000 to enhance the POCA's deterrent effect. Second, according to section 13, a person found guilty of accepting an illegal gratification has to pay the amount he had taken as a bribe in addition to any other punishment imposed by a court. The third and most important feature of the POCA is that it has given the CPIB more powers and a new lease of life. For example, section 15 gives the CPIB officers powers of arrest and search of arrested persons. Furthermore, the CPIB's director and his senior officers are empowered by section 18 to investigate the bank account, share account or purchase account of any person suspected of committing a corruption offence. Section 24 is perhaps the most important asset for the CPIB in its investigation of corruption offences because "the fact that an accused person is in possession, for which he [or she] cannot satisfactorily account, of pecuniary resources or property disproportionate to his [or her] known sources of income" is evidence that he or she had obtained these pecuniary resources or property "corruptly as an inducement or reward" (Quah, 2010, pp. 176-7).To ensure the POCA's continued effectiveness, the PAP government has introduced whenever necessary, amendments or new legislation to deal with unanticipated problems or to plug legal loopholes. For example, in 1966, the POCA was amended so that a person could be found guilty of corruption without actually receiving the bribe as long as he had shown the intention of doing so (section 9). The POCA was also amended in 1966 so that, according to section 37, Singapore citizens working for their government in embassies and other government agencies abroad would be prosecuted for corrupt offences committed outside Singapore and would be dealt with as if such offences had occurred in Singapore. In 1989, the fine for corrupt offences was increased tenfold from S$10,000 to S$100,000 (US$78,730)2. On March 3, 1989, the Corruption (Confiscation of Benefits) Act 1989 was passed to enable the court to issue a confiscation order against the estate of a deceased defendant (Quah, 2010, pp. 177-8).Unlike the British colonial government, the PAP government has also demonstrated its political will in curbing corruption not only by enhancing the CPIB's legal powers but also by providing the CPIB with more personnel and budget during the past 53 years. Table I shows that the CPIB has grown by 17 times from eight officers in 1959 to 138 officers in 2011. Similarly, as indicated in Table II, the CPIB's budget has increased by nearly 20 times from S$1,024,370 in 1978 to S$34,073,400 in 2011 (Quah, 2010, pp. 179-80; Republic of Singapore, 2011, p. 378).In contrast to the situation during 1952-1959, the CPIB has adopted a "total approach to enforcement" by dealing with both "big and small cases" of corruption in both the public and private sectors, "both giver and receiver of bribes" and "other crimes uncovered in the course of [the] corruption investigation" (Soh, 2008a, pp. 1-2). In addition to its emphasis on investigation and enforcement, the CPIB also focuses on corruption prevention by reviewing the procedures and practices in those government agencies, where corruption has occurred and makes recommendations to remove the "loopholes and vulnerabilities." The CPIB employs this review process to "identify potential problem areas and loopholes" in order to minimize the opportunities for corruption (Soh, 2008b, p. 8).Finally, the CPIB's extensive outreach programme is implemented by its Public Education Group, which conducts prevention and education talks for pre-university students, principals, teachers, newly appointed civil servants, law enforcement agencies like the police and immigration department, and the management and staff of major organisations in key industries (Quah, 2010, p. 181). Table III shows that the number of persons attending the CPIB's prevention and education talks has increased from 2,500 in 2005 to 9,193 in 2010. Similarly, the number of foreign visitors to the CPIB has more than doubled from 1,000 to 2,538 during 2005-2010. The number of visitors from local organisations has risen from 20 in 2005 to 424 in 2008, and visits by students also increased from 150 to 791 from 2005-2009. The British introduced meritocracy to its colonies in Africa and Asia with the creation of the PSC, which is the adapted version of the Civil Service Commission in Britain. The raison d'etre of the PSC is twofold: to insulate the civil service from politics; and to accelerate its localisation by replacing the expatriate officers with qualified local staff (Sinker, 1953, p. 206). With the advent of self-government and the increasing control of the civil service in the hands of the local population, the PSC was established in the British colonies to insulate appointments, promotions and discipline in the civil service from politics. Accordingly, the PSC was created in India in 1926, in Ceylon (Sri Lanka) in 1931, in Pakistan in 1947, in Hong Kong in 1950, in Singapore and Nepal in 1951, in Malaya in 1957, and in Bangladesh in 1972 (Quah, 2009, p. 810).Meritocracy was introduced in Singapore with the establishment of the PSC on January 1, 1951. The PSC's origins can be traced to the White Paper (Command Paper No. 197) entitled Organisation of the Colonial Service issued by the British government in 1946. Command Paper No. 197 stressed that progress toward self-government could only be achieved if the public services of the colonies were adapted to local conditions and staffed to the maximum possible extent by local people. More importantly, it recommended the establishment of PSCs in the colonies to ensure that qualified local candidates would be recruited into the public services (Quah, 2010, p. 72).The PSC in Singapore was formed with these two objectives in mind: to keep politics out of the SCS and to accelerate the latter's pace of localisation (Quah, 1982, p. 50). The PSC's second objective is no longer important today because the localisation of the SCS was completed with the attainment of self-government in Singapore in June 1959. However, its primary aim of keeping politics out of the SCS remains relevant as the aim of the PSC's programme as stated in the national budget is "to meet the staffing requirements of the government in accordance with the merit principle" (Republic of Singapore, 1980, p. 78).The PSC's evolution during the past 60 years can be divided into four stages, as depicted in Table IV. During its first 31 years, the PSC was the central personnel agency responsible for selecting and promoting civil servants on the basis of merit, disciplinary control, and the granting of scholarships and training awards. As the PSC members and selection boards relied mainly on personal interviews to perform their functions, their workload increased tremendously during this stage. Indeed, the number of candidates interviewed by them for appointments and promotions increased by nearly 19 times from 556 candidates in 1951 to 10,430 candidates in 1982 (Public Service Commission, 1954, p. 2; Public Service Commission, 1983, p. 5). Similarly, the number of disciplinary cases completed has also risen from 24 to 169 during 1957 to 1982 (Public Service Commission, 1959, p. 8; Public Service Commission, 1983, p. 18). The number of scholarships and training fellowships awarded has also increased from 23 in 1963 to 847 in 1982 (Public Service Commission, 1964, pp. 9, 19; Public Service Commission, 1983, p. 8). Finally, the heavy workload of Singapore's PSC becomes obvious when its output of interviewing 58,712 candidates during 1964-1967 is more than nine times that of the PSC in Ceylon, which interviewed only 6,485 candidates during the same period (Quah, 1971, p. 140).The Public Service Division (PSD) was formed on January 3, 1983 on the recommendation of the Management Services Department to formulate and review personnel policies in the SCS and to ensure that these policies are implemented consistently in the various ministries. However, as the PSD was responsible for all personnel policy matters concerning appraisal, posting, training, schemes of service, service conditions, and welfare, its creation did not reduce the PSC's heavy workload of interviewing 50,274 candidates for appointments and promotions, completing 1,148 disciplinary cases, and granting 1,543 scholarships and training awards during 1983-1989 (Quah, 2010, p. 83).In March 1990, the Constitution of Singapore was amended to help the PSC cope with its heavy workload by increasing its membership from 11 to 15, including the chairman, and by creating two new sub-commissions namely, the Education Service Commission (ESC) for education and the Police and Civil Defence Services Commission (PCDSC) for the police and civil defence services. This move was designed to reduce the PSC's workload as the ESC would be responsible for appointing and promoting 21,000 teachers, and the PCDSC would take care of the appointment and promotion of 10,000 police, narcotics, prisons, and civil defence officers, thus leaving the PSC to deal with the remaining 34,000 civil servants. However, in spite of the establishment of the ESC and PCDSC, the PSC's workload was not reduced significantly as Table V shows that the PSC interviewed 9,993 candidates (67.4 per cent) for appointment during 1990-1994, in contrast to the 4,254 candidates (28.7 per cent) interviewed by the ESC, and the 573 candidates (3.9 per cent) interviewed by the PCDSC.To enhance the SCS's ability to compete with the private sector for talented personnel, the public personnel management system in Singapore was further decentralized in January 1995 with the establishment of a system of 31 personnel boards, as shown in Table VI. As the creation of the ESC and PCDSC did not alleviate significantly the PSC's workload in appointing candidates to the SCS during 1990-1994, the ESC and PCDSC were dissolved and amalgamated into a single PSC on 1 April 1998 (Public Service Commission, 1999, p. 9).Unlike the ESC and PCDSC, the 31 personnel boards have reduced considerably the PSC's workload from 1995 to 2010. Table VII shows that the PSC considered 1,724 candidates for appointment, 308 candidates for promotion, completed 832 disciplinary cases, and granted 2,305 scholarships and training awards from 1995-2010.Finally, the comparison of the PSC's workload during its second, third and fourth stages of development in Table VIII shows that the decentralisation of its functions, which began in 1990 with the formation of the ESC and PCDSC and ended with the creation of the 31 personnel boards in 1995, has been effective because the PSC's workload in interviewing candidates for appointment and promotion has been drastically reduced from 50,274 candidates during 1983-1989 to 18,463 candidates during 1990-1994, and to 2,032 candidates during 1995-2010. Similarly, the number of disciplinary cases completed by the PSC has declined from 1,148 cases during 1983-1989 to 832 cases during 1995-2010. On the other hand, it is not surprising that the number of scholarships and training awards granted by the PSC has increased from 1,543 during 1983-1989 to 2,305 during 1995-2010 because this constitutes the PSC's major function today. To balance the budget, a Cabinet Budget Committee on Expenditure recommended in June 1959 the removal of the variable allowances of Divisions I and II civil servants to save S$10 million and prevent a budget deficit of S$14 million (Quah, 2010, p. 103). The government restored the variable allowance in September 1961 with the improvement of the budgetary situation (Seah, 1971, p. 94).In 1968, the Harvey Commission recommended salary increases for five grades in the Division I superscale salaries. However, the government did not implement this recommendation until 1973 for two reasons: the economy could not afford a major salary revision and the private sector was not considered a serious threat in terms of competing for talent as promotion exercises for senior civil servants were conducted frequently to retain talented personnel in the SCS (Lee, 1995, pp. 21-2).However, the improvement of the Singapore economy in the 1970 s resulted in higher salaries in the private sector and aggravated the brain drain of talented civil servants to the private sector. The National Wages Council (NWC) was formed in February 1972 as an advisory body to formulate general guidelines on wage policies, to recommend annual wage adjustments, and to advise on incentive systems for improving efficiency and productivity (Then, 1998, pp. 220-1). The NWC recommended the payment of the Annual Wage Supplement (AWS) or "13th month pay" from 1972 to minimize the gap between salaries in the public and private sectors.The PAP government has relied on increasing the salaries of ministers and senior civil servants in 1973, 1979, 1982, 1989, and 1994 to reduce the growing differential with private sector salaries. On October 21, 1994, a White Paper on Competitive salaries for competent and honest government was presented to parliament to justify the benchmarking of the salaries of ministers and senior civil servants to the average salaries of the top four earners in six private sector professions namely, accounting, banking, engineering, law, local manufacturing companies, and multi-national corporations. The government accepted this recommendation and public sector salaries were benchmarked accordingly from January 1995 with the salaries of the six professions in the private sector (Quah, 2010, pp. 110-11).The 1997 Asian financial crisis and the subsequent slowing down of the Singapore economy resulted in a 2 per cent decrease in Superscale G and a 7 per cent decrease in Staff Grade I salaries and the reduction of the employers' contribution to the Central Provident Fund (CPF) from 20 per cent to 10 per cent for all employees. The purpose of the CPF reduction was to enhance the Singapore's competitiveness by lowering the cost of doing business. In other words, the reduction in the CPF contribution meant an additional decrease in the salaries of the ministers and senior civil servants.When the Singapore economy recovered in 1999 with a growth rate of 5.4 per cent, and the reduction of retrenchments from 29,100 in 1998 to 14,600 in 1999, wages in the private sector began to rise again. Unemployment fell from 4.3 per cent in December 1999 to 3.4 per cent in March 2000. With the tight labour market in Singapore and the improved conditions in the private sector, Deputy Prime Minister Lee Hsien Loong revealed in parliament on 29 June 2000 that eight administrative officers had resigned in 2000. The government increased the performance bonus component in the public sector salaries and broadened the benchmarking of these salaries to the top eight earners in the six private sector professions. Consequently, in June 2000, the variable component of annual salaries was increased from 30 per cent to 40 per cent of the total annual pay of the superscale administrative officers and ministers (Quah, 2010, pp. 113-14). However, public sector salaries were later reduced by a combined total of 30 per cent in November 2001 and July 2003 because of the recession.In December 2007, the PSD announced that the salaries of ministers and senior civil servants would be increased from 4 per cent to 21 per cent from January 2008. Table IX shows the salaries of the President, Prime Minister, Ministers, Permanent Secretaries, superscale civil servants at the entry grade, and Members of Parliament in 2007 and 2008. However, on 24 November 2008, the PSD announced that the salaries of administrative officers, political, judicial and statutory appointment holders would be decreased by 19 per cent in 2009 because of the economic recession. This means that the President's annual salary has been reduced from S$3.87 million to S$3.14 million. Similarly, the Prime Minister's annual salary has been decreased from S$3.76 million to S$3.04 million (PSD 2008, pp. 1-2). Nevertheless, the Prime Minister's annual salary of US$2,183,516 in 2010 made him the best paid political leader in the world (Economist, 2010a, b).The policy of paying competitive public sector salaries has been effective in curbing the brain drain of political leaders to the private sector as none of them have resigned from political office to work in the private sector before their retirement. The attractive remuneration for the permanent secretaries has also been effective in retaining them in the SCS as none of them have left for private sector jobs before their retirement. However, paying competitive public sector salaries has been less effective in preventing Division I officers from leaving the SCS. An analysis of the resignation rate of Division I officers in the SCS from 1971-1984 shows that the salary increases in 1972, 1973, 1979 and 1982 had failed to curb the exodus of senior civil servants to the private sector (Quah, 2010, pp. 93, 119-20). Have the above policies of changing the attitudes of civil servants, minimizing corruption, maintaining meritocracy in the SCS by decentralizing the PSC, and paying competitive salaries to attract the "best and brightest" citizens to the SCS, resulted in good governance in Singapore? The short answer is "Yes" as will be demonstrated in Singapore's rankings and scores on eight governance indicators.Government effectiveness Singapore has been ranked first for the competence of its public officials from 1999 to 2002 by the Global Competitiveness Report3. Singapore was ranked first among 59 countries in 1999 and 2000 (Schwab et al., 1999, p. 242; Porter et al., 2000, p. 238). Singapore's civil servants were also ranked first among their counterparts in 75 countries in 2001-2002 and 80 countries in 2002-2003 (Schwab et al., 2002, p. 399; Cornelius, 2003, p. 604).The effectiveness of Singapore's government has also been confirmed by Singapore's consistently high ranking on the World Bank's governance indicator on government effectiveness, which is defined as "the quality of public service provision, the quality of the bureaucracy, the competence of civil servants, the independence of the civil service from political pressures, and the credibility of the government's commitment to policies" (Kaufmann et al., 2004, p. 3). Table X shows that government effectiveness in Singapore is very high from 1996 to 2010 and ranges from 93.2 percentile rank in 2002 to 100 percentile rank in 1996, 1998, 2008 and 2010.Similarly, a 2010 survey of 12 Asian countries conducted by the Hong Kong-based Political Economic Risk Consultancy (PERC) found that the bureaucracy in Singapore is the most effective, followed by the bureaucracy in Hong Kong SAR. At the other extreme, the least effective bureaucracies are those in the Philippines, Indonesia and India in descending order. Table XI confirms that Singapore has the most effective bureaucracy among 12 Asian countries from 1998 to 2010. In short, Singapore's government and civil servants are effective and performed well according to the above indicators.Effectiveness in curbing corruption The effectiveness of the POCA and CPIB in curbing corruption is reflected in Singapore's consistently high rankings and scores on these five indicators: Transparency International's Corruption Perceptions Index (CPI) from 1995-2011; Political Economic PERC's annual surveys of corruption from 1995-2011; World Bank's Control of Corruption indicator from 1996-2010; World Bank's Doing Business Surveys from 2007-2012; and the Global Competitiveness Report's indicator on public trust of politicians from 1999 to 2010-2011.Table XII shows that first, Singapore has been consistently ranked as the least corrupt country in Asia from 1995-2009 and 2011, and the least corrupt country in the world in 2010 with Denmark and New Zealand. Second, PERC's annual surveys of corruption from 1995-2011 also confirm Singapore's status as the least corrupt country in Asia. Third, the World Bank's control of corruption indicator from 1996-2010 shows that Singapore has the highest percentile rank among all the Asian countries included in the surveysFourth, Table XIII confirms that red tape is not a problem at all in Singapore as the World Bank's Doing Business Surveys show that Singapore is ranked first among the 175-183 economies included from 2007-2012. The relative absence of red tape in Singapore is an important factor responsible for its low level of corruption.Fifth, Table XIV shows that Singapore has been consistently ranked first in the Global Competitiveness Report's indicator on the public trust of politicians among the 59 to 139 countries included in the surveys from 1999 to 2010/2011. Singapore's superior ranking on this indicator also reflects its high level of public trust in politicians and its low level of corruption.In short, the four policies introduced by the PAP government after assuming office in June 1959 have been effective in promoting good governance in Singapore as demonstrated by Singapore's consistently high rankings and scores on the abovementioned eight governance indicators. Is Singapore's experience in promoting good governance transferable to other Asian countries? As Singapore's success in ensuring good governance is the combined influence of the political will of the PAP government to solve the problems facing the country for the past 53 years and its favourable policy context, it will be difficult to transfer Singapore's experience in toto to other Asian countries because of the lack of political will and the unfavourable policy contexts in many Asian countries.Apart from good leadership, Singapore's favourable policy context has enabled its political leaders to stretch the constraints imposed by its small size and lack of resources by formulating and implementing effective policies to solve its problems during the past 53 years. As public administration Singapore-style is the product of the local policy context and the policies implemented by the PAP government, it would be difficult to replicate these policies to other Asian countries in view of the significant contextual differences between Singapore and these countries (Quah, 2010, pp. 246, 51).Table XV shows clearly the significant contextual differences between Singapore and the other 25 Asian countries. First, in terms of size, Singapore is the second smallest country after the Macao Special Administrative Region, which has a land area of 29.2 sq. km. At the other extreme, are the larger countries of China and India, which are 13,466 and 4,630 times larger respectively than Singapore. A second important contextual difference is Singapore's population, which is only larger than those of Mongolia's population of 2.7 million, Timor-Leste's population of 1.2 million, Bhutan's population of 0.7 million, Macao's population of 0.5 million, and Brunei's population of 0.4 million. On the other hand, Singapore's population of 5.1 million in 2010 is dwarfed by the huge populations of China, India, and Indonesia.The third contextual difference between Singapore and the other Asian countries is its economic affluence as manifested in its GDP per capita of US$40,920 in 2010, which is the second highest among all the 26 countries listed in Table XV. In contrast, the GDP per capita of these four countries are less than US$1,000 namely, Cambodia (US$760), Bangladesh (US$640), Afghanistan (US$517), and Nepal (US$490).In short, Singapore is a city-state, which is richer and smaller in terms of land area and population for the PAP government to govern than most of the other Asian countries. Singapore's favourable policy context has enabled the PAP government, which has been in power since June 1959, to implement policies effectively, to curb corruption, and to ensure the ease of doing business in Singapore, as demonstrated in Singapore's superior ranking on the World Bank's governance indicator on government effectiveness from 1996-2010, the Doing Business Surveys from 2007-2012, Transparency International's CPI from 1995-2011, and PERC's corruption surveys from 1995-2011. In his National Day Rally speech on 19 August 1984, then Prime Minister Lee Kuan Yew attributed Singapore's success to the quality of its political leadership:In the end, whatever the system, it is the quality of the men who run it, that is decisive. For they will decide what to make of the society, and how to get the people to give of their best. The Singapore system has worked. It will continue to work if you vote for honest, able and dedicated men, and you give them your best, for the good of all (Lee, 1984, p. 18).More recently, Lee acknowledged the importance to Singapore's development of attracting the "best and brightest" citizens to join the government and the SCS thus:My experience of developments in Asia has led me to conclude that we need good men to have good government. [...] The single most decisive factor that made for Singapore's development was the ability of its ministers and the high quality of the civil servants who supported them. [...] It was Singapore's good fortune that we had, for a small, developing country, a fair share of talent, because our own [talent] had been reinforced by the talented men and women who came here for their education, and stayed on for employment or business opportunities (Lee, 2000, pp. 735-6).In the same vein, Edgar H. Schein (1996, pp. 221-2) has identified the policy of having "the best and brightest" citizens in government as "probably one of Singapore's major strengths" because "they are potentially the most able to invent what the country needs to survive and grow." Furthermore, he has described Singapore as "one of the few models existing in the world of how a society can progress with a government that attempts to maximize intelligence, skill, and honesty."In sum, Singapore's transformation from a poor third world country in 1959 to an affluent and politically stable first world country today is the result of the ability of its political leaders and civil servants to formulate and implement policies to solve the country's problems during the past 53 years. First, the PAP leaders reorganised the SCS and changed the attitudes of the civil servants by convincing them to contribute to the attainment of national development goals. Second, they continued with the tradition of meritocracy introduced by the British by retaining and enhancing the PSC's effectiveness by reducing its heavy workload by decentralizing its functions of appointment and promotion to the ESC and PCDSC in 1990, and the 31 personnel boards in 1995. Third, they learnt from the mistakes made by the British in curbing corruption by enacting the POCA in June 1960 to enhance the CPIB's effectiveness in combating corruption. Fourth, the PAP government's success in promoting economic growth enabled it to compete for talented personnel with the private sector by paying competitive salaries to ministers and senior civil servants from 1972 onwards to prevent them from leaving for private sector jobs.In the final analysis, whether these four elements of Singapore's success - institutional and attitudinal reform of civil servants; zero-tolerance for corruption; meritocracy in appointing and promoting civil servants; and paying competitive salaries to attract the "best and brightest" citizens to join the government and civil service - can be replicated in other Asian countries depends mainly on whether their political leaders and senior civil servants have the political will and are prepared to pay the high economic and political costs of implementing these policies. Opens in a new window.Table I Growth of CPIB's personnel, 1952-2011 Opens in a new window.Table II Budget of the CPIB, 1978-2011 Opens in a new window.Table III Number of visitors to CPIB, 2005-2010 Opens in a new window.Table IV Evolution of the PSC in Singapore, 1951-2011 Opens in a new window.Table V Workloads of the PSC, ESC and PCDSC in appointing and promoting candidates in the SCS, 1990-1994 Opens in a new window.Table VI The SCS's Personnel Management System in 1995 Opens in a new window.Table VII Workload of the PSC in Singapore, 1995-2010 Opens in a new window.Table VIII Workload of the PSC in Singapore, 1983-2010 Opens in a new window.Table IX Salaries of key appointments in Singapore, 2007-2008 Opens in a new window.Table X Government effectiveness of Singapore, 1996-2010 Opens in a new window.Table XI PERC's evaluation of the effectiveness of bureaucracy in 12 Asian countries, 1998 to 2010 Opens in a new window.Table XII Singapore's performance on CPI, PERC and control of corruption 1995-2011 Opens in a new window.Table XIII Singapore's rank on the World Bank's ease of doing business, 2007-2012 Opens in a new window.Table XIV Public trust of politicians in Singapore, 1999-2010/2011 Opens in a new window.Table XV Policy contexts of 26 Asian countries
|
- The purpose of this paper is to attribute Singapore's good governance to the effective policies implemented by the People's Action Party (PAP) government and contend that it will be difficult to transfer Singapore's experience to other countries because of Singapore's unique circumstances and favourable policy context.
|
[SECTION: Method] The People's Action Party (PAP) won the May 30, 1959 general election and assumed office and attained self-government from Britain on 3 June 1959. Singapore was a different place then because it was a poor third world country, afflicted with a serious housing shortage (half the population was living in squatter huts), an unemployment rate of 14 per cent, political instability, labour unrest, corruption, and a high crime rate. However, today, after 53 years under the PAP government, Singapore has been transformed into a first world country, which is no longer afflicted by the problems it faced in 1959.This article contends that Singapore's ability to solve the problems it encountered after attaining self-government can be attributed to the effectiveness of the various policies introduced by the PAP government since 1959. The PAP government created the Housing and Development Board (HDB) in February 1960 to tackle the housing shortage and the Economic Development Board (EDB) in August 1961 to create jobs by attracting foreign investment to Singapore. These two statutory boards were formed to reduce the workload of the Singapore Civil Service (SCS), which was not equipped to solve the housing shortage or create jobs. Apart from being handicapped by rigid regulations and inflexibility, the civil servants also had a "colonial mentality" and were not attuned to the problems facing Singapore.Accordingly, the PAP government initiated these four policies to solve the country's problems:1. Reorganization and attitudinal reform of the SCS.2. Enactment of the Prevention of Corruption Act (POCA) in June 1960 to empower the Corrupt Practices Investigation Bureau (CPIB) (2011) to curb corruption effectively.3. Maintaining the tradition of meritocracy introduced by the British colonial government by retaining and decentralizing the Public Service Commission (PSC) to enhance its effectiveness.4. Paying competitive salaries to ministers and senior civil servants from 1972 and benchmarking these salaries from 1995 to private sector salaries to attract the "best and brightest" citizens to the SCS and government and to minimize the brain drain to the private sector.These four policies will be analyzed in turn below before providing the governance indicators for Singapore and assessing the transferability of Singapore's experience in promoting good governance to other Asian countries. The PAP government realized on assuming office in June 1959 that it had to transform the colonial bureaucracy it inherited in order to ensure the effective implementation of its socio-economic development programmes. During the 137 years of British colonial rule1, the SCS focused on performing the traditional "housekeeping" functions of maintaining law and order, building public works and collecting taxes. Consequently, the SCS did not play an important role in national development and did not introduce administrative reforms until after the Second World War (Quah, 1996a, pp. 62-3).Its victory in the May 1959 general election gave the PAP government the mandate to introduce comprehensive reforms to the SCS and to change the civil servants' "colonial mentality" by making them more sensitive to the needs of the population. Accordingly, it began by reorganising the SCS into nine ministries including the new ministries of culture and national development, which were created to deal with nation-building and economic development respectively (Seah, 1971, pp. 82-3). The workload of the SCS was reduced by making the HDB and EDB responsible for providing low-cost public housing and attracting foreign investment to Singapore. Unlike the SCS, which could not implement the public housing and industrialisation programmes swiftly because of its regulations and red tape, the HDB and EDB as statutory boards were better equipped to implement these programmes expeditiously (Quah, 1985, pp. 124-6).In addition to reorganising the SCS, the PAP government realized the necessity of changing the civil servants' attitudes and convincing them to participate in the process of attaining national development goals. The Political Study Centre was opened by Prime Minister Lee Kuan Yew on August 15, 1959. In his opening speech, he hoped that the civil servants would change their "colonial mentality" once they were made aware of the problems facing Singapore. Their responsibility was to ensure the efficiency of the SCS. The purpose of the two-week part-time and non-residential course conducted by the Political Study Centre for senior civil servants was to change their attitudes and make them more aware of the local contextual constraints (Quah, 2010, p. 134).In addition to the Political Study Centre, the PAP government also relied on four additional methods to change the attitudes and behaviour of the civil servants. First, civil servants were encouraged to participate voluntarily in mass civic projects during the weekends to enable them to get better acquainted with the political leaders, and to provide them with an opportunity to engage in manual work. Second, Nanyang University graduates, who were Chinese-educated, were recruited from 1960 for the education service to reduce the predominance of the English-educated in the SCS. Third, civil servants who were found guilty of misconduct were disciplined and the Central Complaints Bureau was formed in 1961 to enable the non-English-educated public to submit complaints against rude or incompetent civil servants. Finally, expatriate civil servants who were due for retirement were encouraged to remain in the SCS if they were competent while their incompetent colleagues were retired prematurely (Quah, 2010, pp. 134-5).In short, the PAP government relied on the reorganisation of the SCS and attitudinal reform to change the "colonial mentality" of the senior civil servants because it needed their support to implement its policies effectively. Corruption was a way of life in Singapore during the British colonial period because of the British colonial government's lack of political will and the ineffective anti-corruption measures adopted. Corruption was made illegal in Singapore with the enactment of the Penal Code of the Straits Settlements of Malacca, Penang and Singapore in 1871. However, even though police corruption was rampant and confirmed by the 1879 and 1886 Commissions of Inquiry, the British colonial government ignored their findings and delayed the enactment of the Prevention of Corruption Ordinance (POCO) until December 1937 (Quah, 2007, pp. 9-12).The POCO was ineffective because it limited the powers of arrest, search and investigation of police officers as warrants were required before arrests could be made; and the penalty of two years' imprisonment and/or a fine of S$10,000 for corrupt offenders did not deter corrupt behaviour (Quah, 1978, p. 9). Similarly, the Anti-Corruption Branch (ACB) was ineffective because of the prevalence of police corruption. As the ACB was part of the Criminal Investigation Department (CID) of the Singapore Police Force (SPF), it was not surprising that the ACB was ineffective in curbing police corruption. Furthermore, the ACB was inadequately staffed with only 17 personnel and had to compete with other branches in the CID for limited manpower and other resources (Quah, 2007, pp. 14-15).The British colonial government only realized the folly of making the ACB responsible for curbing corruption when it was discovered that three police detectives and some senior police officers were involved in robbing 1,800lbs of opium worth S$400,000 (US$133,333) in October 1951 (Tan, 1999, p. 59). The Opium Hijacking scandal exposed the ACB's weaknesses and its inability to curb police corruption. Consequently, the British colonial government formed the CPIB as an independent agency in October 1952 to replace the ineffective ACB.During their campaign for the May 1959 general election, the PAP leaders demonstrated their commitment to curbing corruption by exposing the acceptance of S$700,000 by the Minister for Education, Chew Swee Kee, from some American donors (Quah, 2010, p. 175). The Labour Front government led by Chief Minister Lim Yew Hock was described as "being corrupted from head to toe" by a retired architect, Lee Kip Lin (Yap et al., 2009, p. 555). The PAP's revelation of the Chew Swee Kee scandal enabled it to win the May 30, 1959 general election by capturing 43 of the 51 seats and obtaining 53.4 per cent of the votes cast.When the PAP leaders assumed office in June 1959, an immediate priority was to ensure a clean government by adopting a zero-tolerance policy toward corruption. Minister Mentor Lee Kuan Yew explained in his memoirs why he and his colleagues were determined to keep Singapore free from corruption from the outset of their administration:We were sickened by the greed, corruption and decadence of many Asian leaders. [...] We had a deep sense of mission to establish a clean and effective government. When we took the oath of office [...] in June 1959, we all wore white shirts and white slacks to symbolize purity and honesty in our personal behaviour and our public life. [...] We made sure from the day we took office in June 1959 that every dollar in revenue would be properly accounted for and would reach the beneficiaries at the grass roots as one dollar, without being siphoned off along the way. So from the very beginning we gave special attention to the areas where discretionary powers had been exploited for personal gain and sharpened the instruments that could prevent, detect or deter such practices (Lee, 2000, pp. 182-4).As corruption was a way of life in Singapore in June 1959, the PAP leaders learnt from the mistakes made by the British colonial government in curbing corruption and demonstrated their commitment by enacting the POCA on June 17, 1960 to replace the ineffective POCO and to strengthen the CPIB by providing it with more legal powers, personnel and funding.The POCA has three important features to rectify the POCO's weaknesses and to enhance the CPIB's legal powers and increase its personnel. First, the penalty for corruption has been increased to imprisonment for five years and/or a fine of S$10,000 to enhance the POCA's deterrent effect. Second, according to section 13, a person found guilty of accepting an illegal gratification has to pay the amount he had taken as a bribe in addition to any other punishment imposed by a court. The third and most important feature of the POCA is that it has given the CPIB more powers and a new lease of life. For example, section 15 gives the CPIB officers powers of arrest and search of arrested persons. Furthermore, the CPIB's director and his senior officers are empowered by section 18 to investigate the bank account, share account or purchase account of any person suspected of committing a corruption offence. Section 24 is perhaps the most important asset for the CPIB in its investigation of corruption offences because "the fact that an accused person is in possession, for which he [or she] cannot satisfactorily account, of pecuniary resources or property disproportionate to his [or her] known sources of income" is evidence that he or she had obtained these pecuniary resources or property "corruptly as an inducement or reward" (Quah, 2010, pp. 176-7).To ensure the POCA's continued effectiveness, the PAP government has introduced whenever necessary, amendments or new legislation to deal with unanticipated problems or to plug legal loopholes. For example, in 1966, the POCA was amended so that a person could be found guilty of corruption without actually receiving the bribe as long as he had shown the intention of doing so (section 9). The POCA was also amended in 1966 so that, according to section 37, Singapore citizens working for their government in embassies and other government agencies abroad would be prosecuted for corrupt offences committed outside Singapore and would be dealt with as if such offences had occurred in Singapore. In 1989, the fine for corrupt offences was increased tenfold from S$10,000 to S$100,000 (US$78,730)2. On March 3, 1989, the Corruption (Confiscation of Benefits) Act 1989 was passed to enable the court to issue a confiscation order against the estate of a deceased defendant (Quah, 2010, pp. 177-8).Unlike the British colonial government, the PAP government has also demonstrated its political will in curbing corruption not only by enhancing the CPIB's legal powers but also by providing the CPIB with more personnel and budget during the past 53 years. Table I shows that the CPIB has grown by 17 times from eight officers in 1959 to 138 officers in 2011. Similarly, as indicated in Table II, the CPIB's budget has increased by nearly 20 times from S$1,024,370 in 1978 to S$34,073,400 in 2011 (Quah, 2010, pp. 179-80; Republic of Singapore, 2011, p. 378).In contrast to the situation during 1952-1959, the CPIB has adopted a "total approach to enforcement" by dealing with both "big and small cases" of corruption in both the public and private sectors, "both giver and receiver of bribes" and "other crimes uncovered in the course of [the] corruption investigation" (Soh, 2008a, pp. 1-2). In addition to its emphasis on investigation and enforcement, the CPIB also focuses on corruption prevention by reviewing the procedures and practices in those government agencies, where corruption has occurred and makes recommendations to remove the "loopholes and vulnerabilities." The CPIB employs this review process to "identify potential problem areas and loopholes" in order to minimize the opportunities for corruption (Soh, 2008b, p. 8).Finally, the CPIB's extensive outreach programme is implemented by its Public Education Group, which conducts prevention and education talks for pre-university students, principals, teachers, newly appointed civil servants, law enforcement agencies like the police and immigration department, and the management and staff of major organisations in key industries (Quah, 2010, p. 181). Table III shows that the number of persons attending the CPIB's prevention and education talks has increased from 2,500 in 2005 to 9,193 in 2010. Similarly, the number of foreign visitors to the CPIB has more than doubled from 1,000 to 2,538 during 2005-2010. The number of visitors from local organisations has risen from 20 in 2005 to 424 in 2008, and visits by students also increased from 150 to 791 from 2005-2009. The British introduced meritocracy to its colonies in Africa and Asia with the creation of the PSC, which is the adapted version of the Civil Service Commission in Britain. The raison d'etre of the PSC is twofold: to insulate the civil service from politics; and to accelerate its localisation by replacing the expatriate officers with qualified local staff (Sinker, 1953, p. 206). With the advent of self-government and the increasing control of the civil service in the hands of the local population, the PSC was established in the British colonies to insulate appointments, promotions and discipline in the civil service from politics. Accordingly, the PSC was created in India in 1926, in Ceylon (Sri Lanka) in 1931, in Pakistan in 1947, in Hong Kong in 1950, in Singapore and Nepal in 1951, in Malaya in 1957, and in Bangladesh in 1972 (Quah, 2009, p. 810).Meritocracy was introduced in Singapore with the establishment of the PSC on January 1, 1951. The PSC's origins can be traced to the White Paper (Command Paper No. 197) entitled Organisation of the Colonial Service issued by the British government in 1946. Command Paper No. 197 stressed that progress toward self-government could only be achieved if the public services of the colonies were adapted to local conditions and staffed to the maximum possible extent by local people. More importantly, it recommended the establishment of PSCs in the colonies to ensure that qualified local candidates would be recruited into the public services (Quah, 2010, p. 72).The PSC in Singapore was formed with these two objectives in mind: to keep politics out of the SCS and to accelerate the latter's pace of localisation (Quah, 1982, p. 50). The PSC's second objective is no longer important today because the localisation of the SCS was completed with the attainment of self-government in Singapore in June 1959. However, its primary aim of keeping politics out of the SCS remains relevant as the aim of the PSC's programme as stated in the national budget is "to meet the staffing requirements of the government in accordance with the merit principle" (Republic of Singapore, 1980, p. 78).The PSC's evolution during the past 60 years can be divided into four stages, as depicted in Table IV. During its first 31 years, the PSC was the central personnel agency responsible for selecting and promoting civil servants on the basis of merit, disciplinary control, and the granting of scholarships and training awards. As the PSC members and selection boards relied mainly on personal interviews to perform their functions, their workload increased tremendously during this stage. Indeed, the number of candidates interviewed by them for appointments and promotions increased by nearly 19 times from 556 candidates in 1951 to 10,430 candidates in 1982 (Public Service Commission, 1954, p. 2; Public Service Commission, 1983, p. 5). Similarly, the number of disciplinary cases completed has also risen from 24 to 169 during 1957 to 1982 (Public Service Commission, 1959, p. 8; Public Service Commission, 1983, p. 18). The number of scholarships and training fellowships awarded has also increased from 23 in 1963 to 847 in 1982 (Public Service Commission, 1964, pp. 9, 19; Public Service Commission, 1983, p. 8). Finally, the heavy workload of Singapore's PSC becomes obvious when its output of interviewing 58,712 candidates during 1964-1967 is more than nine times that of the PSC in Ceylon, which interviewed only 6,485 candidates during the same period (Quah, 1971, p. 140).The Public Service Division (PSD) was formed on January 3, 1983 on the recommendation of the Management Services Department to formulate and review personnel policies in the SCS and to ensure that these policies are implemented consistently in the various ministries. However, as the PSD was responsible for all personnel policy matters concerning appraisal, posting, training, schemes of service, service conditions, and welfare, its creation did not reduce the PSC's heavy workload of interviewing 50,274 candidates for appointments and promotions, completing 1,148 disciplinary cases, and granting 1,543 scholarships and training awards during 1983-1989 (Quah, 2010, p. 83).In March 1990, the Constitution of Singapore was amended to help the PSC cope with its heavy workload by increasing its membership from 11 to 15, including the chairman, and by creating two new sub-commissions namely, the Education Service Commission (ESC) for education and the Police and Civil Defence Services Commission (PCDSC) for the police and civil defence services. This move was designed to reduce the PSC's workload as the ESC would be responsible for appointing and promoting 21,000 teachers, and the PCDSC would take care of the appointment and promotion of 10,000 police, narcotics, prisons, and civil defence officers, thus leaving the PSC to deal with the remaining 34,000 civil servants. However, in spite of the establishment of the ESC and PCDSC, the PSC's workload was not reduced significantly as Table V shows that the PSC interviewed 9,993 candidates (67.4 per cent) for appointment during 1990-1994, in contrast to the 4,254 candidates (28.7 per cent) interviewed by the ESC, and the 573 candidates (3.9 per cent) interviewed by the PCDSC.To enhance the SCS's ability to compete with the private sector for talented personnel, the public personnel management system in Singapore was further decentralized in January 1995 with the establishment of a system of 31 personnel boards, as shown in Table VI. As the creation of the ESC and PCDSC did not alleviate significantly the PSC's workload in appointing candidates to the SCS during 1990-1994, the ESC and PCDSC were dissolved and amalgamated into a single PSC on 1 April 1998 (Public Service Commission, 1999, p. 9).Unlike the ESC and PCDSC, the 31 personnel boards have reduced considerably the PSC's workload from 1995 to 2010. Table VII shows that the PSC considered 1,724 candidates for appointment, 308 candidates for promotion, completed 832 disciplinary cases, and granted 2,305 scholarships and training awards from 1995-2010.Finally, the comparison of the PSC's workload during its second, third and fourth stages of development in Table VIII shows that the decentralisation of its functions, which began in 1990 with the formation of the ESC and PCDSC and ended with the creation of the 31 personnel boards in 1995, has been effective because the PSC's workload in interviewing candidates for appointment and promotion has been drastically reduced from 50,274 candidates during 1983-1989 to 18,463 candidates during 1990-1994, and to 2,032 candidates during 1995-2010. Similarly, the number of disciplinary cases completed by the PSC has declined from 1,148 cases during 1983-1989 to 832 cases during 1995-2010. On the other hand, it is not surprising that the number of scholarships and training awards granted by the PSC has increased from 1,543 during 1983-1989 to 2,305 during 1995-2010 because this constitutes the PSC's major function today. To balance the budget, a Cabinet Budget Committee on Expenditure recommended in June 1959 the removal of the variable allowances of Divisions I and II civil servants to save S$10 million and prevent a budget deficit of S$14 million (Quah, 2010, p. 103). The government restored the variable allowance in September 1961 with the improvement of the budgetary situation (Seah, 1971, p. 94).In 1968, the Harvey Commission recommended salary increases for five grades in the Division I superscale salaries. However, the government did not implement this recommendation until 1973 for two reasons: the economy could not afford a major salary revision and the private sector was not considered a serious threat in terms of competing for talent as promotion exercises for senior civil servants were conducted frequently to retain talented personnel in the SCS (Lee, 1995, pp. 21-2).However, the improvement of the Singapore economy in the 1970 s resulted in higher salaries in the private sector and aggravated the brain drain of talented civil servants to the private sector. The National Wages Council (NWC) was formed in February 1972 as an advisory body to formulate general guidelines on wage policies, to recommend annual wage adjustments, and to advise on incentive systems for improving efficiency and productivity (Then, 1998, pp. 220-1). The NWC recommended the payment of the Annual Wage Supplement (AWS) or "13th month pay" from 1972 to minimize the gap between salaries in the public and private sectors.The PAP government has relied on increasing the salaries of ministers and senior civil servants in 1973, 1979, 1982, 1989, and 1994 to reduce the growing differential with private sector salaries. On October 21, 1994, a White Paper on Competitive salaries for competent and honest government was presented to parliament to justify the benchmarking of the salaries of ministers and senior civil servants to the average salaries of the top four earners in six private sector professions namely, accounting, banking, engineering, law, local manufacturing companies, and multi-national corporations. The government accepted this recommendation and public sector salaries were benchmarked accordingly from January 1995 with the salaries of the six professions in the private sector (Quah, 2010, pp. 110-11).The 1997 Asian financial crisis and the subsequent slowing down of the Singapore economy resulted in a 2 per cent decrease in Superscale G and a 7 per cent decrease in Staff Grade I salaries and the reduction of the employers' contribution to the Central Provident Fund (CPF) from 20 per cent to 10 per cent for all employees. The purpose of the CPF reduction was to enhance the Singapore's competitiveness by lowering the cost of doing business. In other words, the reduction in the CPF contribution meant an additional decrease in the salaries of the ministers and senior civil servants.When the Singapore economy recovered in 1999 with a growth rate of 5.4 per cent, and the reduction of retrenchments from 29,100 in 1998 to 14,600 in 1999, wages in the private sector began to rise again. Unemployment fell from 4.3 per cent in December 1999 to 3.4 per cent in March 2000. With the tight labour market in Singapore and the improved conditions in the private sector, Deputy Prime Minister Lee Hsien Loong revealed in parliament on 29 June 2000 that eight administrative officers had resigned in 2000. The government increased the performance bonus component in the public sector salaries and broadened the benchmarking of these salaries to the top eight earners in the six private sector professions. Consequently, in June 2000, the variable component of annual salaries was increased from 30 per cent to 40 per cent of the total annual pay of the superscale administrative officers and ministers (Quah, 2010, pp. 113-14). However, public sector salaries were later reduced by a combined total of 30 per cent in November 2001 and July 2003 because of the recession.In December 2007, the PSD announced that the salaries of ministers and senior civil servants would be increased from 4 per cent to 21 per cent from January 2008. Table IX shows the salaries of the President, Prime Minister, Ministers, Permanent Secretaries, superscale civil servants at the entry grade, and Members of Parliament in 2007 and 2008. However, on 24 November 2008, the PSD announced that the salaries of administrative officers, political, judicial and statutory appointment holders would be decreased by 19 per cent in 2009 because of the economic recession. This means that the President's annual salary has been reduced from S$3.87 million to S$3.14 million. Similarly, the Prime Minister's annual salary has been decreased from S$3.76 million to S$3.04 million (PSD 2008, pp. 1-2). Nevertheless, the Prime Minister's annual salary of US$2,183,516 in 2010 made him the best paid political leader in the world (Economist, 2010a, b).The policy of paying competitive public sector salaries has been effective in curbing the brain drain of political leaders to the private sector as none of them have resigned from political office to work in the private sector before their retirement. The attractive remuneration for the permanent secretaries has also been effective in retaining them in the SCS as none of them have left for private sector jobs before their retirement. However, paying competitive public sector salaries has been less effective in preventing Division I officers from leaving the SCS. An analysis of the resignation rate of Division I officers in the SCS from 1971-1984 shows that the salary increases in 1972, 1973, 1979 and 1982 had failed to curb the exodus of senior civil servants to the private sector (Quah, 2010, pp. 93, 119-20). Have the above policies of changing the attitudes of civil servants, minimizing corruption, maintaining meritocracy in the SCS by decentralizing the PSC, and paying competitive salaries to attract the "best and brightest" citizens to the SCS, resulted in good governance in Singapore? The short answer is "Yes" as will be demonstrated in Singapore's rankings and scores on eight governance indicators.Government effectiveness Singapore has been ranked first for the competence of its public officials from 1999 to 2002 by the Global Competitiveness Report3. Singapore was ranked first among 59 countries in 1999 and 2000 (Schwab et al., 1999, p. 242; Porter et al., 2000, p. 238). Singapore's civil servants were also ranked first among their counterparts in 75 countries in 2001-2002 and 80 countries in 2002-2003 (Schwab et al., 2002, p. 399; Cornelius, 2003, p. 604).The effectiveness of Singapore's government has also been confirmed by Singapore's consistently high ranking on the World Bank's governance indicator on government effectiveness, which is defined as "the quality of public service provision, the quality of the bureaucracy, the competence of civil servants, the independence of the civil service from political pressures, and the credibility of the government's commitment to policies" (Kaufmann et al., 2004, p. 3). Table X shows that government effectiveness in Singapore is very high from 1996 to 2010 and ranges from 93.2 percentile rank in 2002 to 100 percentile rank in 1996, 1998, 2008 and 2010.Similarly, a 2010 survey of 12 Asian countries conducted by the Hong Kong-based Political Economic Risk Consultancy (PERC) found that the bureaucracy in Singapore is the most effective, followed by the bureaucracy in Hong Kong SAR. At the other extreme, the least effective bureaucracies are those in the Philippines, Indonesia and India in descending order. Table XI confirms that Singapore has the most effective bureaucracy among 12 Asian countries from 1998 to 2010. In short, Singapore's government and civil servants are effective and performed well according to the above indicators.Effectiveness in curbing corruption The effectiveness of the POCA and CPIB in curbing corruption is reflected in Singapore's consistently high rankings and scores on these five indicators: Transparency International's Corruption Perceptions Index (CPI) from 1995-2011; Political Economic PERC's annual surveys of corruption from 1995-2011; World Bank's Control of Corruption indicator from 1996-2010; World Bank's Doing Business Surveys from 2007-2012; and the Global Competitiveness Report's indicator on public trust of politicians from 1999 to 2010-2011.Table XII shows that first, Singapore has been consistently ranked as the least corrupt country in Asia from 1995-2009 and 2011, and the least corrupt country in the world in 2010 with Denmark and New Zealand. Second, PERC's annual surveys of corruption from 1995-2011 also confirm Singapore's status as the least corrupt country in Asia. Third, the World Bank's control of corruption indicator from 1996-2010 shows that Singapore has the highest percentile rank among all the Asian countries included in the surveysFourth, Table XIII confirms that red tape is not a problem at all in Singapore as the World Bank's Doing Business Surveys show that Singapore is ranked first among the 175-183 economies included from 2007-2012. The relative absence of red tape in Singapore is an important factor responsible for its low level of corruption.Fifth, Table XIV shows that Singapore has been consistently ranked first in the Global Competitiveness Report's indicator on the public trust of politicians among the 59 to 139 countries included in the surveys from 1999 to 2010/2011. Singapore's superior ranking on this indicator also reflects its high level of public trust in politicians and its low level of corruption.In short, the four policies introduced by the PAP government after assuming office in June 1959 have been effective in promoting good governance in Singapore as demonstrated by Singapore's consistently high rankings and scores on the abovementioned eight governance indicators. Is Singapore's experience in promoting good governance transferable to other Asian countries? As Singapore's success in ensuring good governance is the combined influence of the political will of the PAP government to solve the problems facing the country for the past 53 years and its favourable policy context, it will be difficult to transfer Singapore's experience in toto to other Asian countries because of the lack of political will and the unfavourable policy contexts in many Asian countries.Apart from good leadership, Singapore's favourable policy context has enabled its political leaders to stretch the constraints imposed by its small size and lack of resources by formulating and implementing effective policies to solve its problems during the past 53 years. As public administration Singapore-style is the product of the local policy context and the policies implemented by the PAP government, it would be difficult to replicate these policies to other Asian countries in view of the significant contextual differences between Singapore and these countries (Quah, 2010, pp. 246, 51).Table XV shows clearly the significant contextual differences between Singapore and the other 25 Asian countries. First, in terms of size, Singapore is the second smallest country after the Macao Special Administrative Region, which has a land area of 29.2 sq. km. At the other extreme, are the larger countries of China and India, which are 13,466 and 4,630 times larger respectively than Singapore. A second important contextual difference is Singapore's population, which is only larger than those of Mongolia's population of 2.7 million, Timor-Leste's population of 1.2 million, Bhutan's population of 0.7 million, Macao's population of 0.5 million, and Brunei's population of 0.4 million. On the other hand, Singapore's population of 5.1 million in 2010 is dwarfed by the huge populations of China, India, and Indonesia.The third contextual difference between Singapore and the other Asian countries is its economic affluence as manifested in its GDP per capita of US$40,920 in 2010, which is the second highest among all the 26 countries listed in Table XV. In contrast, the GDP per capita of these four countries are less than US$1,000 namely, Cambodia (US$760), Bangladesh (US$640), Afghanistan (US$517), and Nepal (US$490).In short, Singapore is a city-state, which is richer and smaller in terms of land area and population for the PAP government to govern than most of the other Asian countries. Singapore's favourable policy context has enabled the PAP government, which has been in power since June 1959, to implement policies effectively, to curb corruption, and to ensure the ease of doing business in Singapore, as demonstrated in Singapore's superior ranking on the World Bank's governance indicator on government effectiveness from 1996-2010, the Doing Business Surveys from 2007-2012, Transparency International's CPI from 1995-2011, and PERC's corruption surveys from 1995-2011. In his National Day Rally speech on 19 August 1984, then Prime Minister Lee Kuan Yew attributed Singapore's success to the quality of its political leadership:In the end, whatever the system, it is the quality of the men who run it, that is decisive. For they will decide what to make of the society, and how to get the people to give of their best. The Singapore system has worked. It will continue to work if you vote for honest, able and dedicated men, and you give them your best, for the good of all (Lee, 1984, p. 18).More recently, Lee acknowledged the importance to Singapore's development of attracting the "best and brightest" citizens to join the government and the SCS thus:My experience of developments in Asia has led me to conclude that we need good men to have good government. [...] The single most decisive factor that made for Singapore's development was the ability of its ministers and the high quality of the civil servants who supported them. [...] It was Singapore's good fortune that we had, for a small, developing country, a fair share of talent, because our own [talent] had been reinforced by the talented men and women who came here for their education, and stayed on for employment or business opportunities (Lee, 2000, pp. 735-6).In the same vein, Edgar H. Schein (1996, pp. 221-2) has identified the policy of having "the best and brightest" citizens in government as "probably one of Singapore's major strengths" because "they are potentially the most able to invent what the country needs to survive and grow." Furthermore, he has described Singapore as "one of the few models existing in the world of how a society can progress with a government that attempts to maximize intelligence, skill, and honesty."In sum, Singapore's transformation from a poor third world country in 1959 to an affluent and politically stable first world country today is the result of the ability of its political leaders and civil servants to formulate and implement policies to solve the country's problems during the past 53 years. First, the PAP leaders reorganised the SCS and changed the attitudes of the civil servants by convincing them to contribute to the attainment of national development goals. Second, they continued with the tradition of meritocracy introduced by the British by retaining and enhancing the PSC's effectiveness by reducing its heavy workload by decentralizing its functions of appointment and promotion to the ESC and PCDSC in 1990, and the 31 personnel boards in 1995. Third, they learnt from the mistakes made by the British in curbing corruption by enacting the POCA in June 1960 to enhance the CPIB's effectiveness in combating corruption. Fourth, the PAP government's success in promoting economic growth enabled it to compete for talented personnel with the private sector by paying competitive salaries to ministers and senior civil servants from 1972 onwards to prevent them from leaving for private sector jobs.In the final analysis, whether these four elements of Singapore's success - institutional and attitudinal reform of civil servants; zero-tolerance for corruption; meritocracy in appointing and promoting civil servants; and paying competitive salaries to attract the "best and brightest" citizens to join the government and civil service - can be replicated in other Asian countries depends mainly on whether their political leaders and senior civil servants have the political will and are prepared to pay the high economic and political costs of implementing these policies. Opens in a new window.Table I Growth of CPIB's personnel, 1952-2011 Opens in a new window.Table II Budget of the CPIB, 1978-2011 Opens in a new window.Table III Number of visitors to CPIB, 2005-2010 Opens in a new window.Table IV Evolution of the PSC in Singapore, 1951-2011 Opens in a new window.Table V Workloads of the PSC, ESC and PCDSC in appointing and promoting candidates in the SCS, 1990-1994 Opens in a new window.Table VI The SCS's Personnel Management System in 1995 Opens in a new window.Table VII Workload of the PSC in Singapore, 1995-2010 Opens in a new window.Table VIII Workload of the PSC in Singapore, 1983-2010 Opens in a new window.Table IX Salaries of key appointments in Singapore, 2007-2008 Opens in a new window.Table X Government effectiveness of Singapore, 1996-2010 Opens in a new window.Table XI PERC's evaluation of the effectiveness of bureaucracy in 12 Asian countries, 1998 to 2010 Opens in a new window.Table XII Singapore's performance on CPI, PERC and control of corruption 1995-2011 Opens in a new window.Table XIII Singapore's rank on the World Bank's ease of doing business, 2007-2012 Opens in a new window.Table XIV Public trust of politicians in Singapore, 1999-2010/2011 Opens in a new window.Table XV Policy contexts of 26 Asian countries
|
- The paper analyses four policies initiated by the PAP government: comprehensive reform of the Singapore Civil Service; anti-corruption measures; decentralization of the Public Service Commission; and payment of competitive salaries to attract and retain the best candidates to the government. The effectiveness of these policies is assessed by referring to Singapore's performance on eight governance indicators.
|
[SECTION: Findings] The People's Action Party (PAP) won the May 30, 1959 general election and assumed office and attained self-government from Britain on 3 June 1959. Singapore was a different place then because it was a poor third world country, afflicted with a serious housing shortage (half the population was living in squatter huts), an unemployment rate of 14 per cent, political instability, labour unrest, corruption, and a high crime rate. However, today, after 53 years under the PAP government, Singapore has been transformed into a first world country, which is no longer afflicted by the problems it faced in 1959.This article contends that Singapore's ability to solve the problems it encountered after attaining self-government can be attributed to the effectiveness of the various policies introduced by the PAP government since 1959. The PAP government created the Housing and Development Board (HDB) in February 1960 to tackle the housing shortage and the Economic Development Board (EDB) in August 1961 to create jobs by attracting foreign investment to Singapore. These two statutory boards were formed to reduce the workload of the Singapore Civil Service (SCS), which was not equipped to solve the housing shortage or create jobs. Apart from being handicapped by rigid regulations and inflexibility, the civil servants also had a "colonial mentality" and were not attuned to the problems facing Singapore.Accordingly, the PAP government initiated these four policies to solve the country's problems:1. Reorganization and attitudinal reform of the SCS.2. Enactment of the Prevention of Corruption Act (POCA) in June 1960 to empower the Corrupt Practices Investigation Bureau (CPIB) (2011) to curb corruption effectively.3. Maintaining the tradition of meritocracy introduced by the British colonial government by retaining and decentralizing the Public Service Commission (PSC) to enhance its effectiveness.4. Paying competitive salaries to ministers and senior civil servants from 1972 and benchmarking these salaries from 1995 to private sector salaries to attract the "best and brightest" citizens to the SCS and government and to minimize the brain drain to the private sector.These four policies will be analyzed in turn below before providing the governance indicators for Singapore and assessing the transferability of Singapore's experience in promoting good governance to other Asian countries. The PAP government realized on assuming office in June 1959 that it had to transform the colonial bureaucracy it inherited in order to ensure the effective implementation of its socio-economic development programmes. During the 137 years of British colonial rule1, the SCS focused on performing the traditional "housekeeping" functions of maintaining law and order, building public works and collecting taxes. Consequently, the SCS did not play an important role in national development and did not introduce administrative reforms until after the Second World War (Quah, 1996a, pp. 62-3).Its victory in the May 1959 general election gave the PAP government the mandate to introduce comprehensive reforms to the SCS and to change the civil servants' "colonial mentality" by making them more sensitive to the needs of the population. Accordingly, it began by reorganising the SCS into nine ministries including the new ministries of culture and national development, which were created to deal with nation-building and economic development respectively (Seah, 1971, pp. 82-3). The workload of the SCS was reduced by making the HDB and EDB responsible for providing low-cost public housing and attracting foreign investment to Singapore. Unlike the SCS, which could not implement the public housing and industrialisation programmes swiftly because of its regulations and red tape, the HDB and EDB as statutory boards were better equipped to implement these programmes expeditiously (Quah, 1985, pp. 124-6).In addition to reorganising the SCS, the PAP government realized the necessity of changing the civil servants' attitudes and convincing them to participate in the process of attaining national development goals. The Political Study Centre was opened by Prime Minister Lee Kuan Yew on August 15, 1959. In his opening speech, he hoped that the civil servants would change their "colonial mentality" once they were made aware of the problems facing Singapore. Their responsibility was to ensure the efficiency of the SCS. The purpose of the two-week part-time and non-residential course conducted by the Political Study Centre for senior civil servants was to change their attitudes and make them more aware of the local contextual constraints (Quah, 2010, p. 134).In addition to the Political Study Centre, the PAP government also relied on four additional methods to change the attitudes and behaviour of the civil servants. First, civil servants were encouraged to participate voluntarily in mass civic projects during the weekends to enable them to get better acquainted with the political leaders, and to provide them with an opportunity to engage in manual work. Second, Nanyang University graduates, who were Chinese-educated, were recruited from 1960 for the education service to reduce the predominance of the English-educated in the SCS. Third, civil servants who were found guilty of misconduct were disciplined and the Central Complaints Bureau was formed in 1961 to enable the non-English-educated public to submit complaints against rude or incompetent civil servants. Finally, expatriate civil servants who were due for retirement were encouraged to remain in the SCS if they were competent while their incompetent colleagues were retired prematurely (Quah, 2010, pp. 134-5).In short, the PAP government relied on the reorganisation of the SCS and attitudinal reform to change the "colonial mentality" of the senior civil servants because it needed their support to implement its policies effectively. Corruption was a way of life in Singapore during the British colonial period because of the British colonial government's lack of political will and the ineffective anti-corruption measures adopted. Corruption was made illegal in Singapore with the enactment of the Penal Code of the Straits Settlements of Malacca, Penang and Singapore in 1871. However, even though police corruption was rampant and confirmed by the 1879 and 1886 Commissions of Inquiry, the British colonial government ignored their findings and delayed the enactment of the Prevention of Corruption Ordinance (POCO) until December 1937 (Quah, 2007, pp. 9-12).The POCO was ineffective because it limited the powers of arrest, search and investigation of police officers as warrants were required before arrests could be made; and the penalty of two years' imprisonment and/or a fine of S$10,000 for corrupt offenders did not deter corrupt behaviour (Quah, 1978, p. 9). Similarly, the Anti-Corruption Branch (ACB) was ineffective because of the prevalence of police corruption. As the ACB was part of the Criminal Investigation Department (CID) of the Singapore Police Force (SPF), it was not surprising that the ACB was ineffective in curbing police corruption. Furthermore, the ACB was inadequately staffed with only 17 personnel and had to compete with other branches in the CID for limited manpower and other resources (Quah, 2007, pp. 14-15).The British colonial government only realized the folly of making the ACB responsible for curbing corruption when it was discovered that three police detectives and some senior police officers were involved in robbing 1,800lbs of opium worth S$400,000 (US$133,333) in October 1951 (Tan, 1999, p. 59). The Opium Hijacking scandal exposed the ACB's weaknesses and its inability to curb police corruption. Consequently, the British colonial government formed the CPIB as an independent agency in October 1952 to replace the ineffective ACB.During their campaign for the May 1959 general election, the PAP leaders demonstrated their commitment to curbing corruption by exposing the acceptance of S$700,000 by the Minister for Education, Chew Swee Kee, from some American donors (Quah, 2010, p. 175). The Labour Front government led by Chief Minister Lim Yew Hock was described as "being corrupted from head to toe" by a retired architect, Lee Kip Lin (Yap et al., 2009, p. 555). The PAP's revelation of the Chew Swee Kee scandal enabled it to win the May 30, 1959 general election by capturing 43 of the 51 seats and obtaining 53.4 per cent of the votes cast.When the PAP leaders assumed office in June 1959, an immediate priority was to ensure a clean government by adopting a zero-tolerance policy toward corruption. Minister Mentor Lee Kuan Yew explained in his memoirs why he and his colleagues were determined to keep Singapore free from corruption from the outset of their administration:We were sickened by the greed, corruption and decadence of many Asian leaders. [...] We had a deep sense of mission to establish a clean and effective government. When we took the oath of office [...] in June 1959, we all wore white shirts and white slacks to symbolize purity and honesty in our personal behaviour and our public life. [...] We made sure from the day we took office in June 1959 that every dollar in revenue would be properly accounted for and would reach the beneficiaries at the grass roots as one dollar, without being siphoned off along the way. So from the very beginning we gave special attention to the areas where discretionary powers had been exploited for personal gain and sharpened the instruments that could prevent, detect or deter such practices (Lee, 2000, pp. 182-4).As corruption was a way of life in Singapore in June 1959, the PAP leaders learnt from the mistakes made by the British colonial government in curbing corruption and demonstrated their commitment by enacting the POCA on June 17, 1960 to replace the ineffective POCO and to strengthen the CPIB by providing it with more legal powers, personnel and funding.The POCA has three important features to rectify the POCO's weaknesses and to enhance the CPIB's legal powers and increase its personnel. First, the penalty for corruption has been increased to imprisonment for five years and/or a fine of S$10,000 to enhance the POCA's deterrent effect. Second, according to section 13, a person found guilty of accepting an illegal gratification has to pay the amount he had taken as a bribe in addition to any other punishment imposed by a court. The third and most important feature of the POCA is that it has given the CPIB more powers and a new lease of life. For example, section 15 gives the CPIB officers powers of arrest and search of arrested persons. Furthermore, the CPIB's director and his senior officers are empowered by section 18 to investigate the bank account, share account or purchase account of any person suspected of committing a corruption offence. Section 24 is perhaps the most important asset for the CPIB in its investigation of corruption offences because "the fact that an accused person is in possession, for which he [or she] cannot satisfactorily account, of pecuniary resources or property disproportionate to his [or her] known sources of income" is evidence that he or she had obtained these pecuniary resources or property "corruptly as an inducement or reward" (Quah, 2010, pp. 176-7).To ensure the POCA's continued effectiveness, the PAP government has introduced whenever necessary, amendments or new legislation to deal with unanticipated problems or to plug legal loopholes. For example, in 1966, the POCA was amended so that a person could be found guilty of corruption without actually receiving the bribe as long as he had shown the intention of doing so (section 9). The POCA was also amended in 1966 so that, according to section 37, Singapore citizens working for their government in embassies and other government agencies abroad would be prosecuted for corrupt offences committed outside Singapore and would be dealt with as if such offences had occurred in Singapore. In 1989, the fine for corrupt offences was increased tenfold from S$10,000 to S$100,000 (US$78,730)2. On March 3, 1989, the Corruption (Confiscation of Benefits) Act 1989 was passed to enable the court to issue a confiscation order against the estate of a deceased defendant (Quah, 2010, pp. 177-8).Unlike the British colonial government, the PAP government has also demonstrated its political will in curbing corruption not only by enhancing the CPIB's legal powers but also by providing the CPIB with more personnel and budget during the past 53 years. Table I shows that the CPIB has grown by 17 times from eight officers in 1959 to 138 officers in 2011. Similarly, as indicated in Table II, the CPIB's budget has increased by nearly 20 times from S$1,024,370 in 1978 to S$34,073,400 in 2011 (Quah, 2010, pp. 179-80; Republic of Singapore, 2011, p. 378).In contrast to the situation during 1952-1959, the CPIB has adopted a "total approach to enforcement" by dealing with both "big and small cases" of corruption in both the public and private sectors, "both giver and receiver of bribes" and "other crimes uncovered in the course of [the] corruption investigation" (Soh, 2008a, pp. 1-2). In addition to its emphasis on investigation and enforcement, the CPIB also focuses on corruption prevention by reviewing the procedures and practices in those government agencies, where corruption has occurred and makes recommendations to remove the "loopholes and vulnerabilities." The CPIB employs this review process to "identify potential problem areas and loopholes" in order to minimize the opportunities for corruption (Soh, 2008b, p. 8).Finally, the CPIB's extensive outreach programme is implemented by its Public Education Group, which conducts prevention and education talks for pre-university students, principals, teachers, newly appointed civil servants, law enforcement agencies like the police and immigration department, and the management and staff of major organisations in key industries (Quah, 2010, p. 181). Table III shows that the number of persons attending the CPIB's prevention and education talks has increased from 2,500 in 2005 to 9,193 in 2010. Similarly, the number of foreign visitors to the CPIB has more than doubled from 1,000 to 2,538 during 2005-2010. The number of visitors from local organisations has risen from 20 in 2005 to 424 in 2008, and visits by students also increased from 150 to 791 from 2005-2009. The British introduced meritocracy to its colonies in Africa and Asia with the creation of the PSC, which is the adapted version of the Civil Service Commission in Britain. The raison d'etre of the PSC is twofold: to insulate the civil service from politics; and to accelerate its localisation by replacing the expatriate officers with qualified local staff (Sinker, 1953, p. 206). With the advent of self-government and the increasing control of the civil service in the hands of the local population, the PSC was established in the British colonies to insulate appointments, promotions and discipline in the civil service from politics. Accordingly, the PSC was created in India in 1926, in Ceylon (Sri Lanka) in 1931, in Pakistan in 1947, in Hong Kong in 1950, in Singapore and Nepal in 1951, in Malaya in 1957, and in Bangladesh in 1972 (Quah, 2009, p. 810).Meritocracy was introduced in Singapore with the establishment of the PSC on January 1, 1951. The PSC's origins can be traced to the White Paper (Command Paper No. 197) entitled Organisation of the Colonial Service issued by the British government in 1946. Command Paper No. 197 stressed that progress toward self-government could only be achieved if the public services of the colonies were adapted to local conditions and staffed to the maximum possible extent by local people. More importantly, it recommended the establishment of PSCs in the colonies to ensure that qualified local candidates would be recruited into the public services (Quah, 2010, p. 72).The PSC in Singapore was formed with these two objectives in mind: to keep politics out of the SCS and to accelerate the latter's pace of localisation (Quah, 1982, p. 50). The PSC's second objective is no longer important today because the localisation of the SCS was completed with the attainment of self-government in Singapore in June 1959. However, its primary aim of keeping politics out of the SCS remains relevant as the aim of the PSC's programme as stated in the national budget is "to meet the staffing requirements of the government in accordance with the merit principle" (Republic of Singapore, 1980, p. 78).The PSC's evolution during the past 60 years can be divided into four stages, as depicted in Table IV. During its first 31 years, the PSC was the central personnel agency responsible for selecting and promoting civil servants on the basis of merit, disciplinary control, and the granting of scholarships and training awards. As the PSC members and selection boards relied mainly on personal interviews to perform their functions, their workload increased tremendously during this stage. Indeed, the number of candidates interviewed by them for appointments and promotions increased by nearly 19 times from 556 candidates in 1951 to 10,430 candidates in 1982 (Public Service Commission, 1954, p. 2; Public Service Commission, 1983, p. 5). Similarly, the number of disciplinary cases completed has also risen from 24 to 169 during 1957 to 1982 (Public Service Commission, 1959, p. 8; Public Service Commission, 1983, p. 18). The number of scholarships and training fellowships awarded has also increased from 23 in 1963 to 847 in 1982 (Public Service Commission, 1964, pp. 9, 19; Public Service Commission, 1983, p. 8). Finally, the heavy workload of Singapore's PSC becomes obvious when its output of interviewing 58,712 candidates during 1964-1967 is more than nine times that of the PSC in Ceylon, which interviewed only 6,485 candidates during the same period (Quah, 1971, p. 140).The Public Service Division (PSD) was formed on January 3, 1983 on the recommendation of the Management Services Department to formulate and review personnel policies in the SCS and to ensure that these policies are implemented consistently in the various ministries. However, as the PSD was responsible for all personnel policy matters concerning appraisal, posting, training, schemes of service, service conditions, and welfare, its creation did not reduce the PSC's heavy workload of interviewing 50,274 candidates for appointments and promotions, completing 1,148 disciplinary cases, and granting 1,543 scholarships and training awards during 1983-1989 (Quah, 2010, p. 83).In March 1990, the Constitution of Singapore was amended to help the PSC cope with its heavy workload by increasing its membership from 11 to 15, including the chairman, and by creating two new sub-commissions namely, the Education Service Commission (ESC) for education and the Police and Civil Defence Services Commission (PCDSC) for the police and civil defence services. This move was designed to reduce the PSC's workload as the ESC would be responsible for appointing and promoting 21,000 teachers, and the PCDSC would take care of the appointment and promotion of 10,000 police, narcotics, prisons, and civil defence officers, thus leaving the PSC to deal with the remaining 34,000 civil servants. However, in spite of the establishment of the ESC and PCDSC, the PSC's workload was not reduced significantly as Table V shows that the PSC interviewed 9,993 candidates (67.4 per cent) for appointment during 1990-1994, in contrast to the 4,254 candidates (28.7 per cent) interviewed by the ESC, and the 573 candidates (3.9 per cent) interviewed by the PCDSC.To enhance the SCS's ability to compete with the private sector for talented personnel, the public personnel management system in Singapore was further decentralized in January 1995 with the establishment of a system of 31 personnel boards, as shown in Table VI. As the creation of the ESC and PCDSC did not alleviate significantly the PSC's workload in appointing candidates to the SCS during 1990-1994, the ESC and PCDSC were dissolved and amalgamated into a single PSC on 1 April 1998 (Public Service Commission, 1999, p. 9).Unlike the ESC and PCDSC, the 31 personnel boards have reduced considerably the PSC's workload from 1995 to 2010. Table VII shows that the PSC considered 1,724 candidates for appointment, 308 candidates for promotion, completed 832 disciplinary cases, and granted 2,305 scholarships and training awards from 1995-2010.Finally, the comparison of the PSC's workload during its second, third and fourth stages of development in Table VIII shows that the decentralisation of its functions, which began in 1990 with the formation of the ESC and PCDSC and ended with the creation of the 31 personnel boards in 1995, has been effective because the PSC's workload in interviewing candidates for appointment and promotion has been drastically reduced from 50,274 candidates during 1983-1989 to 18,463 candidates during 1990-1994, and to 2,032 candidates during 1995-2010. Similarly, the number of disciplinary cases completed by the PSC has declined from 1,148 cases during 1983-1989 to 832 cases during 1995-2010. On the other hand, it is not surprising that the number of scholarships and training awards granted by the PSC has increased from 1,543 during 1983-1989 to 2,305 during 1995-2010 because this constitutes the PSC's major function today. To balance the budget, a Cabinet Budget Committee on Expenditure recommended in June 1959 the removal of the variable allowances of Divisions I and II civil servants to save S$10 million and prevent a budget deficit of S$14 million (Quah, 2010, p. 103). The government restored the variable allowance in September 1961 with the improvement of the budgetary situation (Seah, 1971, p. 94).In 1968, the Harvey Commission recommended salary increases for five grades in the Division I superscale salaries. However, the government did not implement this recommendation until 1973 for two reasons: the economy could not afford a major salary revision and the private sector was not considered a serious threat in terms of competing for talent as promotion exercises for senior civil servants were conducted frequently to retain talented personnel in the SCS (Lee, 1995, pp. 21-2).However, the improvement of the Singapore economy in the 1970 s resulted in higher salaries in the private sector and aggravated the brain drain of talented civil servants to the private sector. The National Wages Council (NWC) was formed in February 1972 as an advisory body to formulate general guidelines on wage policies, to recommend annual wage adjustments, and to advise on incentive systems for improving efficiency and productivity (Then, 1998, pp. 220-1). The NWC recommended the payment of the Annual Wage Supplement (AWS) or "13th month pay" from 1972 to minimize the gap between salaries in the public and private sectors.The PAP government has relied on increasing the salaries of ministers and senior civil servants in 1973, 1979, 1982, 1989, and 1994 to reduce the growing differential with private sector salaries. On October 21, 1994, a White Paper on Competitive salaries for competent and honest government was presented to parliament to justify the benchmarking of the salaries of ministers and senior civil servants to the average salaries of the top four earners in six private sector professions namely, accounting, banking, engineering, law, local manufacturing companies, and multi-national corporations. The government accepted this recommendation and public sector salaries were benchmarked accordingly from January 1995 with the salaries of the six professions in the private sector (Quah, 2010, pp. 110-11).The 1997 Asian financial crisis and the subsequent slowing down of the Singapore economy resulted in a 2 per cent decrease in Superscale G and a 7 per cent decrease in Staff Grade I salaries and the reduction of the employers' contribution to the Central Provident Fund (CPF) from 20 per cent to 10 per cent for all employees. The purpose of the CPF reduction was to enhance the Singapore's competitiveness by lowering the cost of doing business. In other words, the reduction in the CPF contribution meant an additional decrease in the salaries of the ministers and senior civil servants.When the Singapore economy recovered in 1999 with a growth rate of 5.4 per cent, and the reduction of retrenchments from 29,100 in 1998 to 14,600 in 1999, wages in the private sector began to rise again. Unemployment fell from 4.3 per cent in December 1999 to 3.4 per cent in March 2000. With the tight labour market in Singapore and the improved conditions in the private sector, Deputy Prime Minister Lee Hsien Loong revealed in parliament on 29 June 2000 that eight administrative officers had resigned in 2000. The government increased the performance bonus component in the public sector salaries and broadened the benchmarking of these salaries to the top eight earners in the six private sector professions. Consequently, in June 2000, the variable component of annual salaries was increased from 30 per cent to 40 per cent of the total annual pay of the superscale administrative officers and ministers (Quah, 2010, pp. 113-14). However, public sector salaries were later reduced by a combined total of 30 per cent in November 2001 and July 2003 because of the recession.In December 2007, the PSD announced that the salaries of ministers and senior civil servants would be increased from 4 per cent to 21 per cent from January 2008. Table IX shows the salaries of the President, Prime Minister, Ministers, Permanent Secretaries, superscale civil servants at the entry grade, and Members of Parliament in 2007 and 2008. However, on 24 November 2008, the PSD announced that the salaries of administrative officers, political, judicial and statutory appointment holders would be decreased by 19 per cent in 2009 because of the economic recession. This means that the President's annual salary has been reduced from S$3.87 million to S$3.14 million. Similarly, the Prime Minister's annual salary has been decreased from S$3.76 million to S$3.04 million (PSD 2008, pp. 1-2). Nevertheless, the Prime Minister's annual salary of US$2,183,516 in 2010 made him the best paid political leader in the world (Economist, 2010a, b).The policy of paying competitive public sector salaries has been effective in curbing the brain drain of political leaders to the private sector as none of them have resigned from political office to work in the private sector before their retirement. The attractive remuneration for the permanent secretaries has also been effective in retaining them in the SCS as none of them have left for private sector jobs before their retirement. However, paying competitive public sector salaries has been less effective in preventing Division I officers from leaving the SCS. An analysis of the resignation rate of Division I officers in the SCS from 1971-1984 shows that the salary increases in 1972, 1973, 1979 and 1982 had failed to curb the exodus of senior civil servants to the private sector (Quah, 2010, pp. 93, 119-20). Have the above policies of changing the attitudes of civil servants, minimizing corruption, maintaining meritocracy in the SCS by decentralizing the PSC, and paying competitive salaries to attract the "best and brightest" citizens to the SCS, resulted in good governance in Singapore? The short answer is "Yes" as will be demonstrated in Singapore's rankings and scores on eight governance indicators.Government effectiveness Singapore has been ranked first for the competence of its public officials from 1999 to 2002 by the Global Competitiveness Report3. Singapore was ranked first among 59 countries in 1999 and 2000 (Schwab et al., 1999, p. 242; Porter et al., 2000, p. 238). Singapore's civil servants were also ranked first among their counterparts in 75 countries in 2001-2002 and 80 countries in 2002-2003 (Schwab et al., 2002, p. 399; Cornelius, 2003, p. 604).The effectiveness of Singapore's government has also been confirmed by Singapore's consistently high ranking on the World Bank's governance indicator on government effectiveness, which is defined as "the quality of public service provision, the quality of the bureaucracy, the competence of civil servants, the independence of the civil service from political pressures, and the credibility of the government's commitment to policies" (Kaufmann et al., 2004, p. 3). Table X shows that government effectiveness in Singapore is very high from 1996 to 2010 and ranges from 93.2 percentile rank in 2002 to 100 percentile rank in 1996, 1998, 2008 and 2010.Similarly, a 2010 survey of 12 Asian countries conducted by the Hong Kong-based Political Economic Risk Consultancy (PERC) found that the bureaucracy in Singapore is the most effective, followed by the bureaucracy in Hong Kong SAR. At the other extreme, the least effective bureaucracies are those in the Philippines, Indonesia and India in descending order. Table XI confirms that Singapore has the most effective bureaucracy among 12 Asian countries from 1998 to 2010. In short, Singapore's government and civil servants are effective and performed well according to the above indicators.Effectiveness in curbing corruption The effectiveness of the POCA and CPIB in curbing corruption is reflected in Singapore's consistently high rankings and scores on these five indicators: Transparency International's Corruption Perceptions Index (CPI) from 1995-2011; Political Economic PERC's annual surveys of corruption from 1995-2011; World Bank's Control of Corruption indicator from 1996-2010; World Bank's Doing Business Surveys from 2007-2012; and the Global Competitiveness Report's indicator on public trust of politicians from 1999 to 2010-2011.Table XII shows that first, Singapore has been consistently ranked as the least corrupt country in Asia from 1995-2009 and 2011, and the least corrupt country in the world in 2010 with Denmark and New Zealand. Second, PERC's annual surveys of corruption from 1995-2011 also confirm Singapore's status as the least corrupt country in Asia. Third, the World Bank's control of corruption indicator from 1996-2010 shows that Singapore has the highest percentile rank among all the Asian countries included in the surveysFourth, Table XIII confirms that red tape is not a problem at all in Singapore as the World Bank's Doing Business Surveys show that Singapore is ranked first among the 175-183 economies included from 2007-2012. The relative absence of red tape in Singapore is an important factor responsible for its low level of corruption.Fifth, Table XIV shows that Singapore has been consistently ranked first in the Global Competitiveness Report's indicator on the public trust of politicians among the 59 to 139 countries included in the surveys from 1999 to 2010/2011. Singapore's superior ranking on this indicator also reflects its high level of public trust in politicians and its low level of corruption.In short, the four policies introduced by the PAP government after assuming office in June 1959 have been effective in promoting good governance in Singapore as demonstrated by Singapore's consistently high rankings and scores on the abovementioned eight governance indicators. Is Singapore's experience in promoting good governance transferable to other Asian countries? As Singapore's success in ensuring good governance is the combined influence of the political will of the PAP government to solve the problems facing the country for the past 53 years and its favourable policy context, it will be difficult to transfer Singapore's experience in toto to other Asian countries because of the lack of political will and the unfavourable policy contexts in many Asian countries.Apart from good leadership, Singapore's favourable policy context has enabled its political leaders to stretch the constraints imposed by its small size and lack of resources by formulating and implementing effective policies to solve its problems during the past 53 years. As public administration Singapore-style is the product of the local policy context and the policies implemented by the PAP government, it would be difficult to replicate these policies to other Asian countries in view of the significant contextual differences between Singapore and these countries (Quah, 2010, pp. 246, 51).Table XV shows clearly the significant contextual differences between Singapore and the other 25 Asian countries. First, in terms of size, Singapore is the second smallest country after the Macao Special Administrative Region, which has a land area of 29.2 sq. km. At the other extreme, are the larger countries of China and India, which are 13,466 and 4,630 times larger respectively than Singapore. A second important contextual difference is Singapore's population, which is only larger than those of Mongolia's population of 2.7 million, Timor-Leste's population of 1.2 million, Bhutan's population of 0.7 million, Macao's population of 0.5 million, and Brunei's population of 0.4 million. On the other hand, Singapore's population of 5.1 million in 2010 is dwarfed by the huge populations of China, India, and Indonesia.The third contextual difference between Singapore and the other Asian countries is its economic affluence as manifested in its GDP per capita of US$40,920 in 2010, which is the second highest among all the 26 countries listed in Table XV. In contrast, the GDP per capita of these four countries are less than US$1,000 namely, Cambodia (US$760), Bangladesh (US$640), Afghanistan (US$517), and Nepal (US$490).In short, Singapore is a city-state, which is richer and smaller in terms of land area and population for the PAP government to govern than most of the other Asian countries. Singapore's favourable policy context has enabled the PAP government, which has been in power since June 1959, to implement policies effectively, to curb corruption, and to ensure the ease of doing business in Singapore, as demonstrated in Singapore's superior ranking on the World Bank's governance indicator on government effectiveness from 1996-2010, the Doing Business Surveys from 2007-2012, Transparency International's CPI from 1995-2011, and PERC's corruption surveys from 1995-2011. In his National Day Rally speech on 19 August 1984, then Prime Minister Lee Kuan Yew attributed Singapore's success to the quality of its political leadership:In the end, whatever the system, it is the quality of the men who run it, that is decisive. For they will decide what to make of the society, and how to get the people to give of their best. The Singapore system has worked. It will continue to work if you vote for honest, able and dedicated men, and you give them your best, for the good of all (Lee, 1984, p. 18).More recently, Lee acknowledged the importance to Singapore's development of attracting the "best and brightest" citizens to join the government and the SCS thus:My experience of developments in Asia has led me to conclude that we need good men to have good government. [...] The single most decisive factor that made for Singapore's development was the ability of its ministers and the high quality of the civil servants who supported them. [...] It was Singapore's good fortune that we had, for a small, developing country, a fair share of talent, because our own [talent] had been reinforced by the talented men and women who came here for their education, and stayed on for employment or business opportunities (Lee, 2000, pp. 735-6).In the same vein, Edgar H. Schein (1996, pp. 221-2) has identified the policy of having "the best and brightest" citizens in government as "probably one of Singapore's major strengths" because "they are potentially the most able to invent what the country needs to survive and grow." Furthermore, he has described Singapore as "one of the few models existing in the world of how a society can progress with a government that attempts to maximize intelligence, skill, and honesty."In sum, Singapore's transformation from a poor third world country in 1959 to an affluent and politically stable first world country today is the result of the ability of its political leaders and civil servants to formulate and implement policies to solve the country's problems during the past 53 years. First, the PAP leaders reorganised the SCS and changed the attitudes of the civil servants by convincing them to contribute to the attainment of national development goals. Second, they continued with the tradition of meritocracy introduced by the British by retaining and enhancing the PSC's effectiveness by reducing its heavy workload by decentralizing its functions of appointment and promotion to the ESC and PCDSC in 1990, and the 31 personnel boards in 1995. Third, they learnt from the mistakes made by the British in curbing corruption by enacting the POCA in June 1960 to enhance the CPIB's effectiveness in combating corruption. Fourth, the PAP government's success in promoting economic growth enabled it to compete for talented personnel with the private sector by paying competitive salaries to ministers and senior civil servants from 1972 onwards to prevent them from leaving for private sector jobs.In the final analysis, whether these four elements of Singapore's success - institutional and attitudinal reform of civil servants; zero-tolerance for corruption; meritocracy in appointing and promoting civil servants; and paying competitive salaries to attract the "best and brightest" citizens to join the government and civil service - can be replicated in other Asian countries depends mainly on whether their political leaders and senior civil servants have the political will and are prepared to pay the high economic and political costs of implementing these policies. Opens in a new window.Table I Growth of CPIB's personnel, 1952-2011 Opens in a new window.Table II Budget of the CPIB, 1978-2011 Opens in a new window.Table III Number of visitors to CPIB, 2005-2010 Opens in a new window.Table IV Evolution of the PSC in Singapore, 1951-2011 Opens in a new window.Table V Workloads of the PSC, ESC and PCDSC in appointing and promoting candidates in the SCS, 1990-1994 Opens in a new window.Table VI The SCS's Personnel Management System in 1995 Opens in a new window.Table VII Workload of the PSC in Singapore, 1995-2010 Opens in a new window.Table VIII Workload of the PSC in Singapore, 1983-2010 Opens in a new window.Table IX Salaries of key appointments in Singapore, 2007-2008 Opens in a new window.Table X Government effectiveness of Singapore, 1996-2010 Opens in a new window.Table XI PERC's evaluation of the effectiveness of bureaucracy in 12 Asian countries, 1998 to 2010 Opens in a new window.Table XII Singapore's performance on CPI, PERC and control of corruption 1995-2011 Opens in a new window.Table XIII Singapore's rank on the World Bank's ease of doing business, 2007-2012 Opens in a new window.Table XIV Public trust of politicians in Singapore, 1999-2010/2011 Opens in a new window.Table XV Policy contexts of 26 Asian countries
|
- The four policies are effective, as reflected in Singapore's superior rankings and scores on eight indicators: Global Competitiveness Report's (GCR's) competence of public officials; World Bank's indicator on government effectiveness; Political Economic Risk Consultancy's (PERC's) survey on bureaucratic effectiveness; Transparency International's Corruption Perceptions Index; PERC's survey on corruption; World Bank's indicator on control of corruption; World Bank's ease of doing business survey; and GCR's public trust of politicians survey. However, as Singapore's good governance is the result of the PAP government's political will and its favourable policy context, it is difficult to transfer Singapore's experience elsewhere because of the limited political will and unfavourable policy contexts in many Asian countries.
|
[SECTION: Value] The People's Action Party (PAP) won the May 30, 1959 general election and assumed office and attained self-government from Britain on 3 June 1959. Singapore was a different place then because it was a poor third world country, afflicted with a serious housing shortage (half the population was living in squatter huts), an unemployment rate of 14 per cent, political instability, labour unrest, corruption, and a high crime rate. However, today, after 53 years under the PAP government, Singapore has been transformed into a first world country, which is no longer afflicted by the problems it faced in 1959.This article contends that Singapore's ability to solve the problems it encountered after attaining self-government can be attributed to the effectiveness of the various policies introduced by the PAP government since 1959. The PAP government created the Housing and Development Board (HDB) in February 1960 to tackle the housing shortage and the Economic Development Board (EDB) in August 1961 to create jobs by attracting foreign investment to Singapore. These two statutory boards were formed to reduce the workload of the Singapore Civil Service (SCS), which was not equipped to solve the housing shortage or create jobs. Apart from being handicapped by rigid regulations and inflexibility, the civil servants also had a "colonial mentality" and were not attuned to the problems facing Singapore.Accordingly, the PAP government initiated these four policies to solve the country's problems:1. Reorganization and attitudinal reform of the SCS.2. Enactment of the Prevention of Corruption Act (POCA) in June 1960 to empower the Corrupt Practices Investigation Bureau (CPIB) (2011) to curb corruption effectively.3. Maintaining the tradition of meritocracy introduced by the British colonial government by retaining and decentralizing the Public Service Commission (PSC) to enhance its effectiveness.4. Paying competitive salaries to ministers and senior civil servants from 1972 and benchmarking these salaries from 1995 to private sector salaries to attract the "best and brightest" citizens to the SCS and government and to minimize the brain drain to the private sector.These four policies will be analyzed in turn below before providing the governance indicators for Singapore and assessing the transferability of Singapore's experience in promoting good governance to other Asian countries. The PAP government realized on assuming office in June 1959 that it had to transform the colonial bureaucracy it inherited in order to ensure the effective implementation of its socio-economic development programmes. During the 137 years of British colonial rule1, the SCS focused on performing the traditional "housekeeping" functions of maintaining law and order, building public works and collecting taxes. Consequently, the SCS did not play an important role in national development and did not introduce administrative reforms until after the Second World War (Quah, 1996a, pp. 62-3).Its victory in the May 1959 general election gave the PAP government the mandate to introduce comprehensive reforms to the SCS and to change the civil servants' "colonial mentality" by making them more sensitive to the needs of the population. Accordingly, it began by reorganising the SCS into nine ministries including the new ministries of culture and national development, which were created to deal with nation-building and economic development respectively (Seah, 1971, pp. 82-3). The workload of the SCS was reduced by making the HDB and EDB responsible for providing low-cost public housing and attracting foreign investment to Singapore. Unlike the SCS, which could not implement the public housing and industrialisation programmes swiftly because of its regulations and red tape, the HDB and EDB as statutory boards were better equipped to implement these programmes expeditiously (Quah, 1985, pp. 124-6).In addition to reorganising the SCS, the PAP government realized the necessity of changing the civil servants' attitudes and convincing them to participate in the process of attaining national development goals. The Political Study Centre was opened by Prime Minister Lee Kuan Yew on August 15, 1959. In his opening speech, he hoped that the civil servants would change their "colonial mentality" once they were made aware of the problems facing Singapore. Their responsibility was to ensure the efficiency of the SCS. The purpose of the two-week part-time and non-residential course conducted by the Political Study Centre for senior civil servants was to change their attitudes and make them more aware of the local contextual constraints (Quah, 2010, p. 134).In addition to the Political Study Centre, the PAP government also relied on four additional methods to change the attitudes and behaviour of the civil servants. First, civil servants were encouraged to participate voluntarily in mass civic projects during the weekends to enable them to get better acquainted with the political leaders, and to provide them with an opportunity to engage in manual work. Second, Nanyang University graduates, who were Chinese-educated, were recruited from 1960 for the education service to reduce the predominance of the English-educated in the SCS. Third, civil servants who were found guilty of misconduct were disciplined and the Central Complaints Bureau was formed in 1961 to enable the non-English-educated public to submit complaints against rude or incompetent civil servants. Finally, expatriate civil servants who were due for retirement were encouraged to remain in the SCS if they were competent while their incompetent colleagues were retired prematurely (Quah, 2010, pp. 134-5).In short, the PAP government relied on the reorganisation of the SCS and attitudinal reform to change the "colonial mentality" of the senior civil servants because it needed their support to implement its policies effectively. Corruption was a way of life in Singapore during the British colonial period because of the British colonial government's lack of political will and the ineffective anti-corruption measures adopted. Corruption was made illegal in Singapore with the enactment of the Penal Code of the Straits Settlements of Malacca, Penang and Singapore in 1871. However, even though police corruption was rampant and confirmed by the 1879 and 1886 Commissions of Inquiry, the British colonial government ignored their findings and delayed the enactment of the Prevention of Corruption Ordinance (POCO) until December 1937 (Quah, 2007, pp. 9-12).The POCO was ineffective because it limited the powers of arrest, search and investigation of police officers as warrants were required before arrests could be made; and the penalty of two years' imprisonment and/or a fine of S$10,000 for corrupt offenders did not deter corrupt behaviour (Quah, 1978, p. 9). Similarly, the Anti-Corruption Branch (ACB) was ineffective because of the prevalence of police corruption. As the ACB was part of the Criminal Investigation Department (CID) of the Singapore Police Force (SPF), it was not surprising that the ACB was ineffective in curbing police corruption. Furthermore, the ACB was inadequately staffed with only 17 personnel and had to compete with other branches in the CID for limited manpower and other resources (Quah, 2007, pp. 14-15).The British colonial government only realized the folly of making the ACB responsible for curbing corruption when it was discovered that three police detectives and some senior police officers were involved in robbing 1,800lbs of opium worth S$400,000 (US$133,333) in October 1951 (Tan, 1999, p. 59). The Opium Hijacking scandal exposed the ACB's weaknesses and its inability to curb police corruption. Consequently, the British colonial government formed the CPIB as an independent agency in October 1952 to replace the ineffective ACB.During their campaign for the May 1959 general election, the PAP leaders demonstrated their commitment to curbing corruption by exposing the acceptance of S$700,000 by the Minister for Education, Chew Swee Kee, from some American donors (Quah, 2010, p. 175). The Labour Front government led by Chief Minister Lim Yew Hock was described as "being corrupted from head to toe" by a retired architect, Lee Kip Lin (Yap et al., 2009, p. 555). The PAP's revelation of the Chew Swee Kee scandal enabled it to win the May 30, 1959 general election by capturing 43 of the 51 seats and obtaining 53.4 per cent of the votes cast.When the PAP leaders assumed office in June 1959, an immediate priority was to ensure a clean government by adopting a zero-tolerance policy toward corruption. Minister Mentor Lee Kuan Yew explained in his memoirs why he and his colleagues were determined to keep Singapore free from corruption from the outset of their administration:We were sickened by the greed, corruption and decadence of many Asian leaders. [...] We had a deep sense of mission to establish a clean and effective government. When we took the oath of office [...] in June 1959, we all wore white shirts and white slacks to symbolize purity and honesty in our personal behaviour and our public life. [...] We made sure from the day we took office in June 1959 that every dollar in revenue would be properly accounted for and would reach the beneficiaries at the grass roots as one dollar, without being siphoned off along the way. So from the very beginning we gave special attention to the areas where discretionary powers had been exploited for personal gain and sharpened the instruments that could prevent, detect or deter such practices (Lee, 2000, pp. 182-4).As corruption was a way of life in Singapore in June 1959, the PAP leaders learnt from the mistakes made by the British colonial government in curbing corruption and demonstrated their commitment by enacting the POCA on June 17, 1960 to replace the ineffective POCO and to strengthen the CPIB by providing it with more legal powers, personnel and funding.The POCA has three important features to rectify the POCO's weaknesses and to enhance the CPIB's legal powers and increase its personnel. First, the penalty for corruption has been increased to imprisonment for five years and/or a fine of S$10,000 to enhance the POCA's deterrent effect. Second, according to section 13, a person found guilty of accepting an illegal gratification has to pay the amount he had taken as a bribe in addition to any other punishment imposed by a court. The third and most important feature of the POCA is that it has given the CPIB more powers and a new lease of life. For example, section 15 gives the CPIB officers powers of arrest and search of arrested persons. Furthermore, the CPIB's director and his senior officers are empowered by section 18 to investigate the bank account, share account or purchase account of any person suspected of committing a corruption offence. Section 24 is perhaps the most important asset for the CPIB in its investigation of corruption offences because "the fact that an accused person is in possession, for which he [or she] cannot satisfactorily account, of pecuniary resources or property disproportionate to his [or her] known sources of income" is evidence that he or she had obtained these pecuniary resources or property "corruptly as an inducement or reward" (Quah, 2010, pp. 176-7).To ensure the POCA's continued effectiveness, the PAP government has introduced whenever necessary, amendments or new legislation to deal with unanticipated problems or to plug legal loopholes. For example, in 1966, the POCA was amended so that a person could be found guilty of corruption without actually receiving the bribe as long as he had shown the intention of doing so (section 9). The POCA was also amended in 1966 so that, according to section 37, Singapore citizens working for their government in embassies and other government agencies abroad would be prosecuted for corrupt offences committed outside Singapore and would be dealt with as if such offences had occurred in Singapore. In 1989, the fine for corrupt offences was increased tenfold from S$10,000 to S$100,000 (US$78,730)2. On March 3, 1989, the Corruption (Confiscation of Benefits) Act 1989 was passed to enable the court to issue a confiscation order against the estate of a deceased defendant (Quah, 2010, pp. 177-8).Unlike the British colonial government, the PAP government has also demonstrated its political will in curbing corruption not only by enhancing the CPIB's legal powers but also by providing the CPIB with more personnel and budget during the past 53 years. Table I shows that the CPIB has grown by 17 times from eight officers in 1959 to 138 officers in 2011. Similarly, as indicated in Table II, the CPIB's budget has increased by nearly 20 times from S$1,024,370 in 1978 to S$34,073,400 in 2011 (Quah, 2010, pp. 179-80; Republic of Singapore, 2011, p. 378).In contrast to the situation during 1952-1959, the CPIB has adopted a "total approach to enforcement" by dealing with both "big and small cases" of corruption in both the public and private sectors, "both giver and receiver of bribes" and "other crimes uncovered in the course of [the] corruption investigation" (Soh, 2008a, pp. 1-2). In addition to its emphasis on investigation and enforcement, the CPIB also focuses on corruption prevention by reviewing the procedures and practices in those government agencies, where corruption has occurred and makes recommendations to remove the "loopholes and vulnerabilities." The CPIB employs this review process to "identify potential problem areas and loopholes" in order to minimize the opportunities for corruption (Soh, 2008b, p. 8).Finally, the CPIB's extensive outreach programme is implemented by its Public Education Group, which conducts prevention and education talks for pre-university students, principals, teachers, newly appointed civil servants, law enforcement agencies like the police and immigration department, and the management and staff of major organisations in key industries (Quah, 2010, p. 181). Table III shows that the number of persons attending the CPIB's prevention and education talks has increased from 2,500 in 2005 to 9,193 in 2010. Similarly, the number of foreign visitors to the CPIB has more than doubled from 1,000 to 2,538 during 2005-2010. The number of visitors from local organisations has risen from 20 in 2005 to 424 in 2008, and visits by students also increased from 150 to 791 from 2005-2009. The British introduced meritocracy to its colonies in Africa and Asia with the creation of the PSC, which is the adapted version of the Civil Service Commission in Britain. The raison d'etre of the PSC is twofold: to insulate the civil service from politics; and to accelerate its localisation by replacing the expatriate officers with qualified local staff (Sinker, 1953, p. 206). With the advent of self-government and the increasing control of the civil service in the hands of the local population, the PSC was established in the British colonies to insulate appointments, promotions and discipline in the civil service from politics. Accordingly, the PSC was created in India in 1926, in Ceylon (Sri Lanka) in 1931, in Pakistan in 1947, in Hong Kong in 1950, in Singapore and Nepal in 1951, in Malaya in 1957, and in Bangladesh in 1972 (Quah, 2009, p. 810).Meritocracy was introduced in Singapore with the establishment of the PSC on January 1, 1951. The PSC's origins can be traced to the White Paper (Command Paper No. 197) entitled Organisation of the Colonial Service issued by the British government in 1946. Command Paper No. 197 stressed that progress toward self-government could only be achieved if the public services of the colonies were adapted to local conditions and staffed to the maximum possible extent by local people. More importantly, it recommended the establishment of PSCs in the colonies to ensure that qualified local candidates would be recruited into the public services (Quah, 2010, p. 72).The PSC in Singapore was formed with these two objectives in mind: to keep politics out of the SCS and to accelerate the latter's pace of localisation (Quah, 1982, p. 50). The PSC's second objective is no longer important today because the localisation of the SCS was completed with the attainment of self-government in Singapore in June 1959. However, its primary aim of keeping politics out of the SCS remains relevant as the aim of the PSC's programme as stated in the national budget is "to meet the staffing requirements of the government in accordance with the merit principle" (Republic of Singapore, 1980, p. 78).The PSC's evolution during the past 60 years can be divided into four stages, as depicted in Table IV. During its first 31 years, the PSC was the central personnel agency responsible for selecting and promoting civil servants on the basis of merit, disciplinary control, and the granting of scholarships and training awards. As the PSC members and selection boards relied mainly on personal interviews to perform their functions, their workload increased tremendously during this stage. Indeed, the number of candidates interviewed by them for appointments and promotions increased by nearly 19 times from 556 candidates in 1951 to 10,430 candidates in 1982 (Public Service Commission, 1954, p. 2; Public Service Commission, 1983, p. 5). Similarly, the number of disciplinary cases completed has also risen from 24 to 169 during 1957 to 1982 (Public Service Commission, 1959, p. 8; Public Service Commission, 1983, p. 18). The number of scholarships and training fellowships awarded has also increased from 23 in 1963 to 847 in 1982 (Public Service Commission, 1964, pp. 9, 19; Public Service Commission, 1983, p. 8). Finally, the heavy workload of Singapore's PSC becomes obvious when its output of interviewing 58,712 candidates during 1964-1967 is more than nine times that of the PSC in Ceylon, which interviewed only 6,485 candidates during the same period (Quah, 1971, p. 140).The Public Service Division (PSD) was formed on January 3, 1983 on the recommendation of the Management Services Department to formulate and review personnel policies in the SCS and to ensure that these policies are implemented consistently in the various ministries. However, as the PSD was responsible for all personnel policy matters concerning appraisal, posting, training, schemes of service, service conditions, and welfare, its creation did not reduce the PSC's heavy workload of interviewing 50,274 candidates for appointments and promotions, completing 1,148 disciplinary cases, and granting 1,543 scholarships and training awards during 1983-1989 (Quah, 2010, p. 83).In March 1990, the Constitution of Singapore was amended to help the PSC cope with its heavy workload by increasing its membership from 11 to 15, including the chairman, and by creating two new sub-commissions namely, the Education Service Commission (ESC) for education and the Police and Civil Defence Services Commission (PCDSC) for the police and civil defence services. This move was designed to reduce the PSC's workload as the ESC would be responsible for appointing and promoting 21,000 teachers, and the PCDSC would take care of the appointment and promotion of 10,000 police, narcotics, prisons, and civil defence officers, thus leaving the PSC to deal with the remaining 34,000 civil servants. However, in spite of the establishment of the ESC and PCDSC, the PSC's workload was not reduced significantly as Table V shows that the PSC interviewed 9,993 candidates (67.4 per cent) for appointment during 1990-1994, in contrast to the 4,254 candidates (28.7 per cent) interviewed by the ESC, and the 573 candidates (3.9 per cent) interviewed by the PCDSC.To enhance the SCS's ability to compete with the private sector for talented personnel, the public personnel management system in Singapore was further decentralized in January 1995 with the establishment of a system of 31 personnel boards, as shown in Table VI. As the creation of the ESC and PCDSC did not alleviate significantly the PSC's workload in appointing candidates to the SCS during 1990-1994, the ESC and PCDSC were dissolved and amalgamated into a single PSC on 1 April 1998 (Public Service Commission, 1999, p. 9).Unlike the ESC and PCDSC, the 31 personnel boards have reduced considerably the PSC's workload from 1995 to 2010. Table VII shows that the PSC considered 1,724 candidates for appointment, 308 candidates for promotion, completed 832 disciplinary cases, and granted 2,305 scholarships and training awards from 1995-2010.Finally, the comparison of the PSC's workload during its second, third and fourth stages of development in Table VIII shows that the decentralisation of its functions, which began in 1990 with the formation of the ESC and PCDSC and ended with the creation of the 31 personnel boards in 1995, has been effective because the PSC's workload in interviewing candidates for appointment and promotion has been drastically reduced from 50,274 candidates during 1983-1989 to 18,463 candidates during 1990-1994, and to 2,032 candidates during 1995-2010. Similarly, the number of disciplinary cases completed by the PSC has declined from 1,148 cases during 1983-1989 to 832 cases during 1995-2010. On the other hand, it is not surprising that the number of scholarships and training awards granted by the PSC has increased from 1,543 during 1983-1989 to 2,305 during 1995-2010 because this constitutes the PSC's major function today. To balance the budget, a Cabinet Budget Committee on Expenditure recommended in June 1959 the removal of the variable allowances of Divisions I and II civil servants to save S$10 million and prevent a budget deficit of S$14 million (Quah, 2010, p. 103). The government restored the variable allowance in September 1961 with the improvement of the budgetary situation (Seah, 1971, p. 94).In 1968, the Harvey Commission recommended salary increases for five grades in the Division I superscale salaries. However, the government did not implement this recommendation until 1973 for two reasons: the economy could not afford a major salary revision and the private sector was not considered a serious threat in terms of competing for talent as promotion exercises for senior civil servants were conducted frequently to retain talented personnel in the SCS (Lee, 1995, pp. 21-2).However, the improvement of the Singapore economy in the 1970 s resulted in higher salaries in the private sector and aggravated the brain drain of talented civil servants to the private sector. The National Wages Council (NWC) was formed in February 1972 as an advisory body to formulate general guidelines on wage policies, to recommend annual wage adjustments, and to advise on incentive systems for improving efficiency and productivity (Then, 1998, pp. 220-1). The NWC recommended the payment of the Annual Wage Supplement (AWS) or "13th month pay" from 1972 to minimize the gap between salaries in the public and private sectors.The PAP government has relied on increasing the salaries of ministers and senior civil servants in 1973, 1979, 1982, 1989, and 1994 to reduce the growing differential with private sector salaries. On October 21, 1994, a White Paper on Competitive salaries for competent and honest government was presented to parliament to justify the benchmarking of the salaries of ministers and senior civil servants to the average salaries of the top four earners in six private sector professions namely, accounting, banking, engineering, law, local manufacturing companies, and multi-national corporations. The government accepted this recommendation and public sector salaries were benchmarked accordingly from January 1995 with the salaries of the six professions in the private sector (Quah, 2010, pp. 110-11).The 1997 Asian financial crisis and the subsequent slowing down of the Singapore economy resulted in a 2 per cent decrease in Superscale G and a 7 per cent decrease in Staff Grade I salaries and the reduction of the employers' contribution to the Central Provident Fund (CPF) from 20 per cent to 10 per cent for all employees. The purpose of the CPF reduction was to enhance the Singapore's competitiveness by lowering the cost of doing business. In other words, the reduction in the CPF contribution meant an additional decrease in the salaries of the ministers and senior civil servants.When the Singapore economy recovered in 1999 with a growth rate of 5.4 per cent, and the reduction of retrenchments from 29,100 in 1998 to 14,600 in 1999, wages in the private sector began to rise again. Unemployment fell from 4.3 per cent in December 1999 to 3.4 per cent in March 2000. With the tight labour market in Singapore and the improved conditions in the private sector, Deputy Prime Minister Lee Hsien Loong revealed in parliament on 29 June 2000 that eight administrative officers had resigned in 2000. The government increased the performance bonus component in the public sector salaries and broadened the benchmarking of these salaries to the top eight earners in the six private sector professions. Consequently, in June 2000, the variable component of annual salaries was increased from 30 per cent to 40 per cent of the total annual pay of the superscale administrative officers and ministers (Quah, 2010, pp. 113-14). However, public sector salaries were later reduced by a combined total of 30 per cent in November 2001 and July 2003 because of the recession.In December 2007, the PSD announced that the salaries of ministers and senior civil servants would be increased from 4 per cent to 21 per cent from January 2008. Table IX shows the salaries of the President, Prime Minister, Ministers, Permanent Secretaries, superscale civil servants at the entry grade, and Members of Parliament in 2007 and 2008. However, on 24 November 2008, the PSD announced that the salaries of administrative officers, political, judicial and statutory appointment holders would be decreased by 19 per cent in 2009 because of the economic recession. This means that the President's annual salary has been reduced from S$3.87 million to S$3.14 million. Similarly, the Prime Minister's annual salary has been decreased from S$3.76 million to S$3.04 million (PSD 2008, pp. 1-2). Nevertheless, the Prime Minister's annual salary of US$2,183,516 in 2010 made him the best paid political leader in the world (Economist, 2010a, b).The policy of paying competitive public sector salaries has been effective in curbing the brain drain of political leaders to the private sector as none of them have resigned from political office to work in the private sector before their retirement. The attractive remuneration for the permanent secretaries has also been effective in retaining them in the SCS as none of them have left for private sector jobs before their retirement. However, paying competitive public sector salaries has been less effective in preventing Division I officers from leaving the SCS. An analysis of the resignation rate of Division I officers in the SCS from 1971-1984 shows that the salary increases in 1972, 1973, 1979 and 1982 had failed to curb the exodus of senior civil servants to the private sector (Quah, 2010, pp. 93, 119-20). Have the above policies of changing the attitudes of civil servants, minimizing corruption, maintaining meritocracy in the SCS by decentralizing the PSC, and paying competitive salaries to attract the "best and brightest" citizens to the SCS, resulted in good governance in Singapore? The short answer is "Yes" as will be demonstrated in Singapore's rankings and scores on eight governance indicators.Government effectiveness Singapore has been ranked first for the competence of its public officials from 1999 to 2002 by the Global Competitiveness Report3. Singapore was ranked first among 59 countries in 1999 and 2000 (Schwab et al., 1999, p. 242; Porter et al., 2000, p. 238). Singapore's civil servants were also ranked first among their counterparts in 75 countries in 2001-2002 and 80 countries in 2002-2003 (Schwab et al., 2002, p. 399; Cornelius, 2003, p. 604).The effectiveness of Singapore's government has also been confirmed by Singapore's consistently high ranking on the World Bank's governance indicator on government effectiveness, which is defined as "the quality of public service provision, the quality of the bureaucracy, the competence of civil servants, the independence of the civil service from political pressures, and the credibility of the government's commitment to policies" (Kaufmann et al., 2004, p. 3). Table X shows that government effectiveness in Singapore is very high from 1996 to 2010 and ranges from 93.2 percentile rank in 2002 to 100 percentile rank in 1996, 1998, 2008 and 2010.Similarly, a 2010 survey of 12 Asian countries conducted by the Hong Kong-based Political Economic Risk Consultancy (PERC) found that the bureaucracy in Singapore is the most effective, followed by the bureaucracy in Hong Kong SAR. At the other extreme, the least effective bureaucracies are those in the Philippines, Indonesia and India in descending order. Table XI confirms that Singapore has the most effective bureaucracy among 12 Asian countries from 1998 to 2010. In short, Singapore's government and civil servants are effective and performed well according to the above indicators.Effectiveness in curbing corruption The effectiveness of the POCA and CPIB in curbing corruption is reflected in Singapore's consistently high rankings and scores on these five indicators: Transparency International's Corruption Perceptions Index (CPI) from 1995-2011; Political Economic PERC's annual surveys of corruption from 1995-2011; World Bank's Control of Corruption indicator from 1996-2010; World Bank's Doing Business Surveys from 2007-2012; and the Global Competitiveness Report's indicator on public trust of politicians from 1999 to 2010-2011.Table XII shows that first, Singapore has been consistently ranked as the least corrupt country in Asia from 1995-2009 and 2011, and the least corrupt country in the world in 2010 with Denmark and New Zealand. Second, PERC's annual surveys of corruption from 1995-2011 also confirm Singapore's status as the least corrupt country in Asia. Third, the World Bank's control of corruption indicator from 1996-2010 shows that Singapore has the highest percentile rank among all the Asian countries included in the surveysFourth, Table XIII confirms that red tape is not a problem at all in Singapore as the World Bank's Doing Business Surveys show that Singapore is ranked first among the 175-183 economies included from 2007-2012. The relative absence of red tape in Singapore is an important factor responsible for its low level of corruption.Fifth, Table XIV shows that Singapore has been consistently ranked first in the Global Competitiveness Report's indicator on the public trust of politicians among the 59 to 139 countries included in the surveys from 1999 to 2010/2011. Singapore's superior ranking on this indicator also reflects its high level of public trust in politicians and its low level of corruption.In short, the four policies introduced by the PAP government after assuming office in June 1959 have been effective in promoting good governance in Singapore as demonstrated by Singapore's consistently high rankings and scores on the abovementioned eight governance indicators. Is Singapore's experience in promoting good governance transferable to other Asian countries? As Singapore's success in ensuring good governance is the combined influence of the political will of the PAP government to solve the problems facing the country for the past 53 years and its favourable policy context, it will be difficult to transfer Singapore's experience in toto to other Asian countries because of the lack of political will and the unfavourable policy contexts in many Asian countries.Apart from good leadership, Singapore's favourable policy context has enabled its political leaders to stretch the constraints imposed by its small size and lack of resources by formulating and implementing effective policies to solve its problems during the past 53 years. As public administration Singapore-style is the product of the local policy context and the policies implemented by the PAP government, it would be difficult to replicate these policies to other Asian countries in view of the significant contextual differences between Singapore and these countries (Quah, 2010, pp. 246, 51).Table XV shows clearly the significant contextual differences between Singapore and the other 25 Asian countries. First, in terms of size, Singapore is the second smallest country after the Macao Special Administrative Region, which has a land area of 29.2 sq. km. At the other extreme, are the larger countries of China and India, which are 13,466 and 4,630 times larger respectively than Singapore. A second important contextual difference is Singapore's population, which is only larger than those of Mongolia's population of 2.7 million, Timor-Leste's population of 1.2 million, Bhutan's population of 0.7 million, Macao's population of 0.5 million, and Brunei's population of 0.4 million. On the other hand, Singapore's population of 5.1 million in 2010 is dwarfed by the huge populations of China, India, and Indonesia.The third contextual difference between Singapore and the other Asian countries is its economic affluence as manifested in its GDP per capita of US$40,920 in 2010, which is the second highest among all the 26 countries listed in Table XV. In contrast, the GDP per capita of these four countries are less than US$1,000 namely, Cambodia (US$760), Bangladesh (US$640), Afghanistan (US$517), and Nepal (US$490).In short, Singapore is a city-state, which is richer and smaller in terms of land area and population for the PAP government to govern than most of the other Asian countries. Singapore's favourable policy context has enabled the PAP government, which has been in power since June 1959, to implement policies effectively, to curb corruption, and to ensure the ease of doing business in Singapore, as demonstrated in Singapore's superior ranking on the World Bank's governance indicator on government effectiveness from 1996-2010, the Doing Business Surveys from 2007-2012, Transparency International's CPI from 1995-2011, and PERC's corruption surveys from 1995-2011. In his National Day Rally speech on 19 August 1984, then Prime Minister Lee Kuan Yew attributed Singapore's success to the quality of its political leadership:In the end, whatever the system, it is the quality of the men who run it, that is decisive. For they will decide what to make of the society, and how to get the people to give of their best. The Singapore system has worked. It will continue to work if you vote for honest, able and dedicated men, and you give them your best, for the good of all (Lee, 1984, p. 18).More recently, Lee acknowledged the importance to Singapore's development of attracting the "best and brightest" citizens to join the government and the SCS thus:My experience of developments in Asia has led me to conclude that we need good men to have good government. [...] The single most decisive factor that made for Singapore's development was the ability of its ministers and the high quality of the civil servants who supported them. [...] It was Singapore's good fortune that we had, for a small, developing country, a fair share of talent, because our own [talent] had been reinforced by the talented men and women who came here for their education, and stayed on for employment or business opportunities (Lee, 2000, pp. 735-6).In the same vein, Edgar H. Schein (1996, pp. 221-2) has identified the policy of having "the best and brightest" citizens in government as "probably one of Singapore's major strengths" because "they are potentially the most able to invent what the country needs to survive and grow." Furthermore, he has described Singapore as "one of the few models existing in the world of how a society can progress with a government that attempts to maximize intelligence, skill, and honesty."In sum, Singapore's transformation from a poor third world country in 1959 to an affluent and politically stable first world country today is the result of the ability of its political leaders and civil servants to formulate and implement policies to solve the country's problems during the past 53 years. First, the PAP leaders reorganised the SCS and changed the attitudes of the civil servants by convincing them to contribute to the attainment of national development goals. Second, they continued with the tradition of meritocracy introduced by the British by retaining and enhancing the PSC's effectiveness by reducing its heavy workload by decentralizing its functions of appointment and promotion to the ESC and PCDSC in 1990, and the 31 personnel boards in 1995. Third, they learnt from the mistakes made by the British in curbing corruption by enacting the POCA in June 1960 to enhance the CPIB's effectiveness in combating corruption. Fourth, the PAP government's success in promoting economic growth enabled it to compete for talented personnel with the private sector by paying competitive salaries to ministers and senior civil servants from 1972 onwards to prevent them from leaving for private sector jobs.In the final analysis, whether these four elements of Singapore's success - institutional and attitudinal reform of civil servants; zero-tolerance for corruption; meritocracy in appointing and promoting civil servants; and paying competitive salaries to attract the "best and brightest" citizens to join the government and civil service - can be replicated in other Asian countries depends mainly on whether their political leaders and senior civil servants have the political will and are prepared to pay the high economic and political costs of implementing these policies. Opens in a new window.Table I Growth of CPIB's personnel, 1952-2011 Opens in a new window.Table II Budget of the CPIB, 1978-2011 Opens in a new window.Table III Number of visitors to CPIB, 2005-2010 Opens in a new window.Table IV Evolution of the PSC in Singapore, 1951-2011 Opens in a new window.Table V Workloads of the PSC, ESC and PCDSC in appointing and promoting candidates in the SCS, 1990-1994 Opens in a new window.Table VI The SCS's Personnel Management System in 1995 Opens in a new window.Table VII Workload of the PSC in Singapore, 1995-2010 Opens in a new window.Table VIII Workload of the PSC in Singapore, 1983-2010 Opens in a new window.Table IX Salaries of key appointments in Singapore, 2007-2008 Opens in a new window.Table X Government effectiveness of Singapore, 1996-2010 Opens in a new window.Table XI PERC's evaluation of the effectiveness of bureaucracy in 12 Asian countries, 1998 to 2010 Opens in a new window.Table XII Singapore's performance on CPI, PERC and control of corruption 1995-2011 Opens in a new window.Table XIII Singapore's rank on the World Bank's ease of doing business, 2007-2012 Opens in a new window.Table XIV Public trust of politicians in Singapore, 1999-2010/2011 Opens in a new window.Table XV Policy contexts of 26 Asian countries
|
- This paper will be useful to those interested in learning how Singapore succeeded in promoting good governance.
|
[SECTION: Purpose] A precept of the marketing concept contends that business achieves success by determining and satisfying the needs, wants, and aspirations of target markets. Few would argue that this determination and satisfaction of target market wants and needs is critical for firm success. These concepts, traditionally thought to be part of the marketing function of the firm, have fueled scholars' interest in the role of marketing within the firm (e.g., Becherer et al., 2003; Berthon et al., 2008; Moorman and Rust, 1999; Simpson and Taylor, 2002; Webster, 1981, 1992, 2003; Webster et al., 2003).Scholars have identified significant differences between large and small organizations. Large organizations tend to use a structured framework with a clear hierarchy in decision making. On the other hand, small firms tend to feature processes that begin with and highly involve the entrepreneur or owner. Furthermore, the entrepreneur or owner's personality and style help to shape decision making (Sadler-Smith et al., 2003).The small-firm sector plays a significant role in the world economy. In the United States a recent analysis of employment changes between September 1992 and March 2005 showed that 65 percent of the net new jobs created during that time are attributed to firms with fewer than 500 employees (US Bureau of Labor Statistics, 2005).Research investigating the competitive advantage of small firms has consistently emphasized the importance of marketing, strategic positioning, and entrepreneurship as key factors in business survival and growth. The ability to identify and operate in a particular market niche enables the firm to exploit a range of specializations and offers protection from larger competitors. Yet despite the widespread acceptance of the importance of the marketing concept, the precise marketing activities and competencies that contribute most strongly to business performance must be identified for small and medium-sized enterprises (SMEs).Our study of SMEs revisits many of the questions Homburg et al. (1999) posed in their study of the marketing functions of large firms, but we direct our attention to SMEs. The definition of what precisely constitutes an SME differs widely. For the purposes of this paper, we follow the Journal of Enterprise Culture and the Asia Pacific Economic Cooperative (APEC), who define US-based SMEs as companies with fewer than 499 employees (APEC, 2003). In this research, we seek to answer the following questions:* Relative to other business functions, how important is the marketing function in SMEs?* Does marketing in SMEs enjoy the same influence found in larger firms?* What internal and external factors influence the role of marketing in small firms?This paper reports the results of a study that employed an activity-based approach whereby demonstrable marketing competencies were related to the entrepreneurial orientation evident in the firms. The analysis demonstrated that certain competencies were more strongly associated with a marketing orientation, whereas others were associated with an entrepreneurial orientation. The results show differences in the role of marketing in small firms as compared with the role of marketing in the large firms that were studied by Homburg et al. (1999). SME marketing functions have a way to go before they enjoy the same degree of influence their counterparts have in large firms. Inquiry into departmental influence (such as marketing) is based on Cyert and March's (1963) work on subunit power. Cyert and March (1963) argued that managers have conflicting goals and seek satisfactory solutions as opposed to optimal solutions to business problems. The concept of "dominant coalition" has been used to explain differences in the power of firm subunits (Thompson, 1967). The relative power of subunits is a function of strategic choices made by dominant coalitions (i.e. power members) within an organization. Influence has been defined as "a change in one party that has its origin in another party and this embodies the successful exercise of power" (Stern and Scheer, 1992, p. 260). Our study focuses on the marketing department subunit. Various authors have examined the role and influence of marketing in firms (e.g. Becherer et al., 2003; Berthon et al., 2008; Moorman and Rust, 1999; Simpson and Taylor, 2002; Webster, 1981, 1992, 2003; Webster et al., 2003) and since the 1980s, the marketing department has been shown to have varying levels of influence in the firm. Homburg et al. (1999) directly studied the influence of marketing in large firms and uncovered a number of situational factors that affect the level of marketing's influence. These situational factors include a firm's strategic orientation (differentiation versus cost leadership) and background of the CEO. Although their study helped to define the relative level of influence of the marketing function for business, the authors focused on large organizations - firms with more than $25 million in sales.However, SMEs have characteristics that differentiate them from large organizations (McCartan-Quinn and Carson, 2003). These differences include advantages such as greater flexibility, innovation, and lower overhead costs. In terms of disadvantages, SMEs are limited by their market power, and capital and managerial resources (Motwani et al., 1998). These differences raise the question, "does the marketing function in SMEs enjoy the same influence found in larger firms?" Compared with the literature stream surrounding the role of marketing in larger firms, inquiry on this topic has been more limited for SMEs. Many studies have attempted to define marketing and outcomes of marketing for SMEs. Carson (2001) and Sui and Kirby (1998) traced the evolution of marketing and the various approaches to SME marketing. Other authors have attempted to develop hypothetical and empirical models of marketing for SMEs. Sui et al. (2004), Julien and Ramangalahy (2003) and Berthon et al. (2008) showed how strategic marketing practices such as knowledge of current market conditions and consumer tastes were positively related to SME performance. Becherer et al. (2003) examined internal environmental factors such as the background and decision processes of CEOs. One aspect of marketing, promotional efforts, was found to be a key influence in performance of SMEs (Wood, 2006). Market orientation as a driver of SME business performance has also generated scholar interest (Blankson and Stokes, 2002; Fillis, 2002; Pacitto et al., 2007). Finally, authors have studied underlying reasons for the characteristics of SME marketing practices. Simpson et al. (2006) examined drivers of marketing effort such as the presence of a marketing department and marketing representation at the board level. SMEs lack the resources to compete head-to-head with larger rivals and thus cannot do traditional marketing (Gilmore et al., 2001).Because SMEs lack the resources to compete head-to-head with larger rivals, some scholars have questioned whether SMEs formally practice marketing at all (Carson et al., 1998; Gilmore et al., 2001; Shepherd and Ridnour, 1996). In a typical argument, Hogarth-Scott et al. (1996) considered most marketing theories to be inappropriate for SMEs and not helpful in the understanding of their markets. However, in the same article, the authors found that the marketing function contributed positively to small business success and the ability to think strategically. Clearly, marketing department exert influence in SMEs, especially in SMEs that succeed. Larger firms achieve better competitive positions than smaller firms when they have greater marketing capabilities (Grimes et al., 2007), so SMEs must maintain strong marketing departments to effectively compete.Evidence has shown that although marketing activities in SMEs may be different, marketing departments are still critical to firms' success. Many firms carry out business via highly informal, unstructured, reactive mechanisms, whereas others develop, over time, a proactive and skilled approach where innovation and identification of opportunities gives them a competitive advantage (Fillis, 2007). Romano and Ratnatunga (1995) observed that "small firms face marketing challenges which can and will ultimately determine their future." Furthermore, Hill (2001) found that SMEs still engage in many traditional marketing functions, especially marketing planning. A widely cited marketing activity for SMEs is networking. Wincent (2005) showed that the size of a firm matters in regard to how a company conducts its networking activity. Networks are important during the establishment, development, and growth of SMEs. Siu (2001) found that SMEs rely heavily on their personal contact network in marketing their firms. Traditional economic structures favor size (large firms). However, today's economy is earmarked by relationships, network, and information, which play to some of the characteristics of SMEs.Our research adds to the SME literature stream by examining outputs and attributes of marketing (e.g. sales, percent of sales spent on marketing) and by considering how these are related to firm characteristics such as entrepreneurial orientation. Two theories provide a conceptual framework to explain the differences in the influence of the marketing function: institutional theory and contingency theory. Institutional theory states that organizational structure and design is inexorably linked to social networks and is shaped by conformity and legitimacy pressures (DiMaggio and Powell, 1983; Meyer and Rowan, 1977). Firm structure does not necessarily reflect efficiency but legitimacy (Meyer and Rowan, 1977). Thus the influence of a firm function such as marketing is a function of conformity/legitimacy pressures.Alternatively, contingency theory states that organizational structure and design is a function of external factors. Firms must "fit" to their environment, and characteristics of firms such as subunit power are a function of environmental determinants. Research on subunit power has explored relationships between environmental factors and organizational structures/power (Boeker, 1989; Hambrick, 1981; Hinings et al., 1974; Salancik and Pfeffer, 1974).This research tests elements of both theories. Hypotheses H1 and H2 are based on the contingency theory argument, and hypotheses H3, H4, and H5 are based on institutional theory.Consumer versus industrial markets Firms serving industrial markets have fewer customers (as compared with consumer markets) and have closer partnerships with their customers (Heide and John, 1992). These partnerships involve more aspects of the firm and, consequently, the marketing department no longer serves as the primary link between the firm and its customers (Homburg et al., 1999). In firms serving consumer markets, the marketing function serves as the primary link between the firm and its market. Thus, it stands to reason that the marketing function has greater influence in these firms.H1. Marketing's influence is higher in SMEs selling consumer goods than in firms selling industrial goods.Differentiation strategy Using Porter's (1980) typology, firms can pursue a cost leadership or differentiation strategy. Bennett and Smith (2002) showed that as SMEs grow, they develop greater specialization and differentiation strategies. Although not empirically proven, conventional wisdom has suggested that marketing's influence is more influential for a differentiation strategy than for a cost leadership strategy. For a firm to pursue a differentiation strategy, it must identify customer needs and develop products to fit those needs. Identifying those needs is one of the core competencies of a marketing department and presumably would lead to greater influence of marketing within a firm.H2. Marketing's influence is related positively to a differentiation strategy compared with a cost leadership strategy.Entrepreneurial orientation As Stevenson and Gumpert (1985) noted, firms with an entrepreneurial focus tend to be flat organizations with informal networks of employees and functions. Entrepreneurial organizations typically have an evolving management structure and are more abstract in defining the roles of management (Freeman and Engle, 2007). Firms with a non-entrepreneurial focus are termed administratively oriented and are organizations with clearly defined authority and structure. It is reasonable to infer from Stevenson and Gumpert's work that firms with an administrative focus will have more clearly delineated functions/departments such as marketing. Bureaucracy has a negative effect on entrepreneurial activity in a firm, and this is anathema to an entrepreneurial-oriented firm (Sorensen, 2007). If the marketing function is more clearly defined, we would expect marketing to have greater influence.H3. Marketing's influence is related negatively to the entrepreneurial orientation of the firm.Background of CEO The top managers influence and shape the firm's direction and focus. Within SMEs, the top manager's role is pivotal for growth and expansion (Hutchinson et al., 2006). For smaller firms, this would seem even more relevant where all employees are likely to feel the CEO's influence. The psychologist Duncker (1945) first coined this phenomenon "functional fixity" to refer to the observation that people tend to react to an object in terms of its usually defined function. That is, a hammer is used to drive nails but is not frequently used as a paperweight, although obviously it could be. Business researchers have applied this idea to management and have found that managers are also influenced by their previous training and find it difficult to deviate from their functional training when devising a solution to a problem. For example, engineers seek technical solutions, accounting managers seek financial solutions, and marketing managers look to advertising and promotion to solve problems (Ashton, 1976; Barnes and Webb, 1986; Chang and Birnberg, 1977; Hand, 1990).H4. Marketing's influence is higher in firms where the chief executive officer has a marketing background.Firm size Recent studies have begun to focus on how firm size affects entrepreneurial orientation in organizations (Dholakia and Kshetri, 2004; Gompers et al., 2005; Dobrev and Barnett, 2005; Stuart and Ding, 2006.) As firms grow, they tend to become more bureaucratic. As firms become more bureaucratic, they tend to become less entrepreneurial (Sorensen, 2007). However, as firms grow, they are in a position to hire specialists and add to their "bench strength." As more specialists are hired, especially in the marketing function, it is assumed that marketing gains greater influence. This leads to our final hypothesis:H5. Within SMEs, marketing's influence is positively related to the size of firm.The proceeding hypothesized relationships are summarized in Figure 1. Sample We employed a mail survey of SMEs using two mailing lists available to the authors. The first mailing came from present and past participants in the Small Business Center of a large Midwestern university. This center caters to small businesses in the Midwestern region of the United States. Its mailing list, totaling 150 businesses, was screened to ensure only SMEs were included in the sample. The second mailing list was obtained from the Greenbrier Chamber of Commerce of West Virginia, which serves businesses in the central region of West Virginia. Again, the mailing list, totaling 100 businesses, was screened to ensure only SMEs were included. Combined with surveys from the Family Business Center, the total sample was 250. The surveys were sent to the top executives (e.g. CEOs, presidents, general managers, owners) of the firms. All the addresses from the mailing lists were valid because no mailings were returned as undeliverable. The survey was conducted October to November, 2006. Of the 250 surveys delivered, 100 completed surveys were returned, a response rate of 40 percent. The response rate from each source was the same. No inducements were included for participation in the survey. Instead, the director of the Family Business Center or Chamber of Commerce sent simple cover letters asking members to participate. We did not stimulate the response rate by doing follow-up contacts to nonresponders.Several steps were taken to address nonresponse error and other external validity concerns. First, on receipt of the completed survey instrument, respondents were contacted by telephone to verify that they personally completed the survey and were top managers of their firms. Of the 100 completed responses, we successfully contacted 45 (45 percent) respondents. In these 45 cases, the top managers of the firms acknowledged personally completing the survey and verified their positions as top managers of the firms.Second, comparisons were performed of sample means (t tests) between early and late survey respondents and the two sources for the sample. Across all independent and dependant measures, no statistical difference was found in sample means.Finally, this study's sample was compared against census data. Although this study used a convenience sample and cannot be generalized to all SMEs, a comparison was done to assess representativeness of the sample. As seen in Table I, our sample covered most employee size categories.As seen in Table II, we analyzed our sample in terms of coverage of NAIC categories. The findings suggest that our sample covered all major NAIC categories.Again, these comparisons do not suggest the study results can be generalized to all SMEs but were done to address potential nonresponse concerns.Measures The independent variables for our study included the influence of marketing, presence of a differentiation strategy, entrepreneurial orientation, years in business, number of full-time employees, and primary functional background of the CEO. The measures for this study have been used previously or adapted from previous studies.Influence of marketing was measured via a multi-item measure featuring a 100-point constant sum scale. This scale was adapted from Homburg et al.'s (1999) scale developed to measure the influence of marketing on decision making for ten strategic issues:1. advertising;2. customer satisfaction measurement;3. customer satisfaction improvement;4. expansion into new geographic markets;5. strategic direction of the firm;6. choice of strategic partners;7. new product development;8. design of customer service and support;9. pricing; and10. major capital expenditures.Homburg et al. (1999) chose these issues to represent those decisions classically thought to be handled by marketing (advertising, pricing) and other issues that are not completely under the control of one department. The list was not designed to be an exhaustive measure of all aspects of marketing's influence but a sampling of business decisions typically considered in the marketing domain. Respondents were asked to allocate 100 points in terms of influence of the following groups: marketing, sales, R&D, operations, finance, and general management on each of the ten issues. The option of "general management" was added to the original Homburg et al.'s (1999) scale because of concern that their original scale presumed that a general manager exerts no influence on decision making relating to marketing. We questioned that assumption and included that option into the scale. Respondents were told to allocate no points to a group if it did not exist in their firm. In no case did any respondent assign zero points to marketing, thus eliminating the potential confound of responses from firms without a marketing department. Respondents were also asked to rate the importance of each of the ten issues on a seven-point Likert scale (anchored by relatively low importance to extremely high importance).Marketing influence was calculated by multiplying the importance of each of the ten issues by the influence allocated to marketing for that issue and then adding across all ten issues. The summed figure was divided by the number of issues that the respondent answered.Whether a firm used a differentiation strategy was determined by using a seven-item measure used by Homburg et al. (1999), Kim and Lim (1988), and Dress and Davis (1984). Respondents gave specific measures in answer to the following questions about company activities: To what extent does your business emphasize the following activities?1. competitive advantage through superior products;2. creating superior value through services accompanying the products;3. new product development;4. building up a premium product or brand image;5. obtaining high prices from the market;6. advertising; and7. development of customer specific solutions and products.A seven-point Likert scale anchored by Not at all and A great deal was used to record responses (alpha=0.84).The entrepreneurial orientation of the firm was measured using a seven-item scale developed by Covin and Slevin (1989) and subsequently used by Keh et al. (2007). The specific measures were:1. our culture emphasizes innovation and research and development;2. we have a low rate of new product innovation;3. we are a follower and have a defensive posture in the market;4. we are first to introduce new technologies and products;5. we maintain a competitive posture toward competition;6. we have a strong dislike for high risk, high return projects; and7. when faced with risk, we adopt an aggressive, bold posture.A seven-point Likert scale, anchored by Strongly disagree and Strongly agree was used to record responses. Items (2), (3), and (6) were reverse coded (alpha=0.83).The age of business (years), number of full-time employees, and primary functional background of the CEO were single-item, open-response measures. Regarding the measure of full-time employees, any response of "more than 500 employees" would have been discarded because this study was limited to SMEs. However, we did not receive any surveys with this response. Functional background of CEO was dummy coded (0=non-marketing background; 1=marketing background). The type of market served was dummy coded (0=industrial; 1=consumer).The dependent variables included sales from most recent year, percent of sales spent on marketing, percent of sales spent on sales activities, and percent of sales spent on advertising. The variables were measured via single-item measures. For sales, respondents were given four options: under $1 million, $1-5 million, $5-25 million, and above $25 million. The percent of sales measures (marketing activities, sales activities, and advertising spending) had 13 response options: (0, less than 1 percent, 1 percent, 2 percent, 3 percent, 4 percent, 5 percent, 6 percent, 7 percent, 8 percent, 9 percent, 10 percent, and above 10 percent). The results from the two samples were analyzed (t-tests) and no statistical difference was found in means across all measures. Accordingly, the results were pooled.The role of marketing Conventional wisdom suggests that marketing's role is to manage the relationship between customers and a firm (Moorman and Rust, 1999). Because customers are an important contributor to a firm's financial performance, the department responsible for management of customer relationships should be viewed as an important force; thus, the marketing department that assumes this management function should be deemed important and influential. Homburg et al.'s (1999) study showed the marketing function to be a significant influence in large firms. Turning to this study, as shown in Table III, marketing shares influence with general management and sales on most marketing issues. This suggests that marketing, as function and influence in business, is not as well developed for smaller firms compared with their larger brethren. The sales function has a slightly greater influence over marketing on all issues outside of advertising (customer satisfaction measurement, customer satisfaction improvement, expansion into new geographic markets, strategic direction of the firm, choice of strategic partners, new product development, design of customer service and support, and pricing and major capital expenditures). Decisions regarding advertising are the only issue where marketing enjoys significant and sole influence. We have included comparative data from the Homberg et al. (1999) study, and the data suggest marketing has a way to go before it enjoys the same influence found in larger firms.Hypotheses testing Regression analysis was used to test the hypotheses. We ran three regression analyses: on all issues, "marketing" issues (advertising messages, distribution strategy, design of customer sales and support, customer satisfaction measurement and customer satisfaction improvement) and non-marketing exclusive issues (strategic direction of business unit, choice of strategic partners, new product development, major capital expenditures, and expansion into new geographic markets). Results were the same, and for purposes of discussion, we will focus on the first regression analysis although results from all three analyses are reported in Table IV. Firm size, sales; percent spent of marketing, sales, and advertising activities; source of survey, and early/late responders were included as control variables.H1 proposed that marketing's influence is higher in firms selling consumer goods than in firms selling industrial goods. We found support for this (b = 48.68, p<0.001). Firms in consumer markets tend to do more advertising, considered the exclusive province of marketing, which likely leads to firms having greater "bench strength" in the marketing department. Interestingly, this conclusion differs from Homburg et al.'s (1999) finding for large firms where no difference appeared in marketing's influence between consumer and industrial firms.H3 stated that marketing's influence is related negatively to the entrepreneurial orientation of the firm. We found support for this (b =.-20.14, p<0.001). Less entrepreneurial firms have more influential marketing departments. This suggests a more hierarchical firm has more clearly defined functions leading to greater influence of marketing.We found no support for the remaining hypotheses (H2, H4 and H5). H2 stated that marketing's influence is related positively to a differentiation strategy. Evidently a differentiation strategy does not automatically mean marketing has greater influence. This contrasts with the Homburg et al. (1999) survey where large businesses employing a differentiation strategy had greater marketing influence. H4 stated that marketing's influence is higher where the CEO has a marketing background. Unlike larger firms, the notion that that the CEO's background indicates which firm functions might have influence was not supported. Finally, H5 related to firm size within SMEs. Firm size was not a significant factor in determining marketing's influence. H4 and H5 results were consistent with Homburg et al.'s (1999) study. A comparison of this study's results and the Homburg et al.'s (1999) can be found in Table V. Our data suggest that marketing is not as well developed or influential in SMEs as it is in large firms. Marketing's scope or extent of responsibility is more limited compared with large firms. Two factors, type of market (consumer) and firm orientation (hierarchal), facilitate marketing's influence within a firm. These results indicate that both underlying theories - institutional and contingency - received support and neither theory can be dismissed, suggesting the influence of marketing is a multidimensional construct. Individual traits (e.g. charisma, power, authority) were not considered in this study. However, our findings show marketing's influence can vary systematically as a function of institutional and external factors.Although our analysis sheds light on some internal and external factors that influence marketing's role in a firm, a major portion of variance remains unexplained. Additional research is necessary to explore other potential influences on marketing's role in a firm. These might include personnel factors (background and experience of marketing department personnel) and firm politics. For example, Reijonen and Komppula (2007) studied SMEs in Finland and found the motivation of entrepreneurs are not solely financial and often are based on job satisfaction, control, and family concerns.Although this research fills a gap in the literature relating to the influence of marketing in SMEs, there is much work to be done. We did not address the relationship between marketing's influence and the performance of firms. Analyses done and conclusions reached in this research were based on a very limited sample of SMEs located in one region of the country. Sampling across all regions of the United States would help in generalization of the findings. The results of this study are particularly troubling because marketing resources are one driver of competitive advantage, and although the studied firms have a formal marketing function within the firm, it would appear the function has not proven its value to the firm. For marketing to increase in influence, individuals leading this function must gain a seat at the table. Our findings would suggest that changing the influence of the marketing function might take considerable effort. One barrier to changing marketing's influence is the firm's entrepreneurial orientation, which can be a deep-seated cultural factor that typically requires time to change. An Indian proverb cautions: "Under a bright lamp, there is great darkness." Although marketing departments have a responsibility to market the firm's products and services, the task of marketers marketing themselves to internal stakeholders remains unanswered. Marketing departments must do more to ensure that the voice of marketing is heard when key decisions are being made. Opens in a new window.Figure 1 Conceptual model Opens in a new window.Table I Firms by employee size Opens in a new window.Table II Percentage of firms by industry Opens in a new window.Table III Influence of business functional departments on marketing issues Opens in a new window.Table IV Results of regression analysis Opens in a new window.Table V Comparison of study results with Homberg et al. (1999) study
|
- The purpose of this paper is to examine marketing in small and medium-sized enterprises (SMEs), its role as a driver of competitive advantage, and, therefore, its importance to the firm.
|
[SECTION: Method] A precept of the marketing concept contends that business achieves success by determining and satisfying the needs, wants, and aspirations of target markets. Few would argue that this determination and satisfaction of target market wants and needs is critical for firm success. These concepts, traditionally thought to be part of the marketing function of the firm, have fueled scholars' interest in the role of marketing within the firm (e.g., Becherer et al., 2003; Berthon et al., 2008; Moorman and Rust, 1999; Simpson and Taylor, 2002; Webster, 1981, 1992, 2003; Webster et al., 2003).Scholars have identified significant differences between large and small organizations. Large organizations tend to use a structured framework with a clear hierarchy in decision making. On the other hand, small firms tend to feature processes that begin with and highly involve the entrepreneur or owner. Furthermore, the entrepreneur or owner's personality and style help to shape decision making (Sadler-Smith et al., 2003).The small-firm sector plays a significant role in the world economy. In the United States a recent analysis of employment changes between September 1992 and March 2005 showed that 65 percent of the net new jobs created during that time are attributed to firms with fewer than 500 employees (US Bureau of Labor Statistics, 2005).Research investigating the competitive advantage of small firms has consistently emphasized the importance of marketing, strategic positioning, and entrepreneurship as key factors in business survival and growth. The ability to identify and operate in a particular market niche enables the firm to exploit a range of specializations and offers protection from larger competitors. Yet despite the widespread acceptance of the importance of the marketing concept, the precise marketing activities and competencies that contribute most strongly to business performance must be identified for small and medium-sized enterprises (SMEs).Our study of SMEs revisits many of the questions Homburg et al. (1999) posed in their study of the marketing functions of large firms, but we direct our attention to SMEs. The definition of what precisely constitutes an SME differs widely. For the purposes of this paper, we follow the Journal of Enterprise Culture and the Asia Pacific Economic Cooperative (APEC), who define US-based SMEs as companies with fewer than 499 employees (APEC, 2003). In this research, we seek to answer the following questions:* Relative to other business functions, how important is the marketing function in SMEs?* Does marketing in SMEs enjoy the same influence found in larger firms?* What internal and external factors influence the role of marketing in small firms?This paper reports the results of a study that employed an activity-based approach whereby demonstrable marketing competencies were related to the entrepreneurial orientation evident in the firms. The analysis demonstrated that certain competencies were more strongly associated with a marketing orientation, whereas others were associated with an entrepreneurial orientation. The results show differences in the role of marketing in small firms as compared with the role of marketing in the large firms that were studied by Homburg et al. (1999). SME marketing functions have a way to go before they enjoy the same degree of influence their counterparts have in large firms. Inquiry into departmental influence (such as marketing) is based on Cyert and March's (1963) work on subunit power. Cyert and March (1963) argued that managers have conflicting goals and seek satisfactory solutions as opposed to optimal solutions to business problems. The concept of "dominant coalition" has been used to explain differences in the power of firm subunits (Thompson, 1967). The relative power of subunits is a function of strategic choices made by dominant coalitions (i.e. power members) within an organization. Influence has been defined as "a change in one party that has its origin in another party and this embodies the successful exercise of power" (Stern and Scheer, 1992, p. 260). Our study focuses on the marketing department subunit. Various authors have examined the role and influence of marketing in firms (e.g. Becherer et al., 2003; Berthon et al., 2008; Moorman and Rust, 1999; Simpson and Taylor, 2002; Webster, 1981, 1992, 2003; Webster et al., 2003) and since the 1980s, the marketing department has been shown to have varying levels of influence in the firm. Homburg et al. (1999) directly studied the influence of marketing in large firms and uncovered a number of situational factors that affect the level of marketing's influence. These situational factors include a firm's strategic orientation (differentiation versus cost leadership) and background of the CEO. Although their study helped to define the relative level of influence of the marketing function for business, the authors focused on large organizations - firms with more than $25 million in sales.However, SMEs have characteristics that differentiate them from large organizations (McCartan-Quinn and Carson, 2003). These differences include advantages such as greater flexibility, innovation, and lower overhead costs. In terms of disadvantages, SMEs are limited by their market power, and capital and managerial resources (Motwani et al., 1998). These differences raise the question, "does the marketing function in SMEs enjoy the same influence found in larger firms?" Compared with the literature stream surrounding the role of marketing in larger firms, inquiry on this topic has been more limited for SMEs. Many studies have attempted to define marketing and outcomes of marketing for SMEs. Carson (2001) and Sui and Kirby (1998) traced the evolution of marketing and the various approaches to SME marketing. Other authors have attempted to develop hypothetical and empirical models of marketing for SMEs. Sui et al. (2004), Julien and Ramangalahy (2003) and Berthon et al. (2008) showed how strategic marketing practices such as knowledge of current market conditions and consumer tastes were positively related to SME performance. Becherer et al. (2003) examined internal environmental factors such as the background and decision processes of CEOs. One aspect of marketing, promotional efforts, was found to be a key influence in performance of SMEs (Wood, 2006). Market orientation as a driver of SME business performance has also generated scholar interest (Blankson and Stokes, 2002; Fillis, 2002; Pacitto et al., 2007). Finally, authors have studied underlying reasons for the characteristics of SME marketing practices. Simpson et al. (2006) examined drivers of marketing effort such as the presence of a marketing department and marketing representation at the board level. SMEs lack the resources to compete head-to-head with larger rivals and thus cannot do traditional marketing (Gilmore et al., 2001).Because SMEs lack the resources to compete head-to-head with larger rivals, some scholars have questioned whether SMEs formally practice marketing at all (Carson et al., 1998; Gilmore et al., 2001; Shepherd and Ridnour, 1996). In a typical argument, Hogarth-Scott et al. (1996) considered most marketing theories to be inappropriate for SMEs and not helpful in the understanding of their markets. However, in the same article, the authors found that the marketing function contributed positively to small business success and the ability to think strategically. Clearly, marketing department exert influence in SMEs, especially in SMEs that succeed. Larger firms achieve better competitive positions than smaller firms when they have greater marketing capabilities (Grimes et al., 2007), so SMEs must maintain strong marketing departments to effectively compete.Evidence has shown that although marketing activities in SMEs may be different, marketing departments are still critical to firms' success. Many firms carry out business via highly informal, unstructured, reactive mechanisms, whereas others develop, over time, a proactive and skilled approach where innovation and identification of opportunities gives them a competitive advantage (Fillis, 2007). Romano and Ratnatunga (1995) observed that "small firms face marketing challenges which can and will ultimately determine their future." Furthermore, Hill (2001) found that SMEs still engage in many traditional marketing functions, especially marketing planning. A widely cited marketing activity for SMEs is networking. Wincent (2005) showed that the size of a firm matters in regard to how a company conducts its networking activity. Networks are important during the establishment, development, and growth of SMEs. Siu (2001) found that SMEs rely heavily on their personal contact network in marketing their firms. Traditional economic structures favor size (large firms). However, today's economy is earmarked by relationships, network, and information, which play to some of the characteristics of SMEs.Our research adds to the SME literature stream by examining outputs and attributes of marketing (e.g. sales, percent of sales spent on marketing) and by considering how these are related to firm characteristics such as entrepreneurial orientation. Two theories provide a conceptual framework to explain the differences in the influence of the marketing function: institutional theory and contingency theory. Institutional theory states that organizational structure and design is inexorably linked to social networks and is shaped by conformity and legitimacy pressures (DiMaggio and Powell, 1983; Meyer and Rowan, 1977). Firm structure does not necessarily reflect efficiency but legitimacy (Meyer and Rowan, 1977). Thus the influence of a firm function such as marketing is a function of conformity/legitimacy pressures.Alternatively, contingency theory states that organizational structure and design is a function of external factors. Firms must "fit" to their environment, and characteristics of firms such as subunit power are a function of environmental determinants. Research on subunit power has explored relationships between environmental factors and organizational structures/power (Boeker, 1989; Hambrick, 1981; Hinings et al., 1974; Salancik and Pfeffer, 1974).This research tests elements of both theories. Hypotheses H1 and H2 are based on the contingency theory argument, and hypotheses H3, H4, and H5 are based on institutional theory.Consumer versus industrial markets Firms serving industrial markets have fewer customers (as compared with consumer markets) and have closer partnerships with their customers (Heide and John, 1992). These partnerships involve more aspects of the firm and, consequently, the marketing department no longer serves as the primary link between the firm and its customers (Homburg et al., 1999). In firms serving consumer markets, the marketing function serves as the primary link between the firm and its market. Thus, it stands to reason that the marketing function has greater influence in these firms.H1. Marketing's influence is higher in SMEs selling consumer goods than in firms selling industrial goods.Differentiation strategy Using Porter's (1980) typology, firms can pursue a cost leadership or differentiation strategy. Bennett and Smith (2002) showed that as SMEs grow, they develop greater specialization and differentiation strategies. Although not empirically proven, conventional wisdom has suggested that marketing's influence is more influential for a differentiation strategy than for a cost leadership strategy. For a firm to pursue a differentiation strategy, it must identify customer needs and develop products to fit those needs. Identifying those needs is one of the core competencies of a marketing department and presumably would lead to greater influence of marketing within a firm.H2. Marketing's influence is related positively to a differentiation strategy compared with a cost leadership strategy.Entrepreneurial orientation As Stevenson and Gumpert (1985) noted, firms with an entrepreneurial focus tend to be flat organizations with informal networks of employees and functions. Entrepreneurial organizations typically have an evolving management structure and are more abstract in defining the roles of management (Freeman and Engle, 2007). Firms with a non-entrepreneurial focus are termed administratively oriented and are organizations with clearly defined authority and structure. It is reasonable to infer from Stevenson and Gumpert's work that firms with an administrative focus will have more clearly delineated functions/departments such as marketing. Bureaucracy has a negative effect on entrepreneurial activity in a firm, and this is anathema to an entrepreneurial-oriented firm (Sorensen, 2007). If the marketing function is more clearly defined, we would expect marketing to have greater influence.H3. Marketing's influence is related negatively to the entrepreneurial orientation of the firm.Background of CEO The top managers influence and shape the firm's direction and focus. Within SMEs, the top manager's role is pivotal for growth and expansion (Hutchinson et al., 2006). For smaller firms, this would seem even more relevant where all employees are likely to feel the CEO's influence. The psychologist Duncker (1945) first coined this phenomenon "functional fixity" to refer to the observation that people tend to react to an object in terms of its usually defined function. That is, a hammer is used to drive nails but is not frequently used as a paperweight, although obviously it could be. Business researchers have applied this idea to management and have found that managers are also influenced by their previous training and find it difficult to deviate from their functional training when devising a solution to a problem. For example, engineers seek technical solutions, accounting managers seek financial solutions, and marketing managers look to advertising and promotion to solve problems (Ashton, 1976; Barnes and Webb, 1986; Chang and Birnberg, 1977; Hand, 1990).H4. Marketing's influence is higher in firms where the chief executive officer has a marketing background.Firm size Recent studies have begun to focus on how firm size affects entrepreneurial orientation in organizations (Dholakia and Kshetri, 2004; Gompers et al., 2005; Dobrev and Barnett, 2005; Stuart and Ding, 2006.) As firms grow, they tend to become more bureaucratic. As firms become more bureaucratic, they tend to become less entrepreneurial (Sorensen, 2007). However, as firms grow, they are in a position to hire specialists and add to their "bench strength." As more specialists are hired, especially in the marketing function, it is assumed that marketing gains greater influence. This leads to our final hypothesis:H5. Within SMEs, marketing's influence is positively related to the size of firm.The proceeding hypothesized relationships are summarized in Figure 1. Sample We employed a mail survey of SMEs using two mailing lists available to the authors. The first mailing came from present and past participants in the Small Business Center of a large Midwestern university. This center caters to small businesses in the Midwestern region of the United States. Its mailing list, totaling 150 businesses, was screened to ensure only SMEs were included in the sample. The second mailing list was obtained from the Greenbrier Chamber of Commerce of West Virginia, which serves businesses in the central region of West Virginia. Again, the mailing list, totaling 100 businesses, was screened to ensure only SMEs were included. Combined with surveys from the Family Business Center, the total sample was 250. The surveys were sent to the top executives (e.g. CEOs, presidents, general managers, owners) of the firms. All the addresses from the mailing lists were valid because no mailings were returned as undeliverable. The survey was conducted October to November, 2006. Of the 250 surveys delivered, 100 completed surveys were returned, a response rate of 40 percent. The response rate from each source was the same. No inducements were included for participation in the survey. Instead, the director of the Family Business Center or Chamber of Commerce sent simple cover letters asking members to participate. We did not stimulate the response rate by doing follow-up contacts to nonresponders.Several steps were taken to address nonresponse error and other external validity concerns. First, on receipt of the completed survey instrument, respondents were contacted by telephone to verify that they personally completed the survey and were top managers of their firms. Of the 100 completed responses, we successfully contacted 45 (45 percent) respondents. In these 45 cases, the top managers of the firms acknowledged personally completing the survey and verified their positions as top managers of the firms.Second, comparisons were performed of sample means (t tests) between early and late survey respondents and the two sources for the sample. Across all independent and dependant measures, no statistical difference was found in sample means.Finally, this study's sample was compared against census data. Although this study used a convenience sample and cannot be generalized to all SMEs, a comparison was done to assess representativeness of the sample. As seen in Table I, our sample covered most employee size categories.As seen in Table II, we analyzed our sample in terms of coverage of NAIC categories. The findings suggest that our sample covered all major NAIC categories.Again, these comparisons do not suggest the study results can be generalized to all SMEs but were done to address potential nonresponse concerns.Measures The independent variables for our study included the influence of marketing, presence of a differentiation strategy, entrepreneurial orientation, years in business, number of full-time employees, and primary functional background of the CEO. The measures for this study have been used previously or adapted from previous studies.Influence of marketing was measured via a multi-item measure featuring a 100-point constant sum scale. This scale was adapted from Homburg et al.'s (1999) scale developed to measure the influence of marketing on decision making for ten strategic issues:1. advertising;2. customer satisfaction measurement;3. customer satisfaction improvement;4. expansion into new geographic markets;5. strategic direction of the firm;6. choice of strategic partners;7. new product development;8. design of customer service and support;9. pricing; and10. major capital expenditures.Homburg et al. (1999) chose these issues to represent those decisions classically thought to be handled by marketing (advertising, pricing) and other issues that are not completely under the control of one department. The list was not designed to be an exhaustive measure of all aspects of marketing's influence but a sampling of business decisions typically considered in the marketing domain. Respondents were asked to allocate 100 points in terms of influence of the following groups: marketing, sales, R&D, operations, finance, and general management on each of the ten issues. The option of "general management" was added to the original Homburg et al.'s (1999) scale because of concern that their original scale presumed that a general manager exerts no influence on decision making relating to marketing. We questioned that assumption and included that option into the scale. Respondents were told to allocate no points to a group if it did not exist in their firm. In no case did any respondent assign zero points to marketing, thus eliminating the potential confound of responses from firms without a marketing department. Respondents were also asked to rate the importance of each of the ten issues on a seven-point Likert scale (anchored by relatively low importance to extremely high importance).Marketing influence was calculated by multiplying the importance of each of the ten issues by the influence allocated to marketing for that issue and then adding across all ten issues. The summed figure was divided by the number of issues that the respondent answered.Whether a firm used a differentiation strategy was determined by using a seven-item measure used by Homburg et al. (1999), Kim and Lim (1988), and Dress and Davis (1984). Respondents gave specific measures in answer to the following questions about company activities: To what extent does your business emphasize the following activities?1. competitive advantage through superior products;2. creating superior value through services accompanying the products;3. new product development;4. building up a premium product or brand image;5. obtaining high prices from the market;6. advertising; and7. development of customer specific solutions and products.A seven-point Likert scale anchored by Not at all and A great deal was used to record responses (alpha=0.84).The entrepreneurial orientation of the firm was measured using a seven-item scale developed by Covin and Slevin (1989) and subsequently used by Keh et al. (2007). The specific measures were:1. our culture emphasizes innovation and research and development;2. we have a low rate of new product innovation;3. we are a follower and have a defensive posture in the market;4. we are first to introduce new technologies and products;5. we maintain a competitive posture toward competition;6. we have a strong dislike for high risk, high return projects; and7. when faced with risk, we adopt an aggressive, bold posture.A seven-point Likert scale, anchored by Strongly disagree and Strongly agree was used to record responses. Items (2), (3), and (6) were reverse coded (alpha=0.83).The age of business (years), number of full-time employees, and primary functional background of the CEO were single-item, open-response measures. Regarding the measure of full-time employees, any response of "more than 500 employees" would have been discarded because this study was limited to SMEs. However, we did not receive any surveys with this response. Functional background of CEO was dummy coded (0=non-marketing background; 1=marketing background). The type of market served was dummy coded (0=industrial; 1=consumer).The dependent variables included sales from most recent year, percent of sales spent on marketing, percent of sales spent on sales activities, and percent of sales spent on advertising. The variables were measured via single-item measures. For sales, respondents were given four options: under $1 million, $1-5 million, $5-25 million, and above $25 million. The percent of sales measures (marketing activities, sales activities, and advertising spending) had 13 response options: (0, less than 1 percent, 1 percent, 2 percent, 3 percent, 4 percent, 5 percent, 6 percent, 7 percent, 8 percent, 9 percent, 10 percent, and above 10 percent). The results from the two samples were analyzed (t-tests) and no statistical difference was found in means across all measures. Accordingly, the results were pooled.The role of marketing Conventional wisdom suggests that marketing's role is to manage the relationship between customers and a firm (Moorman and Rust, 1999). Because customers are an important contributor to a firm's financial performance, the department responsible for management of customer relationships should be viewed as an important force; thus, the marketing department that assumes this management function should be deemed important and influential. Homburg et al.'s (1999) study showed the marketing function to be a significant influence in large firms. Turning to this study, as shown in Table III, marketing shares influence with general management and sales on most marketing issues. This suggests that marketing, as function and influence in business, is not as well developed for smaller firms compared with their larger brethren. The sales function has a slightly greater influence over marketing on all issues outside of advertising (customer satisfaction measurement, customer satisfaction improvement, expansion into new geographic markets, strategic direction of the firm, choice of strategic partners, new product development, design of customer service and support, and pricing and major capital expenditures). Decisions regarding advertising are the only issue where marketing enjoys significant and sole influence. We have included comparative data from the Homberg et al. (1999) study, and the data suggest marketing has a way to go before it enjoys the same influence found in larger firms.Hypotheses testing Regression analysis was used to test the hypotheses. We ran three regression analyses: on all issues, "marketing" issues (advertising messages, distribution strategy, design of customer sales and support, customer satisfaction measurement and customer satisfaction improvement) and non-marketing exclusive issues (strategic direction of business unit, choice of strategic partners, new product development, major capital expenditures, and expansion into new geographic markets). Results were the same, and for purposes of discussion, we will focus on the first regression analysis although results from all three analyses are reported in Table IV. Firm size, sales; percent spent of marketing, sales, and advertising activities; source of survey, and early/late responders were included as control variables.H1 proposed that marketing's influence is higher in firms selling consumer goods than in firms selling industrial goods. We found support for this (b = 48.68, p<0.001). Firms in consumer markets tend to do more advertising, considered the exclusive province of marketing, which likely leads to firms having greater "bench strength" in the marketing department. Interestingly, this conclusion differs from Homburg et al.'s (1999) finding for large firms where no difference appeared in marketing's influence between consumer and industrial firms.H3 stated that marketing's influence is related negatively to the entrepreneurial orientation of the firm. We found support for this (b =.-20.14, p<0.001). Less entrepreneurial firms have more influential marketing departments. This suggests a more hierarchical firm has more clearly defined functions leading to greater influence of marketing.We found no support for the remaining hypotheses (H2, H4 and H5). H2 stated that marketing's influence is related positively to a differentiation strategy. Evidently a differentiation strategy does not automatically mean marketing has greater influence. This contrasts with the Homburg et al. (1999) survey where large businesses employing a differentiation strategy had greater marketing influence. H4 stated that marketing's influence is higher where the CEO has a marketing background. Unlike larger firms, the notion that that the CEO's background indicates which firm functions might have influence was not supported. Finally, H5 related to firm size within SMEs. Firm size was not a significant factor in determining marketing's influence. H4 and H5 results were consistent with Homburg et al.'s (1999) study. A comparison of this study's results and the Homburg et al.'s (1999) can be found in Table V. Our data suggest that marketing is not as well developed or influential in SMEs as it is in large firms. Marketing's scope or extent of responsibility is more limited compared with large firms. Two factors, type of market (consumer) and firm orientation (hierarchal), facilitate marketing's influence within a firm. These results indicate that both underlying theories - institutional and contingency - received support and neither theory can be dismissed, suggesting the influence of marketing is a multidimensional construct. Individual traits (e.g. charisma, power, authority) were not considered in this study. However, our findings show marketing's influence can vary systematically as a function of institutional and external factors.Although our analysis sheds light on some internal and external factors that influence marketing's role in a firm, a major portion of variance remains unexplained. Additional research is necessary to explore other potential influences on marketing's role in a firm. These might include personnel factors (background and experience of marketing department personnel) and firm politics. For example, Reijonen and Komppula (2007) studied SMEs in Finland and found the motivation of entrepreneurs are not solely financial and often are based on job satisfaction, control, and family concerns.Although this research fills a gap in the literature relating to the influence of marketing in SMEs, there is much work to be done. We did not address the relationship between marketing's influence and the performance of firms. Analyses done and conclusions reached in this research were based on a very limited sample of SMEs located in one region of the country. Sampling across all regions of the United States would help in generalization of the findings. The results of this study are particularly troubling because marketing resources are one driver of competitive advantage, and although the studied firms have a formal marketing function within the firm, it would appear the function has not proven its value to the firm. For marketing to increase in influence, individuals leading this function must gain a seat at the table. Our findings would suggest that changing the influence of the marketing function might take considerable effort. One barrier to changing marketing's influence is the firm's entrepreneurial orientation, which can be a deep-seated cultural factor that typically requires time to change. An Indian proverb cautions: "Under a bright lamp, there is great darkness." Although marketing departments have a responsibility to market the firm's products and services, the task of marketers marketing themselves to internal stakeholders remains unanswered. Marketing departments must do more to ensure that the voice of marketing is heard when key decisions are being made. Opens in a new window.Figure 1 Conceptual model Opens in a new window.Table I Firms by employee size Opens in a new window.Table II Percentage of firms by industry Opens in a new window.Table III Influence of business functional departments on marketing issues Opens in a new window.Table IV Results of regression analysis Opens in a new window.Table V Comparison of study results with Homberg et al. (1999) study
|
- The study is conducted with 100 SMEs located in the mid-Atlantic region of the United States.
|
[SECTION: Findings] A precept of the marketing concept contends that business achieves success by determining and satisfying the needs, wants, and aspirations of target markets. Few would argue that this determination and satisfaction of target market wants and needs is critical for firm success. These concepts, traditionally thought to be part of the marketing function of the firm, have fueled scholars' interest in the role of marketing within the firm (e.g., Becherer et al., 2003; Berthon et al., 2008; Moorman and Rust, 1999; Simpson and Taylor, 2002; Webster, 1981, 1992, 2003; Webster et al., 2003).Scholars have identified significant differences between large and small organizations. Large organizations tend to use a structured framework with a clear hierarchy in decision making. On the other hand, small firms tend to feature processes that begin with and highly involve the entrepreneur or owner. Furthermore, the entrepreneur or owner's personality and style help to shape decision making (Sadler-Smith et al., 2003).The small-firm sector plays a significant role in the world economy. In the United States a recent analysis of employment changes between September 1992 and March 2005 showed that 65 percent of the net new jobs created during that time are attributed to firms with fewer than 500 employees (US Bureau of Labor Statistics, 2005).Research investigating the competitive advantage of small firms has consistently emphasized the importance of marketing, strategic positioning, and entrepreneurship as key factors in business survival and growth. The ability to identify and operate in a particular market niche enables the firm to exploit a range of specializations and offers protection from larger competitors. Yet despite the widespread acceptance of the importance of the marketing concept, the precise marketing activities and competencies that contribute most strongly to business performance must be identified for small and medium-sized enterprises (SMEs).Our study of SMEs revisits many of the questions Homburg et al. (1999) posed in their study of the marketing functions of large firms, but we direct our attention to SMEs. The definition of what precisely constitutes an SME differs widely. For the purposes of this paper, we follow the Journal of Enterprise Culture and the Asia Pacific Economic Cooperative (APEC), who define US-based SMEs as companies with fewer than 499 employees (APEC, 2003). In this research, we seek to answer the following questions:* Relative to other business functions, how important is the marketing function in SMEs?* Does marketing in SMEs enjoy the same influence found in larger firms?* What internal and external factors influence the role of marketing in small firms?This paper reports the results of a study that employed an activity-based approach whereby demonstrable marketing competencies were related to the entrepreneurial orientation evident in the firms. The analysis demonstrated that certain competencies were more strongly associated with a marketing orientation, whereas others were associated with an entrepreneurial orientation. The results show differences in the role of marketing in small firms as compared with the role of marketing in the large firms that were studied by Homburg et al. (1999). SME marketing functions have a way to go before they enjoy the same degree of influence their counterparts have in large firms. Inquiry into departmental influence (such as marketing) is based on Cyert and March's (1963) work on subunit power. Cyert and March (1963) argued that managers have conflicting goals and seek satisfactory solutions as opposed to optimal solutions to business problems. The concept of "dominant coalition" has been used to explain differences in the power of firm subunits (Thompson, 1967). The relative power of subunits is a function of strategic choices made by dominant coalitions (i.e. power members) within an organization. Influence has been defined as "a change in one party that has its origin in another party and this embodies the successful exercise of power" (Stern and Scheer, 1992, p. 260). Our study focuses on the marketing department subunit. Various authors have examined the role and influence of marketing in firms (e.g. Becherer et al., 2003; Berthon et al., 2008; Moorman and Rust, 1999; Simpson and Taylor, 2002; Webster, 1981, 1992, 2003; Webster et al., 2003) and since the 1980s, the marketing department has been shown to have varying levels of influence in the firm. Homburg et al. (1999) directly studied the influence of marketing in large firms and uncovered a number of situational factors that affect the level of marketing's influence. These situational factors include a firm's strategic orientation (differentiation versus cost leadership) and background of the CEO. Although their study helped to define the relative level of influence of the marketing function for business, the authors focused on large organizations - firms with more than $25 million in sales.However, SMEs have characteristics that differentiate them from large organizations (McCartan-Quinn and Carson, 2003). These differences include advantages such as greater flexibility, innovation, and lower overhead costs. In terms of disadvantages, SMEs are limited by their market power, and capital and managerial resources (Motwani et al., 1998). These differences raise the question, "does the marketing function in SMEs enjoy the same influence found in larger firms?" Compared with the literature stream surrounding the role of marketing in larger firms, inquiry on this topic has been more limited for SMEs. Many studies have attempted to define marketing and outcomes of marketing for SMEs. Carson (2001) and Sui and Kirby (1998) traced the evolution of marketing and the various approaches to SME marketing. Other authors have attempted to develop hypothetical and empirical models of marketing for SMEs. Sui et al. (2004), Julien and Ramangalahy (2003) and Berthon et al. (2008) showed how strategic marketing practices such as knowledge of current market conditions and consumer tastes were positively related to SME performance. Becherer et al. (2003) examined internal environmental factors such as the background and decision processes of CEOs. One aspect of marketing, promotional efforts, was found to be a key influence in performance of SMEs (Wood, 2006). Market orientation as a driver of SME business performance has also generated scholar interest (Blankson and Stokes, 2002; Fillis, 2002; Pacitto et al., 2007). Finally, authors have studied underlying reasons for the characteristics of SME marketing practices. Simpson et al. (2006) examined drivers of marketing effort such as the presence of a marketing department and marketing representation at the board level. SMEs lack the resources to compete head-to-head with larger rivals and thus cannot do traditional marketing (Gilmore et al., 2001).Because SMEs lack the resources to compete head-to-head with larger rivals, some scholars have questioned whether SMEs formally practice marketing at all (Carson et al., 1998; Gilmore et al., 2001; Shepherd and Ridnour, 1996). In a typical argument, Hogarth-Scott et al. (1996) considered most marketing theories to be inappropriate for SMEs and not helpful in the understanding of their markets. However, in the same article, the authors found that the marketing function contributed positively to small business success and the ability to think strategically. Clearly, marketing department exert influence in SMEs, especially in SMEs that succeed. Larger firms achieve better competitive positions than smaller firms when they have greater marketing capabilities (Grimes et al., 2007), so SMEs must maintain strong marketing departments to effectively compete.Evidence has shown that although marketing activities in SMEs may be different, marketing departments are still critical to firms' success. Many firms carry out business via highly informal, unstructured, reactive mechanisms, whereas others develop, over time, a proactive and skilled approach where innovation and identification of opportunities gives them a competitive advantage (Fillis, 2007). Romano and Ratnatunga (1995) observed that "small firms face marketing challenges which can and will ultimately determine their future." Furthermore, Hill (2001) found that SMEs still engage in many traditional marketing functions, especially marketing planning. A widely cited marketing activity for SMEs is networking. Wincent (2005) showed that the size of a firm matters in regard to how a company conducts its networking activity. Networks are important during the establishment, development, and growth of SMEs. Siu (2001) found that SMEs rely heavily on their personal contact network in marketing their firms. Traditional economic structures favor size (large firms). However, today's economy is earmarked by relationships, network, and information, which play to some of the characteristics of SMEs.Our research adds to the SME literature stream by examining outputs and attributes of marketing (e.g. sales, percent of sales spent on marketing) and by considering how these are related to firm characteristics such as entrepreneurial orientation. Two theories provide a conceptual framework to explain the differences in the influence of the marketing function: institutional theory and contingency theory. Institutional theory states that organizational structure and design is inexorably linked to social networks and is shaped by conformity and legitimacy pressures (DiMaggio and Powell, 1983; Meyer and Rowan, 1977). Firm structure does not necessarily reflect efficiency but legitimacy (Meyer and Rowan, 1977). Thus the influence of a firm function such as marketing is a function of conformity/legitimacy pressures.Alternatively, contingency theory states that organizational structure and design is a function of external factors. Firms must "fit" to their environment, and characteristics of firms such as subunit power are a function of environmental determinants. Research on subunit power has explored relationships between environmental factors and organizational structures/power (Boeker, 1989; Hambrick, 1981; Hinings et al., 1974; Salancik and Pfeffer, 1974).This research tests elements of both theories. Hypotheses H1 and H2 are based on the contingency theory argument, and hypotheses H3, H4, and H5 are based on institutional theory.Consumer versus industrial markets Firms serving industrial markets have fewer customers (as compared with consumer markets) and have closer partnerships with their customers (Heide and John, 1992). These partnerships involve more aspects of the firm and, consequently, the marketing department no longer serves as the primary link between the firm and its customers (Homburg et al., 1999). In firms serving consumer markets, the marketing function serves as the primary link between the firm and its market. Thus, it stands to reason that the marketing function has greater influence in these firms.H1. Marketing's influence is higher in SMEs selling consumer goods than in firms selling industrial goods.Differentiation strategy Using Porter's (1980) typology, firms can pursue a cost leadership or differentiation strategy. Bennett and Smith (2002) showed that as SMEs grow, they develop greater specialization and differentiation strategies. Although not empirically proven, conventional wisdom has suggested that marketing's influence is more influential for a differentiation strategy than for a cost leadership strategy. For a firm to pursue a differentiation strategy, it must identify customer needs and develop products to fit those needs. Identifying those needs is one of the core competencies of a marketing department and presumably would lead to greater influence of marketing within a firm.H2. Marketing's influence is related positively to a differentiation strategy compared with a cost leadership strategy.Entrepreneurial orientation As Stevenson and Gumpert (1985) noted, firms with an entrepreneurial focus tend to be flat organizations with informal networks of employees and functions. Entrepreneurial organizations typically have an evolving management structure and are more abstract in defining the roles of management (Freeman and Engle, 2007). Firms with a non-entrepreneurial focus are termed administratively oriented and are organizations with clearly defined authority and structure. It is reasonable to infer from Stevenson and Gumpert's work that firms with an administrative focus will have more clearly delineated functions/departments such as marketing. Bureaucracy has a negative effect on entrepreneurial activity in a firm, and this is anathema to an entrepreneurial-oriented firm (Sorensen, 2007). If the marketing function is more clearly defined, we would expect marketing to have greater influence.H3. Marketing's influence is related negatively to the entrepreneurial orientation of the firm.Background of CEO The top managers influence and shape the firm's direction and focus. Within SMEs, the top manager's role is pivotal for growth and expansion (Hutchinson et al., 2006). For smaller firms, this would seem even more relevant where all employees are likely to feel the CEO's influence. The psychologist Duncker (1945) first coined this phenomenon "functional fixity" to refer to the observation that people tend to react to an object in terms of its usually defined function. That is, a hammer is used to drive nails but is not frequently used as a paperweight, although obviously it could be. Business researchers have applied this idea to management and have found that managers are also influenced by their previous training and find it difficult to deviate from their functional training when devising a solution to a problem. For example, engineers seek technical solutions, accounting managers seek financial solutions, and marketing managers look to advertising and promotion to solve problems (Ashton, 1976; Barnes and Webb, 1986; Chang and Birnberg, 1977; Hand, 1990).H4. Marketing's influence is higher in firms where the chief executive officer has a marketing background.Firm size Recent studies have begun to focus on how firm size affects entrepreneurial orientation in organizations (Dholakia and Kshetri, 2004; Gompers et al., 2005; Dobrev and Barnett, 2005; Stuart and Ding, 2006.) As firms grow, they tend to become more bureaucratic. As firms become more bureaucratic, they tend to become less entrepreneurial (Sorensen, 2007). However, as firms grow, they are in a position to hire specialists and add to their "bench strength." As more specialists are hired, especially in the marketing function, it is assumed that marketing gains greater influence. This leads to our final hypothesis:H5. Within SMEs, marketing's influence is positively related to the size of firm.The proceeding hypothesized relationships are summarized in Figure 1. Sample We employed a mail survey of SMEs using two mailing lists available to the authors. The first mailing came from present and past participants in the Small Business Center of a large Midwestern university. This center caters to small businesses in the Midwestern region of the United States. Its mailing list, totaling 150 businesses, was screened to ensure only SMEs were included in the sample. The second mailing list was obtained from the Greenbrier Chamber of Commerce of West Virginia, which serves businesses in the central region of West Virginia. Again, the mailing list, totaling 100 businesses, was screened to ensure only SMEs were included. Combined with surveys from the Family Business Center, the total sample was 250. The surveys were sent to the top executives (e.g. CEOs, presidents, general managers, owners) of the firms. All the addresses from the mailing lists were valid because no mailings were returned as undeliverable. The survey was conducted October to November, 2006. Of the 250 surveys delivered, 100 completed surveys were returned, a response rate of 40 percent. The response rate from each source was the same. No inducements were included for participation in the survey. Instead, the director of the Family Business Center or Chamber of Commerce sent simple cover letters asking members to participate. We did not stimulate the response rate by doing follow-up contacts to nonresponders.Several steps were taken to address nonresponse error and other external validity concerns. First, on receipt of the completed survey instrument, respondents were contacted by telephone to verify that they personally completed the survey and were top managers of their firms. Of the 100 completed responses, we successfully contacted 45 (45 percent) respondents. In these 45 cases, the top managers of the firms acknowledged personally completing the survey and verified their positions as top managers of the firms.Second, comparisons were performed of sample means (t tests) between early and late survey respondents and the two sources for the sample. Across all independent and dependant measures, no statistical difference was found in sample means.Finally, this study's sample was compared against census data. Although this study used a convenience sample and cannot be generalized to all SMEs, a comparison was done to assess representativeness of the sample. As seen in Table I, our sample covered most employee size categories.As seen in Table II, we analyzed our sample in terms of coverage of NAIC categories. The findings suggest that our sample covered all major NAIC categories.Again, these comparisons do not suggest the study results can be generalized to all SMEs but were done to address potential nonresponse concerns.Measures The independent variables for our study included the influence of marketing, presence of a differentiation strategy, entrepreneurial orientation, years in business, number of full-time employees, and primary functional background of the CEO. The measures for this study have been used previously or adapted from previous studies.Influence of marketing was measured via a multi-item measure featuring a 100-point constant sum scale. This scale was adapted from Homburg et al.'s (1999) scale developed to measure the influence of marketing on decision making for ten strategic issues:1. advertising;2. customer satisfaction measurement;3. customer satisfaction improvement;4. expansion into new geographic markets;5. strategic direction of the firm;6. choice of strategic partners;7. new product development;8. design of customer service and support;9. pricing; and10. major capital expenditures.Homburg et al. (1999) chose these issues to represent those decisions classically thought to be handled by marketing (advertising, pricing) and other issues that are not completely under the control of one department. The list was not designed to be an exhaustive measure of all aspects of marketing's influence but a sampling of business decisions typically considered in the marketing domain. Respondents were asked to allocate 100 points in terms of influence of the following groups: marketing, sales, R&D, operations, finance, and general management on each of the ten issues. The option of "general management" was added to the original Homburg et al.'s (1999) scale because of concern that their original scale presumed that a general manager exerts no influence on decision making relating to marketing. We questioned that assumption and included that option into the scale. Respondents were told to allocate no points to a group if it did not exist in their firm. In no case did any respondent assign zero points to marketing, thus eliminating the potential confound of responses from firms without a marketing department. Respondents were also asked to rate the importance of each of the ten issues on a seven-point Likert scale (anchored by relatively low importance to extremely high importance).Marketing influence was calculated by multiplying the importance of each of the ten issues by the influence allocated to marketing for that issue and then adding across all ten issues. The summed figure was divided by the number of issues that the respondent answered.Whether a firm used a differentiation strategy was determined by using a seven-item measure used by Homburg et al. (1999), Kim and Lim (1988), and Dress and Davis (1984). Respondents gave specific measures in answer to the following questions about company activities: To what extent does your business emphasize the following activities?1. competitive advantage through superior products;2. creating superior value through services accompanying the products;3. new product development;4. building up a premium product or brand image;5. obtaining high prices from the market;6. advertising; and7. development of customer specific solutions and products.A seven-point Likert scale anchored by Not at all and A great deal was used to record responses (alpha=0.84).The entrepreneurial orientation of the firm was measured using a seven-item scale developed by Covin and Slevin (1989) and subsequently used by Keh et al. (2007). The specific measures were:1. our culture emphasizes innovation and research and development;2. we have a low rate of new product innovation;3. we are a follower and have a defensive posture in the market;4. we are first to introduce new technologies and products;5. we maintain a competitive posture toward competition;6. we have a strong dislike for high risk, high return projects; and7. when faced with risk, we adopt an aggressive, bold posture.A seven-point Likert scale, anchored by Strongly disagree and Strongly agree was used to record responses. Items (2), (3), and (6) were reverse coded (alpha=0.83).The age of business (years), number of full-time employees, and primary functional background of the CEO were single-item, open-response measures. Regarding the measure of full-time employees, any response of "more than 500 employees" would have been discarded because this study was limited to SMEs. However, we did not receive any surveys with this response. Functional background of CEO was dummy coded (0=non-marketing background; 1=marketing background). The type of market served was dummy coded (0=industrial; 1=consumer).The dependent variables included sales from most recent year, percent of sales spent on marketing, percent of sales spent on sales activities, and percent of sales spent on advertising. The variables were measured via single-item measures. For sales, respondents were given four options: under $1 million, $1-5 million, $5-25 million, and above $25 million. The percent of sales measures (marketing activities, sales activities, and advertising spending) had 13 response options: (0, less than 1 percent, 1 percent, 2 percent, 3 percent, 4 percent, 5 percent, 6 percent, 7 percent, 8 percent, 9 percent, 10 percent, and above 10 percent). The results from the two samples were analyzed (t-tests) and no statistical difference was found in means across all measures. Accordingly, the results were pooled.The role of marketing Conventional wisdom suggests that marketing's role is to manage the relationship between customers and a firm (Moorman and Rust, 1999). Because customers are an important contributor to a firm's financial performance, the department responsible for management of customer relationships should be viewed as an important force; thus, the marketing department that assumes this management function should be deemed important and influential. Homburg et al.'s (1999) study showed the marketing function to be a significant influence in large firms. Turning to this study, as shown in Table III, marketing shares influence with general management and sales on most marketing issues. This suggests that marketing, as function and influence in business, is not as well developed for smaller firms compared with their larger brethren. The sales function has a slightly greater influence over marketing on all issues outside of advertising (customer satisfaction measurement, customer satisfaction improvement, expansion into new geographic markets, strategic direction of the firm, choice of strategic partners, new product development, design of customer service and support, and pricing and major capital expenditures). Decisions regarding advertising are the only issue where marketing enjoys significant and sole influence. We have included comparative data from the Homberg et al. (1999) study, and the data suggest marketing has a way to go before it enjoys the same influence found in larger firms.Hypotheses testing Regression analysis was used to test the hypotheses. We ran three regression analyses: on all issues, "marketing" issues (advertising messages, distribution strategy, design of customer sales and support, customer satisfaction measurement and customer satisfaction improvement) and non-marketing exclusive issues (strategic direction of business unit, choice of strategic partners, new product development, major capital expenditures, and expansion into new geographic markets). Results were the same, and for purposes of discussion, we will focus on the first regression analysis although results from all three analyses are reported in Table IV. Firm size, sales; percent spent of marketing, sales, and advertising activities; source of survey, and early/late responders were included as control variables.H1 proposed that marketing's influence is higher in firms selling consumer goods than in firms selling industrial goods. We found support for this (b = 48.68, p<0.001). Firms in consumer markets tend to do more advertising, considered the exclusive province of marketing, which likely leads to firms having greater "bench strength" in the marketing department. Interestingly, this conclusion differs from Homburg et al.'s (1999) finding for large firms where no difference appeared in marketing's influence between consumer and industrial firms.H3 stated that marketing's influence is related negatively to the entrepreneurial orientation of the firm. We found support for this (b =.-20.14, p<0.001). Less entrepreneurial firms have more influential marketing departments. This suggests a more hierarchical firm has more clearly defined functions leading to greater influence of marketing.We found no support for the remaining hypotheses (H2, H4 and H5). H2 stated that marketing's influence is related positively to a differentiation strategy. Evidently a differentiation strategy does not automatically mean marketing has greater influence. This contrasts with the Homburg et al. (1999) survey where large businesses employing a differentiation strategy had greater marketing influence. H4 stated that marketing's influence is higher where the CEO has a marketing background. Unlike larger firms, the notion that that the CEO's background indicates which firm functions might have influence was not supported. Finally, H5 related to firm size within SMEs. Firm size was not a significant factor in determining marketing's influence. H4 and H5 results were consistent with Homburg et al.'s (1999) study. A comparison of this study's results and the Homburg et al.'s (1999) can be found in Table V. Our data suggest that marketing is not as well developed or influential in SMEs as it is in large firms. Marketing's scope or extent of responsibility is more limited compared with large firms. Two factors, type of market (consumer) and firm orientation (hierarchal), facilitate marketing's influence within a firm. These results indicate that both underlying theories - institutional and contingency - received support and neither theory can be dismissed, suggesting the influence of marketing is a multidimensional construct. Individual traits (e.g. charisma, power, authority) were not considered in this study. However, our findings show marketing's influence can vary systematically as a function of institutional and external factors.Although our analysis sheds light on some internal and external factors that influence marketing's role in a firm, a major portion of variance remains unexplained. Additional research is necessary to explore other potential influences on marketing's role in a firm. These might include personnel factors (background and experience of marketing department personnel) and firm politics. For example, Reijonen and Komppula (2007) studied SMEs in Finland and found the motivation of entrepreneurs are not solely financial and often are based on job satisfaction, control, and family concerns.Although this research fills a gap in the literature relating to the influence of marketing in SMEs, there is much work to be done. We did not address the relationship between marketing's influence and the performance of firms. Analyses done and conclusions reached in this research were based on a very limited sample of SMEs located in one region of the country. Sampling across all regions of the United States would help in generalization of the findings. The results of this study are particularly troubling because marketing resources are one driver of competitive advantage, and although the studied firms have a formal marketing function within the firm, it would appear the function has not proven its value to the firm. For marketing to increase in influence, individuals leading this function must gain a seat at the table. Our findings would suggest that changing the influence of the marketing function might take considerable effort. One barrier to changing marketing's influence is the firm's entrepreneurial orientation, which can be a deep-seated cultural factor that typically requires time to change. An Indian proverb cautions: "Under a bright lamp, there is great darkness." Although marketing departments have a responsibility to market the firm's products and services, the task of marketers marketing themselves to internal stakeholders remains unanswered. Marketing departments must do more to ensure that the voice of marketing is heard when key decisions are being made. Opens in a new window.Figure 1 Conceptual model Opens in a new window.Table I Firms by employee size Opens in a new window.Table II Percentage of firms by industry Opens in a new window.Table III Influence of business functional departments on marketing issues Opens in a new window.Table IV Results of regression analysis Opens in a new window.Table V Comparison of study results with Homberg et al. (1999) study
|
- The marketing function is not as well developed or influential in SMEs as it is in large corporations. Two environmental factors, type of market (consumer) and firm orientation (hierarchal), facilitate marketing's influence within a firm.
|
[SECTION: Value] A precept of the marketing concept contends that business achieves success by determining and satisfying the needs, wants, and aspirations of target markets. Few would argue that this determination and satisfaction of target market wants and needs is critical for firm success. These concepts, traditionally thought to be part of the marketing function of the firm, have fueled scholars' interest in the role of marketing within the firm (e.g., Becherer et al., 2003; Berthon et al., 2008; Moorman and Rust, 1999; Simpson and Taylor, 2002; Webster, 1981, 1992, 2003; Webster et al., 2003).Scholars have identified significant differences between large and small organizations. Large organizations tend to use a structured framework with a clear hierarchy in decision making. On the other hand, small firms tend to feature processes that begin with and highly involve the entrepreneur or owner. Furthermore, the entrepreneur or owner's personality and style help to shape decision making (Sadler-Smith et al., 2003).The small-firm sector plays a significant role in the world economy. In the United States a recent analysis of employment changes between September 1992 and March 2005 showed that 65 percent of the net new jobs created during that time are attributed to firms with fewer than 500 employees (US Bureau of Labor Statistics, 2005).Research investigating the competitive advantage of small firms has consistently emphasized the importance of marketing, strategic positioning, and entrepreneurship as key factors in business survival and growth. The ability to identify and operate in a particular market niche enables the firm to exploit a range of specializations and offers protection from larger competitors. Yet despite the widespread acceptance of the importance of the marketing concept, the precise marketing activities and competencies that contribute most strongly to business performance must be identified for small and medium-sized enterprises (SMEs).Our study of SMEs revisits many of the questions Homburg et al. (1999) posed in their study of the marketing functions of large firms, but we direct our attention to SMEs. The definition of what precisely constitutes an SME differs widely. For the purposes of this paper, we follow the Journal of Enterprise Culture and the Asia Pacific Economic Cooperative (APEC), who define US-based SMEs as companies with fewer than 499 employees (APEC, 2003). In this research, we seek to answer the following questions:* Relative to other business functions, how important is the marketing function in SMEs?* Does marketing in SMEs enjoy the same influence found in larger firms?* What internal and external factors influence the role of marketing in small firms?This paper reports the results of a study that employed an activity-based approach whereby demonstrable marketing competencies were related to the entrepreneurial orientation evident in the firms. The analysis demonstrated that certain competencies were more strongly associated with a marketing orientation, whereas others were associated with an entrepreneurial orientation. The results show differences in the role of marketing in small firms as compared with the role of marketing in the large firms that were studied by Homburg et al. (1999). SME marketing functions have a way to go before they enjoy the same degree of influence their counterparts have in large firms. Inquiry into departmental influence (such as marketing) is based on Cyert and March's (1963) work on subunit power. Cyert and March (1963) argued that managers have conflicting goals and seek satisfactory solutions as opposed to optimal solutions to business problems. The concept of "dominant coalition" has been used to explain differences in the power of firm subunits (Thompson, 1967). The relative power of subunits is a function of strategic choices made by dominant coalitions (i.e. power members) within an organization. Influence has been defined as "a change in one party that has its origin in another party and this embodies the successful exercise of power" (Stern and Scheer, 1992, p. 260). Our study focuses on the marketing department subunit. Various authors have examined the role and influence of marketing in firms (e.g. Becherer et al., 2003; Berthon et al., 2008; Moorman and Rust, 1999; Simpson and Taylor, 2002; Webster, 1981, 1992, 2003; Webster et al., 2003) and since the 1980s, the marketing department has been shown to have varying levels of influence in the firm. Homburg et al. (1999) directly studied the influence of marketing in large firms and uncovered a number of situational factors that affect the level of marketing's influence. These situational factors include a firm's strategic orientation (differentiation versus cost leadership) and background of the CEO. Although their study helped to define the relative level of influence of the marketing function for business, the authors focused on large organizations - firms with more than $25 million in sales.However, SMEs have characteristics that differentiate them from large organizations (McCartan-Quinn and Carson, 2003). These differences include advantages such as greater flexibility, innovation, and lower overhead costs. In terms of disadvantages, SMEs are limited by their market power, and capital and managerial resources (Motwani et al., 1998). These differences raise the question, "does the marketing function in SMEs enjoy the same influence found in larger firms?" Compared with the literature stream surrounding the role of marketing in larger firms, inquiry on this topic has been more limited for SMEs. Many studies have attempted to define marketing and outcomes of marketing for SMEs. Carson (2001) and Sui and Kirby (1998) traced the evolution of marketing and the various approaches to SME marketing. Other authors have attempted to develop hypothetical and empirical models of marketing for SMEs. Sui et al. (2004), Julien and Ramangalahy (2003) and Berthon et al. (2008) showed how strategic marketing practices such as knowledge of current market conditions and consumer tastes were positively related to SME performance. Becherer et al. (2003) examined internal environmental factors such as the background and decision processes of CEOs. One aspect of marketing, promotional efforts, was found to be a key influence in performance of SMEs (Wood, 2006). Market orientation as a driver of SME business performance has also generated scholar interest (Blankson and Stokes, 2002; Fillis, 2002; Pacitto et al., 2007). Finally, authors have studied underlying reasons for the characteristics of SME marketing practices. Simpson et al. (2006) examined drivers of marketing effort such as the presence of a marketing department and marketing representation at the board level. SMEs lack the resources to compete head-to-head with larger rivals and thus cannot do traditional marketing (Gilmore et al., 2001).Because SMEs lack the resources to compete head-to-head with larger rivals, some scholars have questioned whether SMEs formally practice marketing at all (Carson et al., 1998; Gilmore et al., 2001; Shepherd and Ridnour, 1996). In a typical argument, Hogarth-Scott et al. (1996) considered most marketing theories to be inappropriate for SMEs and not helpful in the understanding of their markets. However, in the same article, the authors found that the marketing function contributed positively to small business success and the ability to think strategically. Clearly, marketing department exert influence in SMEs, especially in SMEs that succeed. Larger firms achieve better competitive positions than smaller firms when they have greater marketing capabilities (Grimes et al., 2007), so SMEs must maintain strong marketing departments to effectively compete.Evidence has shown that although marketing activities in SMEs may be different, marketing departments are still critical to firms' success. Many firms carry out business via highly informal, unstructured, reactive mechanisms, whereas others develop, over time, a proactive and skilled approach where innovation and identification of opportunities gives them a competitive advantage (Fillis, 2007). Romano and Ratnatunga (1995) observed that "small firms face marketing challenges which can and will ultimately determine their future." Furthermore, Hill (2001) found that SMEs still engage in many traditional marketing functions, especially marketing planning. A widely cited marketing activity for SMEs is networking. Wincent (2005) showed that the size of a firm matters in regard to how a company conducts its networking activity. Networks are important during the establishment, development, and growth of SMEs. Siu (2001) found that SMEs rely heavily on their personal contact network in marketing their firms. Traditional economic structures favor size (large firms). However, today's economy is earmarked by relationships, network, and information, which play to some of the characteristics of SMEs.Our research adds to the SME literature stream by examining outputs and attributes of marketing (e.g. sales, percent of sales spent on marketing) and by considering how these are related to firm characteristics such as entrepreneurial orientation. Two theories provide a conceptual framework to explain the differences in the influence of the marketing function: institutional theory and contingency theory. Institutional theory states that organizational structure and design is inexorably linked to social networks and is shaped by conformity and legitimacy pressures (DiMaggio and Powell, 1983; Meyer and Rowan, 1977). Firm structure does not necessarily reflect efficiency but legitimacy (Meyer and Rowan, 1977). Thus the influence of a firm function such as marketing is a function of conformity/legitimacy pressures.Alternatively, contingency theory states that organizational structure and design is a function of external factors. Firms must "fit" to their environment, and characteristics of firms such as subunit power are a function of environmental determinants. Research on subunit power has explored relationships between environmental factors and organizational structures/power (Boeker, 1989; Hambrick, 1981; Hinings et al., 1974; Salancik and Pfeffer, 1974).This research tests elements of both theories. Hypotheses H1 and H2 are based on the contingency theory argument, and hypotheses H3, H4, and H5 are based on institutional theory.Consumer versus industrial markets Firms serving industrial markets have fewer customers (as compared with consumer markets) and have closer partnerships with their customers (Heide and John, 1992). These partnerships involve more aspects of the firm and, consequently, the marketing department no longer serves as the primary link between the firm and its customers (Homburg et al., 1999). In firms serving consumer markets, the marketing function serves as the primary link between the firm and its market. Thus, it stands to reason that the marketing function has greater influence in these firms.H1. Marketing's influence is higher in SMEs selling consumer goods than in firms selling industrial goods.Differentiation strategy Using Porter's (1980) typology, firms can pursue a cost leadership or differentiation strategy. Bennett and Smith (2002) showed that as SMEs grow, they develop greater specialization and differentiation strategies. Although not empirically proven, conventional wisdom has suggested that marketing's influence is more influential for a differentiation strategy than for a cost leadership strategy. For a firm to pursue a differentiation strategy, it must identify customer needs and develop products to fit those needs. Identifying those needs is one of the core competencies of a marketing department and presumably would lead to greater influence of marketing within a firm.H2. Marketing's influence is related positively to a differentiation strategy compared with a cost leadership strategy.Entrepreneurial orientation As Stevenson and Gumpert (1985) noted, firms with an entrepreneurial focus tend to be flat organizations with informal networks of employees and functions. Entrepreneurial organizations typically have an evolving management structure and are more abstract in defining the roles of management (Freeman and Engle, 2007). Firms with a non-entrepreneurial focus are termed administratively oriented and are organizations with clearly defined authority and structure. It is reasonable to infer from Stevenson and Gumpert's work that firms with an administrative focus will have more clearly delineated functions/departments such as marketing. Bureaucracy has a negative effect on entrepreneurial activity in a firm, and this is anathema to an entrepreneurial-oriented firm (Sorensen, 2007). If the marketing function is more clearly defined, we would expect marketing to have greater influence.H3. Marketing's influence is related negatively to the entrepreneurial orientation of the firm.Background of CEO The top managers influence and shape the firm's direction and focus. Within SMEs, the top manager's role is pivotal for growth and expansion (Hutchinson et al., 2006). For smaller firms, this would seem even more relevant where all employees are likely to feel the CEO's influence. The psychologist Duncker (1945) first coined this phenomenon "functional fixity" to refer to the observation that people tend to react to an object in terms of its usually defined function. That is, a hammer is used to drive nails but is not frequently used as a paperweight, although obviously it could be. Business researchers have applied this idea to management and have found that managers are also influenced by their previous training and find it difficult to deviate from their functional training when devising a solution to a problem. For example, engineers seek technical solutions, accounting managers seek financial solutions, and marketing managers look to advertising and promotion to solve problems (Ashton, 1976; Barnes and Webb, 1986; Chang and Birnberg, 1977; Hand, 1990).H4. Marketing's influence is higher in firms where the chief executive officer has a marketing background.Firm size Recent studies have begun to focus on how firm size affects entrepreneurial orientation in organizations (Dholakia and Kshetri, 2004; Gompers et al., 2005; Dobrev and Barnett, 2005; Stuart and Ding, 2006.) As firms grow, they tend to become more bureaucratic. As firms become more bureaucratic, they tend to become less entrepreneurial (Sorensen, 2007). However, as firms grow, they are in a position to hire specialists and add to their "bench strength." As more specialists are hired, especially in the marketing function, it is assumed that marketing gains greater influence. This leads to our final hypothesis:H5. Within SMEs, marketing's influence is positively related to the size of firm.The proceeding hypothesized relationships are summarized in Figure 1. Sample We employed a mail survey of SMEs using two mailing lists available to the authors. The first mailing came from present and past participants in the Small Business Center of a large Midwestern university. This center caters to small businesses in the Midwestern region of the United States. Its mailing list, totaling 150 businesses, was screened to ensure only SMEs were included in the sample. The second mailing list was obtained from the Greenbrier Chamber of Commerce of West Virginia, which serves businesses in the central region of West Virginia. Again, the mailing list, totaling 100 businesses, was screened to ensure only SMEs were included. Combined with surveys from the Family Business Center, the total sample was 250. The surveys were sent to the top executives (e.g. CEOs, presidents, general managers, owners) of the firms. All the addresses from the mailing lists were valid because no mailings were returned as undeliverable. The survey was conducted October to November, 2006. Of the 250 surveys delivered, 100 completed surveys were returned, a response rate of 40 percent. The response rate from each source was the same. No inducements were included for participation in the survey. Instead, the director of the Family Business Center or Chamber of Commerce sent simple cover letters asking members to participate. We did not stimulate the response rate by doing follow-up contacts to nonresponders.Several steps were taken to address nonresponse error and other external validity concerns. First, on receipt of the completed survey instrument, respondents were contacted by telephone to verify that they personally completed the survey and were top managers of their firms. Of the 100 completed responses, we successfully contacted 45 (45 percent) respondents. In these 45 cases, the top managers of the firms acknowledged personally completing the survey and verified their positions as top managers of the firms.Second, comparisons were performed of sample means (t tests) between early and late survey respondents and the two sources for the sample. Across all independent and dependant measures, no statistical difference was found in sample means.Finally, this study's sample was compared against census data. Although this study used a convenience sample and cannot be generalized to all SMEs, a comparison was done to assess representativeness of the sample. As seen in Table I, our sample covered most employee size categories.As seen in Table II, we analyzed our sample in terms of coverage of NAIC categories. The findings suggest that our sample covered all major NAIC categories.Again, these comparisons do not suggest the study results can be generalized to all SMEs but were done to address potential nonresponse concerns.Measures The independent variables for our study included the influence of marketing, presence of a differentiation strategy, entrepreneurial orientation, years in business, number of full-time employees, and primary functional background of the CEO. The measures for this study have been used previously or adapted from previous studies.Influence of marketing was measured via a multi-item measure featuring a 100-point constant sum scale. This scale was adapted from Homburg et al.'s (1999) scale developed to measure the influence of marketing on decision making for ten strategic issues:1. advertising;2. customer satisfaction measurement;3. customer satisfaction improvement;4. expansion into new geographic markets;5. strategic direction of the firm;6. choice of strategic partners;7. new product development;8. design of customer service and support;9. pricing; and10. major capital expenditures.Homburg et al. (1999) chose these issues to represent those decisions classically thought to be handled by marketing (advertising, pricing) and other issues that are not completely under the control of one department. The list was not designed to be an exhaustive measure of all aspects of marketing's influence but a sampling of business decisions typically considered in the marketing domain. Respondents were asked to allocate 100 points in terms of influence of the following groups: marketing, sales, R&D, operations, finance, and general management on each of the ten issues. The option of "general management" was added to the original Homburg et al.'s (1999) scale because of concern that their original scale presumed that a general manager exerts no influence on decision making relating to marketing. We questioned that assumption and included that option into the scale. Respondents were told to allocate no points to a group if it did not exist in their firm. In no case did any respondent assign zero points to marketing, thus eliminating the potential confound of responses from firms without a marketing department. Respondents were also asked to rate the importance of each of the ten issues on a seven-point Likert scale (anchored by relatively low importance to extremely high importance).Marketing influence was calculated by multiplying the importance of each of the ten issues by the influence allocated to marketing for that issue and then adding across all ten issues. The summed figure was divided by the number of issues that the respondent answered.Whether a firm used a differentiation strategy was determined by using a seven-item measure used by Homburg et al. (1999), Kim and Lim (1988), and Dress and Davis (1984). Respondents gave specific measures in answer to the following questions about company activities: To what extent does your business emphasize the following activities?1. competitive advantage through superior products;2. creating superior value through services accompanying the products;3. new product development;4. building up a premium product or brand image;5. obtaining high prices from the market;6. advertising; and7. development of customer specific solutions and products.A seven-point Likert scale anchored by Not at all and A great deal was used to record responses (alpha=0.84).The entrepreneurial orientation of the firm was measured using a seven-item scale developed by Covin and Slevin (1989) and subsequently used by Keh et al. (2007). The specific measures were:1. our culture emphasizes innovation and research and development;2. we have a low rate of new product innovation;3. we are a follower and have a defensive posture in the market;4. we are first to introduce new technologies and products;5. we maintain a competitive posture toward competition;6. we have a strong dislike for high risk, high return projects; and7. when faced with risk, we adopt an aggressive, bold posture.A seven-point Likert scale, anchored by Strongly disagree and Strongly agree was used to record responses. Items (2), (3), and (6) were reverse coded (alpha=0.83).The age of business (years), number of full-time employees, and primary functional background of the CEO were single-item, open-response measures. Regarding the measure of full-time employees, any response of "more than 500 employees" would have been discarded because this study was limited to SMEs. However, we did not receive any surveys with this response. Functional background of CEO was dummy coded (0=non-marketing background; 1=marketing background). The type of market served was dummy coded (0=industrial; 1=consumer).The dependent variables included sales from most recent year, percent of sales spent on marketing, percent of sales spent on sales activities, and percent of sales spent on advertising. The variables were measured via single-item measures. For sales, respondents were given four options: under $1 million, $1-5 million, $5-25 million, and above $25 million. The percent of sales measures (marketing activities, sales activities, and advertising spending) had 13 response options: (0, less than 1 percent, 1 percent, 2 percent, 3 percent, 4 percent, 5 percent, 6 percent, 7 percent, 8 percent, 9 percent, 10 percent, and above 10 percent). The results from the two samples were analyzed (t-tests) and no statistical difference was found in means across all measures. Accordingly, the results were pooled.The role of marketing Conventional wisdom suggests that marketing's role is to manage the relationship between customers and a firm (Moorman and Rust, 1999). Because customers are an important contributor to a firm's financial performance, the department responsible for management of customer relationships should be viewed as an important force; thus, the marketing department that assumes this management function should be deemed important and influential. Homburg et al.'s (1999) study showed the marketing function to be a significant influence in large firms. Turning to this study, as shown in Table III, marketing shares influence with general management and sales on most marketing issues. This suggests that marketing, as function and influence in business, is not as well developed for smaller firms compared with their larger brethren. The sales function has a slightly greater influence over marketing on all issues outside of advertising (customer satisfaction measurement, customer satisfaction improvement, expansion into new geographic markets, strategic direction of the firm, choice of strategic partners, new product development, design of customer service and support, and pricing and major capital expenditures). Decisions regarding advertising are the only issue where marketing enjoys significant and sole influence. We have included comparative data from the Homberg et al. (1999) study, and the data suggest marketing has a way to go before it enjoys the same influence found in larger firms.Hypotheses testing Regression analysis was used to test the hypotheses. We ran three regression analyses: on all issues, "marketing" issues (advertising messages, distribution strategy, design of customer sales and support, customer satisfaction measurement and customer satisfaction improvement) and non-marketing exclusive issues (strategic direction of business unit, choice of strategic partners, new product development, major capital expenditures, and expansion into new geographic markets). Results were the same, and for purposes of discussion, we will focus on the first regression analysis although results from all three analyses are reported in Table IV. Firm size, sales; percent spent of marketing, sales, and advertising activities; source of survey, and early/late responders were included as control variables.H1 proposed that marketing's influence is higher in firms selling consumer goods than in firms selling industrial goods. We found support for this (b = 48.68, p<0.001). Firms in consumer markets tend to do more advertising, considered the exclusive province of marketing, which likely leads to firms having greater "bench strength" in the marketing department. Interestingly, this conclusion differs from Homburg et al.'s (1999) finding for large firms where no difference appeared in marketing's influence between consumer and industrial firms.H3 stated that marketing's influence is related negatively to the entrepreneurial orientation of the firm. We found support for this (b =.-20.14, p<0.001). Less entrepreneurial firms have more influential marketing departments. This suggests a more hierarchical firm has more clearly defined functions leading to greater influence of marketing.We found no support for the remaining hypotheses (H2, H4 and H5). H2 stated that marketing's influence is related positively to a differentiation strategy. Evidently a differentiation strategy does not automatically mean marketing has greater influence. This contrasts with the Homburg et al. (1999) survey where large businesses employing a differentiation strategy had greater marketing influence. H4 stated that marketing's influence is higher where the CEO has a marketing background. Unlike larger firms, the notion that that the CEO's background indicates which firm functions might have influence was not supported. Finally, H5 related to firm size within SMEs. Firm size was not a significant factor in determining marketing's influence. H4 and H5 results were consistent with Homburg et al.'s (1999) study. A comparison of this study's results and the Homburg et al.'s (1999) can be found in Table V. Our data suggest that marketing is not as well developed or influential in SMEs as it is in large firms. Marketing's scope or extent of responsibility is more limited compared with large firms. Two factors, type of market (consumer) and firm orientation (hierarchal), facilitate marketing's influence within a firm. These results indicate that both underlying theories - institutional and contingency - received support and neither theory can be dismissed, suggesting the influence of marketing is a multidimensional construct. Individual traits (e.g. charisma, power, authority) were not considered in this study. However, our findings show marketing's influence can vary systematically as a function of institutional and external factors.Although our analysis sheds light on some internal and external factors that influence marketing's role in a firm, a major portion of variance remains unexplained. Additional research is necessary to explore other potential influences on marketing's role in a firm. These might include personnel factors (background and experience of marketing department personnel) and firm politics. For example, Reijonen and Komppula (2007) studied SMEs in Finland and found the motivation of entrepreneurs are not solely financial and often are based on job satisfaction, control, and family concerns.Although this research fills a gap in the literature relating to the influence of marketing in SMEs, there is much work to be done. We did not address the relationship between marketing's influence and the performance of firms. Analyses done and conclusions reached in this research were based on a very limited sample of SMEs located in one region of the country. Sampling across all regions of the United States would help in generalization of the findings. The results of this study are particularly troubling because marketing resources are one driver of competitive advantage, and although the studied firms have a formal marketing function within the firm, it would appear the function has not proven its value to the firm. For marketing to increase in influence, individuals leading this function must gain a seat at the table. Our findings would suggest that changing the influence of the marketing function might take considerable effort. One barrier to changing marketing's influence is the firm's entrepreneurial orientation, which can be a deep-seated cultural factor that typically requires time to change. An Indian proverb cautions: "Under a bright lamp, there is great darkness." Although marketing departments have a responsibility to market the firm's products and services, the task of marketers marketing themselves to internal stakeholders remains unanswered. Marketing departments must do more to ensure that the voice of marketing is heard when key decisions are being made. Opens in a new window.Figure 1 Conceptual model Opens in a new window.Table I Firms by employee size Opens in a new window.Table II Percentage of firms by industry Opens in a new window.Table III Influence of business functional departments on marketing issues Opens in a new window.Table IV Results of regression analysis Opens in a new window.Table V Comparison of study results with Homberg et al. (1999) study
|
- This research fills a gap in the literature relating to the influence of marketing in SMEs. More importantly, the findings differ from conclusions raised by a similar study conducted for larger firms.
|
[SECTION: Purpose] A single "extreme" observation can make an inequality index estimator uninformative, i.e., meaningless. Frequently, survey data are used to calculate inequality within an economy; we expect that survey data will contain both outliers and so-called contamination points. The main purpose of this research is to provide a new methodology to estimate inequality indices robustly; that is, to provide an estimator of inequality that is not severely affected by atypical data (outliers/contamination points). Our procedure uses information from multiple variables (e.g. socio-economic and geographic variables) to eliminate the effect of atypical observations. To our knowledge, in the inequality literature this is the first methodology that is able to identify outliers and contamination points in more than one direction[1]. Economic theories are often based on inference from data or are, at least, validated by real observations. However, the vast majority of this data is not collected under laboratory conditions. The main source of information to british analyse individual behaviour is a sample survey in which people are asked to recall previous actions[2]. A crucial question, and often ignored in applied work, is whether the data from these surveys are reliable. It is to be expected that asking people to recall past actions will lead to some errors. Both macro and microeconomic models depend crucially on ideas deduced from or corroborated by this survey data and, in turn, government policies are informed by the economic/statistical analysis of this data. If statistical analysis does not take into account that survey data might have as well some "non-sampling error", the amount of resources invested in public policies might not be optimal. Income inequality has become increasingly important within public policy as indices such as Gini, Theil and Coefficient of Variation have made metrics accessible to policy makers and the general public. The availability of data and software has made calculations easier and more readily available in both developed and developing countries. However, ironically one of the main concerns of Dalton (1920) has been almost ignored in the way that inequality indices are used. Dalton was concerned with the practical applicability of his indices and the "risk of errors" when measuring income inequality using "imperfect data". Dalton pointed out correctly that the main weakness of any income inequality index is a bad data set. Nevertheless, the empirical studies of inequality usually overlook this warning. Consequently, it seems essential to propose procedures that can take advantage of the advances in income inequality theory whilst empirically producing reliable estimates, no matter the quality of the data used. This research aims to develop an approach to achieve this. This paper illustrates that one should be very careful when using data "as it is" to estimate income inequality. We will propose a procedure that allows us to discover atypical observations; that is, observations that probably do not belong to the data set in use. By discovering these atypical observations, we will be able to estimate confidently any inequality measure. We catalogue atypical observations into two main categories: contamination points and outliers. Contamination refers to corrupted observations. The contamination can occur during processing or due to inaccurate answers of the respondent. Data contamination is especially relevant for developing countries, since it can be a common phenomena due to financial, technological and logistical constraints. Outliers are atypical observations that, opposed to data contamination, do not come from an error in the survey process but appear to be "very extreme". One possibility is that the survey includes a subgroup of households that come from a different population to that of the majority of the observations. On the other hand, we can have outliers that do not follow the pattern of the rest of the data. Given these challenges, contamination is almost unavoidable in household survey data. Sometimes, applied researches made arbitrary imputations when they believe that some outlieres or contaminated points appear in the sample. This is an ad hoc and questionable "solution". The purpose of this research is to estimate inequality indices robustly and objectively under an unknown distribution using all the information from a multivariate data set. This research proposes a four-step methodology based on robust procedures (minimum covariance determinant (MCD) and least trimmed squares) to remove atypical observations and to estimate inequality indices robustly based on a "clean" subsample. Assumptions about the structure of the data are relaxed. Multidimensional data sets permit the use of economic and social variables that are potentially related to income in order to uncover "masked" atypical observations. Whilst it is impossible to detect those "masked" observations with any univariate procedure, both of our multivariate methodologies can effectively tag them. One of our empirical application is the data set "Household Below Average Income" for the UK. The official Gini estimate for the UK in 1979 is 0.26, which is our estimate using "raw" data. However, with our methodology we observed that by removing only 1.8 per cent of atypical observations, the Gini falls to 0.24. Our other procedure suggests that by dropping 3 per cent of our sample, we find a change of 0.03 points. These findings are worth noting since they represent a change between 10 and 15 per cent in inequality. The changes in the Theil Index and in the Coefficient of Variation are even more drastic, using "raw" data the Theil Inequality Index is 0.1130 and the Coefficient of Variation 0.0709. With our procedures, the Theil Index diminishes up to 0.0912 (a change of 20 per cent) and the Coefficient of Variation up to 0.0589 (a change of 17 per cent). In our second empirical application we use the Mexican data set "Encuesta ESRU de Movilidad Social en Mexico (EMOVI)" for 2006, and 2011. Given that Mexico is a developing country, and likely to experience more data quality problems, it was not surprising that a greater number of outliers and contamination points were tagged in these data sets. Once we remove these very atypical observations, for Mexico 2006, we find a reduction of about 0.07 to 0.08 in the Gini Coefficient. Again the effect in the Theil Index and in the Coefficient of Variation is of greater magnitude. The Theil Index diminishes from 0.3615 up to 0.2481, and the Coefficient of Variation from 0.2091 to 0.1385. In an inequality framework, Cowell and Victoria-Feser (1996) used the concept of the influence function (IF) (Hampel, 1974) to assess the influence of an infinitesimal amount of contamination upon the value of an inequality statistic. They found that most inequality measures are not robust, i.e. a single arbitrary large observation can make an inequality index totally uninformative. This implies that the quality of inequality estimates derived from classical statistical procedures can be quite poor. It is therefore necessary to develop robust procedures in order to estimate confidently how the total income in a given society is distributed. In their seminal paper Cowell and Victoria-Feser (1996) proposed what is so far the only robust procedure that has been used to calculate inequality indices. It is based on the framework develop by Huber (1964). Their method, Optimal Bias-Robust Estimate (OBRE), is the most efficient approach under the right parametric model. However, OBRE has some major limitations: it is based on parametric procedures; and OBRE can only manage a single random variable. Their parametric approach is questionable since it is expected that the sample quality would not be the best. Notably, the multivariate nature of a household survey suggests that we should extend the procedure to a multivariate set up. This is mainly because of two reasons: a multivariate framework allows to control for atypical observations more effectively; and there are many situations in which income is not our only variable of interest. Univariate parametric robust procedures are not appropriate to estimate inequality indices. This kind of procedures are quite restrictive given that survey data has an intrinsic multidimensional nature. Furthermore, it is difficult to approximate a parametric distribution to a data set that presumably contains atypical data. Therefore, taking into account these limitations, it appears that the only robust approach in the literature to inequality measurement is not completely satisfactory. This paper addresses the current weaknesses in the literature: we propose two procedures that do not rely on parametric assumptions, and can work with multivariate data. By Monte Carlo simulations we show that these procedures can identify almost all contamination points. As opposed to the current state of the art methods, our procedures can be easily extended to work with multivariate inequality or multivariate poverty indices. Finally, under very mild assumptions, one of our methodologies produce consistent inequality indices estimators[3]. Our methodology has two main limitations: our procedures are not able to identify a bad leverage outlier[4] from a contamination point; and in the case that the data has no atypical observations, our procedures will tag as atypical a (very small) fraction of observations. The first limitation is not unique to our methodology; regardless of the estimation technique considered, we still could not differentiate these kind of observations[5]. Furthermore, we will argue that neither contamination points nor bad leverage outliers should be included in data sets so this limitation has no empirical consequences. On the other hand, the second limitation is the price we shall pay in order to have a reliable inequality estimator. Before explaining our methodology, let us clarify the concepts of outliers and contamination points. The definitions proposed in the next subsections are considered important for the reader to understand the efficacy and unprecedentedness of the methodology and the main results, which follow. 2.1. Outliers and contamination points For the purpose of this research, we breakdown atypical observations in two categories: outliers and contamination points; and outliers into two further subcategories: good leverage points and bad leverage points[6]: (Equation 1) There is no formal mathematical definition for an outlier, but the literature contains a number of explanatory definitions, as follows: Definition 1. "Outlier": Maguluri and Singh (1997): small fraction of data due to gross errors which are totally inconsistent with the rest of the data; Hampel et al. (1986): data which are far away from the bulk of the data, or more generally, from the pattern set by the majority of the data; Maronna et al. (2006): observations quite atypical by virtue of being far from the bulk of the data; Grubbs (1969): an observation that appears to deviate markedly from other members of the sample in within it occurs; and Barnett and Lewis (1994): a point that stands out in contrast to other observations as an extreme value. There are two common characteristics in these classical definitions of outliers. First, there is a great deal of subjectivity in determining the degree of distance from the bulk of the data at which an observation becomes an outlier; and second, the definition of an outlier will be based on beliefs about the underlying distribution of the population. If an egalitarian distribution is expected, this would lead to the exclusion of observations which would arouse no concern if a more dispersed population was presupposed. Notice that all definitions imply that outlying points stand out, but this might be in any direction. This fact relates to our previous discussion of univariate vs multivariate detection of atypical observations. There is not agreement that the outliers need to lie at the tails of the distribution; in fact, the definition seems to suggest that the outlying observation can happen in any direction. Let us proceed to specify the two types of outliers that we might find: Definition 2. Kinds of outlier (Rousseeuw and Van Driessen, 2006): good leverage point: a real and extreme observation in one or more directions that follows the pattern of the majority of the data; and bad leverage point: a real and extreme observation in one or more directions that does not follow the pattern of the majority of the data. On the other hand, despite the fact that they are intrinsically different, contamination points are usually considered under the outlier umbrella. We would like to make explicit the differences between contamination point and outlier by using the following definition: Definition 3. Contamination point: purely deterministic unreal data, e.g. observations that were coded, recorded or calculated with error. Notice that without any further assumption, it is impossible to identify a contamination point from a bad leverage point. Both should be discordant in one or more directions, and also both can appear at any part of the distribution. This situation motivates us to establish our definition for "masked atypical observation". Consider our household survey data X, where x i =(y i , z 1i , ..., z ki ) where Y is the income, and Z is some socio-economic variables: Definition 4. "Masked" atypical observation: a bad leverage point or contamination point y j =a such that P(a-e < Y < a+e)=c, for a small e and c[?](0, 1), but P(a-e < X < a+e|Z=z j )-0. That is, by just looking at the empirical distribution of income, the observation j appears to be perfectly normal. However, once we conditioned on Z=z j , we "uncover" the aberrant part. These definitions (kinds of outliers, contamination point and "masked" atypical observation) clarify why good leverage points cannot be "masked atypical observations" while bad leverage and contamination points can[7]. Whilst contamination points or bad leverage points would not desirably be included in a data set, decisions to include good leverage points introduce some degree of subjectivity. It is ultimately the researcher's decision, to determine whether a good leverage point is a low-probability observation but still part of the population of study or whether this observation is not part of the intended study population. 2.2. IF and robust statistic A fundamental concept in measuring robustness and efficiency of a statistical functional T (i.e. how a statistic is affected by atypical data) is its IF (Hampel, 1974). The IF measures how an statistic behaves if we add one more observation to a very large sample. If the IF is unbounded, a single outlier may cause trouble: Definition 5. IF: using our framework from the contamination model, the IF of a statistic T at a distribution G is given by: (Equation 2) The IF can be interpreted as a directional derivative[8] of T in the space of distributions. In other words, it gives the effect on T of an infinitesimal perturbation to the data at the point H. See that a bounded IF does not guarantee that the estimator is "good" in any sense, it only assures us that the estimator cannot be terribly bad, i.e. totally uninformative. Therefore, a bounded IF is a minimum condition that we should ask to our estimators. 2.3. Dalton's principle of transfers and its statistical implications Starting in the beginning of the twentieth century, there has been a debate between economists about the description and measurement of economic disparity. Usually, inequality indices should measure the spread of the distribution of incomes with respect to an all-equal distribution. This can be done in several ways, but still the ideas of Lorenz (1905), Gini (1921) and Dalton (1920) are the fundamental concepts in the construction of inequality indices. In fact, the only commonly accepted criterion that these indices should satisfy is the so-called Dalton's (1920) principle of transfers[9]: Definition 6. Dalton's principle of transfers: in a population of n individuals, let y i and y j be the income of the agent i and j, respectively. Suppose y i > y j , and define a=y i - y j . The Dalton principle suggests that inequality is reduced if there is a transfer from a richer person to a poorer person, i.e. if agent i transfers to agent j any amount up to a [10]. 2.3.1. Dalton principle and IF Cowell and Victoria-Feser (1996) showed that classical estimates of inequality indices that satisfy Dalton' s principle of transfers have an unbounded IF. This is a major concern, since they show that these indices can become totally uninformative when small arbitrary contaminations are present. Therefore, with a survey data that contains even only a small subsample of atypical observations, we cannot trust most inequality indices[11]. This raises a major question, can we trust at all in the estimates of such indexes? This research will produce robust inequality estimates by first trimming the "raw" data using robust procedures, and then calculate classical inequality indices using sample moments' analogues in the "clean" subsample (i.e. after dropping atypical data). This is an innovation in the literature. The next subsections explain the methodology applied to do this and the very last subsection in this section briefly discuss some alternative approaches. 3.1. Robust procedures, sorting the data by MCD and LTS The aim of a robust methodology is to produce estimators that are not "greatly" affected by atypical data. The definitions of outlier and contamination point of Section 2.1 suggest us that we should use a robust procedure to rank our observations, and then derive a test with a cut-off value to define atypical observations. Unfortunately, there is not a unique and unambiguous procedure to rank observations from a multivariate data set. In this paper we will order our data using two procedures: sorting residuals from a linear projection (LTS); and sorting observations by their distance to the centre of the data (MCD). Then we will derive a statistical test in order to define whether an observation is atypical. 3.2. LTS sorting Our aim will be to sort our data using a robust methodology (LTS). Our framework is a linear projection model with a response variable y i and a vector of p explanatory variables z i . The equation in matrix form is y=Z b+ e: Definition 7. Least trimmed squares (Rousseeuw, 1984): (Equation 3) where (Equation 4), (Equation 5) are the ordered square residuals, and h[?] is a predetermined threshold such that (Equation 6). The value h can be thought of as the minimum number of points which must not be atypical[12]. Then LTS choose the subset h with the smallest sum of residuals[13]. 3.3. MCD sorting Our aim in this subsection will be to sort our data using a robust methodology (MCD). A different option from LTS sorting is to define a centre of the data and then measure the distance between each observation and the centre. A commonly used measure is the Mahalanobis distance (see Rousseeuw, 1984): (Equation 7) where again x i is a row vector (that includes income y i and some socio-economic variables z i ), and x 0 is usually a variable that represents the sample mean for each variable; finally, G is the covariance matrix. This distance is very similar to the Euclidean distance, but it weights the distances by the covariance matrix. That is, it puts less weight to variables which variance is "big". Notice that R i is a scalar, therefore we can sort our data based on this scalar. Nevertheless, to implement this distance we would have to calculate the sample mean and the sample covariance, (Equation 8). As we have argued the entire paper, neither the sample mean nor the sample covariance are robust to outliers. We just need a single atypical observation to make this method completely uninformative. However, Rousseeuw (1984) proposed a solution to this problem. He suggested to first find a robust estimate of the mean and a robust estimate of the sample covariance. This robustification assures us that the sorting process of our data are not influenced by atypical observations: Definition 8. MCD (Rousseeuw, 1984): The MCD selects the subset of h points that are most tightly clustered. In particular, out of all subsets of size h, it finds the one whose sample covariance determinant is minimal[14]. 3.4. Final step to estimate income inequality indices Our methodology consists of four steps: sort the data robustly, this can be done by one of two procedures: LTS (sorting using robust residuals) or MCD (sorting using robust Mahalanobis distances); define a cut-off value; trim the data in order to remove atypical observations; and use sample analogues to estimate the inequality indices (Gini, Theil and Coefficient of Variation). These steps guarantee a robust estimate of income inequality, which is our main goal. Notice that our methodology is flexible enough that we can modify step 4 to estimate a multivariate inequality index or a poverty index robustly, if desired. Summarising, our four-step methodology can be implemented either by the LTS trimming procedure or by the MCD trimming procedure, and it provides us with a sample that we can confidently use to estimate inequality indices. 3.4.1. Alternative to our general four-step methodology: OBRE As mentioned in the literature review, to the author's knowledge, there is just one robust method (OBRE) that has been applied to the estimation of inequality indices. Cowell and Victoria-Feser (1996) developed a parametric approach, based on Hampel et al. (1986), to estimate inequality indices robustly. Their method, OBRE, is a subclass of the M-estimators. Essentially, they assume the data follows a distribution, then they estimate the parameters of that distribution by a procedure similar to a weighted maximum likelihood estimation. Before applying our methodology to real data, we would like to compare our procedures vs the other robust inequality estimation (OBRE). Therefore, we conduct the same simulation that Cowell and Victoria-Feser (1996) proposed. The next subsection will describe that simulation exercise. We also performed artificial contamination exercises with real data to show that we can effectively tag contamination points, see Appendices 1 and 2 for those results. 4.1. Univariate simulation We will randomly and artificially "contaminate" the data in order to show the extent to which income inequality indices can be distorted through contamination. We will use two types of contamination: Definition 9. "Severe" contamination: an observation for which income is multiplied by 10. Definition 10. "Weekly-Monthly" contamination: an observation for which income is multiplied by 4. Whilst this is only an univariate case[15], and therefore the main advantages of our procedures cannot be exploited, it will still provide useful insights into the effectiveness of the methodology. All codes and data are available in the author's webpage[16]. Simulation g distribution with a=3, l=1: (Equation 9) draws: 200 observations; samples: 100; and contamination: a fix percentage randomly chosen multiplied by 10 ("severe"). Figure 1 contrasts a sample from a g distribution vs a sample from the same g distribution, but with 15 per cent of its observations "severely" contaminated. It shows us that a mildly contaminated sample can appear totally different. The green empirical distribution (the contaminated one) appears to be very heavy-tailed while this is not the case in the uncontaminated red sample. This figure should alert us against the likely failure of robust parametric methods even in the univariate case. Table I displays the results we obtain after estimating the inequality indices using the "raw" (contaminated) data, and it also contains the mean squared error (MSE). It presents with data the great sensitivity of the indices shown theoretically in Section 2.2. Even with only six "severe" contaminated observations (3 per cent) the Coefficient of Variation increases almost seven times, the Theil Index almost doubles and the Gini Index increases by 42 per cent. Table II shows the results from Cowell and Victoria-Feser (1996). This method performs very well for 3 per cent of "severe" contamination, although it starts to exhibit some flaws at 5 per cent. We can compare this table with our results. Tables III and IV show the results of our four-step methodology. Both of our procedures are theoretically equivalent in the univariate case (see Rousseeuw, 1984). However, since we are using algorithms that approximate the exact solution and different methods to define the cut-off, we do expect a small numerical difference. So this exercise is also useful as a test to see how different our procedures are numerically. The results show that our methodology demonstrates a slight improvement (i.e. closer to the actual index prior to contamination) under low levels of contaminations. Remarkably, it performs quite a bit better for a contamination of 5 per cent (where OBRE starts to look questionable). Notice that it also performs admirably well for higher contamination rates. Recall that with OBRE we need to assume that the right distribution is known, this assumption is not needed either for MCD trimming or for LTS trimming. Even for a contamination of 10 or 15 per cent, the values of the indices calculated with our procedures are almost unaffected by the data contamination. Therefore, our methodology looks quite promising for computing robust estimates of inequality measures even with highly contaminated data, at least for the univariate case. The next section will show real multivariate applications for our procedures. 4.2. Multivariate empirical application, UK The first data set ever used for robust inequality estimation is "Households Below Average Income Data Set" (HBAI)[17] (Department for Work and Pensions, 1979), which contrary to the name is representative for the whole UK. This data set was originally used by Cowell and Victoria-Feser (1996). Therefore, it would be useful to make a direct comparison between the results of our method and theirs with this data set. However, they used a (presumably arbitrary) subset of only 746 observations, while we have used the full data set. Unfortunately, their sample was no longer available[18], which means our results are not quite comparable. The original HBAI data set provides information on the incomes and characteristics of 6,711 households based[19] on family expenditure survey data collected in January 1979. Participation in the survey is voluntary and the results are fully anonymised. This data set has played a fundamental role in the analysis of income by both researchers and governmental agencies[20]. Variables available describe the source of income, as well as information about the household such as household's composition (single, couple, children and tenure) and age of the head of the household. Notice that there are not many variables that might help to explain the income levels and which could be used to identify outliers through multivariate procedures. For example, socio-economic variables such as education level, background or ethnicity are not included. This means that this is not an ideal data set to apply our methodology; nevertheless, we expect that some variables like house tenure and social security benefits will help us to "explain" household income, and therefore to identify atypical observations[21]. All codes and data are available in the author's webpage[22]. Tables V and VI show the descriptive statistics for the "raw" data for the variables we used. Figures 2 and 3 show the empirical distribution of income[23]. The green bars represent the distribution from the "raw" data, the red bars show the resulting income distribution after we apply our four-step methodology. As we can see, the LTS trimming removes more observations near to the median than the MCD trimming. This suggests that, even though some income data points are not far away from the media, once we consider socio-economic variables these observations do not follow the pattern of the data; therefore, they are "atypical". In Table VII we compare the results obtained calculating the inequality indices using "raw" data vs using our four-step methodology (both procedures LTS and MCD trimming). Notice that MCD trimming does not remove many observations (about 1.8 per cent), but it does suggest a considerable reduction in the inequality indices: around 7 per cent for the Gini, 13 per cent for the Theil and 11 per cent for the Coefficient of Variation. In contrast, LTS trimming suggests a slightly greater amount of atypical observations and a further reduction in each inequality index. In general, the fact that MCD trimming removed fewer observations than LTS trimming suggests that observations in the data set are closer between each other, but that many observations do not follow a linear pattern. Figure 4 illustrates the first three steps in our four-step methodology, it also contrasts our MCD and LTS trimming procedures. On the x-axis data are sorted according with the MCD procedure, whilst on the y-axis observations are sorted according with the LTS procedure. Recall, the closest the observation is to (0, 0), the less "atypical" it is. If an observation is between the green lines, it is not atypical according with the LTS procedure; and if it is to the left of the vertical red line it is not atypical according to the MCD procedure. Notice that there are observations that are commonly atypical for either procedure (those at the top right corner), and vice versa. Also there are observations that are atypical for one procedure but not for the other; for instance, the group of observations to the right of the red vertical line and between the green lines is atypical for the MCD procedure but not for the LTS. 4.2.1. Public policy implications A reduction in the estimate of inequality has important consequences from a policy maker perspective. First, ceteris paribus, the optimal amount of resources destinated to directly address inequality/poverty[24] will be less than with the estimate of inequality using the "raw" data. Those "extra" resources can be destinated to promote growth. Notice that this is a direct consequence of having a more egalitarian economy than previously thought, this is due to the fact that poor people will actually enjoy a bigger share of any national income increment. This also implies that, in order to reduce poverty, public policies should focus more on economic growth. Some resources that were tagged directly to the poor[25], might be tagged to promote economic growth. Since now we know that the inequality is actually smaller than previously thought. Of course, the amount of those resources will depend on the difference between the inequality estimator calculated from "raw" data and the more precise inequality estimator calculated using our four-step procedure; the bigger the difference, the bigger the amount of resources that might be reallocated. We have another two hypothesis of the implications of our new methodology. First, its advantages will be more evident if we study inequality between subgroups of the population. Second, our methodology will help to confirm the classic argument that inequality is almost constant over time within a country[26]. Our first hypothesis can be thought in Roemerian terms: if we calculate inequality of opportunities, we need to estimate inequality over many subgroups; in each subgroup atypical data will be even more influential (in a IF way) than in the whole population. Therefore, the gains of our methodology will be even more notorious. Finally, in order to analyse our second hypothesis, we will use two cross-sectional data for Mexico in the next subsection. 4.3. Multivariate empirical application, Mexico (2006 and 2011) In order to have a broader perspective on the effect of atypical observations on inequality, specially over time and in a developing country, we use data from a Mexico. We expect to find more outliers in developing country data sets, given the special mechanism of income transmission; and also more contaminated observations, given technology and financial restrictions. We have two cross-sections, then our data allows us to study how atypical observations influence the change in inequality from 2006 to 2011. Another benefit of our data set is that the questionnaire is very rich; it provides socio-economic characteristics for each head of household as well as for their parents. These data sets are ideal for our methodology since we have many socio-economic variables[27] that can help to "explain" household income, therefore this empirical exercise is a unique opportunity. The data sets that we are using are "Encuesta ESRU de Movilidad Social en Mexico (EMOVI)" for 2006, and 2011, which are nationally representative for Mexico. Variables available include household income, household's members and head household characteristics, as well as information about the head of household background such as the place where they use to live and information about their parents[28]. The sample size is 5,902 for 2006 and 6,799 for 2011. All codes and data are available in the author's webpage[29]. Tables VIII and IX show the descriptive statistics for both of our data sets. Notice that we were able to use the same variables in both samples. 4.3.1. Main results Analogously as we did with the UK data set, we implement our four-step methodology for each year. The results are shown in Tables X-XIII. From 2006 to 2011 income inequality was reduced according to all indices. This is also the case in the official Mexican statistics[30]. However, the extent of the reduction is an interesting question. The "raw" data suggests a reduction in the Gini of 0.036, whilst the MCD trimming and the LTS trimming for the same index have a reduction of 0.025 and 0.06, respectively. This fact helps to corrobate the hyphotesis that inequality within a country is stable over time. Applying our methodology, we diminish the potential danger of violent swings in inequality due to data contamination. On the other hand, using the "raw" data we have a 0.06 reduction on the Theil Index, this is the same reduction that we get using the MCD trimming procedure; however, the LTS trimming also suggests a similar reduction. Finally, the Coefficient of Variation behaves similarly to that of the Gini Index. Figure 5 provides us with a unique opportunity to explain the differences between our LTS trimming and MCD trimming procedures and how these differences affect the estimation of the inequality indices. Both the Theil Index and the Coefficient of Variation belong to the same family (generalised entropy). However, the Coefficient of Variation is more sensitive to income differences at the top of the distribution[31]. This is an important feature to explain Figure 5. The 2011 graph in Figure 5 tells us an interesting story. There is an atypical group according to the MCD trimming procedure that is completely normal for the LTS trimming. That is, there is a group of observations that are "far" from the centre of the data, but that do follow the linear pattern of the data (i.e. they can be "explained" with the rest of the data using a linear projection). By reviewing the data we notice that this atypical group actually has high incomes, therefore we expect that the inclusion of this group would have a greater impact on the Coefficient of Variation than on the Theil Index[32]. That is the reason why the reduction from 2006 to 2011 in the MCD trimming procedure is greater in the Coefficient of Variation (0.0326) compared with the LTS trimming (0.0324); whilst the reduction in the Theil Index is greater with the LTS trimming procedure (0.0602) than with the MCD trimming procedure (0.0599). Notice that our comparison is merely descriptive, since we are not calculating the standard errors we cannot formally establish whether those changes are statistically different. 4.3.2. UK and Mexico's results comparison Figure 6 summarises graphically our results for Mexico 2006 and contrast them with our results for UK 1979. Each graph should be read as follows: The closest to (0, 0) the less "atypical"[33] an observation is. If an observation is between the green horizontal lines, then the observation is not "atypical" for one of our methods (LTS trimming). The further away from the lines, the more "atypical". If an observation is on the right of the vertical red line, then it is "atypical" for the other method (MCD trimming). The further right, the more "atypical". Notice that, relative to the UK data set, the Mexican data set contains more observations that are identified as highly atypical using both methodologies. These observations are likely to have a significant impact on inequality indices, yet both of our methodologies suggest that these data do not have any relationship with the rest of the sample. We can of expect this result given our discussion about a developing country. In this paper we consider the estimation of income inequality indices (such as Gini, Theil, Coefficient of Variation) using data with some atypical observations. It is shown that even when a small portion of the data are atypical, classical statistical procedures are not reliable in estimating income inequality indices. We propose two procedures to solve this problem: MCD trimming and LTS trimming. Both methodologies use all the information from a multivariate data set to remove atypical observations in order to estimate inequality indices robustly. We illustrate the problems with classical procedures and the merits of our methodologies by means of Monte Carlo simulations and three empirical applications. Our Monte Carlo simulation showed that even with a small percentage of atypical observations, inequality indices estimated with classical procedures become uninformative. Under the same simulation, our procedures successfully estimated inequality indices even when the data set contained a considerable number of atypical observations. The real magnitude of the problem of atypical observations is illustrated in the empirical applications. We found the presence of atypical observations in the UK, and Mexican data sets. The inequality indices changed drastically when these atypical observations were excluded. As expected, these results were particularly strong for the case of Mexico. We have hypothesised that situation given that Mexico is a developing country, therefore: Mexican surveys face more financial and technological restrictions (prone to contamination); and we expect that the wealth would be more concentrated in Mexico, therefore Mexico is more prone to outliers. The case for our methodology is clear. To strengthen our claims and results, we performed a further Monte Carlo simulation by artificially contaminating real data. We showed that we can successfully identify the artificially contaminated observations. We started this paper by asking a very important question in empirical economics as well as in policy making: can we trust in the estimates of income inequality? We have shown that the answer is: generally not, especially since most of the analysis use household survey data prone to contamination. This situation might be aggravated if the data set is from a developing country. We tried to highlight that household survey data has serious limitations in measuring income inequality, irrespective of the size of the data set. A common posture is to " believe" in the inequality estimate computed from "raw" data without trying to understand if this estimate corresponds to reality. A lesson to learn from this study is that even if you have plenty of data, these data might contain biases that can lead to misleading estimates of income inequality. We show that our methodology is an effective remedy to atypical observations. At least, one should compare the estimates from "raw" data with estimates derived from our procedures. Our methodology does not only remove outliers and contamination points, but it also provides a fundamental tool to answer the challenging, but important question of the true nature of income inequality. 5.1. Policy implications Our methodology provides a useful approach to ensuring income inequality indices are as meaningful as possible for policy makers. Our procedures do not only protect our estimations from atypical observations, but also highlight how the income inequality estimates are affected by groups of observations that are more closely together (MCD trimming) and by groups of observations that follow the same pattern (LTS trimming). This is a crucial tool for policy making, since we can contrast the MCD trimming and the LTS trimming to see if there is any particular subgroup that is atypical for the former but not for the latter. This would help to answer the question of where the inequality lies within the population. We think that this approach might be relevant especially when it comes to changes in inequality over time (as we discussed in the Mexican application). Finally, the simplicity of our procedures makes it realistic that such analysis could be applied and used by policy makers. 5.2. Further research A straight-forward extension of this research is to apply the methodologies proposed here to poverty indices and multivariate inequality indices. That is, once we "clean" our sample we can construct indices that take into account gender, age, ethnicity, education, etc. Further thought is still needed to polish and to extend these procedures, but given the critical importance of methods to measure inequality and other related measures of welfare to policy, our methodology is a relevant and useful addition to this debate. A further extension to our procedures is in the area of public goods provision. The detection and "treatment" of atypical data in the area of the provision and finance of local public goods is essential for analytical purposes within the frame of public economics. A single atypical observation can make an otherwise welfare improving set of local public goods, fiscally unfeasible. Moreover, it could have interesting policy implications in a decentralised political system. Distributive and allocative discrepancies of that type could be manifestations of the Gibbard-Satterthwaite[34] theorem. Our methodology could be combined with policy suggestions so that a local optimum is restored a la Tiebout-Oates[35]. Again, this is the case given that a single atypical observation can distord the household preferences for a tax-service mix. Finally, another related question not covered in this study is whether our procedures can be combined with the theory of optimal currency/fiscal areas. The remedy of atypical observations may be fruitful in enabling the fiscal/monetary authority to provide an optimal policy response. In summary, it is clear that there is a wide range of economic analyses to which our methodology could be usefully applied. Opens in a new window. Figure 1 15 per cent "Severe" contamination empirical Gamma distribution vs "clean" empirical Gamma distribution, 200 observations Opens in a new window. Figure 2 Empiricaldistribution: "raw" vs MCD trimming Opens in a new window. Figure 3 Empiricaldistribution: "raw" vs LTS trimming Opens in a new window. Figure 4 LTS trimming vs MCD trimming Opens in a new window. Figure 5 LTS trimming vs MCD trimming (2006 left, 2011 right) Opens in a new window. Figure 6 Measurement of atypical data, UK 1979 (left) vs Mexico 2006 (right) Opens in a new window. Table I Empirical indices using raw data Opens in a new window. Table II Inequality indices using OBRE Opens in a new window. Table III Inequality indices using MCD Opens in a new window. Table IV Inequality indices using LTS Opens in a new window. Table V Continuous variables Opens in a new window. Table VI Discrete variables Opens in a new window. Table VII Inequality indices using raw data and "clean" data Opens in a new window. Table VIII Inequality indices using MCD trimming with artificially contaminated data Opens in a new window. Table IX Inequality indices using LTS trimming with artificially contaminated data Opens in a new window. Table X Continuous variables Opens in a new window. Table XI Discrete variables Opens in a new window. Table XII Inequality indices using "raw" data and "clean" data Opens in a new window. Table XIII Change in inequality indices from 2006 to 2011 Opens in a new window. Table AI Inequality indices using MCD trimming with artificially contaminated data Opens in a new window. Table AII Inequality indices using LTS trimming with artificially contaminated data Opens in a new window. Equation 1 Opens in a new window. Equation 2 Opens in a new window. Equation 3 Opens in a new window. Equation 4 Opens in a new window. Equation 5 Opens in a new window. Equation 6 Opens in a new window. Equation 7 Opens in a new window. Equation 8 Opens in a new window. Equation 9
|
The data employed to measure income inequality usually come from household surveys, which commonly suffer from atypical observations such as outliers and contamination points. This is of importance since a single atypical observation can make classical inequality indices totally uninformative. To deal with this problem, robust univariate parametric or ad hoc procedures are commonly used; however, neither is fully satisfactory. The purpose of this paper is to propose a methodology to deal with this problem.
|
[SECTION: Method] A single "extreme" observation can make an inequality index estimator uninformative, i.e., meaningless. Frequently, survey data are used to calculate inequality within an economy; we expect that survey data will contain both outliers and so-called contamination points. The main purpose of this research is to provide a new methodology to estimate inequality indices robustly; that is, to provide an estimator of inequality that is not severely affected by atypical data (outliers/contamination points). Our procedure uses information from multiple variables (e.g. socio-economic and geographic variables) to eliminate the effect of atypical observations. To our knowledge, in the inequality literature this is the first methodology that is able to identify outliers and contamination points in more than one direction[1]. Economic theories are often based on inference from data or are, at least, validated by real observations. However, the vast majority of this data is not collected under laboratory conditions. The main source of information to british analyse individual behaviour is a sample survey in which people are asked to recall previous actions[2]. A crucial question, and often ignored in applied work, is whether the data from these surveys are reliable. It is to be expected that asking people to recall past actions will lead to some errors. Both macro and microeconomic models depend crucially on ideas deduced from or corroborated by this survey data and, in turn, government policies are informed by the economic/statistical analysis of this data. If statistical analysis does not take into account that survey data might have as well some "non-sampling error", the amount of resources invested in public policies might not be optimal. Income inequality has become increasingly important within public policy as indices such as Gini, Theil and Coefficient of Variation have made metrics accessible to policy makers and the general public. The availability of data and software has made calculations easier and more readily available in both developed and developing countries. However, ironically one of the main concerns of Dalton (1920) has been almost ignored in the way that inequality indices are used. Dalton was concerned with the practical applicability of his indices and the "risk of errors" when measuring income inequality using "imperfect data". Dalton pointed out correctly that the main weakness of any income inequality index is a bad data set. Nevertheless, the empirical studies of inequality usually overlook this warning. Consequently, it seems essential to propose procedures that can take advantage of the advances in income inequality theory whilst empirically producing reliable estimates, no matter the quality of the data used. This research aims to develop an approach to achieve this. This paper illustrates that one should be very careful when using data "as it is" to estimate income inequality. We will propose a procedure that allows us to discover atypical observations; that is, observations that probably do not belong to the data set in use. By discovering these atypical observations, we will be able to estimate confidently any inequality measure. We catalogue atypical observations into two main categories: contamination points and outliers. Contamination refers to corrupted observations. The contamination can occur during processing or due to inaccurate answers of the respondent. Data contamination is especially relevant for developing countries, since it can be a common phenomena due to financial, technological and logistical constraints. Outliers are atypical observations that, opposed to data contamination, do not come from an error in the survey process but appear to be "very extreme". One possibility is that the survey includes a subgroup of households that come from a different population to that of the majority of the observations. On the other hand, we can have outliers that do not follow the pattern of the rest of the data. Given these challenges, contamination is almost unavoidable in household survey data. Sometimes, applied researches made arbitrary imputations when they believe that some outlieres or contaminated points appear in the sample. This is an ad hoc and questionable "solution". The purpose of this research is to estimate inequality indices robustly and objectively under an unknown distribution using all the information from a multivariate data set. This research proposes a four-step methodology based on robust procedures (minimum covariance determinant (MCD) and least trimmed squares) to remove atypical observations and to estimate inequality indices robustly based on a "clean" subsample. Assumptions about the structure of the data are relaxed. Multidimensional data sets permit the use of economic and social variables that are potentially related to income in order to uncover "masked" atypical observations. Whilst it is impossible to detect those "masked" observations with any univariate procedure, both of our multivariate methodologies can effectively tag them. One of our empirical application is the data set "Household Below Average Income" for the UK. The official Gini estimate for the UK in 1979 is 0.26, which is our estimate using "raw" data. However, with our methodology we observed that by removing only 1.8 per cent of atypical observations, the Gini falls to 0.24. Our other procedure suggests that by dropping 3 per cent of our sample, we find a change of 0.03 points. These findings are worth noting since they represent a change between 10 and 15 per cent in inequality. The changes in the Theil Index and in the Coefficient of Variation are even more drastic, using "raw" data the Theil Inequality Index is 0.1130 and the Coefficient of Variation 0.0709. With our procedures, the Theil Index diminishes up to 0.0912 (a change of 20 per cent) and the Coefficient of Variation up to 0.0589 (a change of 17 per cent). In our second empirical application we use the Mexican data set "Encuesta ESRU de Movilidad Social en Mexico (EMOVI)" for 2006, and 2011. Given that Mexico is a developing country, and likely to experience more data quality problems, it was not surprising that a greater number of outliers and contamination points were tagged in these data sets. Once we remove these very atypical observations, for Mexico 2006, we find a reduction of about 0.07 to 0.08 in the Gini Coefficient. Again the effect in the Theil Index and in the Coefficient of Variation is of greater magnitude. The Theil Index diminishes from 0.3615 up to 0.2481, and the Coefficient of Variation from 0.2091 to 0.1385. In an inequality framework, Cowell and Victoria-Feser (1996) used the concept of the influence function (IF) (Hampel, 1974) to assess the influence of an infinitesimal amount of contamination upon the value of an inequality statistic. They found that most inequality measures are not robust, i.e. a single arbitrary large observation can make an inequality index totally uninformative. This implies that the quality of inequality estimates derived from classical statistical procedures can be quite poor. It is therefore necessary to develop robust procedures in order to estimate confidently how the total income in a given society is distributed. In their seminal paper Cowell and Victoria-Feser (1996) proposed what is so far the only robust procedure that has been used to calculate inequality indices. It is based on the framework develop by Huber (1964). Their method, Optimal Bias-Robust Estimate (OBRE), is the most efficient approach under the right parametric model. However, OBRE has some major limitations: it is based on parametric procedures; and OBRE can only manage a single random variable. Their parametric approach is questionable since it is expected that the sample quality would not be the best. Notably, the multivariate nature of a household survey suggests that we should extend the procedure to a multivariate set up. This is mainly because of two reasons: a multivariate framework allows to control for atypical observations more effectively; and there are many situations in which income is not our only variable of interest. Univariate parametric robust procedures are not appropriate to estimate inequality indices. This kind of procedures are quite restrictive given that survey data has an intrinsic multidimensional nature. Furthermore, it is difficult to approximate a parametric distribution to a data set that presumably contains atypical data. Therefore, taking into account these limitations, it appears that the only robust approach in the literature to inequality measurement is not completely satisfactory. This paper addresses the current weaknesses in the literature: we propose two procedures that do not rely on parametric assumptions, and can work with multivariate data. By Monte Carlo simulations we show that these procedures can identify almost all contamination points. As opposed to the current state of the art methods, our procedures can be easily extended to work with multivariate inequality or multivariate poverty indices. Finally, under very mild assumptions, one of our methodologies produce consistent inequality indices estimators[3]. Our methodology has two main limitations: our procedures are not able to identify a bad leverage outlier[4] from a contamination point; and in the case that the data has no atypical observations, our procedures will tag as atypical a (very small) fraction of observations. The first limitation is not unique to our methodology; regardless of the estimation technique considered, we still could not differentiate these kind of observations[5]. Furthermore, we will argue that neither contamination points nor bad leverage outliers should be included in data sets so this limitation has no empirical consequences. On the other hand, the second limitation is the price we shall pay in order to have a reliable inequality estimator. Before explaining our methodology, let us clarify the concepts of outliers and contamination points. The definitions proposed in the next subsections are considered important for the reader to understand the efficacy and unprecedentedness of the methodology and the main results, which follow. 2.1. Outliers and contamination points For the purpose of this research, we breakdown atypical observations in two categories: outliers and contamination points; and outliers into two further subcategories: good leverage points and bad leverage points[6]: (Equation 1) There is no formal mathematical definition for an outlier, but the literature contains a number of explanatory definitions, as follows: Definition 1. "Outlier": Maguluri and Singh (1997): small fraction of data due to gross errors which are totally inconsistent with the rest of the data; Hampel et al. (1986): data which are far away from the bulk of the data, or more generally, from the pattern set by the majority of the data; Maronna et al. (2006): observations quite atypical by virtue of being far from the bulk of the data; Grubbs (1969): an observation that appears to deviate markedly from other members of the sample in within it occurs; and Barnett and Lewis (1994): a point that stands out in contrast to other observations as an extreme value. There are two common characteristics in these classical definitions of outliers. First, there is a great deal of subjectivity in determining the degree of distance from the bulk of the data at which an observation becomes an outlier; and second, the definition of an outlier will be based on beliefs about the underlying distribution of the population. If an egalitarian distribution is expected, this would lead to the exclusion of observations which would arouse no concern if a more dispersed population was presupposed. Notice that all definitions imply that outlying points stand out, but this might be in any direction. This fact relates to our previous discussion of univariate vs multivariate detection of atypical observations. There is not agreement that the outliers need to lie at the tails of the distribution; in fact, the definition seems to suggest that the outlying observation can happen in any direction. Let us proceed to specify the two types of outliers that we might find: Definition 2. Kinds of outlier (Rousseeuw and Van Driessen, 2006): good leverage point: a real and extreme observation in one or more directions that follows the pattern of the majority of the data; and bad leverage point: a real and extreme observation in one or more directions that does not follow the pattern of the majority of the data. On the other hand, despite the fact that they are intrinsically different, contamination points are usually considered under the outlier umbrella. We would like to make explicit the differences between contamination point and outlier by using the following definition: Definition 3. Contamination point: purely deterministic unreal data, e.g. observations that were coded, recorded or calculated with error. Notice that without any further assumption, it is impossible to identify a contamination point from a bad leverage point. Both should be discordant in one or more directions, and also both can appear at any part of the distribution. This situation motivates us to establish our definition for "masked atypical observation". Consider our household survey data X, where x i =(y i , z 1i , ..., z ki ) where Y is the income, and Z is some socio-economic variables: Definition 4. "Masked" atypical observation: a bad leverage point or contamination point y j =a such that P(a-e < Y < a+e)=c, for a small e and c[?](0, 1), but P(a-e < X < a+e|Z=z j )-0. That is, by just looking at the empirical distribution of income, the observation j appears to be perfectly normal. However, once we conditioned on Z=z j , we "uncover" the aberrant part. These definitions (kinds of outliers, contamination point and "masked" atypical observation) clarify why good leverage points cannot be "masked atypical observations" while bad leverage and contamination points can[7]. Whilst contamination points or bad leverage points would not desirably be included in a data set, decisions to include good leverage points introduce some degree of subjectivity. It is ultimately the researcher's decision, to determine whether a good leverage point is a low-probability observation but still part of the population of study or whether this observation is not part of the intended study population. 2.2. IF and robust statistic A fundamental concept in measuring robustness and efficiency of a statistical functional T (i.e. how a statistic is affected by atypical data) is its IF (Hampel, 1974). The IF measures how an statistic behaves if we add one more observation to a very large sample. If the IF is unbounded, a single outlier may cause trouble: Definition 5. IF: using our framework from the contamination model, the IF of a statistic T at a distribution G is given by: (Equation 2) The IF can be interpreted as a directional derivative[8] of T in the space of distributions. In other words, it gives the effect on T of an infinitesimal perturbation to the data at the point H. See that a bounded IF does not guarantee that the estimator is "good" in any sense, it only assures us that the estimator cannot be terribly bad, i.e. totally uninformative. Therefore, a bounded IF is a minimum condition that we should ask to our estimators. 2.3. Dalton's principle of transfers and its statistical implications Starting in the beginning of the twentieth century, there has been a debate between economists about the description and measurement of economic disparity. Usually, inequality indices should measure the spread of the distribution of incomes with respect to an all-equal distribution. This can be done in several ways, but still the ideas of Lorenz (1905), Gini (1921) and Dalton (1920) are the fundamental concepts in the construction of inequality indices. In fact, the only commonly accepted criterion that these indices should satisfy is the so-called Dalton's (1920) principle of transfers[9]: Definition 6. Dalton's principle of transfers: in a population of n individuals, let y i and y j be the income of the agent i and j, respectively. Suppose y i > y j , and define a=y i - y j . The Dalton principle suggests that inequality is reduced if there is a transfer from a richer person to a poorer person, i.e. if agent i transfers to agent j any amount up to a [10]. 2.3.1. Dalton principle and IF Cowell and Victoria-Feser (1996) showed that classical estimates of inequality indices that satisfy Dalton' s principle of transfers have an unbounded IF. This is a major concern, since they show that these indices can become totally uninformative when small arbitrary contaminations are present. Therefore, with a survey data that contains even only a small subsample of atypical observations, we cannot trust most inequality indices[11]. This raises a major question, can we trust at all in the estimates of such indexes? This research will produce robust inequality estimates by first trimming the "raw" data using robust procedures, and then calculate classical inequality indices using sample moments' analogues in the "clean" subsample (i.e. after dropping atypical data). This is an innovation in the literature. The next subsections explain the methodology applied to do this and the very last subsection in this section briefly discuss some alternative approaches. 3.1. Robust procedures, sorting the data by MCD and LTS The aim of a robust methodology is to produce estimators that are not "greatly" affected by atypical data. The definitions of outlier and contamination point of Section 2.1 suggest us that we should use a robust procedure to rank our observations, and then derive a test with a cut-off value to define atypical observations. Unfortunately, there is not a unique and unambiguous procedure to rank observations from a multivariate data set. In this paper we will order our data using two procedures: sorting residuals from a linear projection (LTS); and sorting observations by their distance to the centre of the data (MCD). Then we will derive a statistical test in order to define whether an observation is atypical. 3.2. LTS sorting Our aim will be to sort our data using a robust methodology (LTS). Our framework is a linear projection model with a response variable y i and a vector of p explanatory variables z i . The equation in matrix form is y=Z b+ e: Definition 7. Least trimmed squares (Rousseeuw, 1984): (Equation 3) where (Equation 4), (Equation 5) are the ordered square residuals, and h[?] is a predetermined threshold such that (Equation 6). The value h can be thought of as the minimum number of points which must not be atypical[12]. Then LTS choose the subset h with the smallest sum of residuals[13]. 3.3. MCD sorting Our aim in this subsection will be to sort our data using a robust methodology (MCD). A different option from LTS sorting is to define a centre of the data and then measure the distance between each observation and the centre. A commonly used measure is the Mahalanobis distance (see Rousseeuw, 1984): (Equation 7) where again x i is a row vector (that includes income y i and some socio-economic variables z i ), and x 0 is usually a variable that represents the sample mean for each variable; finally, G is the covariance matrix. This distance is very similar to the Euclidean distance, but it weights the distances by the covariance matrix. That is, it puts less weight to variables which variance is "big". Notice that R i is a scalar, therefore we can sort our data based on this scalar. Nevertheless, to implement this distance we would have to calculate the sample mean and the sample covariance, (Equation 8). As we have argued the entire paper, neither the sample mean nor the sample covariance are robust to outliers. We just need a single atypical observation to make this method completely uninformative. However, Rousseeuw (1984) proposed a solution to this problem. He suggested to first find a robust estimate of the mean and a robust estimate of the sample covariance. This robustification assures us that the sorting process of our data are not influenced by atypical observations: Definition 8. MCD (Rousseeuw, 1984): The MCD selects the subset of h points that are most tightly clustered. In particular, out of all subsets of size h, it finds the one whose sample covariance determinant is minimal[14]. 3.4. Final step to estimate income inequality indices Our methodology consists of four steps: sort the data robustly, this can be done by one of two procedures: LTS (sorting using robust residuals) or MCD (sorting using robust Mahalanobis distances); define a cut-off value; trim the data in order to remove atypical observations; and use sample analogues to estimate the inequality indices (Gini, Theil and Coefficient of Variation). These steps guarantee a robust estimate of income inequality, which is our main goal. Notice that our methodology is flexible enough that we can modify step 4 to estimate a multivariate inequality index or a poverty index robustly, if desired. Summarising, our four-step methodology can be implemented either by the LTS trimming procedure or by the MCD trimming procedure, and it provides us with a sample that we can confidently use to estimate inequality indices. 3.4.1. Alternative to our general four-step methodology: OBRE As mentioned in the literature review, to the author's knowledge, there is just one robust method (OBRE) that has been applied to the estimation of inequality indices. Cowell and Victoria-Feser (1996) developed a parametric approach, based on Hampel et al. (1986), to estimate inequality indices robustly. Their method, OBRE, is a subclass of the M-estimators. Essentially, they assume the data follows a distribution, then they estimate the parameters of that distribution by a procedure similar to a weighted maximum likelihood estimation. Before applying our methodology to real data, we would like to compare our procedures vs the other robust inequality estimation (OBRE). Therefore, we conduct the same simulation that Cowell and Victoria-Feser (1996) proposed. The next subsection will describe that simulation exercise. We also performed artificial contamination exercises with real data to show that we can effectively tag contamination points, see Appendices 1 and 2 for those results. 4.1. Univariate simulation We will randomly and artificially "contaminate" the data in order to show the extent to which income inequality indices can be distorted through contamination. We will use two types of contamination: Definition 9. "Severe" contamination: an observation for which income is multiplied by 10. Definition 10. "Weekly-Monthly" contamination: an observation for which income is multiplied by 4. Whilst this is only an univariate case[15], and therefore the main advantages of our procedures cannot be exploited, it will still provide useful insights into the effectiveness of the methodology. All codes and data are available in the author's webpage[16]. Simulation g distribution with a=3, l=1: (Equation 9) draws: 200 observations; samples: 100; and contamination: a fix percentage randomly chosen multiplied by 10 ("severe"). Figure 1 contrasts a sample from a g distribution vs a sample from the same g distribution, but with 15 per cent of its observations "severely" contaminated. It shows us that a mildly contaminated sample can appear totally different. The green empirical distribution (the contaminated one) appears to be very heavy-tailed while this is not the case in the uncontaminated red sample. This figure should alert us against the likely failure of robust parametric methods even in the univariate case. Table I displays the results we obtain after estimating the inequality indices using the "raw" (contaminated) data, and it also contains the mean squared error (MSE). It presents with data the great sensitivity of the indices shown theoretically in Section 2.2. Even with only six "severe" contaminated observations (3 per cent) the Coefficient of Variation increases almost seven times, the Theil Index almost doubles and the Gini Index increases by 42 per cent. Table II shows the results from Cowell and Victoria-Feser (1996). This method performs very well for 3 per cent of "severe" contamination, although it starts to exhibit some flaws at 5 per cent. We can compare this table with our results. Tables III and IV show the results of our four-step methodology. Both of our procedures are theoretically equivalent in the univariate case (see Rousseeuw, 1984). However, since we are using algorithms that approximate the exact solution and different methods to define the cut-off, we do expect a small numerical difference. So this exercise is also useful as a test to see how different our procedures are numerically. The results show that our methodology demonstrates a slight improvement (i.e. closer to the actual index prior to contamination) under low levels of contaminations. Remarkably, it performs quite a bit better for a contamination of 5 per cent (where OBRE starts to look questionable). Notice that it also performs admirably well for higher contamination rates. Recall that with OBRE we need to assume that the right distribution is known, this assumption is not needed either for MCD trimming or for LTS trimming. Even for a contamination of 10 or 15 per cent, the values of the indices calculated with our procedures are almost unaffected by the data contamination. Therefore, our methodology looks quite promising for computing robust estimates of inequality measures even with highly contaminated data, at least for the univariate case. The next section will show real multivariate applications for our procedures. 4.2. Multivariate empirical application, UK The first data set ever used for robust inequality estimation is "Households Below Average Income Data Set" (HBAI)[17] (Department for Work and Pensions, 1979), which contrary to the name is representative for the whole UK. This data set was originally used by Cowell and Victoria-Feser (1996). Therefore, it would be useful to make a direct comparison between the results of our method and theirs with this data set. However, they used a (presumably arbitrary) subset of only 746 observations, while we have used the full data set. Unfortunately, their sample was no longer available[18], which means our results are not quite comparable. The original HBAI data set provides information on the incomes and characteristics of 6,711 households based[19] on family expenditure survey data collected in January 1979. Participation in the survey is voluntary and the results are fully anonymised. This data set has played a fundamental role in the analysis of income by both researchers and governmental agencies[20]. Variables available describe the source of income, as well as information about the household such as household's composition (single, couple, children and tenure) and age of the head of the household. Notice that there are not many variables that might help to explain the income levels and which could be used to identify outliers through multivariate procedures. For example, socio-economic variables such as education level, background or ethnicity are not included. This means that this is not an ideal data set to apply our methodology; nevertheless, we expect that some variables like house tenure and social security benefits will help us to "explain" household income, and therefore to identify atypical observations[21]. All codes and data are available in the author's webpage[22]. Tables V and VI show the descriptive statistics for the "raw" data for the variables we used. Figures 2 and 3 show the empirical distribution of income[23]. The green bars represent the distribution from the "raw" data, the red bars show the resulting income distribution after we apply our four-step methodology. As we can see, the LTS trimming removes more observations near to the median than the MCD trimming. This suggests that, even though some income data points are not far away from the media, once we consider socio-economic variables these observations do not follow the pattern of the data; therefore, they are "atypical". In Table VII we compare the results obtained calculating the inequality indices using "raw" data vs using our four-step methodology (both procedures LTS and MCD trimming). Notice that MCD trimming does not remove many observations (about 1.8 per cent), but it does suggest a considerable reduction in the inequality indices: around 7 per cent for the Gini, 13 per cent for the Theil and 11 per cent for the Coefficient of Variation. In contrast, LTS trimming suggests a slightly greater amount of atypical observations and a further reduction in each inequality index. In general, the fact that MCD trimming removed fewer observations than LTS trimming suggests that observations in the data set are closer between each other, but that many observations do not follow a linear pattern. Figure 4 illustrates the first three steps in our four-step methodology, it also contrasts our MCD and LTS trimming procedures. On the x-axis data are sorted according with the MCD procedure, whilst on the y-axis observations are sorted according with the LTS procedure. Recall, the closest the observation is to (0, 0), the less "atypical" it is. If an observation is between the green lines, it is not atypical according with the LTS procedure; and if it is to the left of the vertical red line it is not atypical according to the MCD procedure. Notice that there are observations that are commonly atypical for either procedure (those at the top right corner), and vice versa. Also there are observations that are atypical for one procedure but not for the other; for instance, the group of observations to the right of the red vertical line and between the green lines is atypical for the MCD procedure but not for the LTS. 4.2.1. Public policy implications A reduction in the estimate of inequality has important consequences from a policy maker perspective. First, ceteris paribus, the optimal amount of resources destinated to directly address inequality/poverty[24] will be less than with the estimate of inequality using the "raw" data. Those "extra" resources can be destinated to promote growth. Notice that this is a direct consequence of having a more egalitarian economy than previously thought, this is due to the fact that poor people will actually enjoy a bigger share of any national income increment. This also implies that, in order to reduce poverty, public policies should focus more on economic growth. Some resources that were tagged directly to the poor[25], might be tagged to promote economic growth. Since now we know that the inequality is actually smaller than previously thought. Of course, the amount of those resources will depend on the difference between the inequality estimator calculated from "raw" data and the more precise inequality estimator calculated using our four-step procedure; the bigger the difference, the bigger the amount of resources that might be reallocated. We have another two hypothesis of the implications of our new methodology. First, its advantages will be more evident if we study inequality between subgroups of the population. Second, our methodology will help to confirm the classic argument that inequality is almost constant over time within a country[26]. Our first hypothesis can be thought in Roemerian terms: if we calculate inequality of opportunities, we need to estimate inequality over many subgroups; in each subgroup atypical data will be even more influential (in a IF way) than in the whole population. Therefore, the gains of our methodology will be even more notorious. Finally, in order to analyse our second hypothesis, we will use two cross-sectional data for Mexico in the next subsection. 4.3. Multivariate empirical application, Mexico (2006 and 2011) In order to have a broader perspective on the effect of atypical observations on inequality, specially over time and in a developing country, we use data from a Mexico. We expect to find more outliers in developing country data sets, given the special mechanism of income transmission; and also more contaminated observations, given technology and financial restrictions. We have two cross-sections, then our data allows us to study how atypical observations influence the change in inequality from 2006 to 2011. Another benefit of our data set is that the questionnaire is very rich; it provides socio-economic characteristics for each head of household as well as for their parents. These data sets are ideal for our methodology since we have many socio-economic variables[27] that can help to "explain" household income, therefore this empirical exercise is a unique opportunity. The data sets that we are using are "Encuesta ESRU de Movilidad Social en Mexico (EMOVI)" for 2006, and 2011, which are nationally representative for Mexico. Variables available include household income, household's members and head household characteristics, as well as information about the head of household background such as the place where they use to live and information about their parents[28]. The sample size is 5,902 for 2006 and 6,799 for 2011. All codes and data are available in the author's webpage[29]. Tables VIII and IX show the descriptive statistics for both of our data sets. Notice that we were able to use the same variables in both samples. 4.3.1. Main results Analogously as we did with the UK data set, we implement our four-step methodology for each year. The results are shown in Tables X-XIII. From 2006 to 2011 income inequality was reduced according to all indices. This is also the case in the official Mexican statistics[30]. However, the extent of the reduction is an interesting question. The "raw" data suggests a reduction in the Gini of 0.036, whilst the MCD trimming and the LTS trimming for the same index have a reduction of 0.025 and 0.06, respectively. This fact helps to corrobate the hyphotesis that inequality within a country is stable over time. Applying our methodology, we diminish the potential danger of violent swings in inequality due to data contamination. On the other hand, using the "raw" data we have a 0.06 reduction on the Theil Index, this is the same reduction that we get using the MCD trimming procedure; however, the LTS trimming also suggests a similar reduction. Finally, the Coefficient of Variation behaves similarly to that of the Gini Index. Figure 5 provides us with a unique opportunity to explain the differences between our LTS trimming and MCD trimming procedures and how these differences affect the estimation of the inequality indices. Both the Theil Index and the Coefficient of Variation belong to the same family (generalised entropy). However, the Coefficient of Variation is more sensitive to income differences at the top of the distribution[31]. This is an important feature to explain Figure 5. The 2011 graph in Figure 5 tells us an interesting story. There is an atypical group according to the MCD trimming procedure that is completely normal for the LTS trimming. That is, there is a group of observations that are "far" from the centre of the data, but that do follow the linear pattern of the data (i.e. they can be "explained" with the rest of the data using a linear projection). By reviewing the data we notice that this atypical group actually has high incomes, therefore we expect that the inclusion of this group would have a greater impact on the Coefficient of Variation than on the Theil Index[32]. That is the reason why the reduction from 2006 to 2011 in the MCD trimming procedure is greater in the Coefficient of Variation (0.0326) compared with the LTS trimming (0.0324); whilst the reduction in the Theil Index is greater with the LTS trimming procedure (0.0602) than with the MCD trimming procedure (0.0599). Notice that our comparison is merely descriptive, since we are not calculating the standard errors we cannot formally establish whether those changes are statistically different. 4.3.2. UK and Mexico's results comparison Figure 6 summarises graphically our results for Mexico 2006 and contrast them with our results for UK 1979. Each graph should be read as follows: The closest to (0, 0) the less "atypical"[33] an observation is. If an observation is between the green horizontal lines, then the observation is not "atypical" for one of our methods (LTS trimming). The further away from the lines, the more "atypical". If an observation is on the right of the vertical red line, then it is "atypical" for the other method (MCD trimming). The further right, the more "atypical". Notice that, relative to the UK data set, the Mexican data set contains more observations that are identified as highly atypical using both methodologies. These observations are likely to have a significant impact on inequality indices, yet both of our methodologies suggest that these data do not have any relationship with the rest of the sample. We can of expect this result given our discussion about a developing country. In this paper we consider the estimation of income inequality indices (such as Gini, Theil, Coefficient of Variation) using data with some atypical observations. It is shown that even when a small portion of the data are atypical, classical statistical procedures are not reliable in estimating income inequality indices. We propose two procedures to solve this problem: MCD trimming and LTS trimming. Both methodologies use all the information from a multivariate data set to remove atypical observations in order to estimate inequality indices robustly. We illustrate the problems with classical procedures and the merits of our methodologies by means of Monte Carlo simulations and three empirical applications. Our Monte Carlo simulation showed that even with a small percentage of atypical observations, inequality indices estimated with classical procedures become uninformative. Under the same simulation, our procedures successfully estimated inequality indices even when the data set contained a considerable number of atypical observations. The real magnitude of the problem of atypical observations is illustrated in the empirical applications. We found the presence of atypical observations in the UK, and Mexican data sets. The inequality indices changed drastically when these atypical observations were excluded. As expected, these results were particularly strong for the case of Mexico. We have hypothesised that situation given that Mexico is a developing country, therefore: Mexican surveys face more financial and technological restrictions (prone to contamination); and we expect that the wealth would be more concentrated in Mexico, therefore Mexico is more prone to outliers. The case for our methodology is clear. To strengthen our claims and results, we performed a further Monte Carlo simulation by artificially contaminating real data. We showed that we can successfully identify the artificially contaminated observations. We started this paper by asking a very important question in empirical economics as well as in policy making: can we trust in the estimates of income inequality? We have shown that the answer is: generally not, especially since most of the analysis use household survey data prone to contamination. This situation might be aggravated if the data set is from a developing country. We tried to highlight that household survey data has serious limitations in measuring income inequality, irrespective of the size of the data set. A common posture is to " believe" in the inequality estimate computed from "raw" data without trying to understand if this estimate corresponds to reality. A lesson to learn from this study is that even if you have plenty of data, these data might contain biases that can lead to misleading estimates of income inequality. We show that our methodology is an effective remedy to atypical observations. At least, one should compare the estimates from "raw" data with estimates derived from our procedures. Our methodology does not only remove outliers and contamination points, but it also provides a fundamental tool to answer the challenging, but important question of the true nature of income inequality. 5.1. Policy implications Our methodology provides a useful approach to ensuring income inequality indices are as meaningful as possible for policy makers. Our procedures do not only protect our estimations from atypical observations, but also highlight how the income inequality estimates are affected by groups of observations that are more closely together (MCD trimming) and by groups of observations that follow the same pattern (LTS trimming). This is a crucial tool for policy making, since we can contrast the MCD trimming and the LTS trimming to see if there is any particular subgroup that is atypical for the former but not for the latter. This would help to answer the question of where the inequality lies within the population. We think that this approach might be relevant especially when it comes to changes in inequality over time (as we discussed in the Mexican application). Finally, the simplicity of our procedures makes it realistic that such analysis could be applied and used by policy makers. 5.2. Further research A straight-forward extension of this research is to apply the methodologies proposed here to poverty indices and multivariate inequality indices. That is, once we "clean" our sample we can construct indices that take into account gender, age, ethnicity, education, etc. Further thought is still needed to polish and to extend these procedures, but given the critical importance of methods to measure inequality and other related measures of welfare to policy, our methodology is a relevant and useful addition to this debate. A further extension to our procedures is in the area of public goods provision. The detection and "treatment" of atypical data in the area of the provision and finance of local public goods is essential for analytical purposes within the frame of public economics. A single atypical observation can make an otherwise welfare improving set of local public goods, fiscally unfeasible. Moreover, it could have interesting policy implications in a decentralised political system. Distributive and allocative discrepancies of that type could be manifestations of the Gibbard-Satterthwaite[34] theorem. Our methodology could be combined with policy suggestions so that a local optimum is restored a la Tiebout-Oates[35]. Again, this is the case given that a single atypical observation can distord the household preferences for a tax-service mix. Finally, another related question not covered in this study is whether our procedures can be combined with the theory of optimal currency/fiscal areas. The remedy of atypical observations may be fruitful in enabling the fiscal/monetary authority to provide an optimal policy response. In summary, it is clear that there is a wide range of economic analyses to which our methodology could be usefully applied. Opens in a new window. Figure 1 15 per cent "Severe" contamination empirical Gamma distribution vs "clean" empirical Gamma distribution, 200 observations Opens in a new window. Figure 2 Empiricaldistribution: "raw" vs MCD trimming Opens in a new window. Figure 3 Empiricaldistribution: "raw" vs LTS trimming Opens in a new window. Figure 4 LTS trimming vs MCD trimming Opens in a new window. Figure 5 LTS trimming vs MCD trimming (2006 left, 2011 right) Opens in a new window. Figure 6 Measurement of atypical data, UK 1979 (left) vs Mexico 2006 (right) Opens in a new window. Table I Empirical indices using raw data Opens in a new window. Table II Inequality indices using OBRE Opens in a new window. Table III Inequality indices using MCD Opens in a new window. Table IV Inequality indices using LTS Opens in a new window. Table V Continuous variables Opens in a new window. Table VI Discrete variables Opens in a new window. Table VII Inequality indices using raw data and "clean" data Opens in a new window. Table VIII Inequality indices using MCD trimming with artificially contaminated data Opens in a new window. Table IX Inequality indices using LTS trimming with artificially contaminated data Opens in a new window. Table X Continuous variables Opens in a new window. Table XI Discrete variables Opens in a new window. Table XII Inequality indices using "raw" data and "clean" data Opens in a new window. Table XIII Change in inequality indices from 2006 to 2011 Opens in a new window. Table AI Inequality indices using MCD trimming with artificially contaminated data Opens in a new window. Table AII Inequality indices using LTS trimming with artificially contaminated data Opens in a new window. Equation 1 Opens in a new window. Equation 2 Opens in a new window. Equation 3 Opens in a new window. Equation 4 Opens in a new window. Equation 5 Opens in a new window. Equation 6 Opens in a new window. Equation 7 Opens in a new window. Equation 8 Opens in a new window. Equation 9
|
The author propose two robust procedures to estimate inequality indices that can use all the information from a data set, and neither of them rely on a parametric distributional assumption. The methodology performs well irrespectively of the size and quality of the data set.
|
[SECTION: Findings] A single "extreme" observation can make an inequality index estimator uninformative, i.e., meaningless. Frequently, survey data are used to calculate inequality within an economy; we expect that survey data will contain both outliers and so-called contamination points. The main purpose of this research is to provide a new methodology to estimate inequality indices robustly; that is, to provide an estimator of inequality that is not severely affected by atypical data (outliers/contamination points). Our procedure uses information from multiple variables (e.g. socio-economic and geographic variables) to eliminate the effect of atypical observations. To our knowledge, in the inequality literature this is the first methodology that is able to identify outliers and contamination points in more than one direction[1]. Economic theories are often based on inference from data or are, at least, validated by real observations. However, the vast majority of this data is not collected under laboratory conditions. The main source of information to british analyse individual behaviour is a sample survey in which people are asked to recall previous actions[2]. A crucial question, and often ignored in applied work, is whether the data from these surveys are reliable. It is to be expected that asking people to recall past actions will lead to some errors. Both macro and microeconomic models depend crucially on ideas deduced from or corroborated by this survey data and, in turn, government policies are informed by the economic/statistical analysis of this data. If statistical analysis does not take into account that survey data might have as well some "non-sampling error", the amount of resources invested in public policies might not be optimal. Income inequality has become increasingly important within public policy as indices such as Gini, Theil and Coefficient of Variation have made metrics accessible to policy makers and the general public. The availability of data and software has made calculations easier and more readily available in both developed and developing countries. However, ironically one of the main concerns of Dalton (1920) has been almost ignored in the way that inequality indices are used. Dalton was concerned with the practical applicability of his indices and the "risk of errors" when measuring income inequality using "imperfect data". Dalton pointed out correctly that the main weakness of any income inequality index is a bad data set. Nevertheless, the empirical studies of inequality usually overlook this warning. Consequently, it seems essential to propose procedures that can take advantage of the advances in income inequality theory whilst empirically producing reliable estimates, no matter the quality of the data used. This research aims to develop an approach to achieve this. This paper illustrates that one should be very careful when using data "as it is" to estimate income inequality. We will propose a procedure that allows us to discover atypical observations; that is, observations that probably do not belong to the data set in use. By discovering these atypical observations, we will be able to estimate confidently any inequality measure. We catalogue atypical observations into two main categories: contamination points and outliers. Contamination refers to corrupted observations. The contamination can occur during processing or due to inaccurate answers of the respondent. Data contamination is especially relevant for developing countries, since it can be a common phenomena due to financial, technological and logistical constraints. Outliers are atypical observations that, opposed to data contamination, do not come from an error in the survey process but appear to be "very extreme". One possibility is that the survey includes a subgroup of households that come from a different population to that of the majority of the observations. On the other hand, we can have outliers that do not follow the pattern of the rest of the data. Given these challenges, contamination is almost unavoidable in household survey data. Sometimes, applied researches made arbitrary imputations when they believe that some outlieres or contaminated points appear in the sample. This is an ad hoc and questionable "solution". The purpose of this research is to estimate inequality indices robustly and objectively under an unknown distribution using all the information from a multivariate data set. This research proposes a four-step methodology based on robust procedures (minimum covariance determinant (MCD) and least trimmed squares) to remove atypical observations and to estimate inequality indices robustly based on a "clean" subsample. Assumptions about the structure of the data are relaxed. Multidimensional data sets permit the use of economic and social variables that are potentially related to income in order to uncover "masked" atypical observations. Whilst it is impossible to detect those "masked" observations with any univariate procedure, both of our multivariate methodologies can effectively tag them. One of our empirical application is the data set "Household Below Average Income" for the UK. The official Gini estimate for the UK in 1979 is 0.26, which is our estimate using "raw" data. However, with our methodology we observed that by removing only 1.8 per cent of atypical observations, the Gini falls to 0.24. Our other procedure suggests that by dropping 3 per cent of our sample, we find a change of 0.03 points. These findings are worth noting since they represent a change between 10 and 15 per cent in inequality. The changes in the Theil Index and in the Coefficient of Variation are even more drastic, using "raw" data the Theil Inequality Index is 0.1130 and the Coefficient of Variation 0.0709. With our procedures, the Theil Index diminishes up to 0.0912 (a change of 20 per cent) and the Coefficient of Variation up to 0.0589 (a change of 17 per cent). In our second empirical application we use the Mexican data set "Encuesta ESRU de Movilidad Social en Mexico (EMOVI)" for 2006, and 2011. Given that Mexico is a developing country, and likely to experience more data quality problems, it was not surprising that a greater number of outliers and contamination points were tagged in these data sets. Once we remove these very atypical observations, for Mexico 2006, we find a reduction of about 0.07 to 0.08 in the Gini Coefficient. Again the effect in the Theil Index and in the Coefficient of Variation is of greater magnitude. The Theil Index diminishes from 0.3615 up to 0.2481, and the Coefficient of Variation from 0.2091 to 0.1385. In an inequality framework, Cowell and Victoria-Feser (1996) used the concept of the influence function (IF) (Hampel, 1974) to assess the influence of an infinitesimal amount of contamination upon the value of an inequality statistic. They found that most inequality measures are not robust, i.e. a single arbitrary large observation can make an inequality index totally uninformative. This implies that the quality of inequality estimates derived from classical statistical procedures can be quite poor. It is therefore necessary to develop robust procedures in order to estimate confidently how the total income in a given society is distributed. In their seminal paper Cowell and Victoria-Feser (1996) proposed what is so far the only robust procedure that has been used to calculate inequality indices. It is based on the framework develop by Huber (1964). Their method, Optimal Bias-Robust Estimate (OBRE), is the most efficient approach under the right parametric model. However, OBRE has some major limitations: it is based on parametric procedures; and OBRE can only manage a single random variable. Their parametric approach is questionable since it is expected that the sample quality would not be the best. Notably, the multivariate nature of a household survey suggests that we should extend the procedure to a multivariate set up. This is mainly because of two reasons: a multivariate framework allows to control for atypical observations more effectively; and there are many situations in which income is not our only variable of interest. Univariate parametric robust procedures are not appropriate to estimate inequality indices. This kind of procedures are quite restrictive given that survey data has an intrinsic multidimensional nature. Furthermore, it is difficult to approximate a parametric distribution to a data set that presumably contains atypical data. Therefore, taking into account these limitations, it appears that the only robust approach in the literature to inequality measurement is not completely satisfactory. This paper addresses the current weaknesses in the literature: we propose two procedures that do not rely on parametric assumptions, and can work with multivariate data. By Monte Carlo simulations we show that these procedures can identify almost all contamination points. As opposed to the current state of the art methods, our procedures can be easily extended to work with multivariate inequality or multivariate poverty indices. Finally, under very mild assumptions, one of our methodologies produce consistent inequality indices estimators[3]. Our methodology has two main limitations: our procedures are not able to identify a bad leverage outlier[4] from a contamination point; and in the case that the data has no atypical observations, our procedures will tag as atypical a (very small) fraction of observations. The first limitation is not unique to our methodology; regardless of the estimation technique considered, we still could not differentiate these kind of observations[5]. Furthermore, we will argue that neither contamination points nor bad leverage outliers should be included in data sets so this limitation has no empirical consequences. On the other hand, the second limitation is the price we shall pay in order to have a reliable inequality estimator. Before explaining our methodology, let us clarify the concepts of outliers and contamination points. The definitions proposed in the next subsections are considered important for the reader to understand the efficacy and unprecedentedness of the methodology and the main results, which follow. 2.1. Outliers and contamination points For the purpose of this research, we breakdown atypical observations in two categories: outliers and contamination points; and outliers into two further subcategories: good leverage points and bad leverage points[6]: (Equation 1) There is no formal mathematical definition for an outlier, but the literature contains a number of explanatory definitions, as follows: Definition 1. "Outlier": Maguluri and Singh (1997): small fraction of data due to gross errors which are totally inconsistent with the rest of the data; Hampel et al. (1986): data which are far away from the bulk of the data, or more generally, from the pattern set by the majority of the data; Maronna et al. (2006): observations quite atypical by virtue of being far from the bulk of the data; Grubbs (1969): an observation that appears to deviate markedly from other members of the sample in within it occurs; and Barnett and Lewis (1994): a point that stands out in contrast to other observations as an extreme value. There are two common characteristics in these classical definitions of outliers. First, there is a great deal of subjectivity in determining the degree of distance from the bulk of the data at which an observation becomes an outlier; and second, the definition of an outlier will be based on beliefs about the underlying distribution of the population. If an egalitarian distribution is expected, this would lead to the exclusion of observations which would arouse no concern if a more dispersed population was presupposed. Notice that all definitions imply that outlying points stand out, but this might be in any direction. This fact relates to our previous discussion of univariate vs multivariate detection of atypical observations. There is not agreement that the outliers need to lie at the tails of the distribution; in fact, the definition seems to suggest that the outlying observation can happen in any direction. Let us proceed to specify the two types of outliers that we might find: Definition 2. Kinds of outlier (Rousseeuw and Van Driessen, 2006): good leverage point: a real and extreme observation in one or more directions that follows the pattern of the majority of the data; and bad leverage point: a real and extreme observation in one or more directions that does not follow the pattern of the majority of the data. On the other hand, despite the fact that they are intrinsically different, contamination points are usually considered under the outlier umbrella. We would like to make explicit the differences between contamination point and outlier by using the following definition: Definition 3. Contamination point: purely deterministic unreal data, e.g. observations that were coded, recorded or calculated with error. Notice that without any further assumption, it is impossible to identify a contamination point from a bad leverage point. Both should be discordant in one or more directions, and also both can appear at any part of the distribution. This situation motivates us to establish our definition for "masked atypical observation". Consider our household survey data X, where x i =(y i , z 1i , ..., z ki ) where Y is the income, and Z is some socio-economic variables: Definition 4. "Masked" atypical observation: a bad leverage point or contamination point y j =a such that P(a-e < Y < a+e)=c, for a small e and c[?](0, 1), but P(a-e < X < a+e|Z=z j )-0. That is, by just looking at the empirical distribution of income, the observation j appears to be perfectly normal. However, once we conditioned on Z=z j , we "uncover" the aberrant part. These definitions (kinds of outliers, contamination point and "masked" atypical observation) clarify why good leverage points cannot be "masked atypical observations" while bad leverage and contamination points can[7]. Whilst contamination points or bad leverage points would not desirably be included in a data set, decisions to include good leverage points introduce some degree of subjectivity. It is ultimately the researcher's decision, to determine whether a good leverage point is a low-probability observation but still part of the population of study or whether this observation is not part of the intended study population. 2.2. IF and robust statistic A fundamental concept in measuring robustness and efficiency of a statistical functional T (i.e. how a statistic is affected by atypical data) is its IF (Hampel, 1974). The IF measures how an statistic behaves if we add one more observation to a very large sample. If the IF is unbounded, a single outlier may cause trouble: Definition 5. IF: using our framework from the contamination model, the IF of a statistic T at a distribution G is given by: (Equation 2) The IF can be interpreted as a directional derivative[8] of T in the space of distributions. In other words, it gives the effect on T of an infinitesimal perturbation to the data at the point H. See that a bounded IF does not guarantee that the estimator is "good" in any sense, it only assures us that the estimator cannot be terribly bad, i.e. totally uninformative. Therefore, a bounded IF is a minimum condition that we should ask to our estimators. 2.3. Dalton's principle of transfers and its statistical implications Starting in the beginning of the twentieth century, there has been a debate between economists about the description and measurement of economic disparity. Usually, inequality indices should measure the spread of the distribution of incomes with respect to an all-equal distribution. This can be done in several ways, but still the ideas of Lorenz (1905), Gini (1921) and Dalton (1920) are the fundamental concepts in the construction of inequality indices. In fact, the only commonly accepted criterion that these indices should satisfy is the so-called Dalton's (1920) principle of transfers[9]: Definition 6. Dalton's principle of transfers: in a population of n individuals, let y i and y j be the income of the agent i and j, respectively. Suppose y i > y j , and define a=y i - y j . The Dalton principle suggests that inequality is reduced if there is a transfer from a richer person to a poorer person, i.e. if agent i transfers to agent j any amount up to a [10]. 2.3.1. Dalton principle and IF Cowell and Victoria-Feser (1996) showed that classical estimates of inequality indices that satisfy Dalton' s principle of transfers have an unbounded IF. This is a major concern, since they show that these indices can become totally uninformative when small arbitrary contaminations are present. Therefore, with a survey data that contains even only a small subsample of atypical observations, we cannot trust most inequality indices[11]. This raises a major question, can we trust at all in the estimates of such indexes? This research will produce robust inequality estimates by first trimming the "raw" data using robust procedures, and then calculate classical inequality indices using sample moments' analogues in the "clean" subsample (i.e. after dropping atypical data). This is an innovation in the literature. The next subsections explain the methodology applied to do this and the very last subsection in this section briefly discuss some alternative approaches. 3.1. Robust procedures, sorting the data by MCD and LTS The aim of a robust methodology is to produce estimators that are not "greatly" affected by atypical data. The definitions of outlier and contamination point of Section 2.1 suggest us that we should use a robust procedure to rank our observations, and then derive a test with a cut-off value to define atypical observations. Unfortunately, there is not a unique and unambiguous procedure to rank observations from a multivariate data set. In this paper we will order our data using two procedures: sorting residuals from a linear projection (LTS); and sorting observations by their distance to the centre of the data (MCD). Then we will derive a statistical test in order to define whether an observation is atypical. 3.2. LTS sorting Our aim will be to sort our data using a robust methodology (LTS). Our framework is a linear projection model with a response variable y i and a vector of p explanatory variables z i . The equation in matrix form is y=Z b+ e: Definition 7. Least trimmed squares (Rousseeuw, 1984): (Equation 3) where (Equation 4), (Equation 5) are the ordered square residuals, and h[?] is a predetermined threshold such that (Equation 6). The value h can be thought of as the minimum number of points which must not be atypical[12]. Then LTS choose the subset h with the smallest sum of residuals[13]. 3.3. MCD sorting Our aim in this subsection will be to sort our data using a robust methodology (MCD). A different option from LTS sorting is to define a centre of the data and then measure the distance between each observation and the centre. A commonly used measure is the Mahalanobis distance (see Rousseeuw, 1984): (Equation 7) where again x i is a row vector (that includes income y i and some socio-economic variables z i ), and x 0 is usually a variable that represents the sample mean for each variable; finally, G is the covariance matrix. This distance is very similar to the Euclidean distance, but it weights the distances by the covariance matrix. That is, it puts less weight to variables which variance is "big". Notice that R i is a scalar, therefore we can sort our data based on this scalar. Nevertheless, to implement this distance we would have to calculate the sample mean and the sample covariance, (Equation 8). As we have argued the entire paper, neither the sample mean nor the sample covariance are robust to outliers. We just need a single atypical observation to make this method completely uninformative. However, Rousseeuw (1984) proposed a solution to this problem. He suggested to first find a robust estimate of the mean and a robust estimate of the sample covariance. This robustification assures us that the sorting process of our data are not influenced by atypical observations: Definition 8. MCD (Rousseeuw, 1984): The MCD selects the subset of h points that are most tightly clustered. In particular, out of all subsets of size h, it finds the one whose sample covariance determinant is minimal[14]. 3.4. Final step to estimate income inequality indices Our methodology consists of four steps: sort the data robustly, this can be done by one of two procedures: LTS (sorting using robust residuals) or MCD (sorting using robust Mahalanobis distances); define a cut-off value; trim the data in order to remove atypical observations; and use sample analogues to estimate the inequality indices (Gini, Theil and Coefficient of Variation). These steps guarantee a robust estimate of income inequality, which is our main goal. Notice that our methodology is flexible enough that we can modify step 4 to estimate a multivariate inequality index or a poverty index robustly, if desired. Summarising, our four-step methodology can be implemented either by the LTS trimming procedure or by the MCD trimming procedure, and it provides us with a sample that we can confidently use to estimate inequality indices. 3.4.1. Alternative to our general four-step methodology: OBRE As mentioned in the literature review, to the author's knowledge, there is just one robust method (OBRE) that has been applied to the estimation of inequality indices. Cowell and Victoria-Feser (1996) developed a parametric approach, based on Hampel et al. (1986), to estimate inequality indices robustly. Their method, OBRE, is a subclass of the M-estimators. Essentially, they assume the data follows a distribution, then they estimate the parameters of that distribution by a procedure similar to a weighted maximum likelihood estimation. Before applying our methodology to real data, we would like to compare our procedures vs the other robust inequality estimation (OBRE). Therefore, we conduct the same simulation that Cowell and Victoria-Feser (1996) proposed. The next subsection will describe that simulation exercise. We also performed artificial contamination exercises with real data to show that we can effectively tag contamination points, see Appendices 1 and 2 for those results. 4.1. Univariate simulation We will randomly and artificially "contaminate" the data in order to show the extent to which income inequality indices can be distorted through contamination. We will use two types of contamination: Definition 9. "Severe" contamination: an observation for which income is multiplied by 10. Definition 10. "Weekly-Monthly" contamination: an observation for which income is multiplied by 4. Whilst this is only an univariate case[15], and therefore the main advantages of our procedures cannot be exploited, it will still provide useful insights into the effectiveness of the methodology. All codes and data are available in the author's webpage[16]. Simulation g distribution with a=3, l=1: (Equation 9) draws: 200 observations; samples: 100; and contamination: a fix percentage randomly chosen multiplied by 10 ("severe"). Figure 1 contrasts a sample from a g distribution vs a sample from the same g distribution, but with 15 per cent of its observations "severely" contaminated. It shows us that a mildly contaminated sample can appear totally different. The green empirical distribution (the contaminated one) appears to be very heavy-tailed while this is not the case in the uncontaminated red sample. This figure should alert us against the likely failure of robust parametric methods even in the univariate case. Table I displays the results we obtain after estimating the inequality indices using the "raw" (contaminated) data, and it also contains the mean squared error (MSE). It presents with data the great sensitivity of the indices shown theoretically in Section 2.2. Even with only six "severe" contaminated observations (3 per cent) the Coefficient of Variation increases almost seven times, the Theil Index almost doubles and the Gini Index increases by 42 per cent. Table II shows the results from Cowell and Victoria-Feser (1996). This method performs very well for 3 per cent of "severe" contamination, although it starts to exhibit some flaws at 5 per cent. We can compare this table with our results. Tables III and IV show the results of our four-step methodology. Both of our procedures are theoretically equivalent in the univariate case (see Rousseeuw, 1984). However, since we are using algorithms that approximate the exact solution and different methods to define the cut-off, we do expect a small numerical difference. So this exercise is also useful as a test to see how different our procedures are numerically. The results show that our methodology demonstrates a slight improvement (i.e. closer to the actual index prior to contamination) under low levels of contaminations. Remarkably, it performs quite a bit better for a contamination of 5 per cent (where OBRE starts to look questionable). Notice that it also performs admirably well for higher contamination rates. Recall that with OBRE we need to assume that the right distribution is known, this assumption is not needed either for MCD trimming or for LTS trimming. Even for a contamination of 10 or 15 per cent, the values of the indices calculated with our procedures are almost unaffected by the data contamination. Therefore, our methodology looks quite promising for computing robust estimates of inequality measures even with highly contaminated data, at least for the univariate case. The next section will show real multivariate applications for our procedures. 4.2. Multivariate empirical application, UK The first data set ever used for robust inequality estimation is "Households Below Average Income Data Set" (HBAI)[17] (Department for Work and Pensions, 1979), which contrary to the name is representative for the whole UK. This data set was originally used by Cowell and Victoria-Feser (1996). Therefore, it would be useful to make a direct comparison between the results of our method and theirs with this data set. However, they used a (presumably arbitrary) subset of only 746 observations, while we have used the full data set. Unfortunately, their sample was no longer available[18], which means our results are not quite comparable. The original HBAI data set provides information on the incomes and characteristics of 6,711 households based[19] on family expenditure survey data collected in January 1979. Participation in the survey is voluntary and the results are fully anonymised. This data set has played a fundamental role in the analysis of income by both researchers and governmental agencies[20]. Variables available describe the source of income, as well as information about the household such as household's composition (single, couple, children and tenure) and age of the head of the household. Notice that there are not many variables that might help to explain the income levels and which could be used to identify outliers through multivariate procedures. For example, socio-economic variables such as education level, background or ethnicity are not included. This means that this is not an ideal data set to apply our methodology; nevertheless, we expect that some variables like house tenure and social security benefits will help us to "explain" household income, and therefore to identify atypical observations[21]. All codes and data are available in the author's webpage[22]. Tables V and VI show the descriptive statistics for the "raw" data for the variables we used. Figures 2 and 3 show the empirical distribution of income[23]. The green bars represent the distribution from the "raw" data, the red bars show the resulting income distribution after we apply our four-step methodology. As we can see, the LTS trimming removes more observations near to the median than the MCD trimming. This suggests that, even though some income data points are not far away from the media, once we consider socio-economic variables these observations do not follow the pattern of the data; therefore, they are "atypical". In Table VII we compare the results obtained calculating the inequality indices using "raw" data vs using our four-step methodology (both procedures LTS and MCD trimming). Notice that MCD trimming does not remove many observations (about 1.8 per cent), but it does suggest a considerable reduction in the inequality indices: around 7 per cent for the Gini, 13 per cent for the Theil and 11 per cent for the Coefficient of Variation. In contrast, LTS trimming suggests a slightly greater amount of atypical observations and a further reduction in each inequality index. In general, the fact that MCD trimming removed fewer observations than LTS trimming suggests that observations in the data set are closer between each other, but that many observations do not follow a linear pattern. Figure 4 illustrates the first three steps in our four-step methodology, it also contrasts our MCD and LTS trimming procedures. On the x-axis data are sorted according with the MCD procedure, whilst on the y-axis observations are sorted according with the LTS procedure. Recall, the closest the observation is to (0, 0), the less "atypical" it is. If an observation is between the green lines, it is not atypical according with the LTS procedure; and if it is to the left of the vertical red line it is not atypical according to the MCD procedure. Notice that there are observations that are commonly atypical for either procedure (those at the top right corner), and vice versa. Also there are observations that are atypical for one procedure but not for the other; for instance, the group of observations to the right of the red vertical line and between the green lines is atypical for the MCD procedure but not for the LTS. 4.2.1. Public policy implications A reduction in the estimate of inequality has important consequences from a policy maker perspective. First, ceteris paribus, the optimal amount of resources destinated to directly address inequality/poverty[24] will be less than with the estimate of inequality using the "raw" data. Those "extra" resources can be destinated to promote growth. Notice that this is a direct consequence of having a more egalitarian economy than previously thought, this is due to the fact that poor people will actually enjoy a bigger share of any national income increment. This also implies that, in order to reduce poverty, public policies should focus more on economic growth. Some resources that were tagged directly to the poor[25], might be tagged to promote economic growth. Since now we know that the inequality is actually smaller than previously thought. Of course, the amount of those resources will depend on the difference between the inequality estimator calculated from "raw" data and the more precise inequality estimator calculated using our four-step procedure; the bigger the difference, the bigger the amount of resources that might be reallocated. We have another two hypothesis of the implications of our new methodology. First, its advantages will be more evident if we study inequality between subgroups of the population. Second, our methodology will help to confirm the classic argument that inequality is almost constant over time within a country[26]. Our first hypothesis can be thought in Roemerian terms: if we calculate inequality of opportunities, we need to estimate inequality over many subgroups; in each subgroup atypical data will be even more influential (in a IF way) than in the whole population. Therefore, the gains of our methodology will be even more notorious. Finally, in order to analyse our second hypothesis, we will use two cross-sectional data for Mexico in the next subsection. 4.3. Multivariate empirical application, Mexico (2006 and 2011) In order to have a broader perspective on the effect of atypical observations on inequality, specially over time and in a developing country, we use data from a Mexico. We expect to find more outliers in developing country data sets, given the special mechanism of income transmission; and also more contaminated observations, given technology and financial restrictions. We have two cross-sections, then our data allows us to study how atypical observations influence the change in inequality from 2006 to 2011. Another benefit of our data set is that the questionnaire is very rich; it provides socio-economic characteristics for each head of household as well as for their parents. These data sets are ideal for our methodology since we have many socio-economic variables[27] that can help to "explain" household income, therefore this empirical exercise is a unique opportunity. The data sets that we are using are "Encuesta ESRU de Movilidad Social en Mexico (EMOVI)" for 2006, and 2011, which are nationally representative for Mexico. Variables available include household income, household's members and head household characteristics, as well as information about the head of household background such as the place where they use to live and information about their parents[28]. The sample size is 5,902 for 2006 and 6,799 for 2011. All codes and data are available in the author's webpage[29]. Tables VIII and IX show the descriptive statistics for both of our data sets. Notice that we were able to use the same variables in both samples. 4.3.1. Main results Analogously as we did with the UK data set, we implement our four-step methodology for each year. The results are shown in Tables X-XIII. From 2006 to 2011 income inequality was reduced according to all indices. This is also the case in the official Mexican statistics[30]. However, the extent of the reduction is an interesting question. The "raw" data suggests a reduction in the Gini of 0.036, whilst the MCD trimming and the LTS trimming for the same index have a reduction of 0.025 and 0.06, respectively. This fact helps to corrobate the hyphotesis that inequality within a country is stable over time. Applying our methodology, we diminish the potential danger of violent swings in inequality due to data contamination. On the other hand, using the "raw" data we have a 0.06 reduction on the Theil Index, this is the same reduction that we get using the MCD trimming procedure; however, the LTS trimming also suggests a similar reduction. Finally, the Coefficient of Variation behaves similarly to that of the Gini Index. Figure 5 provides us with a unique opportunity to explain the differences between our LTS trimming and MCD trimming procedures and how these differences affect the estimation of the inequality indices. Both the Theil Index and the Coefficient of Variation belong to the same family (generalised entropy). However, the Coefficient of Variation is more sensitive to income differences at the top of the distribution[31]. This is an important feature to explain Figure 5. The 2011 graph in Figure 5 tells us an interesting story. There is an atypical group according to the MCD trimming procedure that is completely normal for the LTS trimming. That is, there is a group of observations that are "far" from the centre of the data, but that do follow the linear pattern of the data (i.e. they can be "explained" with the rest of the data using a linear projection). By reviewing the data we notice that this atypical group actually has high incomes, therefore we expect that the inclusion of this group would have a greater impact on the Coefficient of Variation than on the Theil Index[32]. That is the reason why the reduction from 2006 to 2011 in the MCD trimming procedure is greater in the Coefficient of Variation (0.0326) compared with the LTS trimming (0.0324); whilst the reduction in the Theil Index is greater with the LTS trimming procedure (0.0602) than with the MCD trimming procedure (0.0599). Notice that our comparison is merely descriptive, since we are not calculating the standard errors we cannot formally establish whether those changes are statistically different. 4.3.2. UK and Mexico's results comparison Figure 6 summarises graphically our results for Mexico 2006 and contrast them with our results for UK 1979. Each graph should be read as follows: The closest to (0, 0) the less "atypical"[33] an observation is. If an observation is between the green horizontal lines, then the observation is not "atypical" for one of our methods (LTS trimming). The further away from the lines, the more "atypical". If an observation is on the right of the vertical red line, then it is "atypical" for the other method (MCD trimming). The further right, the more "atypical". Notice that, relative to the UK data set, the Mexican data set contains more observations that are identified as highly atypical using both methodologies. These observations are likely to have a significant impact on inequality indices, yet both of our methodologies suggest that these data do not have any relationship with the rest of the sample. We can of expect this result given our discussion about a developing country. In this paper we consider the estimation of income inequality indices (such as Gini, Theil, Coefficient of Variation) using data with some atypical observations. It is shown that even when a small portion of the data are atypical, classical statistical procedures are not reliable in estimating income inequality indices. We propose two procedures to solve this problem: MCD trimming and LTS trimming. Both methodologies use all the information from a multivariate data set to remove atypical observations in order to estimate inequality indices robustly. We illustrate the problems with classical procedures and the merits of our methodologies by means of Monte Carlo simulations and three empirical applications. Our Monte Carlo simulation showed that even with a small percentage of atypical observations, inequality indices estimated with classical procedures become uninformative. Under the same simulation, our procedures successfully estimated inequality indices even when the data set contained a considerable number of atypical observations. The real magnitude of the problem of atypical observations is illustrated in the empirical applications. We found the presence of atypical observations in the UK, and Mexican data sets. The inequality indices changed drastically when these atypical observations were excluded. As expected, these results were particularly strong for the case of Mexico. We have hypothesised that situation given that Mexico is a developing country, therefore: Mexican surveys face more financial and technological restrictions (prone to contamination); and we expect that the wealth would be more concentrated in Mexico, therefore Mexico is more prone to outliers. The case for our methodology is clear. To strengthen our claims and results, we performed a further Monte Carlo simulation by artificially contaminating real data. We showed that we can successfully identify the artificially contaminated observations. We started this paper by asking a very important question in empirical economics as well as in policy making: can we trust in the estimates of income inequality? We have shown that the answer is: generally not, especially since most of the analysis use household survey data prone to contamination. This situation might be aggravated if the data set is from a developing country. We tried to highlight that household survey data has serious limitations in measuring income inequality, irrespective of the size of the data set. A common posture is to " believe" in the inequality estimate computed from "raw" data without trying to understand if this estimate corresponds to reality. A lesson to learn from this study is that even if you have plenty of data, these data might contain biases that can lead to misleading estimates of income inequality. We show that our methodology is an effective remedy to atypical observations. At least, one should compare the estimates from "raw" data with estimates derived from our procedures. Our methodology does not only remove outliers and contamination points, but it also provides a fundamental tool to answer the challenging, but important question of the true nature of income inequality. 5.1. Policy implications Our methodology provides a useful approach to ensuring income inequality indices are as meaningful as possible for policy makers. Our procedures do not only protect our estimations from atypical observations, but also highlight how the income inequality estimates are affected by groups of observations that are more closely together (MCD trimming) and by groups of observations that follow the same pattern (LTS trimming). This is a crucial tool for policy making, since we can contrast the MCD trimming and the LTS trimming to see if there is any particular subgroup that is atypical for the former but not for the latter. This would help to answer the question of where the inequality lies within the population. We think that this approach might be relevant especially when it comes to changes in inequality over time (as we discussed in the Mexican application). Finally, the simplicity of our procedures makes it realistic that such analysis could be applied and used by policy makers. 5.2. Further research A straight-forward extension of this research is to apply the methodologies proposed here to poverty indices and multivariate inequality indices. That is, once we "clean" our sample we can construct indices that take into account gender, age, ethnicity, education, etc. Further thought is still needed to polish and to extend these procedures, but given the critical importance of methods to measure inequality and other related measures of welfare to policy, our methodology is a relevant and useful addition to this debate. A further extension to our procedures is in the area of public goods provision. The detection and "treatment" of atypical data in the area of the provision and finance of local public goods is essential for analytical purposes within the frame of public economics. A single atypical observation can make an otherwise welfare improving set of local public goods, fiscally unfeasible. Moreover, it could have interesting policy implications in a decentralised political system. Distributive and allocative discrepancies of that type could be manifestations of the Gibbard-Satterthwaite[34] theorem. Our methodology could be combined with policy suggestions so that a local optimum is restored a la Tiebout-Oates[35]. Again, this is the case given that a single atypical observation can distord the household preferences for a tax-service mix. Finally, another related question not covered in this study is whether our procedures can be combined with the theory of optimal currency/fiscal areas. The remedy of atypical observations may be fruitful in enabling the fiscal/monetary authority to provide an optimal policy response. In summary, it is clear that there is a wide range of economic analyses to which our methodology could be usefully applied. Opens in a new window. Figure 1 15 per cent "Severe" contamination empirical Gamma distribution vs "clean" empirical Gamma distribution, 200 observations Opens in a new window. Figure 2 Empiricaldistribution: "raw" vs MCD trimming Opens in a new window. Figure 3 Empiricaldistribution: "raw" vs LTS trimming Opens in a new window. Figure 4 LTS trimming vs MCD trimming Opens in a new window. Figure 5 LTS trimming vs MCD trimming (2006 left, 2011 right) Opens in a new window. Figure 6 Measurement of atypical data, UK 1979 (left) vs Mexico 2006 (right) Opens in a new window. Table I Empirical indices using raw data Opens in a new window. Table II Inequality indices using OBRE Opens in a new window. Table III Inequality indices using MCD Opens in a new window. Table IV Inequality indices using LTS Opens in a new window. Table V Continuous variables Opens in a new window. Table VI Discrete variables Opens in a new window. Table VII Inequality indices using raw data and "clean" data Opens in a new window. Table VIII Inequality indices using MCD trimming with artificially contaminated data Opens in a new window. Table IX Inequality indices using LTS trimming with artificially contaminated data Opens in a new window. Table X Continuous variables Opens in a new window. Table XI Discrete variables Opens in a new window. Table XII Inequality indices using "raw" data and "clean" data Opens in a new window. Table XIII Change in inequality indices from 2006 to 2011 Opens in a new window. Table AI Inequality indices using MCD trimming with artificially contaminated data Opens in a new window. Table AII Inequality indices using LTS trimming with artificially contaminated data Opens in a new window. Equation 1 Opens in a new window. Equation 2 Opens in a new window. Equation 3 Opens in a new window. Equation 4 Opens in a new window. Equation 5 Opens in a new window. Equation 6 Opens in a new window. Equation 7 Opens in a new window. Equation 8 Opens in a new window. Equation 9
|
Applying these methods to household data for UK (1979) and Mexico (2006 and 2011), the author find that for UK data the Gini, Coefficient of Variation and Theil Inequality Indices are over estimated by between 0.02 and 0.04, while in the case of Mexico the same indices are over estimated more deeply, between 0.1 and almost 0.4. The relevance of including atypical observations that follow the linear pattern of the data are shown using the data from Mexico (2011).
|
[SECTION: Value] A single "extreme" observation can make an inequality index estimator uninformative, i.e., meaningless. Frequently, survey data are used to calculate inequality within an economy; we expect that survey data will contain both outliers and so-called contamination points. The main purpose of this research is to provide a new methodology to estimate inequality indices robustly; that is, to provide an estimator of inequality that is not severely affected by atypical data (outliers/contamination points). Our procedure uses information from multiple variables (e.g. socio-economic and geographic variables) to eliminate the effect of atypical observations. To our knowledge, in the inequality literature this is the first methodology that is able to identify outliers and contamination points in more than one direction[1]. Economic theories are often based on inference from data or are, at least, validated by real observations. However, the vast majority of this data is not collected under laboratory conditions. The main source of information to british analyse individual behaviour is a sample survey in which people are asked to recall previous actions[2]. A crucial question, and often ignored in applied work, is whether the data from these surveys are reliable. It is to be expected that asking people to recall past actions will lead to some errors. Both macro and microeconomic models depend crucially on ideas deduced from or corroborated by this survey data and, in turn, government policies are informed by the economic/statistical analysis of this data. If statistical analysis does not take into account that survey data might have as well some "non-sampling error", the amount of resources invested in public policies might not be optimal. Income inequality has become increasingly important within public policy as indices such as Gini, Theil and Coefficient of Variation have made metrics accessible to policy makers and the general public. The availability of data and software has made calculations easier and more readily available in both developed and developing countries. However, ironically one of the main concerns of Dalton (1920) has been almost ignored in the way that inequality indices are used. Dalton was concerned with the practical applicability of his indices and the "risk of errors" when measuring income inequality using "imperfect data". Dalton pointed out correctly that the main weakness of any income inequality index is a bad data set. Nevertheless, the empirical studies of inequality usually overlook this warning. Consequently, it seems essential to propose procedures that can take advantage of the advances in income inequality theory whilst empirically producing reliable estimates, no matter the quality of the data used. This research aims to develop an approach to achieve this. This paper illustrates that one should be very careful when using data "as it is" to estimate income inequality. We will propose a procedure that allows us to discover atypical observations; that is, observations that probably do not belong to the data set in use. By discovering these atypical observations, we will be able to estimate confidently any inequality measure. We catalogue atypical observations into two main categories: contamination points and outliers. Contamination refers to corrupted observations. The contamination can occur during processing or due to inaccurate answers of the respondent. Data contamination is especially relevant for developing countries, since it can be a common phenomena due to financial, technological and logistical constraints. Outliers are atypical observations that, opposed to data contamination, do not come from an error in the survey process but appear to be "very extreme". One possibility is that the survey includes a subgroup of households that come from a different population to that of the majority of the observations. On the other hand, we can have outliers that do not follow the pattern of the rest of the data. Given these challenges, contamination is almost unavoidable in household survey data. Sometimes, applied researches made arbitrary imputations when they believe that some outlieres or contaminated points appear in the sample. This is an ad hoc and questionable "solution". The purpose of this research is to estimate inequality indices robustly and objectively under an unknown distribution using all the information from a multivariate data set. This research proposes a four-step methodology based on robust procedures (minimum covariance determinant (MCD) and least trimmed squares) to remove atypical observations and to estimate inequality indices robustly based on a "clean" subsample. Assumptions about the structure of the data are relaxed. Multidimensional data sets permit the use of economic and social variables that are potentially related to income in order to uncover "masked" atypical observations. Whilst it is impossible to detect those "masked" observations with any univariate procedure, both of our multivariate methodologies can effectively tag them. One of our empirical application is the data set "Household Below Average Income" for the UK. The official Gini estimate for the UK in 1979 is 0.26, which is our estimate using "raw" data. However, with our methodology we observed that by removing only 1.8 per cent of atypical observations, the Gini falls to 0.24. Our other procedure suggests that by dropping 3 per cent of our sample, we find a change of 0.03 points. These findings are worth noting since they represent a change between 10 and 15 per cent in inequality. The changes in the Theil Index and in the Coefficient of Variation are even more drastic, using "raw" data the Theil Inequality Index is 0.1130 and the Coefficient of Variation 0.0709. With our procedures, the Theil Index diminishes up to 0.0912 (a change of 20 per cent) and the Coefficient of Variation up to 0.0589 (a change of 17 per cent). In our second empirical application we use the Mexican data set "Encuesta ESRU de Movilidad Social en Mexico (EMOVI)" for 2006, and 2011. Given that Mexico is a developing country, and likely to experience more data quality problems, it was not surprising that a greater number of outliers and contamination points were tagged in these data sets. Once we remove these very atypical observations, for Mexico 2006, we find a reduction of about 0.07 to 0.08 in the Gini Coefficient. Again the effect in the Theil Index and in the Coefficient of Variation is of greater magnitude. The Theil Index diminishes from 0.3615 up to 0.2481, and the Coefficient of Variation from 0.2091 to 0.1385. In an inequality framework, Cowell and Victoria-Feser (1996) used the concept of the influence function (IF) (Hampel, 1974) to assess the influence of an infinitesimal amount of contamination upon the value of an inequality statistic. They found that most inequality measures are not robust, i.e. a single arbitrary large observation can make an inequality index totally uninformative. This implies that the quality of inequality estimates derived from classical statistical procedures can be quite poor. It is therefore necessary to develop robust procedures in order to estimate confidently how the total income in a given society is distributed. In their seminal paper Cowell and Victoria-Feser (1996) proposed what is so far the only robust procedure that has been used to calculate inequality indices. It is based on the framework develop by Huber (1964). Their method, Optimal Bias-Robust Estimate (OBRE), is the most efficient approach under the right parametric model. However, OBRE has some major limitations: it is based on parametric procedures; and OBRE can only manage a single random variable. Their parametric approach is questionable since it is expected that the sample quality would not be the best. Notably, the multivariate nature of a household survey suggests that we should extend the procedure to a multivariate set up. This is mainly because of two reasons: a multivariate framework allows to control for atypical observations more effectively; and there are many situations in which income is not our only variable of interest. Univariate parametric robust procedures are not appropriate to estimate inequality indices. This kind of procedures are quite restrictive given that survey data has an intrinsic multidimensional nature. Furthermore, it is difficult to approximate a parametric distribution to a data set that presumably contains atypical data. Therefore, taking into account these limitations, it appears that the only robust approach in the literature to inequality measurement is not completely satisfactory. This paper addresses the current weaknesses in the literature: we propose two procedures that do not rely on parametric assumptions, and can work with multivariate data. By Monte Carlo simulations we show that these procedures can identify almost all contamination points. As opposed to the current state of the art methods, our procedures can be easily extended to work with multivariate inequality or multivariate poverty indices. Finally, under very mild assumptions, one of our methodologies produce consistent inequality indices estimators[3]. Our methodology has two main limitations: our procedures are not able to identify a bad leverage outlier[4] from a contamination point; and in the case that the data has no atypical observations, our procedures will tag as atypical a (very small) fraction of observations. The first limitation is not unique to our methodology; regardless of the estimation technique considered, we still could not differentiate these kind of observations[5]. Furthermore, we will argue that neither contamination points nor bad leverage outliers should be included in data sets so this limitation has no empirical consequences. On the other hand, the second limitation is the price we shall pay in order to have a reliable inequality estimator. Before explaining our methodology, let us clarify the concepts of outliers and contamination points. The definitions proposed in the next subsections are considered important for the reader to understand the efficacy and unprecedentedness of the methodology and the main results, which follow. 2.1. Outliers and contamination points For the purpose of this research, we breakdown atypical observations in two categories: outliers and contamination points; and outliers into two further subcategories: good leverage points and bad leverage points[6]: (Equation 1) There is no formal mathematical definition for an outlier, but the literature contains a number of explanatory definitions, as follows: Definition 1. "Outlier": Maguluri and Singh (1997): small fraction of data due to gross errors which are totally inconsistent with the rest of the data; Hampel et al. (1986): data which are far away from the bulk of the data, or more generally, from the pattern set by the majority of the data; Maronna et al. (2006): observations quite atypical by virtue of being far from the bulk of the data; Grubbs (1969): an observation that appears to deviate markedly from other members of the sample in within it occurs; and Barnett and Lewis (1994): a point that stands out in contrast to other observations as an extreme value. There are two common characteristics in these classical definitions of outliers. First, there is a great deal of subjectivity in determining the degree of distance from the bulk of the data at which an observation becomes an outlier; and second, the definition of an outlier will be based on beliefs about the underlying distribution of the population. If an egalitarian distribution is expected, this would lead to the exclusion of observations which would arouse no concern if a more dispersed population was presupposed. Notice that all definitions imply that outlying points stand out, but this might be in any direction. This fact relates to our previous discussion of univariate vs multivariate detection of atypical observations. There is not agreement that the outliers need to lie at the tails of the distribution; in fact, the definition seems to suggest that the outlying observation can happen in any direction. Let us proceed to specify the two types of outliers that we might find: Definition 2. Kinds of outlier (Rousseeuw and Van Driessen, 2006): good leverage point: a real and extreme observation in one or more directions that follows the pattern of the majority of the data; and bad leverage point: a real and extreme observation in one or more directions that does not follow the pattern of the majority of the data. On the other hand, despite the fact that they are intrinsically different, contamination points are usually considered under the outlier umbrella. We would like to make explicit the differences between contamination point and outlier by using the following definition: Definition 3. Contamination point: purely deterministic unreal data, e.g. observations that were coded, recorded or calculated with error. Notice that without any further assumption, it is impossible to identify a contamination point from a bad leverage point. Both should be discordant in one or more directions, and also both can appear at any part of the distribution. This situation motivates us to establish our definition for "masked atypical observation". Consider our household survey data X, where x i =(y i , z 1i , ..., z ki ) where Y is the income, and Z is some socio-economic variables: Definition 4. "Masked" atypical observation: a bad leverage point or contamination point y j =a such that P(a-e < Y < a+e)=c, for a small e and c[?](0, 1), but P(a-e < X < a+e|Z=z j )-0. That is, by just looking at the empirical distribution of income, the observation j appears to be perfectly normal. However, once we conditioned on Z=z j , we "uncover" the aberrant part. These definitions (kinds of outliers, contamination point and "masked" atypical observation) clarify why good leverage points cannot be "masked atypical observations" while bad leverage and contamination points can[7]. Whilst contamination points or bad leverage points would not desirably be included in a data set, decisions to include good leverage points introduce some degree of subjectivity. It is ultimately the researcher's decision, to determine whether a good leverage point is a low-probability observation but still part of the population of study or whether this observation is not part of the intended study population. 2.2. IF and robust statistic A fundamental concept in measuring robustness and efficiency of a statistical functional T (i.e. how a statistic is affected by atypical data) is its IF (Hampel, 1974). The IF measures how an statistic behaves if we add one more observation to a very large sample. If the IF is unbounded, a single outlier may cause trouble: Definition 5. IF: using our framework from the contamination model, the IF of a statistic T at a distribution G is given by: (Equation 2) The IF can be interpreted as a directional derivative[8] of T in the space of distributions. In other words, it gives the effect on T of an infinitesimal perturbation to the data at the point H. See that a bounded IF does not guarantee that the estimator is "good" in any sense, it only assures us that the estimator cannot be terribly bad, i.e. totally uninformative. Therefore, a bounded IF is a minimum condition that we should ask to our estimators. 2.3. Dalton's principle of transfers and its statistical implications Starting in the beginning of the twentieth century, there has been a debate between economists about the description and measurement of economic disparity. Usually, inequality indices should measure the spread of the distribution of incomes with respect to an all-equal distribution. This can be done in several ways, but still the ideas of Lorenz (1905), Gini (1921) and Dalton (1920) are the fundamental concepts in the construction of inequality indices. In fact, the only commonly accepted criterion that these indices should satisfy is the so-called Dalton's (1920) principle of transfers[9]: Definition 6. Dalton's principle of transfers: in a population of n individuals, let y i and y j be the income of the agent i and j, respectively. Suppose y i > y j , and define a=y i - y j . The Dalton principle suggests that inequality is reduced if there is a transfer from a richer person to a poorer person, i.e. if agent i transfers to agent j any amount up to a [10]. 2.3.1. Dalton principle and IF Cowell and Victoria-Feser (1996) showed that classical estimates of inequality indices that satisfy Dalton' s principle of transfers have an unbounded IF. This is a major concern, since they show that these indices can become totally uninformative when small arbitrary contaminations are present. Therefore, with a survey data that contains even only a small subsample of atypical observations, we cannot trust most inequality indices[11]. This raises a major question, can we trust at all in the estimates of such indexes? This research will produce robust inequality estimates by first trimming the "raw" data using robust procedures, and then calculate classical inequality indices using sample moments' analogues in the "clean" subsample (i.e. after dropping atypical data). This is an innovation in the literature. The next subsections explain the methodology applied to do this and the very last subsection in this section briefly discuss some alternative approaches. 3.1. Robust procedures, sorting the data by MCD and LTS The aim of a robust methodology is to produce estimators that are not "greatly" affected by atypical data. The definitions of outlier and contamination point of Section 2.1 suggest us that we should use a robust procedure to rank our observations, and then derive a test with a cut-off value to define atypical observations. Unfortunately, there is not a unique and unambiguous procedure to rank observations from a multivariate data set. In this paper we will order our data using two procedures: sorting residuals from a linear projection (LTS); and sorting observations by their distance to the centre of the data (MCD). Then we will derive a statistical test in order to define whether an observation is atypical. 3.2. LTS sorting Our aim will be to sort our data using a robust methodology (LTS). Our framework is a linear projection model with a response variable y i and a vector of p explanatory variables z i . The equation in matrix form is y=Z b+ e: Definition 7. Least trimmed squares (Rousseeuw, 1984): (Equation 3) where (Equation 4), (Equation 5) are the ordered square residuals, and h[?] is a predetermined threshold such that (Equation 6). The value h can be thought of as the minimum number of points which must not be atypical[12]. Then LTS choose the subset h with the smallest sum of residuals[13]. 3.3. MCD sorting Our aim in this subsection will be to sort our data using a robust methodology (MCD). A different option from LTS sorting is to define a centre of the data and then measure the distance between each observation and the centre. A commonly used measure is the Mahalanobis distance (see Rousseeuw, 1984): (Equation 7) where again x i is a row vector (that includes income y i and some socio-economic variables z i ), and x 0 is usually a variable that represents the sample mean for each variable; finally, G is the covariance matrix. This distance is very similar to the Euclidean distance, but it weights the distances by the covariance matrix. That is, it puts less weight to variables which variance is "big". Notice that R i is a scalar, therefore we can sort our data based on this scalar. Nevertheless, to implement this distance we would have to calculate the sample mean and the sample covariance, (Equation 8). As we have argued the entire paper, neither the sample mean nor the sample covariance are robust to outliers. We just need a single atypical observation to make this method completely uninformative. However, Rousseeuw (1984) proposed a solution to this problem. He suggested to first find a robust estimate of the mean and a robust estimate of the sample covariance. This robustification assures us that the sorting process of our data are not influenced by atypical observations: Definition 8. MCD (Rousseeuw, 1984): The MCD selects the subset of h points that are most tightly clustered. In particular, out of all subsets of size h, it finds the one whose sample covariance determinant is minimal[14]. 3.4. Final step to estimate income inequality indices Our methodology consists of four steps: sort the data robustly, this can be done by one of two procedures: LTS (sorting using robust residuals) or MCD (sorting using robust Mahalanobis distances); define a cut-off value; trim the data in order to remove atypical observations; and use sample analogues to estimate the inequality indices (Gini, Theil and Coefficient of Variation). These steps guarantee a robust estimate of income inequality, which is our main goal. Notice that our methodology is flexible enough that we can modify step 4 to estimate a multivariate inequality index or a poverty index robustly, if desired. Summarising, our four-step methodology can be implemented either by the LTS trimming procedure or by the MCD trimming procedure, and it provides us with a sample that we can confidently use to estimate inequality indices. 3.4.1. Alternative to our general four-step methodology: OBRE As mentioned in the literature review, to the author's knowledge, there is just one robust method (OBRE) that has been applied to the estimation of inequality indices. Cowell and Victoria-Feser (1996) developed a parametric approach, based on Hampel et al. (1986), to estimate inequality indices robustly. Their method, OBRE, is a subclass of the M-estimators. Essentially, they assume the data follows a distribution, then they estimate the parameters of that distribution by a procedure similar to a weighted maximum likelihood estimation. Before applying our methodology to real data, we would like to compare our procedures vs the other robust inequality estimation (OBRE). Therefore, we conduct the same simulation that Cowell and Victoria-Feser (1996) proposed. The next subsection will describe that simulation exercise. We also performed artificial contamination exercises with real data to show that we can effectively tag contamination points, see Appendices 1 and 2 for those results. 4.1. Univariate simulation We will randomly and artificially "contaminate" the data in order to show the extent to which income inequality indices can be distorted through contamination. We will use two types of contamination: Definition 9. "Severe" contamination: an observation for which income is multiplied by 10. Definition 10. "Weekly-Monthly" contamination: an observation for which income is multiplied by 4. Whilst this is only an univariate case[15], and therefore the main advantages of our procedures cannot be exploited, it will still provide useful insights into the effectiveness of the methodology. All codes and data are available in the author's webpage[16]. Simulation g distribution with a=3, l=1: (Equation 9) draws: 200 observations; samples: 100; and contamination: a fix percentage randomly chosen multiplied by 10 ("severe"). Figure 1 contrasts a sample from a g distribution vs a sample from the same g distribution, but with 15 per cent of its observations "severely" contaminated. It shows us that a mildly contaminated sample can appear totally different. The green empirical distribution (the contaminated one) appears to be very heavy-tailed while this is not the case in the uncontaminated red sample. This figure should alert us against the likely failure of robust parametric methods even in the univariate case. Table I displays the results we obtain after estimating the inequality indices using the "raw" (contaminated) data, and it also contains the mean squared error (MSE). It presents with data the great sensitivity of the indices shown theoretically in Section 2.2. Even with only six "severe" contaminated observations (3 per cent) the Coefficient of Variation increases almost seven times, the Theil Index almost doubles and the Gini Index increases by 42 per cent. Table II shows the results from Cowell and Victoria-Feser (1996). This method performs very well for 3 per cent of "severe" contamination, although it starts to exhibit some flaws at 5 per cent. We can compare this table with our results. Tables III and IV show the results of our four-step methodology. Both of our procedures are theoretically equivalent in the univariate case (see Rousseeuw, 1984). However, since we are using algorithms that approximate the exact solution and different methods to define the cut-off, we do expect a small numerical difference. So this exercise is also useful as a test to see how different our procedures are numerically. The results show that our methodology demonstrates a slight improvement (i.e. closer to the actual index prior to contamination) under low levels of contaminations. Remarkably, it performs quite a bit better for a contamination of 5 per cent (where OBRE starts to look questionable). Notice that it also performs admirably well for higher contamination rates. Recall that with OBRE we need to assume that the right distribution is known, this assumption is not needed either for MCD trimming or for LTS trimming. Even for a contamination of 10 or 15 per cent, the values of the indices calculated with our procedures are almost unaffected by the data contamination. Therefore, our methodology looks quite promising for computing robust estimates of inequality measures even with highly contaminated data, at least for the univariate case. The next section will show real multivariate applications for our procedures. 4.2. Multivariate empirical application, UK The first data set ever used for robust inequality estimation is "Households Below Average Income Data Set" (HBAI)[17] (Department for Work and Pensions, 1979), which contrary to the name is representative for the whole UK. This data set was originally used by Cowell and Victoria-Feser (1996). Therefore, it would be useful to make a direct comparison between the results of our method and theirs with this data set. However, they used a (presumably arbitrary) subset of only 746 observations, while we have used the full data set. Unfortunately, their sample was no longer available[18], which means our results are not quite comparable. The original HBAI data set provides information on the incomes and characteristics of 6,711 households based[19] on family expenditure survey data collected in January 1979. Participation in the survey is voluntary and the results are fully anonymised. This data set has played a fundamental role in the analysis of income by both researchers and governmental agencies[20]. Variables available describe the source of income, as well as information about the household such as household's composition (single, couple, children and tenure) and age of the head of the household. Notice that there are not many variables that might help to explain the income levels and which could be used to identify outliers through multivariate procedures. For example, socio-economic variables such as education level, background or ethnicity are not included. This means that this is not an ideal data set to apply our methodology; nevertheless, we expect that some variables like house tenure and social security benefits will help us to "explain" household income, and therefore to identify atypical observations[21]. All codes and data are available in the author's webpage[22]. Tables V and VI show the descriptive statistics for the "raw" data for the variables we used. Figures 2 and 3 show the empirical distribution of income[23]. The green bars represent the distribution from the "raw" data, the red bars show the resulting income distribution after we apply our four-step methodology. As we can see, the LTS trimming removes more observations near to the median than the MCD trimming. This suggests that, even though some income data points are not far away from the media, once we consider socio-economic variables these observations do not follow the pattern of the data; therefore, they are "atypical". In Table VII we compare the results obtained calculating the inequality indices using "raw" data vs using our four-step methodology (both procedures LTS and MCD trimming). Notice that MCD trimming does not remove many observations (about 1.8 per cent), but it does suggest a considerable reduction in the inequality indices: around 7 per cent for the Gini, 13 per cent for the Theil and 11 per cent for the Coefficient of Variation. In contrast, LTS trimming suggests a slightly greater amount of atypical observations and a further reduction in each inequality index. In general, the fact that MCD trimming removed fewer observations than LTS trimming suggests that observations in the data set are closer between each other, but that many observations do not follow a linear pattern. Figure 4 illustrates the first three steps in our four-step methodology, it also contrasts our MCD and LTS trimming procedures. On the x-axis data are sorted according with the MCD procedure, whilst on the y-axis observations are sorted according with the LTS procedure. Recall, the closest the observation is to (0, 0), the less "atypical" it is. If an observation is between the green lines, it is not atypical according with the LTS procedure; and if it is to the left of the vertical red line it is not atypical according to the MCD procedure. Notice that there are observations that are commonly atypical for either procedure (those at the top right corner), and vice versa. Also there are observations that are atypical for one procedure but not for the other; for instance, the group of observations to the right of the red vertical line and between the green lines is atypical for the MCD procedure but not for the LTS. 4.2.1. Public policy implications A reduction in the estimate of inequality has important consequences from a policy maker perspective. First, ceteris paribus, the optimal amount of resources destinated to directly address inequality/poverty[24] will be less than with the estimate of inequality using the "raw" data. Those "extra" resources can be destinated to promote growth. Notice that this is a direct consequence of having a more egalitarian economy than previously thought, this is due to the fact that poor people will actually enjoy a bigger share of any national income increment. This also implies that, in order to reduce poverty, public policies should focus more on economic growth. Some resources that were tagged directly to the poor[25], might be tagged to promote economic growth. Since now we know that the inequality is actually smaller than previously thought. Of course, the amount of those resources will depend on the difference between the inequality estimator calculated from "raw" data and the more precise inequality estimator calculated using our four-step procedure; the bigger the difference, the bigger the amount of resources that might be reallocated. We have another two hypothesis of the implications of our new methodology. First, its advantages will be more evident if we study inequality between subgroups of the population. Second, our methodology will help to confirm the classic argument that inequality is almost constant over time within a country[26]. Our first hypothesis can be thought in Roemerian terms: if we calculate inequality of opportunities, we need to estimate inequality over many subgroups; in each subgroup atypical data will be even more influential (in a IF way) than in the whole population. Therefore, the gains of our methodology will be even more notorious. Finally, in order to analyse our second hypothesis, we will use two cross-sectional data for Mexico in the next subsection. 4.3. Multivariate empirical application, Mexico (2006 and 2011) In order to have a broader perspective on the effect of atypical observations on inequality, specially over time and in a developing country, we use data from a Mexico. We expect to find more outliers in developing country data sets, given the special mechanism of income transmission; and also more contaminated observations, given technology and financial restrictions. We have two cross-sections, then our data allows us to study how atypical observations influence the change in inequality from 2006 to 2011. Another benefit of our data set is that the questionnaire is very rich; it provides socio-economic characteristics for each head of household as well as for their parents. These data sets are ideal for our methodology since we have many socio-economic variables[27] that can help to "explain" household income, therefore this empirical exercise is a unique opportunity. The data sets that we are using are "Encuesta ESRU de Movilidad Social en Mexico (EMOVI)" for 2006, and 2011, which are nationally representative for Mexico. Variables available include household income, household's members and head household characteristics, as well as information about the head of household background such as the place where they use to live and information about their parents[28]. The sample size is 5,902 for 2006 and 6,799 for 2011. All codes and data are available in the author's webpage[29]. Tables VIII and IX show the descriptive statistics for both of our data sets. Notice that we were able to use the same variables in both samples. 4.3.1. Main results Analogously as we did with the UK data set, we implement our four-step methodology for each year. The results are shown in Tables X-XIII. From 2006 to 2011 income inequality was reduced according to all indices. This is also the case in the official Mexican statistics[30]. However, the extent of the reduction is an interesting question. The "raw" data suggests a reduction in the Gini of 0.036, whilst the MCD trimming and the LTS trimming for the same index have a reduction of 0.025 and 0.06, respectively. This fact helps to corrobate the hyphotesis that inequality within a country is stable over time. Applying our methodology, we diminish the potential danger of violent swings in inequality due to data contamination. On the other hand, using the "raw" data we have a 0.06 reduction on the Theil Index, this is the same reduction that we get using the MCD trimming procedure; however, the LTS trimming also suggests a similar reduction. Finally, the Coefficient of Variation behaves similarly to that of the Gini Index. Figure 5 provides us with a unique opportunity to explain the differences between our LTS trimming and MCD trimming procedures and how these differences affect the estimation of the inequality indices. Both the Theil Index and the Coefficient of Variation belong to the same family (generalised entropy). However, the Coefficient of Variation is more sensitive to income differences at the top of the distribution[31]. This is an important feature to explain Figure 5. The 2011 graph in Figure 5 tells us an interesting story. There is an atypical group according to the MCD trimming procedure that is completely normal for the LTS trimming. That is, there is a group of observations that are "far" from the centre of the data, but that do follow the linear pattern of the data (i.e. they can be "explained" with the rest of the data using a linear projection). By reviewing the data we notice that this atypical group actually has high incomes, therefore we expect that the inclusion of this group would have a greater impact on the Coefficient of Variation than on the Theil Index[32]. That is the reason why the reduction from 2006 to 2011 in the MCD trimming procedure is greater in the Coefficient of Variation (0.0326) compared with the LTS trimming (0.0324); whilst the reduction in the Theil Index is greater with the LTS trimming procedure (0.0602) than with the MCD trimming procedure (0.0599). Notice that our comparison is merely descriptive, since we are not calculating the standard errors we cannot formally establish whether those changes are statistically different. 4.3.2. UK and Mexico's results comparison Figure 6 summarises graphically our results for Mexico 2006 and contrast them with our results for UK 1979. Each graph should be read as follows: The closest to (0, 0) the less "atypical"[33] an observation is. If an observation is between the green horizontal lines, then the observation is not "atypical" for one of our methods (LTS trimming). The further away from the lines, the more "atypical". If an observation is on the right of the vertical red line, then it is "atypical" for the other method (MCD trimming). The further right, the more "atypical". Notice that, relative to the UK data set, the Mexican data set contains more observations that are identified as highly atypical using both methodologies. These observations are likely to have a significant impact on inequality indices, yet both of our methodologies suggest that these data do not have any relationship with the rest of the sample. We can of expect this result given our discussion about a developing country. In this paper we consider the estimation of income inequality indices (such as Gini, Theil, Coefficient of Variation) using data with some atypical observations. It is shown that even when a small portion of the data are atypical, classical statistical procedures are not reliable in estimating income inequality indices. We propose two procedures to solve this problem: MCD trimming and LTS trimming. Both methodologies use all the information from a multivariate data set to remove atypical observations in order to estimate inequality indices robustly. We illustrate the problems with classical procedures and the merits of our methodologies by means of Monte Carlo simulations and three empirical applications. Our Monte Carlo simulation showed that even with a small percentage of atypical observations, inequality indices estimated with classical procedures become uninformative. Under the same simulation, our procedures successfully estimated inequality indices even when the data set contained a considerable number of atypical observations. The real magnitude of the problem of atypical observations is illustrated in the empirical applications. We found the presence of atypical observations in the UK, and Mexican data sets. The inequality indices changed drastically when these atypical observations were excluded. As expected, these results were particularly strong for the case of Mexico. We have hypothesised that situation given that Mexico is a developing country, therefore: Mexican surveys face more financial and technological restrictions (prone to contamination); and we expect that the wealth would be more concentrated in Mexico, therefore Mexico is more prone to outliers. The case for our methodology is clear. To strengthen our claims and results, we performed a further Monte Carlo simulation by artificially contaminating real data. We showed that we can successfully identify the artificially contaminated observations. We started this paper by asking a very important question in empirical economics as well as in policy making: can we trust in the estimates of income inequality? We have shown that the answer is: generally not, especially since most of the analysis use household survey data prone to contamination. This situation might be aggravated if the data set is from a developing country. We tried to highlight that household survey data has serious limitations in measuring income inequality, irrespective of the size of the data set. A common posture is to " believe" in the inequality estimate computed from "raw" data without trying to understand if this estimate corresponds to reality. A lesson to learn from this study is that even if you have plenty of data, these data might contain biases that can lead to misleading estimates of income inequality. We show that our methodology is an effective remedy to atypical observations. At least, one should compare the estimates from "raw" data with estimates derived from our procedures. Our methodology does not only remove outliers and contamination points, but it also provides a fundamental tool to answer the challenging, but important question of the true nature of income inequality. 5.1. Policy implications Our methodology provides a useful approach to ensuring income inequality indices are as meaningful as possible for policy makers. Our procedures do not only protect our estimations from atypical observations, but also highlight how the income inequality estimates are affected by groups of observations that are more closely together (MCD trimming) and by groups of observations that follow the same pattern (LTS trimming). This is a crucial tool for policy making, since we can contrast the MCD trimming and the LTS trimming to see if there is any particular subgroup that is atypical for the former but not for the latter. This would help to answer the question of where the inequality lies within the population. We think that this approach might be relevant especially when it comes to changes in inequality over time (as we discussed in the Mexican application). Finally, the simplicity of our procedures makes it realistic that such analysis could be applied and used by policy makers. 5.2. Further research A straight-forward extension of this research is to apply the methodologies proposed here to poverty indices and multivariate inequality indices. That is, once we "clean" our sample we can construct indices that take into account gender, age, ethnicity, education, etc. Further thought is still needed to polish and to extend these procedures, but given the critical importance of methods to measure inequality and other related measures of welfare to policy, our methodology is a relevant and useful addition to this debate. A further extension to our procedures is in the area of public goods provision. The detection and "treatment" of atypical data in the area of the provision and finance of local public goods is essential for analytical purposes within the frame of public economics. A single atypical observation can make an otherwise welfare improving set of local public goods, fiscally unfeasible. Moreover, it could have interesting policy implications in a decentralised political system. Distributive and allocative discrepancies of that type could be manifestations of the Gibbard-Satterthwaite[34] theorem. Our methodology could be combined with policy suggestions so that a local optimum is restored a la Tiebout-Oates[35]. Again, this is the case given that a single atypical observation can distord the household preferences for a tax-service mix. Finally, another related question not covered in this study is whether our procedures can be combined with the theory of optimal currency/fiscal areas. The remedy of atypical observations may be fruitful in enabling the fiscal/monetary authority to provide an optimal policy response. In summary, it is clear that there is a wide range of economic analyses to which our methodology could be usefully applied. Opens in a new window. Figure 1 15 per cent "Severe" contamination empirical Gamma distribution vs "clean" empirical Gamma distribution, 200 observations Opens in a new window. Figure 2 Empiricaldistribution: "raw" vs MCD trimming Opens in a new window. Figure 3 Empiricaldistribution: "raw" vs LTS trimming Opens in a new window. Figure 4 LTS trimming vs MCD trimming Opens in a new window. Figure 5 LTS trimming vs MCD trimming (2006 left, 2011 right) Opens in a new window. Figure 6 Measurement of atypical data, UK 1979 (left) vs Mexico 2006 (right) Opens in a new window. Table I Empirical indices using raw data Opens in a new window. Table II Inequality indices using OBRE Opens in a new window. Table III Inequality indices using MCD Opens in a new window. Table IV Inequality indices using LTS Opens in a new window. Table V Continuous variables Opens in a new window. Table VI Discrete variables Opens in a new window. Table VII Inequality indices using raw data and "clean" data Opens in a new window. Table VIII Inequality indices using MCD trimming with artificially contaminated data Opens in a new window. Table IX Inequality indices using LTS trimming with artificially contaminated data Opens in a new window. Table X Continuous variables Opens in a new window. Table XI Discrete variables Opens in a new window. Table XII Inequality indices using "raw" data and "clean" data Opens in a new window. Table XIII Change in inequality indices from 2006 to 2011 Opens in a new window. Table AI Inequality indices using MCD trimming with artificially contaminated data Opens in a new window. Table AII Inequality indices using LTS trimming with artificially contaminated data Opens in a new window. Equation 1 Opens in a new window. Equation 2 Opens in a new window. Equation 3 Opens in a new window. Equation 4 Opens in a new window. Equation 5 Opens in a new window. Equation 6 Opens in a new window. Equation 7 Opens in a new window. Equation 8 Opens in a new window. Equation 9
|
To the knowledge, in the inequality literature this is the first methodology that is able to identify outliers and contamination points in more than one direction. That is, not only at the tails of the distribution, but on the whole marginal distribution of income. This is possible via the use of other variables related to income.
|
[SECTION: Purpose] Looking across Swedish higher education, the last few years have seen an increase in the number of environmental courses being offered at the undergraduate and postgraduate levels. According to a survey conducted by the National Agency for Higher Education (2001), 3 of 37 universities and university colleges offer introductory courses in "sustainable development" that are open to all students. Fourteen educational institutions offer specific environmental introductory courses to a select group of undergraduate students and another 16 universities give undergraduate courses that include environmental issues or sustainable development. Moreover, several Swedish universities offer programmes in "Environmental science" leading to a BSc or an MSc degree, and there are also a number of civil engineering programmes that focus on environmental engineering.In this paper, the results of which draw from a doctoral thesis (Lundholm, 2003), students' perspectives on environmental education are presented. It is based on three case studies in which the aim was to explore the students' conceptions of ecology and environmental issues, and the learning process. Learning is viewed from an intentional perspective, which means that the students' aims in their studies, here defined as "projects", are explored through analysis of their actions and utterances in educational settings or interviews. Previous research on learners and learning in environmental education has focused mainly on the learners' knowledge, attitudes and behaviour in relation to environmental phenomena and environmental problems[1]. A number of studies on young people's understanding of environmental problems were concerned with investigating their scientific knowledge (Gomez-Granell and Cervera-March, 1993; Gambro and Switzky, 1996; Ivy et al., 1998). Boyes and Stanisstreet (1993, 1994, 1998) have studied young people's knowledge of the greenhouse effect and ozone depletion in particular.Research on students' learning has focused on the effects of environmental education, and has often been conducted as pre-test and post-test studies. Only a few investigations have been concerned specifically with students' personal interpretations and experiences of environmental education (Lai, 1999; Rickinson, 1999) or with the learning process (Emmons, 1997; Mason and Santi, 1998; Grace and Ratcliffe, 2002).Research on subject-related knowledge in environmental education has had its primary focus on the natural sciences, whereas the social science disciplines have received little attention from researchers in this field (Rickinson, 2001).Within much of this work there is a very strong science education influence. The environment phenomena that have been investigated are ones from school curricula, the research participants are often from science teaching groups, and the implications drawn often take the form of strategies for science teachers (Rickinson, 2001, p. 219). The purpose of the study was to investigate and describe students' interpretations of various tasks concerning environmental issues and of course content in ecology. The purpose was also to investigate the students' learning, defined as a process of differentiation between contexts.The thesis attended to the following research questions:RQ1. How do six first-year civil engineering students interpret a course content in ecology and environmental issues; i.e. how do they contextualise the content? RQ2. How does a group of biology students interpret a task on "environmental reports?" What "problems" are brought to the fore? RQ3. How do six postgraduate students interpret environmental research? What "problems" are brought to the fore? In relation to previous research, the aim of the thesis was to explore the learning process, as opposed to learning outcomes, and the students' interpretations and experiences of environmental education. Learning as a process of differentiation Constructivism has been an established field in educational research ever since the pioneering work of Piaget (1929/1989) and his associates in the 1920s on children's conceptions of the world. Within this field, "conceptual change" has emerged as the dominant description of learning, implying that learning can be described as a process in which commonsense notions of the world are replaced by scientifically accepted ways of conceptualising the world (Posner et al., 1982). Critics of such a view have pointed out that learning is not a matter of replacing less qualified or naive conceptions of phenomena in the world, but is rather a question of understanding in what situations and within which genres different knowledge is useful and appropriate (Caravita and Hallden, 1994; Hallden, 1999; Hallden et al., 2002; Petersson, 2002). Learning is thus seen as a process of differentiating between various contexts.Wistedt (1994), Wistedt and Martinsson (1996), Wistedt et al. (1997) and Hallden (1999) describe the different ways in which students interpret and contextualise tasks assigned in classroom situations. Students' and teachers' interpretations that are brought to the fore in school settings cannot be described merely as being either commonsense or theoretical, but rather as belonging to different theoretical contexts representing different disciplines or genres (Wistedt, 1998). Students' alternative contextualisations can also be seen as expressions of values based on norms and values that are prevalent in society, and the context can therefore be characterised as cultural. An empirical example of this is given in a study by Hallden (1999). The study describes how biology students contextualise a task concerning the ecological relation between animals and plants as a problem pertaining to the extinction of species. The students' interpretation of the task becomes understandable when we consider the ongoing debate in today's society regarding the extinction of species and biological diversity. The students' cultural contextualisation of the task, placing it within an environmental discussion, raises value judgement issues, such as: which species should be saved and which ones should be left to become extinct?Within different academic disciplines, concepts are defined within the framework of the different subjects' theoretical perspectives on phenomena and data. In a theoretical context, for example, an academic discipline, there are theories and explanatory models at a meta-level that explain data and empirical findings at a lower level (Figure 1). In a study by Wistedt and Martinsson (1994) on how pupils solve mathematics tasks presented to them in a formal educational setting, the findings show how the pupils communicate and make explicit the premises for their interpretations of the task. When the pupils articulate their thoughts at a meta-level and relate theories to an empirical level, or relate phenomena and concepts to an empirical level, they are elaborating within a context. If different contexts are brought to the fore by the learners, and are articulated as different theories or premises for interpretations at a meta-level, this is described as elaborating between contexts.Task problem and project Hallden (1982, 1988, 2001) and Wistedt (1987) use the term project to describe what students are doing and aiming at achieving when working in educational settings. Often in education students are given tasks to work on and solve. Hallden (1982, 1988, 2001) and Wistedt (1987) have investigated the way students interpret tasks given to them as problems. Hallden defines task as "what is presented to the pupils by the teacher with the intention that they are to do something and/or to learn something" (1988, p. 125). The terms task and problem distinguishe what the teacher has presented to the students, written or orally, from the students' interpretation of the very same. When we analyse a student's interpretation of a task we can describe the problem he/she is trying to solve. Problem is a term that describes what the students are trying to accomplish. Furthermore, the problem he or she is trying to solve can be interpreted in a wider sense by taking into account the student's aims and goals. By ascribing a student a plausible goal that he/she is trying to realise, defined here as project, the students' interpretation of the task becomes intelligible. Understanding what the students are doing is, therefore, looking for the meaning that they ascribe the task. Case studies Even though there are no clear-cut or absolute boundaries between different kinds of case studies, quantitative and qualitative case studies are often described as being exploratory, descriptive or explanatory (Yin, 1994). Deciding on what kind of case study can be constructively used in a particular research project inevitably entails questions of research design that have to do with the type of research question, the extent of control over actual behavioural events and to what extent the focus will be on contemporary as opposed to historical events. Stake (1994) distinguishes intrinsic case studies from instrumental and collective case studies. In an intrinsic case study the aim is to obtain knowledge about a particular case. An instrumental case study is used for enhancing understanding of an issue or for refining a theory. So, in an instrumental case study the case is of secondary interest since it is examined for the purpose of gaining knowledge about something else. When the interest is to theorise further about a phenomenon or a population, and the particular case is of less interest, the term "collective case studies" is used. Collective case studies thus serve the same purpose as instrumental case studies, but are numerous.The three case studies[2] in this thesis on students' experiences and learning in environmental education can be analysed in view of the above terminology. With an interest in elaborating a theory on learning, the cases can be seen as both instrumental and explorative. The term "explore" indicates that something is unknown and needs to be investigated and, considering previous research on environmental education, this seems highly appropriate. The research on students and environmental education has for the most part concerned students' attitudes, behaviours, knowledge and learning outcomes (Rickinson, 2001). However, the research presented here had as its main focus students' learning processes and interpretations of a particular course content. The case studies in the thesis vary with respect to the content to be learnt. The ecology course contains various natural science subjects, and the task on environmental reports that the biology students work with can be related to the social sciences, whereas the postgraduates' research work comprises aspects of both natural and social science.In the first case study lectures in a compulsory course in ecology (four credit points) taken by first-year civil engineering students were tape-recorded and observed. The course was given during the spring term of the academic year of 2000, from March to June, at the Royal Institute of Technology (KTH) in Stockholm. There were 103 students in the class, and six of these students, two women and four men, were selected for interview. The interviews lasted between 50 and 70 minutes. The interview questions concerned the students' reasons for studying an engineering programme and becoming a civil engineer, their conceptions of their first year at university and the ecology course in general - what they had learnt and how they related the course content to their future profession.In the second case study observations were made of the lectures in a course at Stockholm University. The course was a ten-credit point supplementary course, "Environmental control for biologists", conducted in 1998. This course, which is multi-disciplinary in character, is still offered by the university and adopts a social perspective on environmental issues with lectures on politics, economy, environment and technology, environmental law and international environmental control. At the end of the course the students were offered five different themes to work with. In the case study reported on in the thesis, a group of four students, three women and one man, chose to work on a task relating to "environmental reports". The students' group work was observed, hand-written notes were made of their activities and the discussions were tape-recorded.The third case study was conducted in 1999, also at KTH. Six postgraduate students, four women and two men, were interviewed concerning their environmental research, choice of topic and interest in pursuing postgraduate studies. Three of the students had just begun, or were in the process of planning for, their postgraduate studies and the other three intended to present their licentiate thesis within the coming year. The interviews lasted between 60 and 90 minutes. All of the interviews were transcribed in full.Interpretation of the data - understanding students' utterances and actions This research is based on the assumption that in order to understand students' interpretations and learning we must acknowledge the students' reasons for studying and their personal study aims. How do we account for that, and how is it done when analysing data? Von Wright (1971) has written about intentionality in action as a presupposition for understanding the meaning of peoples' actions. Without such a presupposition, action becomes mere behaviour, that is, muscular movement. Von Wright stated:Behaviour gets its intentional character from being seen by the agent himself or by an outside observer in a wider perspective, from being set in a context of aims and cognitions. This is what happens when we construe a practical inference to match it, as premises match a given conclusion (von Wright, 1971, pp. 114-15).A practical inference, or syllogism, implies that action is understood as teleological rather than causal. The "causes" of an action reside in the future and in the mind of the person we are trying to understand. The following may serve as an example of such a teleological inference:A person P intends to bring about x.P believes bringing about x would require the fulfilment of y.Hence: P does y (in order to bring about x).In the syllogism the first premise states the person's goal, what he/she intends, and the second premise states the ways and the means by which that person believes that the goal can be achieved. The premises, and the conclusion, are logically and not causally related.This kind of interpretation of what people are doing and saying is, in a sense, trivial since all of us in our daily lives constantly ascribe intentions to people's actions; if we see a man running towards a bus we assume that his intention is to get on the bus. If we know this man well enough, we can also make the assumption that he is running towards the bus in order to get to the train station and catch a particular train. This interpretation involves putting the act of running towards the bus within the more far-reaching goal of catching the train. Hallden (1982, 1988, 2001) and Wistedt (1987) use the term project to describe what students are doing and what they are trying to achieve when working in an educational setting. When analysing a student's interpretation of a task, it is possible to describe the problem that he/she is trying to solve. The problem that the student is working on can also be interpreted as part of a future goal, a project, for example, to pass a particular exam or to successfully complete a whole course or obtain a degree. Analysing data from such a project-oriented viewpoint means asking questions about what kinds of problems the students are working on. In light of which projects do the students study activities become meaningful? Such questions give the researcher a chance to understand what the students are saying and doing, and may reveal possible explanations as to why they interpreted a certain content or task in a certain way. The engineering students' interpretations can be described in terms of a theoretical context in the sense that the students related the content to concepts and different academic subjects within the natural sciences[3]. The content also raised issues concerning human beings in relation to nature. The students discussed the notion of humankind as being within or outside the eco-system, and this seems to have led them to consider the moral aspects of human actions and the effects of these actions on nature. This means that the students' interpretations of the course content can be said to be embedded within a cultural context of values and norms. The students seemed to believe that the teacher had, implicitly, expressed the viewpoint that human beings have affected nature only in a negative way and that the teacher had neglected to highlight the positive effect which that impact has had on humans. In talking about these aspects of values, the students showed irritation and indignation and seemed to find the whole issue quite provoking.In the analysis of the different ways in which the students interpreted the course content, the results show that during the lectures the engineering students posed questions to the teacher concerning possible solutions to various environmental problems and whether these solutions were good or bad and could eventually lead to improving the environment. Finally, the students also focused on different aspects of the course content, such as the water cycle; they also stated that learning about such matters as the chemical formulas of the phosphorus cycle was irrelevant for their future profession as civil engineers.The students' different contextualisations of the course content, as well as their questions and focused aspects of the content, were analysed from an intentional perspective. The students' actions and utterances become meaningful if we interpret them in relation to their future goals, i.e. their projects. Their interpretations and reactions to the course content become intelligible and possible to understand in view of the professional project of becoming a civil engineer. As civil engineers they will inevitably affect nature in some way, good or bad. An ecocentric perspective that places a higher value on nature than on human beings, together with a view that implies that the effect humans have on nature is harmful, may come into conflict with the very notion of being a civil engineer. The questions that were posed in the classroom concerning solutions to environmental problems can also be understood as being both part of a professional project and a possible project related to a more general interest in wanting to know what could be done to improve the environment. Finally, the two students who stated that certain aspects of the content were irrelevant were arguing with reference to the civil engineering profession. When the students were working on a task concerning environmental reports, different contextualisations as well as different problems were brought to the fore. The students decided to work on a problem about environmental reports as a general phenomenon and to collect reports from different companies and analyse and study the text they contained. While the students were working on that problem, another problem, this one concerning the environmental work of certain companies was brought to fore. In the students' discussion, this problem meant that the students analysed and discussed the companies' work as such, and not the texts presented in the reports. In particular, one of the students in the group, Hans, clarified and stressed this difference.The students' work on analysing the texts was contextualised into an academic and cultural context. By setting up 15 criteria for analysing the 14 reports, the students' aim was to describe the reports' content concerning the companies' environmental work and how these routines were presented to the reader. While solving the problem, the students' personal judgements about, for example, the environmental education given to the staff became a topic for discussion. One student in particular emphasised the importance of keeping judgements out of the discussion, and argued that the group should try to describe the reports in terms of "what" content was being presented and "how" it was being presented to the reader. He also stressed that the group should not evaluate the different aspects presented in the reports, for example, concerning environmental education, as being good or bad. Eventually, on the last day of working on the task, the students discussed the difficulty of trying to keep their values and judgements as well as their feelings about the different companies separate from the analysis.In analysing the students' interpretations of the task from an intentional perspective, the different contextualisations, as well as the different problems, can be discussed in relation to the students' projects. A possible project that can be ascribed the students is a concern for the environment and an interest in promoting and working towards a better environment. Whether this is a project of interest or of profession, or perhaps both, is difficult to say. The students' apparent interest in working with a problem that concerned the companies' business as such, and not just what was stated in the reports, is understandable in light of an environmental interest in finding out what is being done within the companies and if this will eventually improve the environment. The students' difficulty in differentiating between, on one hand, a cultural context of values and judgements about the companies' environmental work as it is described in the reports, and, on the other, an academic context can also be explained in relation to a project. If the students' project of interest is to promote environmental change, their will to not only describe environmental reports, but also analyse and judge them in terms of good and bad, becomes reasonable. For the six postgraduate students the task of writing a doctoral thesis raised different problems, which were possible to analyse in relation to the students' different projects. Firstly, the thesis was interpreted as an assignment to be attended to within the university and the scientific community. This problem can be understood as being part of the students' educational project. Second, for five of the six postgraduate students, the task of writing a thesis was commissioned by a company outside the university, and the research results were to serve as a means for companies to take action; writing a thesis was thus interpreted as a question of more or less producing a plan of action. Third, writing the thesis was a project of personal interest in the sense that the students felt they could contribute to improving the environment. Three of the students also stated that their postgraduate studies would benefit them in their professional work and that this was a contributory reason for their taking on postgraduate studies.In solving the problem of writing a doctoral thesis within the educational project, some difficulties arose. To be able to answer their research questions, two of the students, Elisabeth and Sofia, had to acquaint themselves with subject areas in the social sciences. These subjects were quite new to them, as were the qualitative methods they were expected to use in their research. They realised that knowledge about how to conduct qualitative research was scarce at KTH. Therefore, the students had to learn on their own and try to overcome this obstacle in different ways. Another student, Mikael, stressed that he had chosen to narrow down his research and exclude questions that would lead him too far afield into other subject areas. He argued that too wide a research question could not be handled in a doctoral research project.The field of environmental research evoked thoughts about identity and feelings of uncertainty. Elisabeth discussed the fact the her background was in civil engineering and that civil engineers traditionally had not engaged in the kind of environmental research that she was now conducting. Sofia, on the other hand, was not a civil engineer and was worried that her natural science background would pose problems. However, she stated that any student entering her research field and having to learn a new subject, in her case macroeconomics, would have been confronted with the same difficulties regardless of whether their backgrounds were in civil engineering or natural science.In working on her thesis, Elisabeth had come to realise that she and the Hydropower Company that had given her the assignment interpreted the concept of "sustainability" in different ways. She also discussed the fact that she had personal beliefs and values about the issues she was writing about and was worried that these would influence her research in an inappropriate way. Contextualisations and differentiation When comparing the students' interpretations of course content and different tasks, the results show that a diversity of contexts was brought to the fore. In all three cases the students interpreted a cultural context concerned with values. In confronting course content in ecology, the civil engineering students were provoked by the value judgement that human beings are generally harmful to nature. When the biology students solved the task concerning environmental reports, values and personal judgements about the various companies and their environmental development work were also brought to the fore. The third case study, finally, makes it clear that values and beliefs about the research topic can become an issue for postgraduate students in their doctoral research.What can be said about the students' learning as a process of differentiation? In the case study focusing on the civil engineering students' interpretations of the course content, the students did not talk about their values and perspective on nature as values. They, therefore, did not differentiate them as values in contrast with, for example, descriptions of nature. The biology students, on the other hand, discussed the fact that they were judging the companies on emotional grounds and tried to differentiate between an academic context, where they described the different companies, and a value-laden context.The postgraduate students' interpretations of the task to write a doctoral thesis brought a range of problems to the fore. One of the problems concerned viewing a thesis as a scientific product, and two of the students, Mikael and Sofia, discussed methodology and ways of approaching new subjects in the social sciences. Their statements can be analysed with a focus on tendencies to elaborate within and between different contexts. In Mikael's discussion on what methodologies can be used to investigate a certain phenomenon at an empirical level, stressing this as a matter of choice to consider in research may be described as a form of elaboration within a scientific context. Sofia's reasoning about the subject of political science and the fact that she thought it important to be cautious when using concepts within a new discipline suggests that she was aware that a meta-level defines concepts within different subjects.Elisabeth brought to the fore a difficulty that concerns values and how to deal with this aspect. She reflected on the fact that she had beliefs and different interpretations about the concept of sustainability. The fact that she had realised this and tried to distinguish between her personal definitions, definitions that would be appropriate to use in the scientific community and definitions embraced by her project sponsors, also indicates that a differentiation has taken place. The findings raise the question of how affects can influence the learning process, and the students' possibilities of differentiating between various conceptual frameworks. In relation to research on cognitive development as a process of conceptual change where the student abandons conceptions due to "dissatisfaction" (Posner et al., 1982, p. 214) and finds new conceptions "intelligible" and "plausible" (Posner et al., 1982, p. 214), it can be questioned if learning is such a rational and intellectual process. If we consider an intentional perspective when looking at learners and learning we can perhaps better understand the rationale and logic of their way of reasoning and understanding environmental issues.The results can be discussed in relation to educational practice and the content, as well as the purpose, of environmental education. If environmental education can raise these kinds of moral and ethical conceptions concerning nature and man's behaviour, as well as judgements about companies, ethics or discussions concerning values should be part of the curriculum or in some way highlighted and recognised. Perhaps the engineering students would have benefited from a discussion on different ethical perspectives on nature as well as the ethics of professional engineers and possible dilemmas they will meet in their future career. Furthermore, the students' focus on solving and finding solutions to environmental problems is an issue to consider in education. Whether the content includes aspects of the ways in which environmental problems are dealt with in society or not, this focus of interest will probably affect the learning process. Opens in a new window.Figure 1 Different contexts for interpretation
|
- To present results from a research project on postgraduate and undergraduate students' learning about environmental issues in education.
|
[SECTION: Method] Looking across Swedish higher education, the last few years have seen an increase in the number of environmental courses being offered at the undergraduate and postgraduate levels. According to a survey conducted by the National Agency for Higher Education (2001), 3 of 37 universities and university colleges offer introductory courses in "sustainable development" that are open to all students. Fourteen educational institutions offer specific environmental introductory courses to a select group of undergraduate students and another 16 universities give undergraduate courses that include environmental issues or sustainable development. Moreover, several Swedish universities offer programmes in "Environmental science" leading to a BSc or an MSc degree, and there are also a number of civil engineering programmes that focus on environmental engineering.In this paper, the results of which draw from a doctoral thesis (Lundholm, 2003), students' perspectives on environmental education are presented. It is based on three case studies in which the aim was to explore the students' conceptions of ecology and environmental issues, and the learning process. Learning is viewed from an intentional perspective, which means that the students' aims in their studies, here defined as "projects", are explored through analysis of their actions and utterances in educational settings or interviews. Previous research on learners and learning in environmental education has focused mainly on the learners' knowledge, attitudes and behaviour in relation to environmental phenomena and environmental problems[1]. A number of studies on young people's understanding of environmental problems were concerned with investigating their scientific knowledge (Gomez-Granell and Cervera-March, 1993; Gambro and Switzky, 1996; Ivy et al., 1998). Boyes and Stanisstreet (1993, 1994, 1998) have studied young people's knowledge of the greenhouse effect and ozone depletion in particular.Research on students' learning has focused on the effects of environmental education, and has often been conducted as pre-test and post-test studies. Only a few investigations have been concerned specifically with students' personal interpretations and experiences of environmental education (Lai, 1999; Rickinson, 1999) or with the learning process (Emmons, 1997; Mason and Santi, 1998; Grace and Ratcliffe, 2002).Research on subject-related knowledge in environmental education has had its primary focus on the natural sciences, whereas the social science disciplines have received little attention from researchers in this field (Rickinson, 2001).Within much of this work there is a very strong science education influence. The environment phenomena that have been investigated are ones from school curricula, the research participants are often from science teaching groups, and the implications drawn often take the form of strategies for science teachers (Rickinson, 2001, p. 219). The purpose of the study was to investigate and describe students' interpretations of various tasks concerning environmental issues and of course content in ecology. The purpose was also to investigate the students' learning, defined as a process of differentiation between contexts.The thesis attended to the following research questions:RQ1. How do six first-year civil engineering students interpret a course content in ecology and environmental issues; i.e. how do they contextualise the content? RQ2. How does a group of biology students interpret a task on "environmental reports?" What "problems" are brought to the fore? RQ3. How do six postgraduate students interpret environmental research? What "problems" are brought to the fore? In relation to previous research, the aim of the thesis was to explore the learning process, as opposed to learning outcomes, and the students' interpretations and experiences of environmental education. Learning as a process of differentiation Constructivism has been an established field in educational research ever since the pioneering work of Piaget (1929/1989) and his associates in the 1920s on children's conceptions of the world. Within this field, "conceptual change" has emerged as the dominant description of learning, implying that learning can be described as a process in which commonsense notions of the world are replaced by scientifically accepted ways of conceptualising the world (Posner et al., 1982). Critics of such a view have pointed out that learning is not a matter of replacing less qualified or naive conceptions of phenomena in the world, but is rather a question of understanding in what situations and within which genres different knowledge is useful and appropriate (Caravita and Hallden, 1994; Hallden, 1999; Hallden et al., 2002; Petersson, 2002). Learning is thus seen as a process of differentiating between various contexts.Wistedt (1994), Wistedt and Martinsson (1996), Wistedt et al. (1997) and Hallden (1999) describe the different ways in which students interpret and contextualise tasks assigned in classroom situations. Students' and teachers' interpretations that are brought to the fore in school settings cannot be described merely as being either commonsense or theoretical, but rather as belonging to different theoretical contexts representing different disciplines or genres (Wistedt, 1998). Students' alternative contextualisations can also be seen as expressions of values based on norms and values that are prevalent in society, and the context can therefore be characterised as cultural. An empirical example of this is given in a study by Hallden (1999). The study describes how biology students contextualise a task concerning the ecological relation between animals and plants as a problem pertaining to the extinction of species. The students' interpretation of the task becomes understandable when we consider the ongoing debate in today's society regarding the extinction of species and biological diversity. The students' cultural contextualisation of the task, placing it within an environmental discussion, raises value judgement issues, such as: which species should be saved and which ones should be left to become extinct?Within different academic disciplines, concepts are defined within the framework of the different subjects' theoretical perspectives on phenomena and data. In a theoretical context, for example, an academic discipline, there are theories and explanatory models at a meta-level that explain data and empirical findings at a lower level (Figure 1). In a study by Wistedt and Martinsson (1994) on how pupils solve mathematics tasks presented to them in a formal educational setting, the findings show how the pupils communicate and make explicit the premises for their interpretations of the task. When the pupils articulate their thoughts at a meta-level and relate theories to an empirical level, or relate phenomena and concepts to an empirical level, they are elaborating within a context. If different contexts are brought to the fore by the learners, and are articulated as different theories or premises for interpretations at a meta-level, this is described as elaborating between contexts.Task problem and project Hallden (1982, 1988, 2001) and Wistedt (1987) use the term project to describe what students are doing and aiming at achieving when working in educational settings. Often in education students are given tasks to work on and solve. Hallden (1982, 1988, 2001) and Wistedt (1987) have investigated the way students interpret tasks given to them as problems. Hallden defines task as "what is presented to the pupils by the teacher with the intention that they are to do something and/or to learn something" (1988, p. 125). The terms task and problem distinguishe what the teacher has presented to the students, written or orally, from the students' interpretation of the very same. When we analyse a student's interpretation of a task we can describe the problem he/she is trying to solve. Problem is a term that describes what the students are trying to accomplish. Furthermore, the problem he or she is trying to solve can be interpreted in a wider sense by taking into account the student's aims and goals. By ascribing a student a plausible goal that he/she is trying to realise, defined here as project, the students' interpretation of the task becomes intelligible. Understanding what the students are doing is, therefore, looking for the meaning that they ascribe the task. Case studies Even though there are no clear-cut or absolute boundaries between different kinds of case studies, quantitative and qualitative case studies are often described as being exploratory, descriptive or explanatory (Yin, 1994). Deciding on what kind of case study can be constructively used in a particular research project inevitably entails questions of research design that have to do with the type of research question, the extent of control over actual behavioural events and to what extent the focus will be on contemporary as opposed to historical events. Stake (1994) distinguishes intrinsic case studies from instrumental and collective case studies. In an intrinsic case study the aim is to obtain knowledge about a particular case. An instrumental case study is used for enhancing understanding of an issue or for refining a theory. So, in an instrumental case study the case is of secondary interest since it is examined for the purpose of gaining knowledge about something else. When the interest is to theorise further about a phenomenon or a population, and the particular case is of less interest, the term "collective case studies" is used. Collective case studies thus serve the same purpose as instrumental case studies, but are numerous.The three case studies[2] in this thesis on students' experiences and learning in environmental education can be analysed in view of the above terminology. With an interest in elaborating a theory on learning, the cases can be seen as both instrumental and explorative. The term "explore" indicates that something is unknown and needs to be investigated and, considering previous research on environmental education, this seems highly appropriate. The research on students and environmental education has for the most part concerned students' attitudes, behaviours, knowledge and learning outcomes (Rickinson, 2001). However, the research presented here had as its main focus students' learning processes and interpretations of a particular course content. The case studies in the thesis vary with respect to the content to be learnt. The ecology course contains various natural science subjects, and the task on environmental reports that the biology students work with can be related to the social sciences, whereas the postgraduates' research work comprises aspects of both natural and social science.In the first case study lectures in a compulsory course in ecology (four credit points) taken by first-year civil engineering students were tape-recorded and observed. The course was given during the spring term of the academic year of 2000, from March to June, at the Royal Institute of Technology (KTH) in Stockholm. There were 103 students in the class, and six of these students, two women and four men, were selected for interview. The interviews lasted between 50 and 70 minutes. The interview questions concerned the students' reasons for studying an engineering programme and becoming a civil engineer, their conceptions of their first year at university and the ecology course in general - what they had learnt and how they related the course content to their future profession.In the second case study observations were made of the lectures in a course at Stockholm University. The course was a ten-credit point supplementary course, "Environmental control for biologists", conducted in 1998. This course, which is multi-disciplinary in character, is still offered by the university and adopts a social perspective on environmental issues with lectures on politics, economy, environment and technology, environmental law and international environmental control. At the end of the course the students were offered five different themes to work with. In the case study reported on in the thesis, a group of four students, three women and one man, chose to work on a task relating to "environmental reports". The students' group work was observed, hand-written notes were made of their activities and the discussions were tape-recorded.The third case study was conducted in 1999, also at KTH. Six postgraduate students, four women and two men, were interviewed concerning their environmental research, choice of topic and interest in pursuing postgraduate studies. Three of the students had just begun, or were in the process of planning for, their postgraduate studies and the other three intended to present their licentiate thesis within the coming year. The interviews lasted between 60 and 90 minutes. All of the interviews were transcribed in full.Interpretation of the data - understanding students' utterances and actions This research is based on the assumption that in order to understand students' interpretations and learning we must acknowledge the students' reasons for studying and their personal study aims. How do we account for that, and how is it done when analysing data? Von Wright (1971) has written about intentionality in action as a presupposition for understanding the meaning of peoples' actions. Without such a presupposition, action becomes mere behaviour, that is, muscular movement. Von Wright stated:Behaviour gets its intentional character from being seen by the agent himself or by an outside observer in a wider perspective, from being set in a context of aims and cognitions. This is what happens when we construe a practical inference to match it, as premises match a given conclusion (von Wright, 1971, pp. 114-15).A practical inference, or syllogism, implies that action is understood as teleological rather than causal. The "causes" of an action reside in the future and in the mind of the person we are trying to understand. The following may serve as an example of such a teleological inference:A person P intends to bring about x.P believes bringing about x would require the fulfilment of y.Hence: P does y (in order to bring about x).In the syllogism the first premise states the person's goal, what he/she intends, and the second premise states the ways and the means by which that person believes that the goal can be achieved. The premises, and the conclusion, are logically and not causally related.This kind of interpretation of what people are doing and saying is, in a sense, trivial since all of us in our daily lives constantly ascribe intentions to people's actions; if we see a man running towards a bus we assume that his intention is to get on the bus. If we know this man well enough, we can also make the assumption that he is running towards the bus in order to get to the train station and catch a particular train. This interpretation involves putting the act of running towards the bus within the more far-reaching goal of catching the train. Hallden (1982, 1988, 2001) and Wistedt (1987) use the term project to describe what students are doing and what they are trying to achieve when working in an educational setting. When analysing a student's interpretation of a task, it is possible to describe the problem that he/she is trying to solve. The problem that the student is working on can also be interpreted as part of a future goal, a project, for example, to pass a particular exam or to successfully complete a whole course or obtain a degree. Analysing data from such a project-oriented viewpoint means asking questions about what kinds of problems the students are working on. In light of which projects do the students study activities become meaningful? Such questions give the researcher a chance to understand what the students are saying and doing, and may reveal possible explanations as to why they interpreted a certain content or task in a certain way. The engineering students' interpretations can be described in terms of a theoretical context in the sense that the students related the content to concepts and different academic subjects within the natural sciences[3]. The content also raised issues concerning human beings in relation to nature. The students discussed the notion of humankind as being within or outside the eco-system, and this seems to have led them to consider the moral aspects of human actions and the effects of these actions on nature. This means that the students' interpretations of the course content can be said to be embedded within a cultural context of values and norms. The students seemed to believe that the teacher had, implicitly, expressed the viewpoint that human beings have affected nature only in a negative way and that the teacher had neglected to highlight the positive effect which that impact has had on humans. In talking about these aspects of values, the students showed irritation and indignation and seemed to find the whole issue quite provoking.In the analysis of the different ways in which the students interpreted the course content, the results show that during the lectures the engineering students posed questions to the teacher concerning possible solutions to various environmental problems and whether these solutions were good or bad and could eventually lead to improving the environment. Finally, the students also focused on different aspects of the course content, such as the water cycle; they also stated that learning about such matters as the chemical formulas of the phosphorus cycle was irrelevant for their future profession as civil engineers.The students' different contextualisations of the course content, as well as their questions and focused aspects of the content, were analysed from an intentional perspective. The students' actions and utterances become meaningful if we interpret them in relation to their future goals, i.e. their projects. Their interpretations and reactions to the course content become intelligible and possible to understand in view of the professional project of becoming a civil engineer. As civil engineers they will inevitably affect nature in some way, good or bad. An ecocentric perspective that places a higher value on nature than on human beings, together with a view that implies that the effect humans have on nature is harmful, may come into conflict with the very notion of being a civil engineer. The questions that were posed in the classroom concerning solutions to environmental problems can also be understood as being both part of a professional project and a possible project related to a more general interest in wanting to know what could be done to improve the environment. Finally, the two students who stated that certain aspects of the content were irrelevant were arguing with reference to the civil engineering profession. When the students were working on a task concerning environmental reports, different contextualisations as well as different problems were brought to the fore. The students decided to work on a problem about environmental reports as a general phenomenon and to collect reports from different companies and analyse and study the text they contained. While the students were working on that problem, another problem, this one concerning the environmental work of certain companies was brought to fore. In the students' discussion, this problem meant that the students analysed and discussed the companies' work as such, and not the texts presented in the reports. In particular, one of the students in the group, Hans, clarified and stressed this difference.The students' work on analysing the texts was contextualised into an academic and cultural context. By setting up 15 criteria for analysing the 14 reports, the students' aim was to describe the reports' content concerning the companies' environmental work and how these routines were presented to the reader. While solving the problem, the students' personal judgements about, for example, the environmental education given to the staff became a topic for discussion. One student in particular emphasised the importance of keeping judgements out of the discussion, and argued that the group should try to describe the reports in terms of "what" content was being presented and "how" it was being presented to the reader. He also stressed that the group should not evaluate the different aspects presented in the reports, for example, concerning environmental education, as being good or bad. Eventually, on the last day of working on the task, the students discussed the difficulty of trying to keep their values and judgements as well as their feelings about the different companies separate from the analysis.In analysing the students' interpretations of the task from an intentional perspective, the different contextualisations, as well as the different problems, can be discussed in relation to the students' projects. A possible project that can be ascribed the students is a concern for the environment and an interest in promoting and working towards a better environment. Whether this is a project of interest or of profession, or perhaps both, is difficult to say. The students' apparent interest in working with a problem that concerned the companies' business as such, and not just what was stated in the reports, is understandable in light of an environmental interest in finding out what is being done within the companies and if this will eventually improve the environment. The students' difficulty in differentiating between, on one hand, a cultural context of values and judgements about the companies' environmental work as it is described in the reports, and, on the other, an academic context can also be explained in relation to a project. If the students' project of interest is to promote environmental change, their will to not only describe environmental reports, but also analyse and judge them in terms of good and bad, becomes reasonable. For the six postgraduate students the task of writing a doctoral thesis raised different problems, which were possible to analyse in relation to the students' different projects. Firstly, the thesis was interpreted as an assignment to be attended to within the university and the scientific community. This problem can be understood as being part of the students' educational project. Second, for five of the six postgraduate students, the task of writing a thesis was commissioned by a company outside the university, and the research results were to serve as a means for companies to take action; writing a thesis was thus interpreted as a question of more or less producing a plan of action. Third, writing the thesis was a project of personal interest in the sense that the students felt they could contribute to improving the environment. Three of the students also stated that their postgraduate studies would benefit them in their professional work and that this was a contributory reason for their taking on postgraduate studies.In solving the problem of writing a doctoral thesis within the educational project, some difficulties arose. To be able to answer their research questions, two of the students, Elisabeth and Sofia, had to acquaint themselves with subject areas in the social sciences. These subjects were quite new to them, as were the qualitative methods they were expected to use in their research. They realised that knowledge about how to conduct qualitative research was scarce at KTH. Therefore, the students had to learn on their own and try to overcome this obstacle in different ways. Another student, Mikael, stressed that he had chosen to narrow down his research and exclude questions that would lead him too far afield into other subject areas. He argued that too wide a research question could not be handled in a doctoral research project.The field of environmental research evoked thoughts about identity and feelings of uncertainty. Elisabeth discussed the fact the her background was in civil engineering and that civil engineers traditionally had not engaged in the kind of environmental research that she was now conducting. Sofia, on the other hand, was not a civil engineer and was worried that her natural science background would pose problems. However, she stated that any student entering her research field and having to learn a new subject, in her case macroeconomics, would have been confronted with the same difficulties regardless of whether their backgrounds were in civil engineering or natural science.In working on her thesis, Elisabeth had come to realise that she and the Hydropower Company that had given her the assignment interpreted the concept of "sustainability" in different ways. She also discussed the fact that she had personal beliefs and values about the issues she was writing about and was worried that these would influence her research in an inappropriate way. Contextualisations and differentiation When comparing the students' interpretations of course content and different tasks, the results show that a diversity of contexts was brought to the fore. In all three cases the students interpreted a cultural context concerned with values. In confronting course content in ecology, the civil engineering students were provoked by the value judgement that human beings are generally harmful to nature. When the biology students solved the task concerning environmental reports, values and personal judgements about the various companies and their environmental development work were also brought to the fore. The third case study, finally, makes it clear that values and beliefs about the research topic can become an issue for postgraduate students in their doctoral research.What can be said about the students' learning as a process of differentiation? In the case study focusing on the civil engineering students' interpretations of the course content, the students did not talk about their values and perspective on nature as values. They, therefore, did not differentiate them as values in contrast with, for example, descriptions of nature. The biology students, on the other hand, discussed the fact that they were judging the companies on emotional grounds and tried to differentiate between an academic context, where they described the different companies, and a value-laden context.The postgraduate students' interpretations of the task to write a doctoral thesis brought a range of problems to the fore. One of the problems concerned viewing a thesis as a scientific product, and two of the students, Mikael and Sofia, discussed methodology and ways of approaching new subjects in the social sciences. Their statements can be analysed with a focus on tendencies to elaborate within and between different contexts. In Mikael's discussion on what methodologies can be used to investigate a certain phenomenon at an empirical level, stressing this as a matter of choice to consider in research may be described as a form of elaboration within a scientific context. Sofia's reasoning about the subject of political science and the fact that she thought it important to be cautious when using concepts within a new discipline suggests that she was aware that a meta-level defines concepts within different subjects.Elisabeth brought to the fore a difficulty that concerns values and how to deal with this aspect. She reflected on the fact that she had beliefs and different interpretations about the concept of sustainability. The fact that she had realised this and tried to distinguish between her personal definitions, definitions that would be appropriate to use in the scientific community and definitions embraced by her project sponsors, also indicates that a differentiation has taken place. The findings raise the question of how affects can influence the learning process, and the students' possibilities of differentiating between various conceptual frameworks. In relation to research on cognitive development as a process of conceptual change where the student abandons conceptions due to "dissatisfaction" (Posner et al., 1982, p. 214) and finds new conceptions "intelligible" and "plausible" (Posner et al., 1982, p. 214), it can be questioned if learning is such a rational and intellectual process. If we consider an intentional perspective when looking at learners and learning we can perhaps better understand the rationale and logic of their way of reasoning and understanding environmental issues.The results can be discussed in relation to educational practice and the content, as well as the purpose, of environmental education. If environmental education can raise these kinds of moral and ethical conceptions concerning nature and man's behaviour, as well as judgements about companies, ethics or discussions concerning values should be part of the curriculum or in some way highlighted and recognised. Perhaps the engineering students would have benefited from a discussion on different ethical perspectives on nature as well as the ethics of professional engineers and possible dilemmas they will meet in their future career. Furthermore, the students' focus on solving and finding solutions to environmental problems is an issue to consider in education. Whether the content includes aspects of the ways in which environmental problems are dealt with in society or not, this focus of interest will probably affect the learning process. Opens in a new window.Figure 1 Different contexts for interpretation
|
- Three cases were carried out with civil engineering students, biology students and postgraduate students. Discussions in classroom were tape-recorded, as well as discussions while working with assignments, and interviews were carried out.
|
[SECTION: Findings] Looking across Swedish higher education, the last few years have seen an increase in the number of environmental courses being offered at the undergraduate and postgraduate levels. According to a survey conducted by the National Agency for Higher Education (2001), 3 of 37 universities and university colleges offer introductory courses in "sustainable development" that are open to all students. Fourteen educational institutions offer specific environmental introductory courses to a select group of undergraduate students and another 16 universities give undergraduate courses that include environmental issues or sustainable development. Moreover, several Swedish universities offer programmes in "Environmental science" leading to a BSc or an MSc degree, and there are also a number of civil engineering programmes that focus on environmental engineering.In this paper, the results of which draw from a doctoral thesis (Lundholm, 2003), students' perspectives on environmental education are presented. It is based on three case studies in which the aim was to explore the students' conceptions of ecology and environmental issues, and the learning process. Learning is viewed from an intentional perspective, which means that the students' aims in their studies, here defined as "projects", are explored through analysis of their actions and utterances in educational settings or interviews. Previous research on learners and learning in environmental education has focused mainly on the learners' knowledge, attitudes and behaviour in relation to environmental phenomena and environmental problems[1]. A number of studies on young people's understanding of environmental problems were concerned with investigating their scientific knowledge (Gomez-Granell and Cervera-March, 1993; Gambro and Switzky, 1996; Ivy et al., 1998). Boyes and Stanisstreet (1993, 1994, 1998) have studied young people's knowledge of the greenhouse effect and ozone depletion in particular.Research on students' learning has focused on the effects of environmental education, and has often been conducted as pre-test and post-test studies. Only a few investigations have been concerned specifically with students' personal interpretations and experiences of environmental education (Lai, 1999; Rickinson, 1999) or with the learning process (Emmons, 1997; Mason and Santi, 1998; Grace and Ratcliffe, 2002).Research on subject-related knowledge in environmental education has had its primary focus on the natural sciences, whereas the social science disciplines have received little attention from researchers in this field (Rickinson, 2001).Within much of this work there is a very strong science education influence. The environment phenomena that have been investigated are ones from school curricula, the research participants are often from science teaching groups, and the implications drawn often take the form of strategies for science teachers (Rickinson, 2001, p. 219). The purpose of the study was to investigate and describe students' interpretations of various tasks concerning environmental issues and of course content in ecology. The purpose was also to investigate the students' learning, defined as a process of differentiation between contexts.The thesis attended to the following research questions:RQ1. How do six first-year civil engineering students interpret a course content in ecology and environmental issues; i.e. how do they contextualise the content? RQ2. How does a group of biology students interpret a task on "environmental reports?" What "problems" are brought to the fore? RQ3. How do six postgraduate students interpret environmental research? What "problems" are brought to the fore? In relation to previous research, the aim of the thesis was to explore the learning process, as opposed to learning outcomes, and the students' interpretations and experiences of environmental education. Learning as a process of differentiation Constructivism has been an established field in educational research ever since the pioneering work of Piaget (1929/1989) and his associates in the 1920s on children's conceptions of the world. Within this field, "conceptual change" has emerged as the dominant description of learning, implying that learning can be described as a process in which commonsense notions of the world are replaced by scientifically accepted ways of conceptualising the world (Posner et al., 1982). Critics of such a view have pointed out that learning is not a matter of replacing less qualified or naive conceptions of phenomena in the world, but is rather a question of understanding in what situations and within which genres different knowledge is useful and appropriate (Caravita and Hallden, 1994; Hallden, 1999; Hallden et al., 2002; Petersson, 2002). Learning is thus seen as a process of differentiating between various contexts.Wistedt (1994), Wistedt and Martinsson (1996), Wistedt et al. (1997) and Hallden (1999) describe the different ways in which students interpret and contextualise tasks assigned in classroom situations. Students' and teachers' interpretations that are brought to the fore in school settings cannot be described merely as being either commonsense or theoretical, but rather as belonging to different theoretical contexts representing different disciplines or genres (Wistedt, 1998). Students' alternative contextualisations can also be seen as expressions of values based on norms and values that are prevalent in society, and the context can therefore be characterised as cultural. An empirical example of this is given in a study by Hallden (1999). The study describes how biology students contextualise a task concerning the ecological relation between animals and plants as a problem pertaining to the extinction of species. The students' interpretation of the task becomes understandable when we consider the ongoing debate in today's society regarding the extinction of species and biological diversity. The students' cultural contextualisation of the task, placing it within an environmental discussion, raises value judgement issues, such as: which species should be saved and which ones should be left to become extinct?Within different academic disciplines, concepts are defined within the framework of the different subjects' theoretical perspectives on phenomena and data. In a theoretical context, for example, an academic discipline, there are theories and explanatory models at a meta-level that explain data and empirical findings at a lower level (Figure 1). In a study by Wistedt and Martinsson (1994) on how pupils solve mathematics tasks presented to them in a formal educational setting, the findings show how the pupils communicate and make explicit the premises for their interpretations of the task. When the pupils articulate their thoughts at a meta-level and relate theories to an empirical level, or relate phenomena and concepts to an empirical level, they are elaborating within a context. If different contexts are brought to the fore by the learners, and are articulated as different theories or premises for interpretations at a meta-level, this is described as elaborating between contexts.Task problem and project Hallden (1982, 1988, 2001) and Wistedt (1987) use the term project to describe what students are doing and aiming at achieving when working in educational settings. Often in education students are given tasks to work on and solve. Hallden (1982, 1988, 2001) and Wistedt (1987) have investigated the way students interpret tasks given to them as problems. Hallden defines task as "what is presented to the pupils by the teacher with the intention that they are to do something and/or to learn something" (1988, p. 125). The terms task and problem distinguishe what the teacher has presented to the students, written or orally, from the students' interpretation of the very same. When we analyse a student's interpretation of a task we can describe the problem he/she is trying to solve. Problem is a term that describes what the students are trying to accomplish. Furthermore, the problem he or she is trying to solve can be interpreted in a wider sense by taking into account the student's aims and goals. By ascribing a student a plausible goal that he/she is trying to realise, defined here as project, the students' interpretation of the task becomes intelligible. Understanding what the students are doing is, therefore, looking for the meaning that they ascribe the task. Case studies Even though there are no clear-cut or absolute boundaries between different kinds of case studies, quantitative and qualitative case studies are often described as being exploratory, descriptive or explanatory (Yin, 1994). Deciding on what kind of case study can be constructively used in a particular research project inevitably entails questions of research design that have to do with the type of research question, the extent of control over actual behavioural events and to what extent the focus will be on contemporary as opposed to historical events. Stake (1994) distinguishes intrinsic case studies from instrumental and collective case studies. In an intrinsic case study the aim is to obtain knowledge about a particular case. An instrumental case study is used for enhancing understanding of an issue or for refining a theory. So, in an instrumental case study the case is of secondary interest since it is examined for the purpose of gaining knowledge about something else. When the interest is to theorise further about a phenomenon or a population, and the particular case is of less interest, the term "collective case studies" is used. Collective case studies thus serve the same purpose as instrumental case studies, but are numerous.The three case studies[2] in this thesis on students' experiences and learning in environmental education can be analysed in view of the above terminology. With an interest in elaborating a theory on learning, the cases can be seen as both instrumental and explorative. The term "explore" indicates that something is unknown and needs to be investigated and, considering previous research on environmental education, this seems highly appropriate. The research on students and environmental education has for the most part concerned students' attitudes, behaviours, knowledge and learning outcomes (Rickinson, 2001). However, the research presented here had as its main focus students' learning processes and interpretations of a particular course content. The case studies in the thesis vary with respect to the content to be learnt. The ecology course contains various natural science subjects, and the task on environmental reports that the biology students work with can be related to the social sciences, whereas the postgraduates' research work comprises aspects of both natural and social science.In the first case study lectures in a compulsory course in ecology (four credit points) taken by first-year civil engineering students were tape-recorded and observed. The course was given during the spring term of the academic year of 2000, from March to June, at the Royal Institute of Technology (KTH) in Stockholm. There were 103 students in the class, and six of these students, two women and four men, were selected for interview. The interviews lasted between 50 and 70 minutes. The interview questions concerned the students' reasons for studying an engineering programme and becoming a civil engineer, their conceptions of their first year at university and the ecology course in general - what they had learnt and how they related the course content to their future profession.In the second case study observations were made of the lectures in a course at Stockholm University. The course was a ten-credit point supplementary course, "Environmental control for biologists", conducted in 1998. This course, which is multi-disciplinary in character, is still offered by the university and adopts a social perspective on environmental issues with lectures on politics, economy, environment and technology, environmental law and international environmental control. At the end of the course the students were offered five different themes to work with. In the case study reported on in the thesis, a group of four students, three women and one man, chose to work on a task relating to "environmental reports". The students' group work was observed, hand-written notes were made of their activities and the discussions were tape-recorded.The third case study was conducted in 1999, also at KTH. Six postgraduate students, four women and two men, were interviewed concerning their environmental research, choice of topic and interest in pursuing postgraduate studies. Three of the students had just begun, or were in the process of planning for, their postgraduate studies and the other three intended to present their licentiate thesis within the coming year. The interviews lasted between 60 and 90 minutes. All of the interviews were transcribed in full.Interpretation of the data - understanding students' utterances and actions This research is based on the assumption that in order to understand students' interpretations and learning we must acknowledge the students' reasons for studying and their personal study aims. How do we account for that, and how is it done when analysing data? Von Wright (1971) has written about intentionality in action as a presupposition for understanding the meaning of peoples' actions. Without such a presupposition, action becomes mere behaviour, that is, muscular movement. Von Wright stated:Behaviour gets its intentional character from being seen by the agent himself or by an outside observer in a wider perspective, from being set in a context of aims and cognitions. This is what happens when we construe a practical inference to match it, as premises match a given conclusion (von Wright, 1971, pp. 114-15).A practical inference, or syllogism, implies that action is understood as teleological rather than causal. The "causes" of an action reside in the future and in the mind of the person we are trying to understand. The following may serve as an example of such a teleological inference:A person P intends to bring about x.P believes bringing about x would require the fulfilment of y.Hence: P does y (in order to bring about x).In the syllogism the first premise states the person's goal, what he/she intends, and the second premise states the ways and the means by which that person believes that the goal can be achieved. The premises, and the conclusion, are logically and not causally related.This kind of interpretation of what people are doing and saying is, in a sense, trivial since all of us in our daily lives constantly ascribe intentions to people's actions; if we see a man running towards a bus we assume that his intention is to get on the bus. If we know this man well enough, we can also make the assumption that he is running towards the bus in order to get to the train station and catch a particular train. This interpretation involves putting the act of running towards the bus within the more far-reaching goal of catching the train. Hallden (1982, 1988, 2001) and Wistedt (1987) use the term project to describe what students are doing and what they are trying to achieve when working in an educational setting. When analysing a student's interpretation of a task, it is possible to describe the problem that he/she is trying to solve. The problem that the student is working on can also be interpreted as part of a future goal, a project, for example, to pass a particular exam or to successfully complete a whole course or obtain a degree. Analysing data from such a project-oriented viewpoint means asking questions about what kinds of problems the students are working on. In light of which projects do the students study activities become meaningful? Such questions give the researcher a chance to understand what the students are saying and doing, and may reveal possible explanations as to why they interpreted a certain content or task in a certain way. The engineering students' interpretations can be described in terms of a theoretical context in the sense that the students related the content to concepts and different academic subjects within the natural sciences[3]. The content also raised issues concerning human beings in relation to nature. The students discussed the notion of humankind as being within or outside the eco-system, and this seems to have led them to consider the moral aspects of human actions and the effects of these actions on nature. This means that the students' interpretations of the course content can be said to be embedded within a cultural context of values and norms. The students seemed to believe that the teacher had, implicitly, expressed the viewpoint that human beings have affected nature only in a negative way and that the teacher had neglected to highlight the positive effect which that impact has had on humans. In talking about these aspects of values, the students showed irritation and indignation and seemed to find the whole issue quite provoking.In the analysis of the different ways in which the students interpreted the course content, the results show that during the lectures the engineering students posed questions to the teacher concerning possible solutions to various environmental problems and whether these solutions were good or bad and could eventually lead to improving the environment. Finally, the students also focused on different aspects of the course content, such as the water cycle; they also stated that learning about such matters as the chemical formulas of the phosphorus cycle was irrelevant for their future profession as civil engineers.The students' different contextualisations of the course content, as well as their questions and focused aspects of the content, were analysed from an intentional perspective. The students' actions and utterances become meaningful if we interpret them in relation to their future goals, i.e. their projects. Their interpretations and reactions to the course content become intelligible and possible to understand in view of the professional project of becoming a civil engineer. As civil engineers they will inevitably affect nature in some way, good or bad. An ecocentric perspective that places a higher value on nature than on human beings, together with a view that implies that the effect humans have on nature is harmful, may come into conflict with the very notion of being a civil engineer. The questions that were posed in the classroom concerning solutions to environmental problems can also be understood as being both part of a professional project and a possible project related to a more general interest in wanting to know what could be done to improve the environment. Finally, the two students who stated that certain aspects of the content were irrelevant were arguing with reference to the civil engineering profession. When the students were working on a task concerning environmental reports, different contextualisations as well as different problems were brought to the fore. The students decided to work on a problem about environmental reports as a general phenomenon and to collect reports from different companies and analyse and study the text they contained. While the students were working on that problem, another problem, this one concerning the environmental work of certain companies was brought to fore. In the students' discussion, this problem meant that the students analysed and discussed the companies' work as such, and not the texts presented in the reports. In particular, one of the students in the group, Hans, clarified and stressed this difference.The students' work on analysing the texts was contextualised into an academic and cultural context. By setting up 15 criteria for analysing the 14 reports, the students' aim was to describe the reports' content concerning the companies' environmental work and how these routines were presented to the reader. While solving the problem, the students' personal judgements about, for example, the environmental education given to the staff became a topic for discussion. One student in particular emphasised the importance of keeping judgements out of the discussion, and argued that the group should try to describe the reports in terms of "what" content was being presented and "how" it was being presented to the reader. He also stressed that the group should not evaluate the different aspects presented in the reports, for example, concerning environmental education, as being good or bad. Eventually, on the last day of working on the task, the students discussed the difficulty of trying to keep their values and judgements as well as their feelings about the different companies separate from the analysis.In analysing the students' interpretations of the task from an intentional perspective, the different contextualisations, as well as the different problems, can be discussed in relation to the students' projects. A possible project that can be ascribed the students is a concern for the environment and an interest in promoting and working towards a better environment. Whether this is a project of interest or of profession, or perhaps both, is difficult to say. The students' apparent interest in working with a problem that concerned the companies' business as such, and not just what was stated in the reports, is understandable in light of an environmental interest in finding out what is being done within the companies and if this will eventually improve the environment. The students' difficulty in differentiating between, on one hand, a cultural context of values and judgements about the companies' environmental work as it is described in the reports, and, on the other, an academic context can also be explained in relation to a project. If the students' project of interest is to promote environmental change, their will to not only describe environmental reports, but also analyse and judge them in terms of good and bad, becomes reasonable. For the six postgraduate students the task of writing a doctoral thesis raised different problems, which were possible to analyse in relation to the students' different projects. Firstly, the thesis was interpreted as an assignment to be attended to within the university and the scientific community. This problem can be understood as being part of the students' educational project. Second, for five of the six postgraduate students, the task of writing a thesis was commissioned by a company outside the university, and the research results were to serve as a means for companies to take action; writing a thesis was thus interpreted as a question of more or less producing a plan of action. Third, writing the thesis was a project of personal interest in the sense that the students felt they could contribute to improving the environment. Three of the students also stated that their postgraduate studies would benefit them in their professional work and that this was a contributory reason for their taking on postgraduate studies.In solving the problem of writing a doctoral thesis within the educational project, some difficulties arose. To be able to answer their research questions, two of the students, Elisabeth and Sofia, had to acquaint themselves with subject areas in the social sciences. These subjects were quite new to them, as were the qualitative methods they were expected to use in their research. They realised that knowledge about how to conduct qualitative research was scarce at KTH. Therefore, the students had to learn on their own and try to overcome this obstacle in different ways. Another student, Mikael, stressed that he had chosen to narrow down his research and exclude questions that would lead him too far afield into other subject areas. He argued that too wide a research question could not be handled in a doctoral research project.The field of environmental research evoked thoughts about identity and feelings of uncertainty. Elisabeth discussed the fact the her background was in civil engineering and that civil engineers traditionally had not engaged in the kind of environmental research that she was now conducting. Sofia, on the other hand, was not a civil engineer and was worried that her natural science background would pose problems. However, she stated that any student entering her research field and having to learn a new subject, in her case macroeconomics, would have been confronted with the same difficulties regardless of whether their backgrounds were in civil engineering or natural science.In working on her thesis, Elisabeth had come to realise that she and the Hydropower Company that had given her the assignment interpreted the concept of "sustainability" in different ways. She also discussed the fact that she had personal beliefs and values about the issues she was writing about and was worried that these would influence her research in an inappropriate way. Contextualisations and differentiation When comparing the students' interpretations of course content and different tasks, the results show that a diversity of contexts was brought to the fore. In all three cases the students interpreted a cultural context concerned with values. In confronting course content in ecology, the civil engineering students were provoked by the value judgement that human beings are generally harmful to nature. When the biology students solved the task concerning environmental reports, values and personal judgements about the various companies and their environmental development work were also brought to the fore. The third case study, finally, makes it clear that values and beliefs about the research topic can become an issue for postgraduate students in their doctoral research.What can be said about the students' learning as a process of differentiation? In the case study focusing on the civil engineering students' interpretations of the course content, the students did not talk about their values and perspective on nature as values. They, therefore, did not differentiate them as values in contrast with, for example, descriptions of nature. The biology students, on the other hand, discussed the fact that they were judging the companies on emotional grounds and tried to differentiate between an academic context, where they described the different companies, and a value-laden context.The postgraduate students' interpretations of the task to write a doctoral thesis brought a range of problems to the fore. One of the problems concerned viewing a thesis as a scientific product, and two of the students, Mikael and Sofia, discussed methodology and ways of approaching new subjects in the social sciences. Their statements can be analysed with a focus on tendencies to elaborate within and between different contexts. In Mikael's discussion on what methodologies can be used to investigate a certain phenomenon at an empirical level, stressing this as a matter of choice to consider in research may be described as a form of elaboration within a scientific context. Sofia's reasoning about the subject of political science and the fact that she thought it important to be cautious when using concepts within a new discipline suggests that she was aware that a meta-level defines concepts within different subjects.Elisabeth brought to the fore a difficulty that concerns values and how to deal with this aspect. She reflected on the fact that she had beliefs and different interpretations about the concept of sustainability. The fact that she had realised this and tried to distinguish between her personal definitions, definitions that would be appropriate to use in the scientific community and definitions embraced by her project sponsors, also indicates that a differentiation has taken place. The findings raise the question of how affects can influence the learning process, and the students' possibilities of differentiating between various conceptual frameworks. In relation to research on cognitive development as a process of conceptual change where the student abandons conceptions due to "dissatisfaction" (Posner et al., 1982, p. 214) and finds new conceptions "intelligible" and "plausible" (Posner et al., 1982, p. 214), it can be questioned if learning is such a rational and intellectual process. If we consider an intentional perspective when looking at learners and learning we can perhaps better understand the rationale and logic of their way of reasoning and understanding environmental issues.The results can be discussed in relation to educational practice and the content, as well as the purpose, of environmental education. If environmental education can raise these kinds of moral and ethical conceptions concerning nature and man's behaviour, as well as judgements about companies, ethics or discussions concerning values should be part of the curriculum or in some way highlighted and recognised. Perhaps the engineering students would have benefited from a discussion on different ethical perspectives on nature as well as the ethics of professional engineers and possible dilemmas they will meet in their future career. Furthermore, the students' focus on solving and finding solutions to environmental problems is an issue to consider in education. Whether the content includes aspects of the ways in which environmental problems are dealt with in society or not, this focus of interest will probably affect the learning process. Opens in a new window.Figure 1 Different contexts for interpretation
|
- Shows how differently environmental issues can be interpreted, i.e. scientifically, existentially and politically, and the way values and emotions become an aspect of the learning process and reveal the students' difficulties in differentiating between values and descriptions of phenomena. This is analysed and explained in relation to the students' various projects that come into conflict in the educational setting.
|
[SECTION: Value] Looking across Swedish higher education, the last few years have seen an increase in the number of environmental courses being offered at the undergraduate and postgraduate levels. According to a survey conducted by the National Agency for Higher Education (2001), 3 of 37 universities and university colleges offer introductory courses in "sustainable development" that are open to all students. Fourteen educational institutions offer specific environmental introductory courses to a select group of undergraduate students and another 16 universities give undergraduate courses that include environmental issues or sustainable development. Moreover, several Swedish universities offer programmes in "Environmental science" leading to a BSc or an MSc degree, and there are also a number of civil engineering programmes that focus on environmental engineering.In this paper, the results of which draw from a doctoral thesis (Lundholm, 2003), students' perspectives on environmental education are presented. It is based on three case studies in which the aim was to explore the students' conceptions of ecology and environmental issues, and the learning process. Learning is viewed from an intentional perspective, which means that the students' aims in their studies, here defined as "projects", are explored through analysis of their actions and utterances in educational settings or interviews. Previous research on learners and learning in environmental education has focused mainly on the learners' knowledge, attitudes and behaviour in relation to environmental phenomena and environmental problems[1]. A number of studies on young people's understanding of environmental problems were concerned with investigating their scientific knowledge (Gomez-Granell and Cervera-March, 1993; Gambro and Switzky, 1996; Ivy et al., 1998). Boyes and Stanisstreet (1993, 1994, 1998) have studied young people's knowledge of the greenhouse effect and ozone depletion in particular.Research on students' learning has focused on the effects of environmental education, and has often been conducted as pre-test and post-test studies. Only a few investigations have been concerned specifically with students' personal interpretations and experiences of environmental education (Lai, 1999; Rickinson, 1999) or with the learning process (Emmons, 1997; Mason and Santi, 1998; Grace and Ratcliffe, 2002).Research on subject-related knowledge in environmental education has had its primary focus on the natural sciences, whereas the social science disciplines have received little attention from researchers in this field (Rickinson, 2001).Within much of this work there is a very strong science education influence. The environment phenomena that have been investigated are ones from school curricula, the research participants are often from science teaching groups, and the implications drawn often take the form of strategies for science teachers (Rickinson, 2001, p. 219). The purpose of the study was to investigate and describe students' interpretations of various tasks concerning environmental issues and of course content in ecology. The purpose was also to investigate the students' learning, defined as a process of differentiation between contexts.The thesis attended to the following research questions:RQ1. How do six first-year civil engineering students interpret a course content in ecology and environmental issues; i.e. how do they contextualise the content? RQ2. How does a group of biology students interpret a task on "environmental reports?" What "problems" are brought to the fore? RQ3. How do six postgraduate students interpret environmental research? What "problems" are brought to the fore? In relation to previous research, the aim of the thesis was to explore the learning process, as opposed to learning outcomes, and the students' interpretations and experiences of environmental education. Learning as a process of differentiation Constructivism has been an established field in educational research ever since the pioneering work of Piaget (1929/1989) and his associates in the 1920s on children's conceptions of the world. Within this field, "conceptual change" has emerged as the dominant description of learning, implying that learning can be described as a process in which commonsense notions of the world are replaced by scientifically accepted ways of conceptualising the world (Posner et al., 1982). Critics of such a view have pointed out that learning is not a matter of replacing less qualified or naive conceptions of phenomena in the world, but is rather a question of understanding in what situations and within which genres different knowledge is useful and appropriate (Caravita and Hallden, 1994; Hallden, 1999; Hallden et al., 2002; Petersson, 2002). Learning is thus seen as a process of differentiating between various contexts.Wistedt (1994), Wistedt and Martinsson (1996), Wistedt et al. (1997) and Hallden (1999) describe the different ways in which students interpret and contextualise tasks assigned in classroom situations. Students' and teachers' interpretations that are brought to the fore in school settings cannot be described merely as being either commonsense or theoretical, but rather as belonging to different theoretical contexts representing different disciplines or genres (Wistedt, 1998). Students' alternative contextualisations can also be seen as expressions of values based on norms and values that are prevalent in society, and the context can therefore be characterised as cultural. An empirical example of this is given in a study by Hallden (1999). The study describes how biology students contextualise a task concerning the ecological relation between animals and plants as a problem pertaining to the extinction of species. The students' interpretation of the task becomes understandable when we consider the ongoing debate in today's society regarding the extinction of species and biological diversity. The students' cultural contextualisation of the task, placing it within an environmental discussion, raises value judgement issues, such as: which species should be saved and which ones should be left to become extinct?Within different academic disciplines, concepts are defined within the framework of the different subjects' theoretical perspectives on phenomena and data. In a theoretical context, for example, an academic discipline, there are theories and explanatory models at a meta-level that explain data and empirical findings at a lower level (Figure 1). In a study by Wistedt and Martinsson (1994) on how pupils solve mathematics tasks presented to them in a formal educational setting, the findings show how the pupils communicate and make explicit the premises for their interpretations of the task. When the pupils articulate their thoughts at a meta-level and relate theories to an empirical level, or relate phenomena and concepts to an empirical level, they are elaborating within a context. If different contexts are brought to the fore by the learners, and are articulated as different theories or premises for interpretations at a meta-level, this is described as elaborating between contexts.Task problem and project Hallden (1982, 1988, 2001) and Wistedt (1987) use the term project to describe what students are doing and aiming at achieving when working in educational settings. Often in education students are given tasks to work on and solve. Hallden (1982, 1988, 2001) and Wistedt (1987) have investigated the way students interpret tasks given to them as problems. Hallden defines task as "what is presented to the pupils by the teacher with the intention that they are to do something and/or to learn something" (1988, p. 125). The terms task and problem distinguishe what the teacher has presented to the students, written or orally, from the students' interpretation of the very same. When we analyse a student's interpretation of a task we can describe the problem he/she is trying to solve. Problem is a term that describes what the students are trying to accomplish. Furthermore, the problem he or she is trying to solve can be interpreted in a wider sense by taking into account the student's aims and goals. By ascribing a student a plausible goal that he/she is trying to realise, defined here as project, the students' interpretation of the task becomes intelligible. Understanding what the students are doing is, therefore, looking for the meaning that they ascribe the task. Case studies Even though there are no clear-cut or absolute boundaries between different kinds of case studies, quantitative and qualitative case studies are often described as being exploratory, descriptive or explanatory (Yin, 1994). Deciding on what kind of case study can be constructively used in a particular research project inevitably entails questions of research design that have to do with the type of research question, the extent of control over actual behavioural events and to what extent the focus will be on contemporary as opposed to historical events. Stake (1994) distinguishes intrinsic case studies from instrumental and collective case studies. In an intrinsic case study the aim is to obtain knowledge about a particular case. An instrumental case study is used for enhancing understanding of an issue or for refining a theory. So, in an instrumental case study the case is of secondary interest since it is examined for the purpose of gaining knowledge about something else. When the interest is to theorise further about a phenomenon or a population, and the particular case is of less interest, the term "collective case studies" is used. Collective case studies thus serve the same purpose as instrumental case studies, but are numerous.The three case studies[2] in this thesis on students' experiences and learning in environmental education can be analysed in view of the above terminology. With an interest in elaborating a theory on learning, the cases can be seen as both instrumental and explorative. The term "explore" indicates that something is unknown and needs to be investigated and, considering previous research on environmental education, this seems highly appropriate. The research on students and environmental education has for the most part concerned students' attitudes, behaviours, knowledge and learning outcomes (Rickinson, 2001). However, the research presented here had as its main focus students' learning processes and interpretations of a particular course content. The case studies in the thesis vary with respect to the content to be learnt. The ecology course contains various natural science subjects, and the task on environmental reports that the biology students work with can be related to the social sciences, whereas the postgraduates' research work comprises aspects of both natural and social science.In the first case study lectures in a compulsory course in ecology (four credit points) taken by first-year civil engineering students were tape-recorded and observed. The course was given during the spring term of the academic year of 2000, from March to June, at the Royal Institute of Technology (KTH) in Stockholm. There were 103 students in the class, and six of these students, two women and four men, were selected for interview. The interviews lasted between 50 and 70 minutes. The interview questions concerned the students' reasons for studying an engineering programme and becoming a civil engineer, their conceptions of their first year at university and the ecology course in general - what they had learnt and how they related the course content to their future profession.In the second case study observations were made of the lectures in a course at Stockholm University. The course was a ten-credit point supplementary course, "Environmental control for biologists", conducted in 1998. This course, which is multi-disciplinary in character, is still offered by the university and adopts a social perspective on environmental issues with lectures on politics, economy, environment and technology, environmental law and international environmental control. At the end of the course the students were offered five different themes to work with. In the case study reported on in the thesis, a group of four students, three women and one man, chose to work on a task relating to "environmental reports". The students' group work was observed, hand-written notes were made of their activities and the discussions were tape-recorded.The third case study was conducted in 1999, also at KTH. Six postgraduate students, four women and two men, were interviewed concerning their environmental research, choice of topic and interest in pursuing postgraduate studies. Three of the students had just begun, or were in the process of planning for, their postgraduate studies and the other three intended to present their licentiate thesis within the coming year. The interviews lasted between 60 and 90 minutes. All of the interviews were transcribed in full.Interpretation of the data - understanding students' utterances and actions This research is based on the assumption that in order to understand students' interpretations and learning we must acknowledge the students' reasons for studying and their personal study aims. How do we account for that, and how is it done when analysing data? Von Wright (1971) has written about intentionality in action as a presupposition for understanding the meaning of peoples' actions. Without such a presupposition, action becomes mere behaviour, that is, muscular movement. Von Wright stated:Behaviour gets its intentional character from being seen by the agent himself or by an outside observer in a wider perspective, from being set in a context of aims and cognitions. This is what happens when we construe a practical inference to match it, as premises match a given conclusion (von Wright, 1971, pp. 114-15).A practical inference, or syllogism, implies that action is understood as teleological rather than causal. The "causes" of an action reside in the future and in the mind of the person we are trying to understand. The following may serve as an example of such a teleological inference:A person P intends to bring about x.P believes bringing about x would require the fulfilment of y.Hence: P does y (in order to bring about x).In the syllogism the first premise states the person's goal, what he/she intends, and the second premise states the ways and the means by which that person believes that the goal can be achieved. The premises, and the conclusion, are logically and not causally related.This kind of interpretation of what people are doing and saying is, in a sense, trivial since all of us in our daily lives constantly ascribe intentions to people's actions; if we see a man running towards a bus we assume that his intention is to get on the bus. If we know this man well enough, we can also make the assumption that he is running towards the bus in order to get to the train station and catch a particular train. This interpretation involves putting the act of running towards the bus within the more far-reaching goal of catching the train. Hallden (1982, 1988, 2001) and Wistedt (1987) use the term project to describe what students are doing and what they are trying to achieve when working in an educational setting. When analysing a student's interpretation of a task, it is possible to describe the problem that he/she is trying to solve. The problem that the student is working on can also be interpreted as part of a future goal, a project, for example, to pass a particular exam or to successfully complete a whole course or obtain a degree. Analysing data from such a project-oriented viewpoint means asking questions about what kinds of problems the students are working on. In light of which projects do the students study activities become meaningful? Such questions give the researcher a chance to understand what the students are saying and doing, and may reveal possible explanations as to why they interpreted a certain content or task in a certain way. The engineering students' interpretations can be described in terms of a theoretical context in the sense that the students related the content to concepts and different academic subjects within the natural sciences[3]. The content also raised issues concerning human beings in relation to nature. The students discussed the notion of humankind as being within or outside the eco-system, and this seems to have led them to consider the moral aspects of human actions and the effects of these actions on nature. This means that the students' interpretations of the course content can be said to be embedded within a cultural context of values and norms. The students seemed to believe that the teacher had, implicitly, expressed the viewpoint that human beings have affected nature only in a negative way and that the teacher had neglected to highlight the positive effect which that impact has had on humans. In talking about these aspects of values, the students showed irritation and indignation and seemed to find the whole issue quite provoking.In the analysis of the different ways in which the students interpreted the course content, the results show that during the lectures the engineering students posed questions to the teacher concerning possible solutions to various environmental problems and whether these solutions were good or bad and could eventually lead to improving the environment. Finally, the students also focused on different aspects of the course content, such as the water cycle; they also stated that learning about such matters as the chemical formulas of the phosphorus cycle was irrelevant for their future profession as civil engineers.The students' different contextualisations of the course content, as well as their questions and focused aspects of the content, were analysed from an intentional perspective. The students' actions and utterances become meaningful if we interpret them in relation to their future goals, i.e. their projects. Their interpretations and reactions to the course content become intelligible and possible to understand in view of the professional project of becoming a civil engineer. As civil engineers they will inevitably affect nature in some way, good or bad. An ecocentric perspective that places a higher value on nature than on human beings, together with a view that implies that the effect humans have on nature is harmful, may come into conflict with the very notion of being a civil engineer. The questions that were posed in the classroom concerning solutions to environmental problems can also be understood as being both part of a professional project and a possible project related to a more general interest in wanting to know what could be done to improve the environment. Finally, the two students who stated that certain aspects of the content were irrelevant were arguing with reference to the civil engineering profession. When the students were working on a task concerning environmental reports, different contextualisations as well as different problems were brought to the fore. The students decided to work on a problem about environmental reports as a general phenomenon and to collect reports from different companies and analyse and study the text they contained. While the students were working on that problem, another problem, this one concerning the environmental work of certain companies was brought to fore. In the students' discussion, this problem meant that the students analysed and discussed the companies' work as such, and not the texts presented in the reports. In particular, one of the students in the group, Hans, clarified and stressed this difference.The students' work on analysing the texts was contextualised into an academic and cultural context. By setting up 15 criteria for analysing the 14 reports, the students' aim was to describe the reports' content concerning the companies' environmental work and how these routines were presented to the reader. While solving the problem, the students' personal judgements about, for example, the environmental education given to the staff became a topic for discussion. One student in particular emphasised the importance of keeping judgements out of the discussion, and argued that the group should try to describe the reports in terms of "what" content was being presented and "how" it was being presented to the reader. He also stressed that the group should not evaluate the different aspects presented in the reports, for example, concerning environmental education, as being good or bad. Eventually, on the last day of working on the task, the students discussed the difficulty of trying to keep their values and judgements as well as their feelings about the different companies separate from the analysis.In analysing the students' interpretations of the task from an intentional perspective, the different contextualisations, as well as the different problems, can be discussed in relation to the students' projects. A possible project that can be ascribed the students is a concern for the environment and an interest in promoting and working towards a better environment. Whether this is a project of interest or of profession, or perhaps both, is difficult to say. The students' apparent interest in working with a problem that concerned the companies' business as such, and not just what was stated in the reports, is understandable in light of an environmental interest in finding out what is being done within the companies and if this will eventually improve the environment. The students' difficulty in differentiating between, on one hand, a cultural context of values and judgements about the companies' environmental work as it is described in the reports, and, on the other, an academic context can also be explained in relation to a project. If the students' project of interest is to promote environmental change, their will to not only describe environmental reports, but also analyse and judge them in terms of good and bad, becomes reasonable. For the six postgraduate students the task of writing a doctoral thesis raised different problems, which were possible to analyse in relation to the students' different projects. Firstly, the thesis was interpreted as an assignment to be attended to within the university and the scientific community. This problem can be understood as being part of the students' educational project. Second, for five of the six postgraduate students, the task of writing a thesis was commissioned by a company outside the university, and the research results were to serve as a means for companies to take action; writing a thesis was thus interpreted as a question of more or less producing a plan of action. Third, writing the thesis was a project of personal interest in the sense that the students felt they could contribute to improving the environment. Three of the students also stated that their postgraduate studies would benefit them in their professional work and that this was a contributory reason for their taking on postgraduate studies.In solving the problem of writing a doctoral thesis within the educational project, some difficulties arose. To be able to answer their research questions, two of the students, Elisabeth and Sofia, had to acquaint themselves with subject areas in the social sciences. These subjects were quite new to them, as were the qualitative methods they were expected to use in their research. They realised that knowledge about how to conduct qualitative research was scarce at KTH. Therefore, the students had to learn on their own and try to overcome this obstacle in different ways. Another student, Mikael, stressed that he had chosen to narrow down his research and exclude questions that would lead him too far afield into other subject areas. He argued that too wide a research question could not be handled in a doctoral research project.The field of environmental research evoked thoughts about identity and feelings of uncertainty. Elisabeth discussed the fact the her background was in civil engineering and that civil engineers traditionally had not engaged in the kind of environmental research that she was now conducting. Sofia, on the other hand, was not a civil engineer and was worried that her natural science background would pose problems. However, she stated that any student entering her research field and having to learn a new subject, in her case macroeconomics, would have been confronted with the same difficulties regardless of whether their backgrounds were in civil engineering or natural science.In working on her thesis, Elisabeth had come to realise that she and the Hydropower Company that had given her the assignment interpreted the concept of "sustainability" in different ways. She also discussed the fact that she had personal beliefs and values about the issues she was writing about and was worried that these would influence her research in an inappropriate way. Contextualisations and differentiation When comparing the students' interpretations of course content and different tasks, the results show that a diversity of contexts was brought to the fore. In all three cases the students interpreted a cultural context concerned with values. In confronting course content in ecology, the civil engineering students were provoked by the value judgement that human beings are generally harmful to nature. When the biology students solved the task concerning environmental reports, values and personal judgements about the various companies and their environmental development work were also brought to the fore. The third case study, finally, makes it clear that values and beliefs about the research topic can become an issue for postgraduate students in their doctoral research.What can be said about the students' learning as a process of differentiation? In the case study focusing on the civil engineering students' interpretations of the course content, the students did not talk about their values and perspective on nature as values. They, therefore, did not differentiate them as values in contrast with, for example, descriptions of nature. The biology students, on the other hand, discussed the fact that they were judging the companies on emotional grounds and tried to differentiate between an academic context, where they described the different companies, and a value-laden context.The postgraduate students' interpretations of the task to write a doctoral thesis brought a range of problems to the fore. One of the problems concerned viewing a thesis as a scientific product, and two of the students, Mikael and Sofia, discussed methodology and ways of approaching new subjects in the social sciences. Their statements can be analysed with a focus on tendencies to elaborate within and between different contexts. In Mikael's discussion on what methodologies can be used to investigate a certain phenomenon at an empirical level, stressing this as a matter of choice to consider in research may be described as a form of elaboration within a scientific context. Sofia's reasoning about the subject of political science and the fact that she thought it important to be cautious when using concepts within a new discipline suggests that she was aware that a meta-level defines concepts within different subjects.Elisabeth brought to the fore a difficulty that concerns values and how to deal with this aspect. She reflected on the fact that she had beliefs and different interpretations about the concept of sustainability. The fact that she had realised this and tried to distinguish between her personal definitions, definitions that would be appropriate to use in the scientific community and definitions embraced by her project sponsors, also indicates that a differentiation has taken place. The findings raise the question of how affects can influence the learning process, and the students' possibilities of differentiating between various conceptual frameworks. In relation to research on cognitive development as a process of conceptual change where the student abandons conceptions due to "dissatisfaction" (Posner et al., 1982, p. 214) and finds new conceptions "intelligible" and "plausible" (Posner et al., 1982, p. 214), it can be questioned if learning is such a rational and intellectual process. If we consider an intentional perspective when looking at learners and learning we can perhaps better understand the rationale and logic of their way of reasoning and understanding environmental issues.The results can be discussed in relation to educational practice and the content, as well as the purpose, of environmental education. If environmental education can raise these kinds of moral and ethical conceptions concerning nature and man's behaviour, as well as judgements about companies, ethics or discussions concerning values should be part of the curriculum or in some way highlighted and recognised. Perhaps the engineering students would have benefited from a discussion on different ethical perspectives on nature as well as the ethics of professional engineers and possible dilemmas they will meet in their future career. Furthermore, the students' focus on solving and finding solutions to environmental problems is an issue to consider in education. Whether the content includes aspects of the ways in which environmental problems are dealt with in society or not, this focus of interest will probably affect the learning process. Opens in a new window.Figure 1 Different contexts for interpretation
|
- Shows how differently environmental issues can be interpreted, i.e. scientifically, existentially and politically, and the way values and emotions become an aspect of the learning process and reveal the students' difficulties in differentiating between values and descriptions of phenomena. This is analysed and explained in relation to the students' various projects that come into conflict in the educational setting.
|
[SECTION: Purpose] Service firms are searching for ways to deliver higher levels of information technology (IT) service as well as demanding more from their information systems (IS) groups, expecting quick responses to new business opportunities, to support critical processes efficiently and to satisfy their customers and internal staff. Hence there is a growing interest in adopting IT service management (ITSM) and best practices such as the Information Technology Infrastructure Library (ITIL) (Galup et al., 2009).ITSM is defined as "the implementation and management of Quality IT Services that meet the needs of the business" (ITIL, 2007). ITSM is a discipline for managing IT service operations, which is process orientated, sharing this theme with other quality improvement methods and as such it is different from traditional technology-oriented approaches due to its focus on customer relationships (Galup et al., 2009). A main benefit reported by ITSM adopters is an increase in customer satisfaction (Cater-Steel et al., 2009). ITSM concepts are often implemented through the ITIL framework; a sector specific quality and best practice guide on the management of information technologies (Iden, 2012). The operational benefits expected from ITIL implementations are the alignment of IT services with business needs, improved quality of the IT services themselves, and a reduction in the long term costs of service provision (McNaughton et al., 2010).One of the main challenges faced by companies during the adoption of ITIL is the organizational change required to attain a service-oriented culture (Cater-Steel et al., 2007; Spafford and Holup, 2010). Various practices to facilitate organizational change are proposed in the academic and professional literature (Buchanan et al., 2005; Kanter, 2001; Kerber, 2001; Kotter, 1995; Spafford and Holup, 2010), and most of this literature argues that the use of change management practices (CMPs) has a positive effect on the speed and quality of the change process and on results for the organization. Nevertheless, there is little published in terms of these practices as they relate to ITIL implementations.Therefore, this paper presents the findings of a multiple case study investigation, which examines the experience of four Peruvian financial companies implementing the ITIL framework. The cases offer insight into how the organizational change management processes supported the ITIL implementation and identifies which organizational CMPs were more frequently used.The paper is organized as follows; next Section 2 introduces the ITIL framework, followed by Section 3, a review of the literature on CMPs concluding with the study's research questions. Section 4 provides details of the methodology used. Section 5 provides the case profiles, the findings and the cross-case analysis. Finally, the conclusion sums up the findings, presents the studies limitations and offers a direction for future research. ITIL is a framework that describes best practices and provides guidance on the management of IT processes, functions, roles and responsibilities in ITSM. It was first developed in the 1980s by the UK Government agency, the Office of Government Commerce (OGC) in response to the government's growing dependence on information technologies (Galup et al., 2009), to promote efficient IT operations and to improve IT service delivery and operations within government controlled computing centers (Salle, 2004).Version one of ITIL consisted of 40 volumes covering "best practices" in different areas of IT service provision. In 2000, ITIL version one was replaced by the seven volume ITIL version two (ITILv2) consolidating the practices within an overall framework. The two primary components of ITILv2 are service delivery and service support which consist of ten core processes and the service desk function (Table I).In 2007, ITILv2 was further enhanced to become ITIL version three (ITILv3), now consisting of five volumes: strategy, design, transition, operation and continuous improvement, which extend the ITILv2 model and processes by organizing them around a service lifecycle model.2.1 ITIL adoption and implementation The ITILv2 has become the de facto framework to implement the ITSM concepts (Cartlidge et al., 2007; Cater-Steel et al., 2009) despite the recent release of ITILv3 (Pollard and Cater-Steel, 2009).ITIL defines a glossary to ensure a common vocabulary and defines a set of conceptual processes to outline best practices (Galup et al., 2009). ITIL does not specify how the implementation should be performed and recent studies on the adoption of ITIL in organizations have focused mainly on the verification of the benefits of the program and the study of the implementations themselves, including success factors and barriers.In terms of benefits, Hochstein et al. (2005) found, an increase in customer orientation and service and the efficiency and transparency of IT processes the main benefits. They also suggest that the biggest challenge in the adoption of ITIL is resistance in the form of reluctance to accept new processes and a lack of understanding of why these processes are necessary. Cater-Steel et al. (2006) identified that ITIL implementation provided firms with a more predictable infrastructure, improved consultations between the IT groups within the organization, agility in negotiating service level agreements and consistent services; also identifying that cultural change is a significant challenge when adopting ITIL. Cater-Steel and Tan (2005a, b) found that the IT staffs skills, adaptability, and quality of training were considered more important for successful ITIL implementation than issues related to clients, external consultants and technology. The importance of IT staff adaptability and change management was in turn identified in a subsequent investigation by Cater-Steel et al. (2007).In terms of ITIL implementation, Tan et al. (2009) identified that a change management strategy that aims to transform the organizational culture towards being service-oriented, is a critical success factor. While, Pollard and Cater-Steel (2009) reported the difficulty of changing the culture of the organization when adopting the new ITIL processes and they identified three further critical success factors for ITIL implementations; creating an ITIL friendly culture, giving priority to the processes and the design of customer-focused metrics. Cater-Steel and McBride (2007) findings highlight the importance of involving staff and the promoting interest in ITIL as a way of overcoming resistance to the new processes; facilitating an improved understanding of them. Spafford and Holub (2010) found that resistance to organizational culture change is a major impediment in implementing ITIL, such that improved processes are achieved in less time and less cost when a formal change management plan is implemented than for projects where the organizational change component is neglected. Kerber (2001) identifies three approaches used in organizations to enact organizational change:1. Direct change, which is motivated from senior management, wielded by authority and achieved through compliance.2. Planned change, which originates from any level in the organization but is managed through the top management layer and that seeks the commitment and involvement of the organization through CMPs that mitigate resistance.3. Guided change, arising from the within the organization and the through commitment and the contribution of staff toward the organizational objectives.Kerber and Buono (2005) argue that the effectiveness of the approach to the management of change depends on contextual variables such as:* the complexity of the business environment;* socio-technical uncertainty of the tasks or problems;* the ability to change the organization; and* the risks associated with the alternatives of not changing or changing slowly.Other factors may add complexity to the management of change, for example the cultural context (Martinsons et al., 2009) and the interaction between different agents of change namely; senior management, middle management, external consultants and work teams who have different experiences and perspectives (Caldwell, 2003; Choi et al., 2011). On the other hand, management support is considered key to the success of organizational change (Abraham et al., 1999; Spafford and Holup, 2010), as is the allocation of adequate resources (Becker, 2010). Although, following a change program is not a guarantee of successful outcomes (Choi et al., 2011).Kotter (2007) presents a set of eight sequential CMPs that allow for successful organizational change. Kotter's approach is well known and has been applied frequently (Aiken and Keller, 2009; Smith, 2011) (Table II).CMPs are commonly proposed in temporal sequences such as preparation, implementation and consolidation (Raineri, 2011). Research has found that change agents show a preference for CMPs related to the preparation stage because; it is easier to prevent errors in early stages, which lessens implementation failure at later stages (Holt et al., 2007); they have the ability to place more resources at the beginning of a change project (Raineri, 2011) and; the preparation stage requires more analytical skills of which managers tend to have more training in, compared with the more political and interpersonal skills required in the later stages of implementation (Shipper, 1999). Similarly, Raineri (2011) found a higher frequency of preparation practices in comparison with the implementation stage and that the use of CMPs has a significant impact on objectives and deadlines achievement. Abraham et al. (1999) found that the key factor in achieving change towards a culture of quality is top management support, with the lesser factors of the clarity of vision, participation, communication and appropriate resources. Becker (2010) found that the understanding the need for change, the level of organizational support and training, measuring the change, the positive experience and informal support, a history of organizational changes and the outlook previous expectations and individual feelings are factors that influence adoption.Considering the lack of focused research on ITIL implementation and managed change it would seem useful to investigate this phenomenon by studying firms that have implemented the ITIL framework and assess which organizational CMPs more frequently used and whether their influence on the ITIL implementation outcomes can be observed. Therefore, two research questions were formulated for this study:RQ1. How did financial companies in Peru that have implemented the ITIL framework use organizational CMPs?RQ2. How are the levels of ITIL compliance achieved by financial companies in Peru influenced by the CMPs that were used? The study employed the multiple-case design method proposed by Yin (2009) as the principal method to gather data to answer the research questions. The case study is an appropriate research method to analyze a phenomenon in its natural environment when the researcher has no influence over events and when the context is considered relevant (Benbasat et al., 1987; Yin, 2009). The case study methodology allows for in-depth questioning and the capture of important aspects of the complexity of an organization. But it is recognized that the conclusions of an investigation by the case study method can be limited, with the generalization of the findings being a common criticism (Gable, 2010; Lee, 1989). However, the problems of validity and generalisability can be addressed by using competent research methodologies (Meredith, 1998; Eisenhardt, 1989).The case study method was considered appropriate for this study as it allowed the phenomenon, the ITIL implementation, to be studied in its natural setting (Meredith, 1998), to have questions of the "how and why" type posed and answered (Yin, 2009) and it is an appropriate methodology for studying new topic areas (Eisenhardt, 1989). The design of the investigation involved four sites each studied in depth using multiple data sources and a range of collection methods (Eisenhardt, 1989).A single industry was selected for the study, in order to control the influence of contexts and for different environments and resource availabilities across industries. Four companies from the Peruvian financial sector that had implemented the ITIL framework were identified. The focus of this study is on the ITILv2 implementation for while the ITILv3 * was recently released, ITIL v2 is still the most commonly used (Pollard and Cater-Steel, 2009). The four companies had each implemented the 11 processes of ITILv2 and in all the cases, the ITIL implementation projects were supported by management and had been allocated sufficient of resources for completion.Data regarding the implementation were collected through separate, semi-structured interviews with various staff involved in the project including, top management, operations and IT managers and users of the IT services. Confirmatory data were gathered through secondary sources such as internal documents and progress reports.Through a review of the literature on research related to ITIL adoption and practice of organizational change management a list of questions based on the change management framework of Kotter (2007) was developed. The survey instrument was developed in the Spanish language by native Spanish speakers and all of the respondents speak and write in Spanish as a normal part of their work. The interviewees were given the set of open-ended questions about their strategy, motivations, problems and use CMPs used during the ITIL implementation, before the interview. During the interview, a discussion following the question format was initiated with the answers to these questions recorded digitally with additional notes taken by the researchers. Raineri (2011) found a likely bias from interviewees who were also project implementers that they would be more inclined to say they had used CMPs. Therefore, to reduce this bias the data were corroborated with other participants of the ITIL implementation.The interviews were transcribed verbatim and underwent content analysis. Responses were coded against the question structure and stages of CMPs implementation following the seven steps allowing the researchers to gauge CMP usage for each case. To guard against bias in the data analysis, each interview transcript was content analyzed twice to verify the results. Each case was then written up and subjected to a cross-case analysis, identifying the similarities and noting differences in themes and patterns across the cases.To determine the level ITIL compliance the researchers selected the itSMF self-assessment tool (itSMF, 2003). The itSMF instrument compares an organizations performance with the ITIL best practices, assessing the level of compliance with the ITIL processes, producing its results based on a framework of capabilities through a five point scale (McNaughton et al., 2010). Case firm IT managers were each given an itSMF questionnaire to complete. The questionnaire results provided each case with a score, between 0 and 5 for each of the ITIL core component's compliance to best practice, enabling case comparison. 5.1 Case profiles The cases sites are four financial institutions operating in Peru coded as A, B, C and D. Table III presents the profiles of the case sites and data on their ITIL implementations. All four cases confirm there was support of management and commitment of adequate resources to complete the implementations. All cases placed an emphasis on training and development during the project implementations. Cases C and D opted to use external consultants as part of their ITIL implementations. The cases chose phased execution as their preferred implementation method.5.2 Case results Table IV presents data on the frequency of the use of the seven CMPs. The four cases show relatively high use of change practices in the preparation phases, with fewer practices being used in the later stages of implementation and consolidation. Case D seemed to have institutional features that supported the use across the phases of the ITIL implementation. For example, case D has internal policies and practices that led to a significantly higher use of the empowerment practice than all of the other cases combined. In contrast case B showed the least commitment in terms of resource allocation and staff stimulus in later phases of the change process. In general there is one group where there seems to be stronger institutionalization (cases C and D), and another group (cases A and B) where there is less evidence of change practices consistently being used.The division between the two groups is equally apparent when comparing the case ITIL compliance scores. Table V presents the ITIL compliance scores, on a scale of 0-5. Table V shows that overall cases C and D achieved higher ITIL compliances than cases A and B. Case D consistently scores high across the categories, especially in those processes involving customer contacts such as service desk and incident management. Both cases A and B tend to score lower in the service support level functions despite having moderate scores for their ITSM level.5.3 Cross-case analysis In general, the analysis of the case data shows that the preparation stage practices are the more frequently change practices used by these Peruvian financial firms. Figure 1 shows the frequency of the CMPs employed in each stage across the cases. The most frequent are: "forming a powerful guiding coalition", "creating a vision" and "planning for and creating short-term wins", and to a lesser extent and "communicating the vision". The practice "establishing a sense of urgency" was used significantly more by cases C and D. The CMPs less frequently used were "institutionalizing new approaches", "consolidating improvements and producing still more change" and "empowering others to act on the vision". The two groups found in the case analysis are easily seen in Figure 1, with cases A and B providing little evidence of activities to communicate, empower and progress the implementations in terms of using CMPs.The most frequently used practices are related to the change preparation stage in contrast to the implementation stage practices. This result is consistent with Raineri (2001, 2011) findings, where the agents of change showed a preference to the use of CMPs related to the preparation stage.However, the results also show a key implementation practice "planning for and creating short-term wins" is frequently used. This may be due to the implementation method chosen by the cases. All cases used the phased implementation approach that may have facilitated the use of this CMP - for as one respondent stated "successful results on the initial phases of the implementation were used to motivate the implementation of the rest of the processes". Spafford and Holup (2010) also suggest the phased strategy provides benefits sooner and thus contributes to a gradual reduction of resistance to change.The compliance data reveals that in general, the greater the use of CMPs the higher the level of ITIL compliance achieved, suggesting that the use of CMPs has a positive influence on the compliance outcomes. Figure 2 shows the score for ITIL implementation compliance against the number of CMPs used on each process implementation. Cases D and C score high in change practice use and the levels of ITIL compliance, with cases A and B that used less practices have lower compliance scores.Cases A and B that have a low use of implementation and consolidation phase practices also reported change resistance problems. This observation is consistent with Siegal et al. (1996) who reported that resistance to change problems can be attributed to deficiencies in the management of the change process or in part are a result of the actions and inactions of change agents (Ford et al., 2008; Tavakoli, 2010).There are some institutional factors that have some bearing on the patterns revealed in the cross-case analysis. First, the use of outside consultants may have had a significant impact on project success. Pollard and Cater-Steel (2009) attribute the use of consultants to be a critical success factor for ITIL implementations. That cases C and D opted for consultants may have strengthened their ability to achieve higher compliance scores and to undertake more CMPs. Second, the formation of teams and leadership of the project was undertaken differently across the cases. At cases A and B the team of leaders was formed by those at the functional level, while at cases C and D the leaders involved tended to have more influence within the organization. While it is difficult to directly attribute this factor to the lower compliance scores, the demonstrated the support of senior managers to uphold change initiatives and reduce change resistance is well documented (Spafford and Holup, 2010; Abraham et al., 1999; Kotter, 2007). This study aimed to investigate the importance of applying managed change as part of an ITIL implementation. To this end four Peruvian financial sector firms were chosen as cases and a multi case study investigation was undertaken. The study undertook to ascertain which CMPs were being used and their level of influence on ITIL compliance.This study finds that not all firms take full advantage of CMPs while implementing the ITIL framework. The cases showed that higher levels of CMPs achieved higher levels of ITIL implementation compliance. This result suggests that the use of CMPs has a positive impact on the outcomes of ITIL implementations. It was also found that the firms that used less CMPs reported the resistance to change during the implementation. However, it important not to discount that the higher scoring firms may have had valuable or rare resources and competencies which they utilized in times of change giving them an advantage over others (Barney, 1991).The case studies also demonstrate that more change practices related to preparation and fewer change practices related to implementation and consolidation were used during the ITIL implementation process. This result may be attributable to institutional strengths of the firm, for as case D showed, it seemed to have institutional policies and practices that supported the ITIL implementation across all of its phases, and case B seemingly having the least commitment in terms of resource allocation, sustainable change and staff involvement. Likewise, the participation of an external consultant during the adoption of ITIL may favor the more frequent use of practices related to all three change stages (Pollard and Cater-Steel, 2009) allowing those firms that used consultants to use more CMPs. In addition, the choice of implementation methodology may contribute to some practices being more evident than others. All of the four case studies presented applied a phased implementations; a strategy, which favors the use of the change practice the "generation of short-term results".The main limitation of this study is the small number of cases analyzed, so the findings should be viewed with caution. The cases support the theory that managed change assists the implementation of ITIL processes. However, it was undertaken in a single industry, which tends to have homogenous processes and systems and the firms were all of a similar size. However, when examining cases of different size, system processes and industries divergent results may be produced, but these results are necessary to properly extend the case theory in question (Meredith, 1998). Future research may wish to extend the themes of this paper to design and undertake research on which CMPs provide the most impact to the levels of ITIL compliance using a suitable quantitative research instrument and method.In conclusion this paper makes three contributions. First it provides an exploratory study of the links between ITIL implementations and the use of CMPs. Second in terms of results it provides evidence of a link between ITIL implementation and managed change, which is important information for practitioners who may be planning future ITIL projects. Finally, this study adds to the literature of ITIL implementations, particularly with respect to organizational change. Opens in a new window.Figure 1 Use of CMPs Opens in a new window.Figure 2 CMPs and levels of ITIL compliance Opens in a new window.Table I Core components of ITILv2 Opens in a new window.Table II Change management practices Opens in a new window.Table III Profile of cases Opens in a new window.Table IV Change practices and frequency of use Opens in a new window.Table V ITIL compliance scores
|
- Although various practices to facilitate organizational change are proposed in the literature, very little is known about how these practices impact on ITIL adoptions. Therefore, the purpose of this study is to address this gap in the literature by offering insights on how change management practices impact on levels of ITIL compliance.
|
[SECTION: Method] Service firms are searching for ways to deliver higher levels of information technology (IT) service as well as demanding more from their information systems (IS) groups, expecting quick responses to new business opportunities, to support critical processes efficiently and to satisfy their customers and internal staff. Hence there is a growing interest in adopting IT service management (ITSM) and best practices such as the Information Technology Infrastructure Library (ITIL) (Galup et al., 2009).ITSM is defined as "the implementation and management of Quality IT Services that meet the needs of the business" (ITIL, 2007). ITSM is a discipline for managing IT service operations, which is process orientated, sharing this theme with other quality improvement methods and as such it is different from traditional technology-oriented approaches due to its focus on customer relationships (Galup et al., 2009). A main benefit reported by ITSM adopters is an increase in customer satisfaction (Cater-Steel et al., 2009). ITSM concepts are often implemented through the ITIL framework; a sector specific quality and best practice guide on the management of information technologies (Iden, 2012). The operational benefits expected from ITIL implementations are the alignment of IT services with business needs, improved quality of the IT services themselves, and a reduction in the long term costs of service provision (McNaughton et al., 2010).One of the main challenges faced by companies during the adoption of ITIL is the organizational change required to attain a service-oriented culture (Cater-Steel et al., 2007; Spafford and Holup, 2010). Various practices to facilitate organizational change are proposed in the academic and professional literature (Buchanan et al., 2005; Kanter, 2001; Kerber, 2001; Kotter, 1995; Spafford and Holup, 2010), and most of this literature argues that the use of change management practices (CMPs) has a positive effect on the speed and quality of the change process and on results for the organization. Nevertheless, there is little published in terms of these practices as they relate to ITIL implementations.Therefore, this paper presents the findings of a multiple case study investigation, which examines the experience of four Peruvian financial companies implementing the ITIL framework. The cases offer insight into how the organizational change management processes supported the ITIL implementation and identifies which organizational CMPs were more frequently used.The paper is organized as follows; next Section 2 introduces the ITIL framework, followed by Section 3, a review of the literature on CMPs concluding with the study's research questions. Section 4 provides details of the methodology used. Section 5 provides the case profiles, the findings and the cross-case analysis. Finally, the conclusion sums up the findings, presents the studies limitations and offers a direction for future research. ITIL is a framework that describes best practices and provides guidance on the management of IT processes, functions, roles and responsibilities in ITSM. It was first developed in the 1980s by the UK Government agency, the Office of Government Commerce (OGC) in response to the government's growing dependence on information technologies (Galup et al., 2009), to promote efficient IT operations and to improve IT service delivery and operations within government controlled computing centers (Salle, 2004).Version one of ITIL consisted of 40 volumes covering "best practices" in different areas of IT service provision. In 2000, ITIL version one was replaced by the seven volume ITIL version two (ITILv2) consolidating the practices within an overall framework. The two primary components of ITILv2 are service delivery and service support which consist of ten core processes and the service desk function (Table I).In 2007, ITILv2 was further enhanced to become ITIL version three (ITILv3), now consisting of five volumes: strategy, design, transition, operation and continuous improvement, which extend the ITILv2 model and processes by organizing them around a service lifecycle model.2.1 ITIL adoption and implementation The ITILv2 has become the de facto framework to implement the ITSM concepts (Cartlidge et al., 2007; Cater-Steel et al., 2009) despite the recent release of ITILv3 (Pollard and Cater-Steel, 2009).ITIL defines a glossary to ensure a common vocabulary and defines a set of conceptual processes to outline best practices (Galup et al., 2009). ITIL does not specify how the implementation should be performed and recent studies on the adoption of ITIL in organizations have focused mainly on the verification of the benefits of the program and the study of the implementations themselves, including success factors and barriers.In terms of benefits, Hochstein et al. (2005) found, an increase in customer orientation and service and the efficiency and transparency of IT processes the main benefits. They also suggest that the biggest challenge in the adoption of ITIL is resistance in the form of reluctance to accept new processes and a lack of understanding of why these processes are necessary. Cater-Steel et al. (2006) identified that ITIL implementation provided firms with a more predictable infrastructure, improved consultations between the IT groups within the organization, agility in negotiating service level agreements and consistent services; also identifying that cultural change is a significant challenge when adopting ITIL. Cater-Steel and Tan (2005a, b) found that the IT staffs skills, adaptability, and quality of training were considered more important for successful ITIL implementation than issues related to clients, external consultants and technology. The importance of IT staff adaptability and change management was in turn identified in a subsequent investigation by Cater-Steel et al. (2007).In terms of ITIL implementation, Tan et al. (2009) identified that a change management strategy that aims to transform the organizational culture towards being service-oriented, is a critical success factor. While, Pollard and Cater-Steel (2009) reported the difficulty of changing the culture of the organization when adopting the new ITIL processes and they identified three further critical success factors for ITIL implementations; creating an ITIL friendly culture, giving priority to the processes and the design of customer-focused metrics. Cater-Steel and McBride (2007) findings highlight the importance of involving staff and the promoting interest in ITIL as a way of overcoming resistance to the new processes; facilitating an improved understanding of them. Spafford and Holub (2010) found that resistance to organizational culture change is a major impediment in implementing ITIL, such that improved processes are achieved in less time and less cost when a formal change management plan is implemented than for projects where the organizational change component is neglected. Kerber (2001) identifies three approaches used in organizations to enact organizational change:1. Direct change, which is motivated from senior management, wielded by authority and achieved through compliance.2. Planned change, which originates from any level in the organization but is managed through the top management layer and that seeks the commitment and involvement of the organization through CMPs that mitigate resistance.3. Guided change, arising from the within the organization and the through commitment and the contribution of staff toward the organizational objectives.Kerber and Buono (2005) argue that the effectiveness of the approach to the management of change depends on contextual variables such as:* the complexity of the business environment;* socio-technical uncertainty of the tasks or problems;* the ability to change the organization; and* the risks associated with the alternatives of not changing or changing slowly.Other factors may add complexity to the management of change, for example the cultural context (Martinsons et al., 2009) and the interaction between different agents of change namely; senior management, middle management, external consultants and work teams who have different experiences and perspectives (Caldwell, 2003; Choi et al., 2011). On the other hand, management support is considered key to the success of organizational change (Abraham et al., 1999; Spafford and Holup, 2010), as is the allocation of adequate resources (Becker, 2010). Although, following a change program is not a guarantee of successful outcomes (Choi et al., 2011).Kotter (2007) presents a set of eight sequential CMPs that allow for successful organizational change. Kotter's approach is well known and has been applied frequently (Aiken and Keller, 2009; Smith, 2011) (Table II).CMPs are commonly proposed in temporal sequences such as preparation, implementation and consolidation (Raineri, 2011). Research has found that change agents show a preference for CMPs related to the preparation stage because; it is easier to prevent errors in early stages, which lessens implementation failure at later stages (Holt et al., 2007); they have the ability to place more resources at the beginning of a change project (Raineri, 2011) and; the preparation stage requires more analytical skills of which managers tend to have more training in, compared with the more political and interpersonal skills required in the later stages of implementation (Shipper, 1999). Similarly, Raineri (2011) found a higher frequency of preparation practices in comparison with the implementation stage and that the use of CMPs has a significant impact on objectives and deadlines achievement. Abraham et al. (1999) found that the key factor in achieving change towards a culture of quality is top management support, with the lesser factors of the clarity of vision, participation, communication and appropriate resources. Becker (2010) found that the understanding the need for change, the level of organizational support and training, measuring the change, the positive experience and informal support, a history of organizational changes and the outlook previous expectations and individual feelings are factors that influence adoption.Considering the lack of focused research on ITIL implementation and managed change it would seem useful to investigate this phenomenon by studying firms that have implemented the ITIL framework and assess which organizational CMPs more frequently used and whether their influence on the ITIL implementation outcomes can be observed. Therefore, two research questions were formulated for this study:RQ1. How did financial companies in Peru that have implemented the ITIL framework use organizational CMPs?RQ2. How are the levels of ITIL compliance achieved by financial companies in Peru influenced by the CMPs that were used? The study employed the multiple-case design method proposed by Yin (2009) as the principal method to gather data to answer the research questions. The case study is an appropriate research method to analyze a phenomenon in its natural environment when the researcher has no influence over events and when the context is considered relevant (Benbasat et al., 1987; Yin, 2009). The case study methodology allows for in-depth questioning and the capture of important aspects of the complexity of an organization. But it is recognized that the conclusions of an investigation by the case study method can be limited, with the generalization of the findings being a common criticism (Gable, 2010; Lee, 1989). However, the problems of validity and generalisability can be addressed by using competent research methodologies (Meredith, 1998; Eisenhardt, 1989).The case study method was considered appropriate for this study as it allowed the phenomenon, the ITIL implementation, to be studied in its natural setting (Meredith, 1998), to have questions of the "how and why" type posed and answered (Yin, 2009) and it is an appropriate methodology for studying new topic areas (Eisenhardt, 1989). The design of the investigation involved four sites each studied in depth using multiple data sources and a range of collection methods (Eisenhardt, 1989).A single industry was selected for the study, in order to control the influence of contexts and for different environments and resource availabilities across industries. Four companies from the Peruvian financial sector that had implemented the ITIL framework were identified. The focus of this study is on the ITILv2 implementation for while the ITILv3 * was recently released, ITIL v2 is still the most commonly used (Pollard and Cater-Steel, 2009). The four companies had each implemented the 11 processes of ITILv2 and in all the cases, the ITIL implementation projects were supported by management and had been allocated sufficient of resources for completion.Data regarding the implementation were collected through separate, semi-structured interviews with various staff involved in the project including, top management, operations and IT managers and users of the IT services. Confirmatory data were gathered through secondary sources such as internal documents and progress reports.Through a review of the literature on research related to ITIL adoption and practice of organizational change management a list of questions based on the change management framework of Kotter (2007) was developed. The survey instrument was developed in the Spanish language by native Spanish speakers and all of the respondents speak and write in Spanish as a normal part of their work. The interviewees were given the set of open-ended questions about their strategy, motivations, problems and use CMPs used during the ITIL implementation, before the interview. During the interview, a discussion following the question format was initiated with the answers to these questions recorded digitally with additional notes taken by the researchers. Raineri (2011) found a likely bias from interviewees who were also project implementers that they would be more inclined to say they had used CMPs. Therefore, to reduce this bias the data were corroborated with other participants of the ITIL implementation.The interviews were transcribed verbatim and underwent content analysis. Responses were coded against the question structure and stages of CMPs implementation following the seven steps allowing the researchers to gauge CMP usage for each case. To guard against bias in the data analysis, each interview transcript was content analyzed twice to verify the results. Each case was then written up and subjected to a cross-case analysis, identifying the similarities and noting differences in themes and patterns across the cases.To determine the level ITIL compliance the researchers selected the itSMF self-assessment tool (itSMF, 2003). The itSMF instrument compares an organizations performance with the ITIL best practices, assessing the level of compliance with the ITIL processes, producing its results based on a framework of capabilities through a five point scale (McNaughton et al., 2010). Case firm IT managers were each given an itSMF questionnaire to complete. The questionnaire results provided each case with a score, between 0 and 5 for each of the ITIL core component's compliance to best practice, enabling case comparison. 5.1 Case profiles The cases sites are four financial institutions operating in Peru coded as A, B, C and D. Table III presents the profiles of the case sites and data on their ITIL implementations. All four cases confirm there was support of management and commitment of adequate resources to complete the implementations. All cases placed an emphasis on training and development during the project implementations. Cases C and D opted to use external consultants as part of their ITIL implementations. The cases chose phased execution as their preferred implementation method.5.2 Case results Table IV presents data on the frequency of the use of the seven CMPs. The four cases show relatively high use of change practices in the preparation phases, with fewer practices being used in the later stages of implementation and consolidation. Case D seemed to have institutional features that supported the use across the phases of the ITIL implementation. For example, case D has internal policies and practices that led to a significantly higher use of the empowerment practice than all of the other cases combined. In contrast case B showed the least commitment in terms of resource allocation and staff stimulus in later phases of the change process. In general there is one group where there seems to be stronger institutionalization (cases C and D), and another group (cases A and B) where there is less evidence of change practices consistently being used.The division between the two groups is equally apparent when comparing the case ITIL compliance scores. Table V presents the ITIL compliance scores, on a scale of 0-5. Table V shows that overall cases C and D achieved higher ITIL compliances than cases A and B. Case D consistently scores high across the categories, especially in those processes involving customer contacts such as service desk and incident management. Both cases A and B tend to score lower in the service support level functions despite having moderate scores for their ITSM level.5.3 Cross-case analysis In general, the analysis of the case data shows that the preparation stage practices are the more frequently change practices used by these Peruvian financial firms. Figure 1 shows the frequency of the CMPs employed in each stage across the cases. The most frequent are: "forming a powerful guiding coalition", "creating a vision" and "planning for and creating short-term wins", and to a lesser extent and "communicating the vision". The practice "establishing a sense of urgency" was used significantly more by cases C and D. The CMPs less frequently used were "institutionalizing new approaches", "consolidating improvements and producing still more change" and "empowering others to act on the vision". The two groups found in the case analysis are easily seen in Figure 1, with cases A and B providing little evidence of activities to communicate, empower and progress the implementations in terms of using CMPs.The most frequently used practices are related to the change preparation stage in contrast to the implementation stage practices. This result is consistent with Raineri (2001, 2011) findings, where the agents of change showed a preference to the use of CMPs related to the preparation stage.However, the results also show a key implementation practice "planning for and creating short-term wins" is frequently used. This may be due to the implementation method chosen by the cases. All cases used the phased implementation approach that may have facilitated the use of this CMP - for as one respondent stated "successful results on the initial phases of the implementation were used to motivate the implementation of the rest of the processes". Spafford and Holup (2010) also suggest the phased strategy provides benefits sooner and thus contributes to a gradual reduction of resistance to change.The compliance data reveals that in general, the greater the use of CMPs the higher the level of ITIL compliance achieved, suggesting that the use of CMPs has a positive influence on the compliance outcomes. Figure 2 shows the score for ITIL implementation compliance against the number of CMPs used on each process implementation. Cases D and C score high in change practice use and the levels of ITIL compliance, with cases A and B that used less practices have lower compliance scores.Cases A and B that have a low use of implementation and consolidation phase practices also reported change resistance problems. This observation is consistent with Siegal et al. (1996) who reported that resistance to change problems can be attributed to deficiencies in the management of the change process or in part are a result of the actions and inactions of change agents (Ford et al., 2008; Tavakoli, 2010).There are some institutional factors that have some bearing on the patterns revealed in the cross-case analysis. First, the use of outside consultants may have had a significant impact on project success. Pollard and Cater-Steel (2009) attribute the use of consultants to be a critical success factor for ITIL implementations. That cases C and D opted for consultants may have strengthened their ability to achieve higher compliance scores and to undertake more CMPs. Second, the formation of teams and leadership of the project was undertaken differently across the cases. At cases A and B the team of leaders was formed by those at the functional level, while at cases C and D the leaders involved tended to have more influence within the organization. While it is difficult to directly attribute this factor to the lower compliance scores, the demonstrated the support of senior managers to uphold change initiatives and reduce change resistance is well documented (Spafford and Holup, 2010; Abraham et al., 1999; Kotter, 2007). This study aimed to investigate the importance of applying managed change as part of an ITIL implementation. To this end four Peruvian financial sector firms were chosen as cases and a multi case study investigation was undertaken. The study undertook to ascertain which CMPs were being used and their level of influence on ITIL compliance.This study finds that not all firms take full advantage of CMPs while implementing the ITIL framework. The cases showed that higher levels of CMPs achieved higher levels of ITIL implementation compliance. This result suggests that the use of CMPs has a positive impact on the outcomes of ITIL implementations. It was also found that the firms that used less CMPs reported the resistance to change during the implementation. However, it important not to discount that the higher scoring firms may have had valuable or rare resources and competencies which they utilized in times of change giving them an advantage over others (Barney, 1991).The case studies also demonstrate that more change practices related to preparation and fewer change practices related to implementation and consolidation were used during the ITIL implementation process. This result may be attributable to institutional strengths of the firm, for as case D showed, it seemed to have institutional policies and practices that supported the ITIL implementation across all of its phases, and case B seemingly having the least commitment in terms of resource allocation, sustainable change and staff involvement. Likewise, the participation of an external consultant during the adoption of ITIL may favor the more frequent use of practices related to all three change stages (Pollard and Cater-Steel, 2009) allowing those firms that used consultants to use more CMPs. In addition, the choice of implementation methodology may contribute to some practices being more evident than others. All of the four case studies presented applied a phased implementations; a strategy, which favors the use of the change practice the "generation of short-term results".The main limitation of this study is the small number of cases analyzed, so the findings should be viewed with caution. The cases support the theory that managed change assists the implementation of ITIL processes. However, it was undertaken in a single industry, which tends to have homogenous processes and systems and the firms were all of a similar size. However, when examining cases of different size, system processes and industries divergent results may be produced, but these results are necessary to properly extend the case theory in question (Meredith, 1998). Future research may wish to extend the themes of this paper to design and undertake research on which CMPs provide the most impact to the levels of ITIL compliance using a suitable quantitative research instrument and method.In conclusion this paper makes three contributions. First it provides an exploratory study of the links between ITIL implementations and the use of CMPs. Second in terms of results it provides evidence of a link between ITIL implementation and managed change, which is important information for practitioners who may be planning future ITIL projects. Finally, this study adds to the literature of ITIL implementations, particularly with respect to organizational change. Opens in a new window.Figure 1 Use of CMPs Opens in a new window.Figure 2 CMPs and levels of ITIL compliance Opens in a new window.Table I Core components of ITILv2 Opens in a new window.Table II Change management practices Opens in a new window.Table III Profile of cases Opens in a new window.Table IV Change practices and frequency of use Opens in a new window.Table V ITIL compliance scores
|
- This exploratory research reports on four case studies of completed implementations of IT service management using the process - based ITILv2 framework. The firms studied are from the Peruvian financial sector. Data were gained primarily through semi-structured in-depth interviews with managers involved in the implementation process. Compliance was measured using the itSMF self assessment questionnaire, which assesses the overall process capability.
|
[SECTION: Findings] Service firms are searching for ways to deliver higher levels of information technology (IT) service as well as demanding more from their information systems (IS) groups, expecting quick responses to new business opportunities, to support critical processes efficiently and to satisfy their customers and internal staff. Hence there is a growing interest in adopting IT service management (ITSM) and best practices such as the Information Technology Infrastructure Library (ITIL) (Galup et al., 2009).ITSM is defined as "the implementation and management of Quality IT Services that meet the needs of the business" (ITIL, 2007). ITSM is a discipline for managing IT service operations, which is process orientated, sharing this theme with other quality improvement methods and as such it is different from traditional technology-oriented approaches due to its focus on customer relationships (Galup et al., 2009). A main benefit reported by ITSM adopters is an increase in customer satisfaction (Cater-Steel et al., 2009). ITSM concepts are often implemented through the ITIL framework; a sector specific quality and best practice guide on the management of information technologies (Iden, 2012). The operational benefits expected from ITIL implementations are the alignment of IT services with business needs, improved quality of the IT services themselves, and a reduction in the long term costs of service provision (McNaughton et al., 2010).One of the main challenges faced by companies during the adoption of ITIL is the organizational change required to attain a service-oriented culture (Cater-Steel et al., 2007; Spafford and Holup, 2010). Various practices to facilitate organizational change are proposed in the academic and professional literature (Buchanan et al., 2005; Kanter, 2001; Kerber, 2001; Kotter, 1995; Spafford and Holup, 2010), and most of this literature argues that the use of change management practices (CMPs) has a positive effect on the speed and quality of the change process and on results for the organization. Nevertheless, there is little published in terms of these practices as they relate to ITIL implementations.Therefore, this paper presents the findings of a multiple case study investigation, which examines the experience of four Peruvian financial companies implementing the ITIL framework. The cases offer insight into how the organizational change management processes supported the ITIL implementation and identifies which organizational CMPs were more frequently used.The paper is organized as follows; next Section 2 introduces the ITIL framework, followed by Section 3, a review of the literature on CMPs concluding with the study's research questions. Section 4 provides details of the methodology used. Section 5 provides the case profiles, the findings and the cross-case analysis. Finally, the conclusion sums up the findings, presents the studies limitations and offers a direction for future research. ITIL is a framework that describes best practices and provides guidance on the management of IT processes, functions, roles and responsibilities in ITSM. It was first developed in the 1980s by the UK Government agency, the Office of Government Commerce (OGC) in response to the government's growing dependence on information technologies (Galup et al., 2009), to promote efficient IT operations and to improve IT service delivery and operations within government controlled computing centers (Salle, 2004).Version one of ITIL consisted of 40 volumes covering "best practices" in different areas of IT service provision. In 2000, ITIL version one was replaced by the seven volume ITIL version two (ITILv2) consolidating the practices within an overall framework. The two primary components of ITILv2 are service delivery and service support which consist of ten core processes and the service desk function (Table I).In 2007, ITILv2 was further enhanced to become ITIL version three (ITILv3), now consisting of five volumes: strategy, design, transition, operation and continuous improvement, which extend the ITILv2 model and processes by organizing them around a service lifecycle model.2.1 ITIL adoption and implementation The ITILv2 has become the de facto framework to implement the ITSM concepts (Cartlidge et al., 2007; Cater-Steel et al., 2009) despite the recent release of ITILv3 (Pollard and Cater-Steel, 2009).ITIL defines a glossary to ensure a common vocabulary and defines a set of conceptual processes to outline best practices (Galup et al., 2009). ITIL does not specify how the implementation should be performed and recent studies on the adoption of ITIL in organizations have focused mainly on the verification of the benefits of the program and the study of the implementations themselves, including success factors and barriers.In terms of benefits, Hochstein et al. (2005) found, an increase in customer orientation and service and the efficiency and transparency of IT processes the main benefits. They also suggest that the biggest challenge in the adoption of ITIL is resistance in the form of reluctance to accept new processes and a lack of understanding of why these processes are necessary. Cater-Steel et al. (2006) identified that ITIL implementation provided firms with a more predictable infrastructure, improved consultations between the IT groups within the organization, agility in negotiating service level agreements and consistent services; also identifying that cultural change is a significant challenge when adopting ITIL. Cater-Steel and Tan (2005a, b) found that the IT staffs skills, adaptability, and quality of training were considered more important for successful ITIL implementation than issues related to clients, external consultants and technology. The importance of IT staff adaptability and change management was in turn identified in a subsequent investigation by Cater-Steel et al. (2007).In terms of ITIL implementation, Tan et al. (2009) identified that a change management strategy that aims to transform the organizational culture towards being service-oriented, is a critical success factor. While, Pollard and Cater-Steel (2009) reported the difficulty of changing the culture of the organization when adopting the new ITIL processes and they identified three further critical success factors for ITIL implementations; creating an ITIL friendly culture, giving priority to the processes and the design of customer-focused metrics. Cater-Steel and McBride (2007) findings highlight the importance of involving staff and the promoting interest in ITIL as a way of overcoming resistance to the new processes; facilitating an improved understanding of them. Spafford and Holub (2010) found that resistance to organizational culture change is a major impediment in implementing ITIL, such that improved processes are achieved in less time and less cost when a formal change management plan is implemented than for projects where the organizational change component is neglected. Kerber (2001) identifies three approaches used in organizations to enact organizational change:1. Direct change, which is motivated from senior management, wielded by authority and achieved through compliance.2. Planned change, which originates from any level in the organization but is managed through the top management layer and that seeks the commitment and involvement of the organization through CMPs that mitigate resistance.3. Guided change, arising from the within the organization and the through commitment and the contribution of staff toward the organizational objectives.Kerber and Buono (2005) argue that the effectiveness of the approach to the management of change depends on contextual variables such as:* the complexity of the business environment;* socio-technical uncertainty of the tasks or problems;* the ability to change the organization; and* the risks associated with the alternatives of not changing or changing slowly.Other factors may add complexity to the management of change, for example the cultural context (Martinsons et al., 2009) and the interaction between different agents of change namely; senior management, middle management, external consultants and work teams who have different experiences and perspectives (Caldwell, 2003; Choi et al., 2011). On the other hand, management support is considered key to the success of organizational change (Abraham et al., 1999; Spafford and Holup, 2010), as is the allocation of adequate resources (Becker, 2010). Although, following a change program is not a guarantee of successful outcomes (Choi et al., 2011).Kotter (2007) presents a set of eight sequential CMPs that allow for successful organizational change. Kotter's approach is well known and has been applied frequently (Aiken and Keller, 2009; Smith, 2011) (Table II).CMPs are commonly proposed in temporal sequences such as preparation, implementation and consolidation (Raineri, 2011). Research has found that change agents show a preference for CMPs related to the preparation stage because; it is easier to prevent errors in early stages, which lessens implementation failure at later stages (Holt et al., 2007); they have the ability to place more resources at the beginning of a change project (Raineri, 2011) and; the preparation stage requires more analytical skills of which managers tend to have more training in, compared with the more political and interpersonal skills required in the later stages of implementation (Shipper, 1999). Similarly, Raineri (2011) found a higher frequency of preparation practices in comparison with the implementation stage and that the use of CMPs has a significant impact on objectives and deadlines achievement. Abraham et al. (1999) found that the key factor in achieving change towards a culture of quality is top management support, with the lesser factors of the clarity of vision, participation, communication and appropriate resources. Becker (2010) found that the understanding the need for change, the level of organizational support and training, measuring the change, the positive experience and informal support, a history of organizational changes and the outlook previous expectations and individual feelings are factors that influence adoption.Considering the lack of focused research on ITIL implementation and managed change it would seem useful to investigate this phenomenon by studying firms that have implemented the ITIL framework and assess which organizational CMPs more frequently used and whether their influence on the ITIL implementation outcomes can be observed. Therefore, two research questions were formulated for this study:RQ1. How did financial companies in Peru that have implemented the ITIL framework use organizational CMPs?RQ2. How are the levels of ITIL compliance achieved by financial companies in Peru influenced by the CMPs that were used? The study employed the multiple-case design method proposed by Yin (2009) as the principal method to gather data to answer the research questions. The case study is an appropriate research method to analyze a phenomenon in its natural environment when the researcher has no influence over events and when the context is considered relevant (Benbasat et al., 1987; Yin, 2009). The case study methodology allows for in-depth questioning and the capture of important aspects of the complexity of an organization. But it is recognized that the conclusions of an investigation by the case study method can be limited, with the generalization of the findings being a common criticism (Gable, 2010; Lee, 1989). However, the problems of validity and generalisability can be addressed by using competent research methodologies (Meredith, 1998; Eisenhardt, 1989).The case study method was considered appropriate for this study as it allowed the phenomenon, the ITIL implementation, to be studied in its natural setting (Meredith, 1998), to have questions of the "how and why" type posed and answered (Yin, 2009) and it is an appropriate methodology for studying new topic areas (Eisenhardt, 1989). The design of the investigation involved four sites each studied in depth using multiple data sources and a range of collection methods (Eisenhardt, 1989).A single industry was selected for the study, in order to control the influence of contexts and for different environments and resource availabilities across industries. Four companies from the Peruvian financial sector that had implemented the ITIL framework were identified. The focus of this study is on the ITILv2 implementation for while the ITILv3 * was recently released, ITIL v2 is still the most commonly used (Pollard and Cater-Steel, 2009). The four companies had each implemented the 11 processes of ITILv2 and in all the cases, the ITIL implementation projects were supported by management and had been allocated sufficient of resources for completion.Data regarding the implementation were collected through separate, semi-structured interviews with various staff involved in the project including, top management, operations and IT managers and users of the IT services. Confirmatory data were gathered through secondary sources such as internal documents and progress reports.Through a review of the literature on research related to ITIL adoption and practice of organizational change management a list of questions based on the change management framework of Kotter (2007) was developed. The survey instrument was developed in the Spanish language by native Spanish speakers and all of the respondents speak and write in Spanish as a normal part of their work. The interviewees were given the set of open-ended questions about their strategy, motivations, problems and use CMPs used during the ITIL implementation, before the interview. During the interview, a discussion following the question format was initiated with the answers to these questions recorded digitally with additional notes taken by the researchers. Raineri (2011) found a likely bias from interviewees who were also project implementers that they would be more inclined to say they had used CMPs. Therefore, to reduce this bias the data were corroborated with other participants of the ITIL implementation.The interviews were transcribed verbatim and underwent content analysis. Responses were coded against the question structure and stages of CMPs implementation following the seven steps allowing the researchers to gauge CMP usage for each case. To guard against bias in the data analysis, each interview transcript was content analyzed twice to verify the results. Each case was then written up and subjected to a cross-case analysis, identifying the similarities and noting differences in themes and patterns across the cases.To determine the level ITIL compliance the researchers selected the itSMF self-assessment tool (itSMF, 2003). The itSMF instrument compares an organizations performance with the ITIL best practices, assessing the level of compliance with the ITIL processes, producing its results based on a framework of capabilities through a five point scale (McNaughton et al., 2010). Case firm IT managers were each given an itSMF questionnaire to complete. The questionnaire results provided each case with a score, between 0 and 5 for each of the ITIL core component's compliance to best practice, enabling case comparison. 5.1 Case profiles The cases sites are four financial institutions operating in Peru coded as A, B, C and D. Table III presents the profiles of the case sites and data on their ITIL implementations. All four cases confirm there was support of management and commitment of adequate resources to complete the implementations. All cases placed an emphasis on training and development during the project implementations. Cases C and D opted to use external consultants as part of their ITIL implementations. The cases chose phased execution as their preferred implementation method.5.2 Case results Table IV presents data on the frequency of the use of the seven CMPs. The four cases show relatively high use of change practices in the preparation phases, with fewer practices being used in the later stages of implementation and consolidation. Case D seemed to have institutional features that supported the use across the phases of the ITIL implementation. For example, case D has internal policies and practices that led to a significantly higher use of the empowerment practice than all of the other cases combined. In contrast case B showed the least commitment in terms of resource allocation and staff stimulus in later phases of the change process. In general there is one group where there seems to be stronger institutionalization (cases C and D), and another group (cases A and B) where there is less evidence of change practices consistently being used.The division between the two groups is equally apparent when comparing the case ITIL compliance scores. Table V presents the ITIL compliance scores, on a scale of 0-5. Table V shows that overall cases C and D achieved higher ITIL compliances than cases A and B. Case D consistently scores high across the categories, especially in those processes involving customer contacts such as service desk and incident management. Both cases A and B tend to score lower in the service support level functions despite having moderate scores for their ITSM level.5.3 Cross-case analysis In general, the analysis of the case data shows that the preparation stage practices are the more frequently change practices used by these Peruvian financial firms. Figure 1 shows the frequency of the CMPs employed in each stage across the cases. The most frequent are: "forming a powerful guiding coalition", "creating a vision" and "planning for and creating short-term wins", and to a lesser extent and "communicating the vision". The practice "establishing a sense of urgency" was used significantly more by cases C and D. The CMPs less frequently used were "institutionalizing new approaches", "consolidating improvements and producing still more change" and "empowering others to act on the vision". The two groups found in the case analysis are easily seen in Figure 1, with cases A and B providing little evidence of activities to communicate, empower and progress the implementations in terms of using CMPs.The most frequently used practices are related to the change preparation stage in contrast to the implementation stage practices. This result is consistent with Raineri (2001, 2011) findings, where the agents of change showed a preference to the use of CMPs related to the preparation stage.However, the results also show a key implementation practice "planning for and creating short-term wins" is frequently used. This may be due to the implementation method chosen by the cases. All cases used the phased implementation approach that may have facilitated the use of this CMP - for as one respondent stated "successful results on the initial phases of the implementation were used to motivate the implementation of the rest of the processes". Spafford and Holup (2010) also suggest the phased strategy provides benefits sooner and thus contributes to a gradual reduction of resistance to change.The compliance data reveals that in general, the greater the use of CMPs the higher the level of ITIL compliance achieved, suggesting that the use of CMPs has a positive influence on the compliance outcomes. Figure 2 shows the score for ITIL implementation compliance against the number of CMPs used on each process implementation. Cases D and C score high in change practice use and the levels of ITIL compliance, with cases A and B that used less practices have lower compliance scores.Cases A and B that have a low use of implementation and consolidation phase practices also reported change resistance problems. This observation is consistent with Siegal et al. (1996) who reported that resistance to change problems can be attributed to deficiencies in the management of the change process or in part are a result of the actions and inactions of change agents (Ford et al., 2008; Tavakoli, 2010).There are some institutional factors that have some bearing on the patterns revealed in the cross-case analysis. First, the use of outside consultants may have had a significant impact on project success. Pollard and Cater-Steel (2009) attribute the use of consultants to be a critical success factor for ITIL implementations. That cases C and D opted for consultants may have strengthened their ability to achieve higher compliance scores and to undertake more CMPs. Second, the formation of teams and leadership of the project was undertaken differently across the cases. At cases A and B the team of leaders was formed by those at the functional level, while at cases C and D the leaders involved tended to have more influence within the organization. While it is difficult to directly attribute this factor to the lower compliance scores, the demonstrated the support of senior managers to uphold change initiatives and reduce change resistance is well documented (Spafford and Holup, 2010; Abraham et al., 1999; Kotter, 2007). This study aimed to investigate the importance of applying managed change as part of an ITIL implementation. To this end four Peruvian financial sector firms were chosen as cases and a multi case study investigation was undertaken. The study undertook to ascertain which CMPs were being used and their level of influence on ITIL compliance.This study finds that not all firms take full advantage of CMPs while implementing the ITIL framework. The cases showed that higher levels of CMPs achieved higher levels of ITIL implementation compliance. This result suggests that the use of CMPs has a positive impact on the outcomes of ITIL implementations. It was also found that the firms that used less CMPs reported the resistance to change during the implementation. However, it important not to discount that the higher scoring firms may have had valuable or rare resources and competencies which they utilized in times of change giving them an advantage over others (Barney, 1991).The case studies also demonstrate that more change practices related to preparation and fewer change practices related to implementation and consolidation were used during the ITIL implementation process. This result may be attributable to institutional strengths of the firm, for as case D showed, it seemed to have institutional policies and practices that supported the ITIL implementation across all of its phases, and case B seemingly having the least commitment in terms of resource allocation, sustainable change and staff involvement. Likewise, the participation of an external consultant during the adoption of ITIL may favor the more frequent use of practices related to all three change stages (Pollard and Cater-Steel, 2009) allowing those firms that used consultants to use more CMPs. In addition, the choice of implementation methodology may contribute to some practices being more evident than others. All of the four case studies presented applied a phased implementations; a strategy, which favors the use of the change practice the "generation of short-term results".The main limitation of this study is the small number of cases analyzed, so the findings should be viewed with caution. The cases support the theory that managed change assists the implementation of ITIL processes. However, it was undertaken in a single industry, which tends to have homogenous processes and systems and the firms were all of a similar size. However, when examining cases of different size, system processes and industries divergent results may be produced, but these results are necessary to properly extend the case theory in question (Meredith, 1998). Future research may wish to extend the themes of this paper to design and undertake research on which CMPs provide the most impact to the levels of ITIL compliance using a suitable quantitative research instrument and method.In conclusion this paper makes three contributions. First it provides an exploratory study of the links between ITIL implementations and the use of CMPs. Second in terms of results it provides evidence of a link between ITIL implementation and managed change, which is important information for practitioners who may be planning future ITIL projects. Finally, this study adds to the literature of ITIL implementations, particularly with respect to organizational change. Opens in a new window.Figure 1 Use of CMPs Opens in a new window.Figure 2 CMPs and levels of ITIL compliance Opens in a new window.Table I Core components of ITILv2 Opens in a new window.Table II Change management practices Opens in a new window.Table III Profile of cases Opens in a new window.Table IV Change practices and frequency of use Opens in a new window.Table V ITIL compliance scores
|
- This study finds that not all of the case firms took full advantage of change management practices while implementing the ITIL framework. The results show that the firms with a higher use of change management practices achieved higher levels of ITIL compliance. Additionally, change practices related to change preparation were used more frequently and fewer practices related to the implementation and consolidation stages were reported.
|
[SECTION: Value] Service firms are searching for ways to deliver higher levels of information technology (IT) service as well as demanding more from their information systems (IS) groups, expecting quick responses to new business opportunities, to support critical processes efficiently and to satisfy their customers and internal staff. Hence there is a growing interest in adopting IT service management (ITSM) and best practices such as the Information Technology Infrastructure Library (ITIL) (Galup et al., 2009).ITSM is defined as "the implementation and management of Quality IT Services that meet the needs of the business" (ITIL, 2007). ITSM is a discipline for managing IT service operations, which is process orientated, sharing this theme with other quality improvement methods and as such it is different from traditional technology-oriented approaches due to its focus on customer relationships (Galup et al., 2009). A main benefit reported by ITSM adopters is an increase in customer satisfaction (Cater-Steel et al., 2009). ITSM concepts are often implemented through the ITIL framework; a sector specific quality and best practice guide on the management of information technologies (Iden, 2012). The operational benefits expected from ITIL implementations are the alignment of IT services with business needs, improved quality of the IT services themselves, and a reduction in the long term costs of service provision (McNaughton et al., 2010).One of the main challenges faced by companies during the adoption of ITIL is the organizational change required to attain a service-oriented culture (Cater-Steel et al., 2007; Spafford and Holup, 2010). Various practices to facilitate organizational change are proposed in the academic and professional literature (Buchanan et al., 2005; Kanter, 2001; Kerber, 2001; Kotter, 1995; Spafford and Holup, 2010), and most of this literature argues that the use of change management practices (CMPs) has a positive effect on the speed and quality of the change process and on results for the organization. Nevertheless, there is little published in terms of these practices as they relate to ITIL implementations.Therefore, this paper presents the findings of a multiple case study investigation, which examines the experience of four Peruvian financial companies implementing the ITIL framework. The cases offer insight into how the organizational change management processes supported the ITIL implementation and identifies which organizational CMPs were more frequently used.The paper is organized as follows; next Section 2 introduces the ITIL framework, followed by Section 3, a review of the literature on CMPs concluding with the study's research questions. Section 4 provides details of the methodology used. Section 5 provides the case profiles, the findings and the cross-case analysis. Finally, the conclusion sums up the findings, presents the studies limitations and offers a direction for future research. ITIL is a framework that describes best practices and provides guidance on the management of IT processes, functions, roles and responsibilities in ITSM. It was first developed in the 1980s by the UK Government agency, the Office of Government Commerce (OGC) in response to the government's growing dependence on information technologies (Galup et al., 2009), to promote efficient IT operations and to improve IT service delivery and operations within government controlled computing centers (Salle, 2004).Version one of ITIL consisted of 40 volumes covering "best practices" in different areas of IT service provision. In 2000, ITIL version one was replaced by the seven volume ITIL version two (ITILv2) consolidating the practices within an overall framework. The two primary components of ITILv2 are service delivery and service support which consist of ten core processes and the service desk function (Table I).In 2007, ITILv2 was further enhanced to become ITIL version three (ITILv3), now consisting of five volumes: strategy, design, transition, operation and continuous improvement, which extend the ITILv2 model and processes by organizing them around a service lifecycle model.2.1 ITIL adoption and implementation The ITILv2 has become the de facto framework to implement the ITSM concepts (Cartlidge et al., 2007; Cater-Steel et al., 2009) despite the recent release of ITILv3 (Pollard and Cater-Steel, 2009).ITIL defines a glossary to ensure a common vocabulary and defines a set of conceptual processes to outline best practices (Galup et al., 2009). ITIL does not specify how the implementation should be performed and recent studies on the adoption of ITIL in organizations have focused mainly on the verification of the benefits of the program and the study of the implementations themselves, including success factors and barriers.In terms of benefits, Hochstein et al. (2005) found, an increase in customer orientation and service and the efficiency and transparency of IT processes the main benefits. They also suggest that the biggest challenge in the adoption of ITIL is resistance in the form of reluctance to accept new processes and a lack of understanding of why these processes are necessary. Cater-Steel et al. (2006) identified that ITIL implementation provided firms with a more predictable infrastructure, improved consultations between the IT groups within the organization, agility in negotiating service level agreements and consistent services; also identifying that cultural change is a significant challenge when adopting ITIL. Cater-Steel and Tan (2005a, b) found that the IT staffs skills, adaptability, and quality of training were considered more important for successful ITIL implementation than issues related to clients, external consultants and technology. The importance of IT staff adaptability and change management was in turn identified in a subsequent investigation by Cater-Steel et al. (2007).In terms of ITIL implementation, Tan et al. (2009) identified that a change management strategy that aims to transform the organizational culture towards being service-oriented, is a critical success factor. While, Pollard and Cater-Steel (2009) reported the difficulty of changing the culture of the organization when adopting the new ITIL processes and they identified three further critical success factors for ITIL implementations; creating an ITIL friendly culture, giving priority to the processes and the design of customer-focused metrics. Cater-Steel and McBride (2007) findings highlight the importance of involving staff and the promoting interest in ITIL as a way of overcoming resistance to the new processes; facilitating an improved understanding of them. Spafford and Holub (2010) found that resistance to organizational culture change is a major impediment in implementing ITIL, such that improved processes are achieved in less time and less cost when a formal change management plan is implemented than for projects where the organizational change component is neglected. Kerber (2001) identifies three approaches used in organizations to enact organizational change:1. Direct change, which is motivated from senior management, wielded by authority and achieved through compliance.2. Planned change, which originates from any level in the organization but is managed through the top management layer and that seeks the commitment and involvement of the organization through CMPs that mitigate resistance.3. Guided change, arising from the within the organization and the through commitment and the contribution of staff toward the organizational objectives.Kerber and Buono (2005) argue that the effectiveness of the approach to the management of change depends on contextual variables such as:* the complexity of the business environment;* socio-technical uncertainty of the tasks or problems;* the ability to change the organization; and* the risks associated with the alternatives of not changing or changing slowly.Other factors may add complexity to the management of change, for example the cultural context (Martinsons et al., 2009) and the interaction between different agents of change namely; senior management, middle management, external consultants and work teams who have different experiences and perspectives (Caldwell, 2003; Choi et al., 2011). On the other hand, management support is considered key to the success of organizational change (Abraham et al., 1999; Spafford and Holup, 2010), as is the allocation of adequate resources (Becker, 2010). Although, following a change program is not a guarantee of successful outcomes (Choi et al., 2011).Kotter (2007) presents a set of eight sequential CMPs that allow for successful organizational change. Kotter's approach is well known and has been applied frequently (Aiken and Keller, 2009; Smith, 2011) (Table II).CMPs are commonly proposed in temporal sequences such as preparation, implementation and consolidation (Raineri, 2011). Research has found that change agents show a preference for CMPs related to the preparation stage because; it is easier to prevent errors in early stages, which lessens implementation failure at later stages (Holt et al., 2007); they have the ability to place more resources at the beginning of a change project (Raineri, 2011) and; the preparation stage requires more analytical skills of which managers tend to have more training in, compared with the more political and interpersonal skills required in the later stages of implementation (Shipper, 1999). Similarly, Raineri (2011) found a higher frequency of preparation practices in comparison with the implementation stage and that the use of CMPs has a significant impact on objectives and deadlines achievement. Abraham et al. (1999) found that the key factor in achieving change towards a culture of quality is top management support, with the lesser factors of the clarity of vision, participation, communication and appropriate resources. Becker (2010) found that the understanding the need for change, the level of organizational support and training, measuring the change, the positive experience and informal support, a history of organizational changes and the outlook previous expectations and individual feelings are factors that influence adoption.Considering the lack of focused research on ITIL implementation and managed change it would seem useful to investigate this phenomenon by studying firms that have implemented the ITIL framework and assess which organizational CMPs more frequently used and whether their influence on the ITIL implementation outcomes can be observed. Therefore, two research questions were formulated for this study:RQ1. How did financial companies in Peru that have implemented the ITIL framework use organizational CMPs?RQ2. How are the levels of ITIL compliance achieved by financial companies in Peru influenced by the CMPs that were used? The study employed the multiple-case design method proposed by Yin (2009) as the principal method to gather data to answer the research questions. The case study is an appropriate research method to analyze a phenomenon in its natural environment when the researcher has no influence over events and when the context is considered relevant (Benbasat et al., 1987; Yin, 2009). The case study methodology allows for in-depth questioning and the capture of important aspects of the complexity of an organization. But it is recognized that the conclusions of an investigation by the case study method can be limited, with the generalization of the findings being a common criticism (Gable, 2010; Lee, 1989). However, the problems of validity and generalisability can be addressed by using competent research methodologies (Meredith, 1998; Eisenhardt, 1989).The case study method was considered appropriate for this study as it allowed the phenomenon, the ITIL implementation, to be studied in its natural setting (Meredith, 1998), to have questions of the "how and why" type posed and answered (Yin, 2009) and it is an appropriate methodology for studying new topic areas (Eisenhardt, 1989). The design of the investigation involved four sites each studied in depth using multiple data sources and a range of collection methods (Eisenhardt, 1989).A single industry was selected for the study, in order to control the influence of contexts and for different environments and resource availabilities across industries. Four companies from the Peruvian financial sector that had implemented the ITIL framework were identified. The focus of this study is on the ITILv2 implementation for while the ITILv3 * was recently released, ITIL v2 is still the most commonly used (Pollard and Cater-Steel, 2009). The four companies had each implemented the 11 processes of ITILv2 and in all the cases, the ITIL implementation projects were supported by management and had been allocated sufficient of resources for completion.Data regarding the implementation were collected through separate, semi-structured interviews with various staff involved in the project including, top management, operations and IT managers and users of the IT services. Confirmatory data were gathered through secondary sources such as internal documents and progress reports.Through a review of the literature on research related to ITIL adoption and practice of organizational change management a list of questions based on the change management framework of Kotter (2007) was developed. The survey instrument was developed in the Spanish language by native Spanish speakers and all of the respondents speak and write in Spanish as a normal part of their work. The interviewees were given the set of open-ended questions about their strategy, motivations, problems and use CMPs used during the ITIL implementation, before the interview. During the interview, a discussion following the question format was initiated with the answers to these questions recorded digitally with additional notes taken by the researchers. Raineri (2011) found a likely bias from interviewees who were also project implementers that they would be more inclined to say they had used CMPs. Therefore, to reduce this bias the data were corroborated with other participants of the ITIL implementation.The interviews were transcribed verbatim and underwent content analysis. Responses were coded against the question structure and stages of CMPs implementation following the seven steps allowing the researchers to gauge CMP usage for each case. To guard against bias in the data analysis, each interview transcript was content analyzed twice to verify the results. Each case was then written up and subjected to a cross-case analysis, identifying the similarities and noting differences in themes and patterns across the cases.To determine the level ITIL compliance the researchers selected the itSMF self-assessment tool (itSMF, 2003). The itSMF instrument compares an organizations performance with the ITIL best practices, assessing the level of compliance with the ITIL processes, producing its results based on a framework of capabilities through a five point scale (McNaughton et al., 2010). Case firm IT managers were each given an itSMF questionnaire to complete. The questionnaire results provided each case with a score, between 0 and 5 for each of the ITIL core component's compliance to best practice, enabling case comparison. 5.1 Case profiles The cases sites are four financial institutions operating in Peru coded as A, B, C and D. Table III presents the profiles of the case sites and data on their ITIL implementations. All four cases confirm there was support of management and commitment of adequate resources to complete the implementations. All cases placed an emphasis on training and development during the project implementations. Cases C and D opted to use external consultants as part of their ITIL implementations. The cases chose phased execution as their preferred implementation method.5.2 Case results Table IV presents data on the frequency of the use of the seven CMPs. The four cases show relatively high use of change practices in the preparation phases, with fewer practices being used in the later stages of implementation and consolidation. Case D seemed to have institutional features that supported the use across the phases of the ITIL implementation. For example, case D has internal policies and practices that led to a significantly higher use of the empowerment practice than all of the other cases combined. In contrast case B showed the least commitment in terms of resource allocation and staff stimulus in later phases of the change process. In general there is one group where there seems to be stronger institutionalization (cases C and D), and another group (cases A and B) where there is less evidence of change practices consistently being used.The division between the two groups is equally apparent when comparing the case ITIL compliance scores. Table V presents the ITIL compliance scores, on a scale of 0-5. Table V shows that overall cases C and D achieved higher ITIL compliances than cases A and B. Case D consistently scores high across the categories, especially in those processes involving customer contacts such as service desk and incident management. Both cases A and B tend to score lower in the service support level functions despite having moderate scores for their ITSM level.5.3 Cross-case analysis In general, the analysis of the case data shows that the preparation stage practices are the more frequently change practices used by these Peruvian financial firms. Figure 1 shows the frequency of the CMPs employed in each stage across the cases. The most frequent are: "forming a powerful guiding coalition", "creating a vision" and "planning for and creating short-term wins", and to a lesser extent and "communicating the vision". The practice "establishing a sense of urgency" was used significantly more by cases C and D. The CMPs less frequently used were "institutionalizing new approaches", "consolidating improvements and producing still more change" and "empowering others to act on the vision". The two groups found in the case analysis are easily seen in Figure 1, with cases A and B providing little evidence of activities to communicate, empower and progress the implementations in terms of using CMPs.The most frequently used practices are related to the change preparation stage in contrast to the implementation stage practices. This result is consistent with Raineri (2001, 2011) findings, where the agents of change showed a preference to the use of CMPs related to the preparation stage.However, the results also show a key implementation practice "planning for and creating short-term wins" is frequently used. This may be due to the implementation method chosen by the cases. All cases used the phased implementation approach that may have facilitated the use of this CMP - for as one respondent stated "successful results on the initial phases of the implementation were used to motivate the implementation of the rest of the processes". Spafford and Holup (2010) also suggest the phased strategy provides benefits sooner and thus contributes to a gradual reduction of resistance to change.The compliance data reveals that in general, the greater the use of CMPs the higher the level of ITIL compliance achieved, suggesting that the use of CMPs has a positive influence on the compliance outcomes. Figure 2 shows the score for ITIL implementation compliance against the number of CMPs used on each process implementation. Cases D and C score high in change practice use and the levels of ITIL compliance, with cases A and B that used less practices have lower compliance scores.Cases A and B that have a low use of implementation and consolidation phase practices also reported change resistance problems. This observation is consistent with Siegal et al. (1996) who reported that resistance to change problems can be attributed to deficiencies in the management of the change process or in part are a result of the actions and inactions of change agents (Ford et al., 2008; Tavakoli, 2010).There are some institutional factors that have some bearing on the patterns revealed in the cross-case analysis. First, the use of outside consultants may have had a significant impact on project success. Pollard and Cater-Steel (2009) attribute the use of consultants to be a critical success factor for ITIL implementations. That cases C and D opted for consultants may have strengthened their ability to achieve higher compliance scores and to undertake more CMPs. Second, the formation of teams and leadership of the project was undertaken differently across the cases. At cases A and B the team of leaders was formed by those at the functional level, while at cases C and D the leaders involved tended to have more influence within the organization. While it is difficult to directly attribute this factor to the lower compliance scores, the demonstrated the support of senior managers to uphold change initiatives and reduce change resistance is well documented (Spafford and Holup, 2010; Abraham et al., 1999; Kotter, 2007). This study aimed to investigate the importance of applying managed change as part of an ITIL implementation. To this end four Peruvian financial sector firms were chosen as cases and a multi case study investigation was undertaken. The study undertook to ascertain which CMPs were being used and their level of influence on ITIL compliance.This study finds that not all firms take full advantage of CMPs while implementing the ITIL framework. The cases showed that higher levels of CMPs achieved higher levels of ITIL implementation compliance. This result suggests that the use of CMPs has a positive impact on the outcomes of ITIL implementations. It was also found that the firms that used less CMPs reported the resistance to change during the implementation. However, it important not to discount that the higher scoring firms may have had valuable or rare resources and competencies which they utilized in times of change giving them an advantage over others (Barney, 1991).The case studies also demonstrate that more change practices related to preparation and fewer change practices related to implementation and consolidation were used during the ITIL implementation process. This result may be attributable to institutional strengths of the firm, for as case D showed, it seemed to have institutional policies and practices that supported the ITIL implementation across all of its phases, and case B seemingly having the least commitment in terms of resource allocation, sustainable change and staff involvement. Likewise, the participation of an external consultant during the adoption of ITIL may favor the more frequent use of practices related to all three change stages (Pollard and Cater-Steel, 2009) allowing those firms that used consultants to use more CMPs. In addition, the choice of implementation methodology may contribute to some practices being more evident than others. All of the four case studies presented applied a phased implementations; a strategy, which favors the use of the change practice the "generation of short-term results".The main limitation of this study is the small number of cases analyzed, so the findings should be viewed with caution. The cases support the theory that managed change assists the implementation of ITIL processes. However, it was undertaken in a single industry, which tends to have homogenous processes and systems and the firms were all of a similar size. However, when examining cases of different size, system processes and industries divergent results may be produced, but these results are necessary to properly extend the case theory in question (Meredith, 1998). Future research may wish to extend the themes of this paper to design and undertake research on which CMPs provide the most impact to the levels of ITIL compliance using a suitable quantitative research instrument and method.In conclusion this paper makes three contributions. First it provides an exploratory study of the links between ITIL implementations and the use of CMPs. Second in terms of results it provides evidence of a link between ITIL implementation and managed change, which is important information for practitioners who may be planning future ITIL projects. Finally, this study adds to the literature of ITIL implementations, particularly with respect to organizational change. Opens in a new window.Figure 1 Use of CMPs Opens in a new window.Figure 2 CMPs and levels of ITIL compliance Opens in a new window.Table I Core components of ITILv2 Opens in a new window.Table II Change management practices Opens in a new window.Table III Profile of cases Opens in a new window.Table IV Change practices and frequency of use Opens in a new window.Table V ITIL compliance scores
|
- Few studies have examined change management practices in the context of ITIL implementation projects. This study also uses the levels of ITIL compliance as an outcome measure.
|
[SECTION: Purpose] Rapid changes and unpredictable events occur in the business environment. These changes are, in part, the result of market growth or technology development, and they create turbulence that can destroy the existing competencies of an organization (Tushman and Anderson, 1986). In such a changing environment, organizations find that their previous strategies, core competencies (Prahalad and Hamel, 1990), beliefs, values, and cultures (Moorman and Miner, 1997; Sproull, 1981) are becoming less effective, or are no longer effective at all. These core competencies, which might have taken many years to develop and refine, can become core rigidities (Leonard-Barton, 1995), and can hinder an organization's ability to compete and win. Hedberg (1981) has noted that knowledge becomes obsolete under turbulent environments and must be renewed. Hedberg (1981) has defined this renewing activity as "unlearning," and has emphasized that an inability to unlearn is a critical weakness of many organizations. Hamel and Prahalad (1994, p. 71) have also noted, "Companies are going to have to unlearn a lot of their past - and also to forget it! The future will not be an extrapolation of the past."Unlearning is vital for organizations to learn how to survive and compete in the competitive landscape (Nystrom and Starbuck, 1984; Hedberg, 1981). Specifically, organizations make it more difficult to learn without first unlearning (Hedberg, 1981). Organizations have widely established and accepted beliefs and methods that persuade them to neglect important new technologies and markets, because they have a great emotional investment in old ways of working. Those established beliefs and methods create rules and competency traps that negatively affect the operations of organizations (Mezias et al., 2001). In particular, those procedures, methods and beliefs that inhibit the reception and evaluation of new market and technology information, and reduce the value of perceived new information. For instance, Day (1994, p. 24) has stated that: "The presumed correctness of past actions and interpretations is reinforced by repeated success, and ensuing complacency breeds rejection of information that conflicts with conventional wisdom." Also, people focus on information that supports their current beliefs and methods. However, incorrect beliefs and methods assimilate errors in judgment and actions because people alter the perception of reality to fit their beliefs and methods (Rousseau, 2001). Fixed beliefs lead to perception rigidity or inaccurate causal attributions, and these result in organizations becoming slower to recognize changes (Dickson, 1992). For instance, Starbuck (1996) points out that:People in organizations find it very difficult to deal effectively with information that conflicts with their current beliefs and methods. They do not know how to accommodate dissonant information and they find it difficult to change a few elements of their interdependent beliefs and methods.Even though unlearning is important, it is problematic due to difficulty of conceptualizing, operationalizing and testing it in the current scholarship. As Crossan et al. (1995) have pointed out, the "unlearning" concept has generated great interest, but it has received limited acceptance in the literature due to confusion regarding the terms "learning" and "unlearning" both in theory and in practice. In particular, confusion was enhanced when the term "unlearning" was used as the reverse of "learning." One of the reasons for this confusion is that most of the arguments on the unlearning has been conducted in the organizational learning literature (Akgun et al., 2003). However, the unlearning concept is not just bounded by or restricted to the organizational learning literature. Since, March's organizational adaptation theory (Cyert and March, 1963) and Lewin's (1951) organizational change theory, unlearning has made claim as the heart of the organizational change process (Walsh and Ungson, 1991). For example, scholars in the school of organizational change and memory pointed out that unlearning and memory are closely related concepts, because memory creates competency traps due to fixed and well-accepted routines and values in organizations (Levitt and March, 1988; Moorman and Miner, 1997). Accordingly, an understanding of unlearning can be leveraged, reducing the confusion between unlearning and learning by reviewing organizational change and memory literature.The purpose of this paper therefore is to shed light on the unlearning concept based on the organizational memory and change literature. Specifically, the research questions addressed in this study are:RQ1. How the unlearning concept can be conceptualized and operationalized for future empirical investigations.RQ2. What types of unlearning can be revealed based on the organizational memory and change literature to enhance the organizational learning and change management scholarship.In the sections that follow, this study:1. explains the concept of unlearning by reviewing different streams of research;2. argues that unlearning reflects two underlying components, changes in beliefs and routines; and then proposes that the combination of these components of unlearning creates distinct types of organizational unlearning based on environmental contingencies; and3. provides guidelines for future research. The concept of "unlearning" has attracted many researchers from diverse fields for the last 40 years with numerous studies coming from the individual learning and cognitive psychology literature. At the individual level, unlearning has been discussed as part of verbal learning psychology and individual cognition. In the verbal learning psychology literature, unlearning has primarily been viewed as deleting and replacing old stimuli by interpolated learning, which is also called interference theory (Postman et al., 1965). Scholars in this stream of research (Postman et al., 1965) have mentioned that unlearning is the gradual weakening of associations between stimuli and responses by retroactive inhibition.In addition to the verbal learning psychology, the cognitive science literature has viewed unlearning as changes in belief structure (Fiske and Taylor, 1984), mental model (Johnson-Laird, 1983), frame of reference (Shrivastava and Schneider, 1984), or cognitive maps (Walsh, 1988). Moreover, most of the studies attributable to unlearning in individual psychology and cognition have focused on memory loss (Hull, 1943), deterioration of the trace in memory (Koffka, 1935), intentional forgetting (Freud, 1943), and retroactive inhibition (Postman et al., 1965). Scholars have thus viewed unlearning as "memory eliminating" in a system (that is, the individual) (Greeno et al., 1971), and that this elimination helps bring about new learning, e.g. acceptance of novelty (new knowledge).Apart from studies at the individual level, unlearning has also been studied at the organizational level, especially in the scholarship on organizational change and memory (Lewin, 1951; Huber, 1991; Bartunek, 1988). Studies on unlearning at the organizational level reveal a similar, although not identical, pattern to that found in the scholarship on individual unlearning - that is, the elimination of memories (Walsh and Ungson, 1991). Scholars in the school of organizational memory have viewed organizational unlearning in various ways as:* changing acquisition processes and possible retrieval processes in the organizational memory (Walsh and Ungson, 1991);* disruption and re-creation of portions of the organization's memory (Anand et al., 1998);* purposely eliminating memories (Stein, 1995);* dissembling and disconfirming the causal connections in the organizational memory (Nicollini and Meznar, 1995); and* disintegrating the community's collective infrastructure of routines (Blackler et al., 1999).Thus, the common theme, to-date, for the unlearning, both in individual and organizational level studies, is:* eliminating memory via disconfirmation;* the disassembly of the connections and mechanisms of memory; and/or* changing how memory is manifested. Because unlearning has been conceptualized as memory eliminating, an investigation of how memory is formed and manifested could help in understanding and operationalizing unlearning in organizations. Moorman and Miner (1997), for instance, proposed that organizational memory is manifested in three forms:1. organizational beliefs that include knowledge, frame of references, models, values, and norms;2. formal and informal behavioral routines, procedures, and scripts that include standard operating procedural, managerial, and technical systems, capabilities, and information-sharing mechanisms; and3. the organization's physical artifacts including tools, programming, assembly-line layout, and features of products and product lines (such as product design, materials, packaging, and logos).Therefore, logically, unlearning must include eliminating beliefs, routines, and physical artifacts in organizations based on the insights garnered from the cognitive psychology, organizational memory and change literature. However, this definition is too broad for a parsimonious operationalization of unlearning, and thereby raises three concerns, namely ontological, construct validity, and pragmatistic. These three concerns clarify the operationalization of the unlearning concept in a systematic fashion. Specifically, these concerns:1. provide scrutiny and scientific explanations, which have the stature and explanatory significance assumed for unlearning;2. demonstrate what makes up the unlearning concept that are observed both in academia and practice; and3. explain variables or constructs of which the unlearning is a function by applying the scientific principles of system thinking, i.e. these concerns are mutually dependent.Ontological concern The term "eliminating" is often referred to as replacing, discarding, removing, and reducing in unlearning studies. Each verb has different meaning in english and encompasses a complex mental structure that is integrative and contains diverse reference points. For instance, when members of an organization replace their current belief with a new belief in the current organizational setting, they decide to use new belief, because it is expected to be more valid and trustworthy. However, the old belief still remains in the new organizational setting, but it is blocked or no longer considered. Unlike belief replacing, which has cognitive ramifications (Weick, 1979), discarding the current belief indicates that current belief is no longer available in the new organizational setting. The organizational actors decide to stop accepting current beliefs, and are unable to recall them from the resultant belief state represented by the new organizational setting. Nevertheless, the common theme of belief replacing, reducing, and discarding is that the stimulus or response assemblies are incongruent with the current organizational schema, and either an incremental or revolutionary schema change is needed. Indeed, belief replacing (i.e. another belief takes the place of the original) indicates an incremental schema change, whereas belief eliminating (the whole memory structure does not exist any more) shows a revolutionary schema change. In this sense, while adapting only one definition partially explains the unlearning concept, using different terms confounds the operationalization of unlearning. Therefore, for a gestalt understanding for the unlearning concept, all the terms should be combined into a task verb to define real activities or experiences of unlearning (Sandelands and Drazin, 1989). It is interesting to note that the ontological properties of unlearning - discarding, replacing and reducing, converge on the task verb of "changing" as noted by the research in artificial intelligence (AI) (Boutilier, 1998; Burns and Gomolinska, 2001) and social and cognitive psychology literature (Fiske and Taylor, 1991; Akgun et al., 2003)[1]. According to AI researchers, the change of an agent's belief state reflects a response to a change or changes in the world as well as what is perceived to be an incorrect or incomplete prior belief, leading to a new belief state (Boutilier, 1998). One might suggest that replacing, reducing or discarding of a belief is a semantic change of a belief state. Also, in the field of social psychology, Crocker et al. (1984, p. 217) state that, "The schematic approach to belief change is concerned with changes in the content and structure of social knowledge in response to information that is incongruent with the schema."The similar rationalization of discarding, replacing, and reducing for routines, and physical artifacts can be made in light of the production and operations management (Tranfield and Smith, 1998) and management literature (Pentland and Rueter, 1994). For instance, Tranfield and Smith (1998, p. 119) wrote that, "Reviewing, revising, redesigning and replacing routines either as automatic responses or effortful accomplishments lies at the heart of change management for routines." This is further supported by Mezias et al. (2001) where they stated that unlearning old rationales embedded in routines is essential to changing those routines. Accordingly, the ontological concern indicates that unlearning is the changing of beliefs, routines and physical artifacts. However, these three forms of unlearning (that is, a change in beliefs, routines, and physical artifacts) pose problems from the perspective of construct validity, leading to the second concern.Construct validity concern Since, unlearning is a cognitive process, changes in organizational artifacts, which are not cognitive, should be coupled with changes in beliefs and routines to facilitate unlearning. For instance, if a manufacturing group changes its beliefs regarding the operational methods and techniques for the production of a particular product, the group also needs to change the "machinery" in that process to remove the previous habits and knowledge. This suggests that a more comprehensive memory structure is needed to operationalize organizational unlearning. It is of interest to note that these three forms of memory structures (i.e. beliefs, routines, and physical artifacts) converge on what cognitive psychologists call "declarative and procedural memory" (Anderson, 1993). Declarative memory is related to "know-what" in organizations - that is, a memory of what needs to be done. This is a memory for facts, events, and propositions (Moorman and Miner, 1998). Moorman and Miner (1998) have indicated that a key characteristic of declarative memory is the variety of uses to which it can be put - which is similar to knowledge structure, frames of reference, mental models, values, and norms. In essence, these constitute beliefs, because beliefs explain the world in terms of cause-and-effect relationships, in interpretative schemata that map the experience of the world as mental models, and cognitive frames (Sproull, 1981; Allard-Poesi, 1998).Apart from declarative memory, another memory structure is procedural memory. Procedural memory is related to "know-how" - that is, how to do it or how things are done (Cohen and Bacdayan, 1994; Moorman and Miner, 1998). Procedural memory involves skills or routines, such as behavioral routines, procedures, tools, programming, and so on (Cohen and Bacdayan, 1994; Pentland and Rueter, 1994). Winter (1986, p. 165), for instance, has pointed out that the word "routine" is a term used generically and includes decision-making, techniques, skills, standard operational procedures, management practices, policies, strategies, information systems, information structures, programs, scripts, and organizational forms.A change in beliefs and routines as unlearning is also connected to the theories on organizational learning. Scholars (Levitt and March, 1988) have indicated that routines are a source of organizational learning. For instance, Hendry (1996, p. 636) wrote that "A way round this whole problem of learning/not learning may be through the notion of organizational routines. Routines offer a reference point to link different perspectives on organizational learning." This is particularly apropos because routines, the fabric of organizational life, shape how organizations define and solve problems and learn. Feldman (2000) also demonstrated organizational routines as a process of organizational learning. In addition to routines, organizational learning implies shared knowledge, perceptions, and beliefs. Specifically, since organizational beliefs filter environmental stimuli, they affect the way people perceive, interpret and evaluate events, how they connect the organizational capabilities with external environment, and govern "what is good and what is bad." As noted by Schein (1993), learning processes can be influenced by organizational beliefs. Unlike the beliefs and routines, which are the important part of the learning process, physical artifacts are the results of prior learning, as noted by Moorman and Miner (1997, p. 93), and are closely related to the organizational knowing rather than learning (Orlikowski, 2002). Kuh and Whitt (1988), for example, noted that physical artifacts surround people physically and provide them with the immediate sensory stimuli and memory. Accordingly, construct validity concern denotes that unlearning is the changing beliefs and routines. As demonstrated empirically, the components of unlearning should occur together for an effective unlearning (Starbuck, 1996), leading to the third concern.Pragmatistic The pragmatistic concern indicates that beliefs and routines should occur simultaneously. From a statistical point of view, changes in beliefs and routines are covariates. For instance, changing the organizational beliefs, such as anticipated market requirements, initiates unlearning. However, changes in manufacturing, distribution or logistics, should also be sanctioned by the organizations for effective organizational unlearning (Sinkula, 2002). In a real sense, changes in beliefs per se cause the deterioration of the routines and paralyze the organization, albeit temporarily, because the existing patterns of actions as indicated by routines are less likely to be congruent with the new set of beliefs (Sinkula, 2002). Similarly, substituting routines with new ones without changes in organizational beliefs also partially initiate organizational unlearning. However, since the initial filters of incoming external stimuli (existing organizational beliefs) remain, i.e. sensing and funneling the same set of stimuli to organizational members, there will be a dissonance between relevant information processing and actual behaviors.Based on the above discussion, "organizational unlearning can be operationalized as covariately changing declarative memories (beliefs) and procedural memories (routines) in organizations"[2]. Since, organizational unlearning is operationalized as changes in beliefs and routines, this operationalization is confused with organizational change as well. Indeed, organizational change and learning occurs through revision of organizational scripts, cognitive schema establishing behavior and routines, or behavioral recipes involving beliefs as indicated in action learning (Johnson, 1990), neo-institutional theory of organization (Greenwood and Hinings, 1996), social cognition (Akgun et al., 2003), and social psychology (Weick, 1979). Unlearning shares similar etiological properties with organizational change (Johnson, 1990). However, note that organizational change is a generic term and is a broad concept involving "analytical, educational, learning and political process as well as a process which combines rational, political and cultural elements" (Hendry, 1996). In essence, organizational change is an end state of a transformation process (Tsoukas and Chia, 2002). Nevertheless, unlearning in particular focuses on the memory eliminating. Organizational unlearning is a change in collective cognition and routines that coordinate organizational change process. Specifically, unlearning is a stage or catalyst (one that causes a process or an event to happen) in the change process to make it a dynamic process, as indicated by organizational change theories (Mezias et al., 2001). For instance, Lewin's (1951) model for change involves three steps:1. unfreezing, which is suspending the current structure and involves disconfirmation of expectations, learning anxiety, and provision of psychological safety;2. transition, which is changing the mental structure and involves cognitive restructuring, semantic redefinition, and new standards of judgment; and3. refreezing, which is adapting the new mental structure and involves creation of supportive social norms, and making change congruent with personality.The second stage of Lewin's model, transition, is indicative of the unlearning phenomenon. Gemmill and Smith (1985), also explained the elements of system changes as:* disequilibrium, which is the result of external and internal forces, such as crisis and chaos;* symmetry breaking, which refers to breaking down of existing patterns of interactions or system habits;* experimentation, which refers to creating different new forms or configuration to reformulate the system; and* reformulation, which is selecting a new configuration.In their model, the system breaking stage denotes the unlearning concept.For that reason, we purport that "unlearning is embedded in organizational change process. However, the aim of unlearning is not performance improvement per se; rather it is a catalyst for the change process." We argued that unlearning involves the combination of the changes in beliefs and routines and these two components of unlearning must exist in tandem for unlearning to occur effectively. Nevertheless, in practice, the magnitude of the changes in beliefs and routines may vary. For instance, an organization is likely to put a higher emphasis on changes in routines than on changes in beliefs. One of the reasons would be the environmental conditions, in particular, environmental turbulence (Schein, 1993; Hedberg, 1981; Nystrom and Starbuck, 1981). However, considering environmental turbulence as a unidimensional construct narrows our understanding of the dynamic unlearning process, requiring a multidimensional view of environmental turbulence. Environmental turbulence refers to the degree of change and unpredictability of an environment (Glazer and Weiss, 1993, p. 510). The degree of change involves more information, leading to multiple and conflicted interpretations about an organizational situation. This type of contingency requires more interpretation and a variety of schemas or belief structures. Unpredictability of environment is related to the time-sensitivity of the information. Information in a given period loses its value in subsequent periods. Since, different environmental conditions require organizations to use the appropriate unlearning process, we propose that environmental turbulence results in four types of unlearning process as shown in Table I. The labels used for unlearning are borrowed from Gnyawali and Stewart (2003). Reinventive unlearning occurs when organizations put high emphasis on both changes in beliefs and routines. Formative unlearning is when more emphasis is placed on the beliefs and less on the routines. High emphasis on the changes in routines and low on the changes in beliefs results in adjustive unlearning, and low emphasis on both changes in beliefs and routines results in operative unlearning.It is interesting to note that these types of unlearning make good connections with the literature on planned (discontinuous) change. The proposed types of unlearning provide a meeting point between theories of planned change and theories of continuous change as Weick and Quinn (1999) stated. For instance, Weick and Quinn (1999, pp. 381-2) pointed out that:Recent work suggests, ironically, that to understand organizational change one must first understand organizational inertia, its content, its tenacity, its interdependencies. Recent work also suggests that change is not an on-off phenomenon nor is its effectiveness contingent on the degree to which it is planned. Furthermore, the trajectory of change is more often spiral or open ended than linear. All of these insights are more likely to be kept in play if researchers focus on "changing" rather than "change." A shift in vocabulary from "change" to "changing" directs attention to actions of substituting one thing for another, of making one thing into another thing, or of attracting one thing to become other than it was.Indeed, the types of unlearning indicate that change is an on-going process, never ceasing, and provide a continuum between continuous change and planned (discontinuous) change. Also, the magnitude of changes in beliefs and routines, in each type of unlearning, directs one's attention to the actions that can substitute one thing for another.Reinventive unlearning Reinventive unlearning indicates the fundamental changes in the existing beliefs and routines of an organization. This type of unlearning is at the heart of radical organizational change and transformation. Reinventive unlearning provides a strategic reorientation (Greenwood and Hinings, 1996), and takes organizations outside their familiar domains (Starbuck, 1996), such as radical shift in business strategy. In particular, reinventive unlearning is needed if an organization is to compete in the next generation of products or in the context of new disruptive technologies (Christensen, 1997). For instance, in the computer disk drive industry, manufacturers of 8 inch disk drives invested in process innovation to such a degree that when 5 inch disk drives began to capture the market, these manufacturers were not able to transform their production and operation procedures and beliefs (Christensen, 1997).The situation that triggers the reinventive unlearning is the high-level environmental change and unpredictability, such as "hyper" turbulent environments as indicated by D'Aveni (1984). In this type of environment, there exists a vast amount of information and a concurrent number of resulting interpretations. Also, information enters and leaves in numerous directions over short periods of time, loosing its validity in an unpredictable way.Reinventive unlearning is difficult and risky, and the outcome of reinventive unlearning is far from certain. It is expected that reinventive unlearning will be challenged when there exists:* myopia or inability of the company to look into the future with clarity;* inadequate strategic vision or lack of clear commitment of top management to changes; and* departmental politics or resistance from departments.Also, the more complacent firms are to their competitive and institutional environments, the more difficult it is for them to achieve reinventive unlearning. Tripsas and Gavetti (2000), for instance, showed how Polaroid Corporation failed to respond to the ongoing shift from analog to digital imagining. They showed how managers' beliefs, following the Gillette "razor blade" business model, delayed the commercialization of a stand-alone digital camera product.Formative unlearning Formative unlearning refers to the creation of new belief structures in organizations, with incremental changes in organizational routines. Organizations create new shared schemas and use these schemas to reinterpret information. In this type of unlearning, organizations develop new sense-making capability. For instance, Royal Dutch/Shell benefited from unlearning during the 1973 oil crisis by changing its managers' conceptual frames of references - developing new sense-making capabilities. Choo (1998, p. 16) notes that:Shell's managers have been assuming that oil demand will continue to grow at rates higher than GNP in a calm political environment where oil supply was unproblematic, a set of norms that managers had taken for granted for some time. The planning scenarios forced them to challenge these norms and to think about a low-growth world where oil consumption was increasing more slowly than GNP, where oil producers were reaching the limits of their capabilities and were reluctant to raise output further because they were unable to absorb the additional revenues.The high level of environmental change, which has numerous information sources, and more predictable change of information over time, triggers formative unlearning. In this type of environment, organizations will tend modify their strategies and rearrange their personnel.However, formative unlearning will be challenged when there exists denial or refusal to accept new belief structures. For instance, in his seminal book, Mastering the Dynamics of Innovation, Utterback (1994) noted how beliefs regarding natural ice harvesters (know what) impeded a change to machine-made ice production, because producers believed that people would prefer ice from nature rather than ice made by machines. Actually, ice harvesters were employing machines for their production and operation, indicating incremental changes in organizational routines, such as applying steam-powered circular saws for cutting blocks from rivers and ponds, and installing mechanical conveyors to provide continuous hauling of ice cakes from ponds to ice houses.Adjustive unlearning Adjustive unlearning refers to incremental changes in the existing knowledge structure of the organization, whereas fundamental changes occur in the organizational routines. This type of unlearning is related to evolutionary innovation and change and promotes new product lines, and new strategic business units. This type of unlearning includes the engineering intervention and work processes redesigns to improve the speed and quality of production. For instance, in a well known article by Imai et al. (1988) in which they discussed Honda's product-development strategy, to choose between a modified version of their Civic model or a totally new design. Even though the Civic was a successful product in the market, Honda changed its product-development mentality and designed a completely new car. During the development stage of the new Accord model, Honda carried out a design contest to identify the best car for its customers, and asked engineers and marketers around the world to help with the early stages of planning by creating a global knowledge network (Nonaka and Takeuchi, 1995).Environmental events that necessitate adjustive learning are those that have high unpredictability and low uncertainty. This type of environment indicates that information changes quickly and randomly. Organizations need to collect information rapidly about environmental events, and to distribute the information throughout the organization using new media and sensing units.Adjustive unlearning will be inhibited when there exists direct costs of change and cannibalization costs, which means change that brings success to a product but at the same time brings losses to a sister product, so it requires some kind of sacrifice.Operative unlearning Operative unlearning refers to the small-scale changes in beliefs and routines of an organization. This type of unlearning is related to incremental and continuous improvement and changes. This type of unlearning involves a limited departure from structural coherence. It helps organization maintain internal reliability and involves adjustments in systems, processes, or structures, but it does not involve fundamental change in strategy, core values, or corporate identity. Operative unlearning is most likely necessary during times of relative environmental stability and is likely to take place over extended periods of time.The environment, which stimulates operative learning, is low on uncertainty and is unpredictable. There is little information emerging and its direction is less unpredictable. In this sense, organizational change is a gradual and ongoing process which is aimed at fine tuning of the fit or match between the organizations' current strategy, structure, people and processes, and external environment.A typical example for an operative unlearning is the incremental product development activities in organizations. For instance, the HP DeskJet printer series, e.g. HP-800, developed from the adjustments to the production and commercialization processes, small-scale changes in production and design routines, and product improvement strategies to sustain the current market against the competitors - an on-going belief change efforts of management to adapt changes in competitive environment. As shown in Figure 1, this study:* denotes that unlearning can be operationalized as declarative and procedural memory eliminating - changing beliefs and routines in organizations based on the organizational change and memory literature; and* identifies different types of unlearning contingent upon the environmental conditions.Also, unlike that suggested by "common wisdom," which suggests that organizational unlearning is the opposite of organizational learning, Figure 1 shows that unlearning is an important sub-process of the organizational learning process and makes unique contributions to organizational learning literature (Akgun et al., 2003). Specifically, since, key contributions of unlearning in the organizational learning process have long been ignored in the literature, the legitimization of unlearning in the learning process was not forthcoming. Based on the organizational change and memory literature, we recognize that, unlearning:* catalyzes organizational learning process to foster a dynamic learning process;* provides a platform for shifting single-loop learning to double-loop learning; and* connects organizational learning and organizational change processes.As noted above, unlearning "catalyzes" the learning processes. The implication of the word, "catalyze" is that unlearning makes learning a dynamic process. As described by writers (Huber, 1991; Akgun et al., 2003), the organizational learning process encompasses continuous, reflexive and reciprocal cycle of activities that include memory, sensemaking, intelligence, thinking, emotions, improvisation, information acquisition, dissemination, and implementation. Since, it is a continuous process, no particular stage or sub-component can be taken as the starting point, therefore, missing or ignoring a component of the process halts the learning process in general (Akgun et al., 2003). In other words, all sub-components or processes work together rather than selectively. Unlearning facilitates a fluid and reflexive process of learning by changing beliefs and routines (Wijnhoven, 2001). For instance, organizational beliefs, which were created based on the positive feedback from a firm's success:* impose an interpretative inertia; and* facilitate inadequate reception of action signals.In addition, organizational routines, which indicate a motor, fixed or automatic action and behavior, pose an organizational inertia in changing and novel conditions. Routines reinforce the status quo (Nelson and Winter, 1982) and inhibit active seeking of alternatives. In this regard, unlearning makes the change and learning process unalterable, because old beliefs and routines were altered and replaced by a new set knowledge.In addition to the dynamic learning process, unlearning secures the learning levels. Specifically, unlearning is a means to shift from lower- (i.e. single-loop) to higher-level learning (i.e. double-loop). According to the organizational learning literature, single-loop learning or adaptive learning consists of detecting performance gaps and their elimination in line with standard operation procedures (Argyris and Schon, 1978). Here, an organization does not change its beliefs and routines. However, based on the environmental feedback, if an organization changes its beliefs and routines, one may say that double-loop learning has occurred. Here, unlearning provides a way for this transition, changing beliefs about cause-effect relations and operating procedures (Baker and Sinkula, 1999; Pawlowski, 2001). Wang and Pervaiz (2003), for instance, take a stronger stance and argue that organizational unlearning is needed to create quantum leaps.Finally, another critical role of unlearning is the interlocking of the organizational change and learning processes. There exists a common agreement that organizational change and learning are closely related concepts. For example, Pratt and Barnett (1997) have argued that learning is not a stand-alone concept; rather, it belongs to a family of change processes that encompasses unlearning and relearning (that is, adapting new mental models, procedures and routines). Indeed, based on contingency theory, the relation between an organization and its environment is the main concern of the theoretical perspectives of organizational change (Pawlowsky, 2001). At the same time, organizational learning requires changes in the relationships between organization and its environment (Hedberg and Wolff, 2001). Change involves some degree of learning, and learning provides a potential for change rather than any guarantee of its realization (Child and Heavens, 2001). Further, unlearning is one of the common dominators of change and learning process. For instance, Huy (2001) noted that beliefs and routines are the content of the change. Additionally, organizational learning concentrates on the process that brings such change about (Krebsbach-Gnath, 2001) and unlearning is one of the vital constructs of the learning process. In this perspective, unlearning acts as a linking pin between organizational learning and change processes. Unlearning is an important construct of organizational change and learning process, and thereby warrants an in-depth investigation. However, since the term "unlearning" was accompanied by the term "learning," such as "learning and unlearning," it was often seen as a demode construct or study area due to semantic usage of the terms and that it really is something different from learning, not simply its opposite. In this study, we further explored unlearning by reviewing organizational change and memory literature. However, our study is just one of a few attempts to discover the unlearning concept, see also Hedberg (1981), Nystrom and Starbuck (1984) and others. In future studies, we suggest that:1. Unlearning be investigated in the nomonological web of other organizational learning constructs. The recent research slightly mentions the niche of unlearning in the organizational learning literature (Akgun et al., 2003; Huber, 1991). For instance, according to Akgun et al. (2003), organizational learning is a process of the reciprocal relations among the information processing (i.e. memory, unlearning, sensemaking, etc.), intelligence and emotions. An empirical investigation of interwoven relations among the sub-components of learning would give a greater understanding of the process of learning at the organizational level.2. The types of unlearning can be investigated in detail using a longitudinal study. Specifically, the:* triggers or antecedents;* primary role of managers;* behavioral aspects of the practical (and unexpected) use; and* behavioral and operational consequences of each types of unlearning can be studied in case studies.3. Time processing orientation and agreement preferences should be investigated in unlearning process. Specifically, conflicts arose during changes in routines and beliefs, and resolution of those conflicts, majority and minority issues and power structures during the changes in beliefs and routines should be studied.4. How unlearning occurs should be investigated and empirically tested in groups, such as in new product development team. For instance, literature on new product development demonstrates that the groups that have strong memories are least able to deviate from previous routines during the new product development processes (Moorman and Miner, 1998). Therefore, eliminating group/team memory or unlearning, is critical for new product performance in turbulent environments/conditions and begs an empirical study. Also, empirically investigating unlearning in cross-functional new product development teams may illuminate researchers at the organizational level. Specifically, it has been asserted that new product development teams show the similar behavioral patterns as organizations do or represent microcosms of organizations based on the literature on the fractal geometry and images in system thinking (Capra, 1996). The basic philosophy of fractal images is that all characteristic patterns of the system are found repeatedly at the descending scales (Capra, 1996, p. 138). This is similar to holographic systems, which create processes where the whole can be encoded in all of the parts, so that "each and every part represents the whole" (Morgan, 1998, p. 71). In this sake, new product development team unlearning process empirically sheds light on the organizational unlearning process based on the functional isomorphism rather than homomorphic conceptualization of unlearning process, because NPD team members exist on more than one level simultaneously, act differently as units and influence each other across levels (Weiss, 1993, p. 278), and NPD process provides a more controllable, identifiable, and accurate unit definability (Meyers and Wilemon, 1986). Opens in a new window.Figure 1 Organizational unlearning Opens in a new window.Table I Types of organizational unlearning
|
- Organizational learning and unlearning is a popular and important topic in business as well as academia. Even though there is a plethora of studies on organizational learning, surprisingly little is known about the conceptualization and operationalization of organizational unlearning. The purpose of this paper is to discuss organizational unlearning based on the organizational change and memory literature enhancing the organizational learning and change scholarship.
|
[SECTION: Method] Rapid changes and unpredictable events occur in the business environment. These changes are, in part, the result of market growth or technology development, and they create turbulence that can destroy the existing competencies of an organization (Tushman and Anderson, 1986). In such a changing environment, organizations find that their previous strategies, core competencies (Prahalad and Hamel, 1990), beliefs, values, and cultures (Moorman and Miner, 1997; Sproull, 1981) are becoming less effective, or are no longer effective at all. These core competencies, which might have taken many years to develop and refine, can become core rigidities (Leonard-Barton, 1995), and can hinder an organization's ability to compete and win. Hedberg (1981) has noted that knowledge becomes obsolete under turbulent environments and must be renewed. Hedberg (1981) has defined this renewing activity as "unlearning," and has emphasized that an inability to unlearn is a critical weakness of many organizations. Hamel and Prahalad (1994, p. 71) have also noted, "Companies are going to have to unlearn a lot of their past - and also to forget it! The future will not be an extrapolation of the past."Unlearning is vital for organizations to learn how to survive and compete in the competitive landscape (Nystrom and Starbuck, 1984; Hedberg, 1981). Specifically, organizations make it more difficult to learn without first unlearning (Hedberg, 1981). Organizations have widely established and accepted beliefs and methods that persuade them to neglect important new technologies and markets, because they have a great emotional investment in old ways of working. Those established beliefs and methods create rules and competency traps that negatively affect the operations of organizations (Mezias et al., 2001). In particular, those procedures, methods and beliefs that inhibit the reception and evaluation of new market and technology information, and reduce the value of perceived new information. For instance, Day (1994, p. 24) has stated that: "The presumed correctness of past actions and interpretations is reinforced by repeated success, and ensuing complacency breeds rejection of information that conflicts with conventional wisdom." Also, people focus on information that supports their current beliefs and methods. However, incorrect beliefs and methods assimilate errors in judgment and actions because people alter the perception of reality to fit their beliefs and methods (Rousseau, 2001). Fixed beliefs lead to perception rigidity or inaccurate causal attributions, and these result in organizations becoming slower to recognize changes (Dickson, 1992). For instance, Starbuck (1996) points out that:People in organizations find it very difficult to deal effectively with information that conflicts with their current beliefs and methods. They do not know how to accommodate dissonant information and they find it difficult to change a few elements of their interdependent beliefs and methods.Even though unlearning is important, it is problematic due to difficulty of conceptualizing, operationalizing and testing it in the current scholarship. As Crossan et al. (1995) have pointed out, the "unlearning" concept has generated great interest, but it has received limited acceptance in the literature due to confusion regarding the terms "learning" and "unlearning" both in theory and in practice. In particular, confusion was enhanced when the term "unlearning" was used as the reverse of "learning." One of the reasons for this confusion is that most of the arguments on the unlearning has been conducted in the organizational learning literature (Akgun et al., 2003). However, the unlearning concept is not just bounded by or restricted to the organizational learning literature. Since, March's organizational adaptation theory (Cyert and March, 1963) and Lewin's (1951) organizational change theory, unlearning has made claim as the heart of the organizational change process (Walsh and Ungson, 1991). For example, scholars in the school of organizational change and memory pointed out that unlearning and memory are closely related concepts, because memory creates competency traps due to fixed and well-accepted routines and values in organizations (Levitt and March, 1988; Moorman and Miner, 1997). Accordingly, an understanding of unlearning can be leveraged, reducing the confusion between unlearning and learning by reviewing organizational change and memory literature.The purpose of this paper therefore is to shed light on the unlearning concept based on the organizational memory and change literature. Specifically, the research questions addressed in this study are:RQ1. How the unlearning concept can be conceptualized and operationalized for future empirical investigations.RQ2. What types of unlearning can be revealed based on the organizational memory and change literature to enhance the organizational learning and change management scholarship.In the sections that follow, this study:1. explains the concept of unlearning by reviewing different streams of research;2. argues that unlearning reflects two underlying components, changes in beliefs and routines; and then proposes that the combination of these components of unlearning creates distinct types of organizational unlearning based on environmental contingencies; and3. provides guidelines for future research. The concept of "unlearning" has attracted many researchers from diverse fields for the last 40 years with numerous studies coming from the individual learning and cognitive psychology literature. At the individual level, unlearning has been discussed as part of verbal learning psychology and individual cognition. In the verbal learning psychology literature, unlearning has primarily been viewed as deleting and replacing old stimuli by interpolated learning, which is also called interference theory (Postman et al., 1965). Scholars in this stream of research (Postman et al., 1965) have mentioned that unlearning is the gradual weakening of associations between stimuli and responses by retroactive inhibition.In addition to the verbal learning psychology, the cognitive science literature has viewed unlearning as changes in belief structure (Fiske and Taylor, 1984), mental model (Johnson-Laird, 1983), frame of reference (Shrivastava and Schneider, 1984), or cognitive maps (Walsh, 1988). Moreover, most of the studies attributable to unlearning in individual psychology and cognition have focused on memory loss (Hull, 1943), deterioration of the trace in memory (Koffka, 1935), intentional forgetting (Freud, 1943), and retroactive inhibition (Postman et al., 1965). Scholars have thus viewed unlearning as "memory eliminating" in a system (that is, the individual) (Greeno et al., 1971), and that this elimination helps bring about new learning, e.g. acceptance of novelty (new knowledge).Apart from studies at the individual level, unlearning has also been studied at the organizational level, especially in the scholarship on organizational change and memory (Lewin, 1951; Huber, 1991; Bartunek, 1988). Studies on unlearning at the organizational level reveal a similar, although not identical, pattern to that found in the scholarship on individual unlearning - that is, the elimination of memories (Walsh and Ungson, 1991). Scholars in the school of organizational memory have viewed organizational unlearning in various ways as:* changing acquisition processes and possible retrieval processes in the organizational memory (Walsh and Ungson, 1991);* disruption and re-creation of portions of the organization's memory (Anand et al., 1998);* purposely eliminating memories (Stein, 1995);* dissembling and disconfirming the causal connections in the organizational memory (Nicollini and Meznar, 1995); and* disintegrating the community's collective infrastructure of routines (Blackler et al., 1999).Thus, the common theme, to-date, for the unlearning, both in individual and organizational level studies, is:* eliminating memory via disconfirmation;* the disassembly of the connections and mechanisms of memory; and/or* changing how memory is manifested. Because unlearning has been conceptualized as memory eliminating, an investigation of how memory is formed and manifested could help in understanding and operationalizing unlearning in organizations. Moorman and Miner (1997), for instance, proposed that organizational memory is manifested in three forms:1. organizational beliefs that include knowledge, frame of references, models, values, and norms;2. formal and informal behavioral routines, procedures, and scripts that include standard operating procedural, managerial, and technical systems, capabilities, and information-sharing mechanisms; and3. the organization's physical artifacts including tools, programming, assembly-line layout, and features of products and product lines (such as product design, materials, packaging, and logos).Therefore, logically, unlearning must include eliminating beliefs, routines, and physical artifacts in organizations based on the insights garnered from the cognitive psychology, organizational memory and change literature. However, this definition is too broad for a parsimonious operationalization of unlearning, and thereby raises three concerns, namely ontological, construct validity, and pragmatistic. These three concerns clarify the operationalization of the unlearning concept in a systematic fashion. Specifically, these concerns:1. provide scrutiny and scientific explanations, which have the stature and explanatory significance assumed for unlearning;2. demonstrate what makes up the unlearning concept that are observed both in academia and practice; and3. explain variables or constructs of which the unlearning is a function by applying the scientific principles of system thinking, i.e. these concerns are mutually dependent.Ontological concern The term "eliminating" is often referred to as replacing, discarding, removing, and reducing in unlearning studies. Each verb has different meaning in english and encompasses a complex mental structure that is integrative and contains diverse reference points. For instance, when members of an organization replace their current belief with a new belief in the current organizational setting, they decide to use new belief, because it is expected to be more valid and trustworthy. However, the old belief still remains in the new organizational setting, but it is blocked or no longer considered. Unlike belief replacing, which has cognitive ramifications (Weick, 1979), discarding the current belief indicates that current belief is no longer available in the new organizational setting. The organizational actors decide to stop accepting current beliefs, and are unable to recall them from the resultant belief state represented by the new organizational setting. Nevertheless, the common theme of belief replacing, reducing, and discarding is that the stimulus or response assemblies are incongruent with the current organizational schema, and either an incremental or revolutionary schema change is needed. Indeed, belief replacing (i.e. another belief takes the place of the original) indicates an incremental schema change, whereas belief eliminating (the whole memory structure does not exist any more) shows a revolutionary schema change. In this sense, while adapting only one definition partially explains the unlearning concept, using different terms confounds the operationalization of unlearning. Therefore, for a gestalt understanding for the unlearning concept, all the terms should be combined into a task verb to define real activities or experiences of unlearning (Sandelands and Drazin, 1989). It is interesting to note that the ontological properties of unlearning - discarding, replacing and reducing, converge on the task verb of "changing" as noted by the research in artificial intelligence (AI) (Boutilier, 1998; Burns and Gomolinska, 2001) and social and cognitive psychology literature (Fiske and Taylor, 1991; Akgun et al., 2003)[1]. According to AI researchers, the change of an agent's belief state reflects a response to a change or changes in the world as well as what is perceived to be an incorrect or incomplete prior belief, leading to a new belief state (Boutilier, 1998). One might suggest that replacing, reducing or discarding of a belief is a semantic change of a belief state. Also, in the field of social psychology, Crocker et al. (1984, p. 217) state that, "The schematic approach to belief change is concerned with changes in the content and structure of social knowledge in response to information that is incongruent with the schema."The similar rationalization of discarding, replacing, and reducing for routines, and physical artifacts can be made in light of the production and operations management (Tranfield and Smith, 1998) and management literature (Pentland and Rueter, 1994). For instance, Tranfield and Smith (1998, p. 119) wrote that, "Reviewing, revising, redesigning and replacing routines either as automatic responses or effortful accomplishments lies at the heart of change management for routines." This is further supported by Mezias et al. (2001) where they stated that unlearning old rationales embedded in routines is essential to changing those routines. Accordingly, the ontological concern indicates that unlearning is the changing of beliefs, routines and physical artifacts. However, these three forms of unlearning (that is, a change in beliefs, routines, and physical artifacts) pose problems from the perspective of construct validity, leading to the second concern.Construct validity concern Since, unlearning is a cognitive process, changes in organizational artifacts, which are not cognitive, should be coupled with changes in beliefs and routines to facilitate unlearning. For instance, if a manufacturing group changes its beliefs regarding the operational methods and techniques for the production of a particular product, the group also needs to change the "machinery" in that process to remove the previous habits and knowledge. This suggests that a more comprehensive memory structure is needed to operationalize organizational unlearning. It is of interest to note that these three forms of memory structures (i.e. beliefs, routines, and physical artifacts) converge on what cognitive psychologists call "declarative and procedural memory" (Anderson, 1993). Declarative memory is related to "know-what" in organizations - that is, a memory of what needs to be done. This is a memory for facts, events, and propositions (Moorman and Miner, 1998). Moorman and Miner (1998) have indicated that a key characteristic of declarative memory is the variety of uses to which it can be put - which is similar to knowledge structure, frames of reference, mental models, values, and norms. In essence, these constitute beliefs, because beliefs explain the world in terms of cause-and-effect relationships, in interpretative schemata that map the experience of the world as mental models, and cognitive frames (Sproull, 1981; Allard-Poesi, 1998).Apart from declarative memory, another memory structure is procedural memory. Procedural memory is related to "know-how" - that is, how to do it or how things are done (Cohen and Bacdayan, 1994; Moorman and Miner, 1998). Procedural memory involves skills or routines, such as behavioral routines, procedures, tools, programming, and so on (Cohen and Bacdayan, 1994; Pentland and Rueter, 1994). Winter (1986, p. 165), for instance, has pointed out that the word "routine" is a term used generically and includes decision-making, techniques, skills, standard operational procedures, management practices, policies, strategies, information systems, information structures, programs, scripts, and organizational forms.A change in beliefs and routines as unlearning is also connected to the theories on organizational learning. Scholars (Levitt and March, 1988) have indicated that routines are a source of organizational learning. For instance, Hendry (1996, p. 636) wrote that "A way round this whole problem of learning/not learning may be through the notion of organizational routines. Routines offer a reference point to link different perspectives on organizational learning." This is particularly apropos because routines, the fabric of organizational life, shape how organizations define and solve problems and learn. Feldman (2000) also demonstrated organizational routines as a process of organizational learning. In addition to routines, organizational learning implies shared knowledge, perceptions, and beliefs. Specifically, since organizational beliefs filter environmental stimuli, they affect the way people perceive, interpret and evaluate events, how they connect the organizational capabilities with external environment, and govern "what is good and what is bad." As noted by Schein (1993), learning processes can be influenced by organizational beliefs. Unlike the beliefs and routines, which are the important part of the learning process, physical artifacts are the results of prior learning, as noted by Moorman and Miner (1997, p. 93), and are closely related to the organizational knowing rather than learning (Orlikowski, 2002). Kuh and Whitt (1988), for example, noted that physical artifacts surround people physically and provide them with the immediate sensory stimuli and memory. Accordingly, construct validity concern denotes that unlearning is the changing beliefs and routines. As demonstrated empirically, the components of unlearning should occur together for an effective unlearning (Starbuck, 1996), leading to the third concern.Pragmatistic The pragmatistic concern indicates that beliefs and routines should occur simultaneously. From a statistical point of view, changes in beliefs and routines are covariates. For instance, changing the organizational beliefs, such as anticipated market requirements, initiates unlearning. However, changes in manufacturing, distribution or logistics, should also be sanctioned by the organizations for effective organizational unlearning (Sinkula, 2002). In a real sense, changes in beliefs per se cause the deterioration of the routines and paralyze the organization, albeit temporarily, because the existing patterns of actions as indicated by routines are less likely to be congruent with the new set of beliefs (Sinkula, 2002). Similarly, substituting routines with new ones without changes in organizational beliefs also partially initiate organizational unlearning. However, since the initial filters of incoming external stimuli (existing organizational beliefs) remain, i.e. sensing and funneling the same set of stimuli to organizational members, there will be a dissonance between relevant information processing and actual behaviors.Based on the above discussion, "organizational unlearning can be operationalized as covariately changing declarative memories (beliefs) and procedural memories (routines) in organizations"[2]. Since, organizational unlearning is operationalized as changes in beliefs and routines, this operationalization is confused with organizational change as well. Indeed, organizational change and learning occurs through revision of organizational scripts, cognitive schema establishing behavior and routines, or behavioral recipes involving beliefs as indicated in action learning (Johnson, 1990), neo-institutional theory of organization (Greenwood and Hinings, 1996), social cognition (Akgun et al., 2003), and social psychology (Weick, 1979). Unlearning shares similar etiological properties with organizational change (Johnson, 1990). However, note that organizational change is a generic term and is a broad concept involving "analytical, educational, learning and political process as well as a process which combines rational, political and cultural elements" (Hendry, 1996). In essence, organizational change is an end state of a transformation process (Tsoukas and Chia, 2002). Nevertheless, unlearning in particular focuses on the memory eliminating. Organizational unlearning is a change in collective cognition and routines that coordinate organizational change process. Specifically, unlearning is a stage or catalyst (one that causes a process or an event to happen) in the change process to make it a dynamic process, as indicated by organizational change theories (Mezias et al., 2001). For instance, Lewin's (1951) model for change involves three steps:1. unfreezing, which is suspending the current structure and involves disconfirmation of expectations, learning anxiety, and provision of psychological safety;2. transition, which is changing the mental structure and involves cognitive restructuring, semantic redefinition, and new standards of judgment; and3. refreezing, which is adapting the new mental structure and involves creation of supportive social norms, and making change congruent with personality.The second stage of Lewin's model, transition, is indicative of the unlearning phenomenon. Gemmill and Smith (1985), also explained the elements of system changes as:* disequilibrium, which is the result of external and internal forces, such as crisis and chaos;* symmetry breaking, which refers to breaking down of existing patterns of interactions or system habits;* experimentation, which refers to creating different new forms or configuration to reformulate the system; and* reformulation, which is selecting a new configuration.In their model, the system breaking stage denotes the unlearning concept.For that reason, we purport that "unlearning is embedded in organizational change process. However, the aim of unlearning is not performance improvement per se; rather it is a catalyst for the change process." We argued that unlearning involves the combination of the changes in beliefs and routines and these two components of unlearning must exist in tandem for unlearning to occur effectively. Nevertheless, in practice, the magnitude of the changes in beliefs and routines may vary. For instance, an organization is likely to put a higher emphasis on changes in routines than on changes in beliefs. One of the reasons would be the environmental conditions, in particular, environmental turbulence (Schein, 1993; Hedberg, 1981; Nystrom and Starbuck, 1981). However, considering environmental turbulence as a unidimensional construct narrows our understanding of the dynamic unlearning process, requiring a multidimensional view of environmental turbulence. Environmental turbulence refers to the degree of change and unpredictability of an environment (Glazer and Weiss, 1993, p. 510). The degree of change involves more information, leading to multiple and conflicted interpretations about an organizational situation. This type of contingency requires more interpretation and a variety of schemas or belief structures. Unpredictability of environment is related to the time-sensitivity of the information. Information in a given period loses its value in subsequent periods. Since, different environmental conditions require organizations to use the appropriate unlearning process, we propose that environmental turbulence results in four types of unlearning process as shown in Table I. The labels used for unlearning are borrowed from Gnyawali and Stewart (2003). Reinventive unlearning occurs when organizations put high emphasis on both changes in beliefs and routines. Formative unlearning is when more emphasis is placed on the beliefs and less on the routines. High emphasis on the changes in routines and low on the changes in beliefs results in adjustive unlearning, and low emphasis on both changes in beliefs and routines results in operative unlearning.It is interesting to note that these types of unlearning make good connections with the literature on planned (discontinuous) change. The proposed types of unlearning provide a meeting point between theories of planned change and theories of continuous change as Weick and Quinn (1999) stated. For instance, Weick and Quinn (1999, pp. 381-2) pointed out that:Recent work suggests, ironically, that to understand organizational change one must first understand organizational inertia, its content, its tenacity, its interdependencies. Recent work also suggests that change is not an on-off phenomenon nor is its effectiveness contingent on the degree to which it is planned. Furthermore, the trajectory of change is more often spiral or open ended than linear. All of these insights are more likely to be kept in play if researchers focus on "changing" rather than "change." A shift in vocabulary from "change" to "changing" directs attention to actions of substituting one thing for another, of making one thing into another thing, or of attracting one thing to become other than it was.Indeed, the types of unlearning indicate that change is an on-going process, never ceasing, and provide a continuum between continuous change and planned (discontinuous) change. Also, the magnitude of changes in beliefs and routines, in each type of unlearning, directs one's attention to the actions that can substitute one thing for another.Reinventive unlearning Reinventive unlearning indicates the fundamental changes in the existing beliefs and routines of an organization. This type of unlearning is at the heart of radical organizational change and transformation. Reinventive unlearning provides a strategic reorientation (Greenwood and Hinings, 1996), and takes organizations outside their familiar domains (Starbuck, 1996), such as radical shift in business strategy. In particular, reinventive unlearning is needed if an organization is to compete in the next generation of products or in the context of new disruptive technologies (Christensen, 1997). For instance, in the computer disk drive industry, manufacturers of 8 inch disk drives invested in process innovation to such a degree that when 5 inch disk drives began to capture the market, these manufacturers were not able to transform their production and operation procedures and beliefs (Christensen, 1997).The situation that triggers the reinventive unlearning is the high-level environmental change and unpredictability, such as "hyper" turbulent environments as indicated by D'Aveni (1984). In this type of environment, there exists a vast amount of information and a concurrent number of resulting interpretations. Also, information enters and leaves in numerous directions over short periods of time, loosing its validity in an unpredictable way.Reinventive unlearning is difficult and risky, and the outcome of reinventive unlearning is far from certain. It is expected that reinventive unlearning will be challenged when there exists:* myopia or inability of the company to look into the future with clarity;* inadequate strategic vision or lack of clear commitment of top management to changes; and* departmental politics or resistance from departments.Also, the more complacent firms are to their competitive and institutional environments, the more difficult it is for them to achieve reinventive unlearning. Tripsas and Gavetti (2000), for instance, showed how Polaroid Corporation failed to respond to the ongoing shift from analog to digital imagining. They showed how managers' beliefs, following the Gillette "razor blade" business model, delayed the commercialization of a stand-alone digital camera product.Formative unlearning Formative unlearning refers to the creation of new belief structures in organizations, with incremental changes in organizational routines. Organizations create new shared schemas and use these schemas to reinterpret information. In this type of unlearning, organizations develop new sense-making capability. For instance, Royal Dutch/Shell benefited from unlearning during the 1973 oil crisis by changing its managers' conceptual frames of references - developing new sense-making capabilities. Choo (1998, p. 16) notes that:Shell's managers have been assuming that oil demand will continue to grow at rates higher than GNP in a calm political environment where oil supply was unproblematic, a set of norms that managers had taken for granted for some time. The planning scenarios forced them to challenge these norms and to think about a low-growth world where oil consumption was increasing more slowly than GNP, where oil producers were reaching the limits of their capabilities and were reluctant to raise output further because they were unable to absorb the additional revenues.The high level of environmental change, which has numerous information sources, and more predictable change of information over time, triggers formative unlearning. In this type of environment, organizations will tend modify their strategies and rearrange their personnel.However, formative unlearning will be challenged when there exists denial or refusal to accept new belief structures. For instance, in his seminal book, Mastering the Dynamics of Innovation, Utterback (1994) noted how beliefs regarding natural ice harvesters (know what) impeded a change to machine-made ice production, because producers believed that people would prefer ice from nature rather than ice made by machines. Actually, ice harvesters were employing machines for their production and operation, indicating incremental changes in organizational routines, such as applying steam-powered circular saws for cutting blocks from rivers and ponds, and installing mechanical conveyors to provide continuous hauling of ice cakes from ponds to ice houses.Adjustive unlearning Adjustive unlearning refers to incremental changes in the existing knowledge structure of the organization, whereas fundamental changes occur in the organizational routines. This type of unlearning is related to evolutionary innovation and change and promotes new product lines, and new strategic business units. This type of unlearning includes the engineering intervention and work processes redesigns to improve the speed and quality of production. For instance, in a well known article by Imai et al. (1988) in which they discussed Honda's product-development strategy, to choose between a modified version of their Civic model or a totally new design. Even though the Civic was a successful product in the market, Honda changed its product-development mentality and designed a completely new car. During the development stage of the new Accord model, Honda carried out a design contest to identify the best car for its customers, and asked engineers and marketers around the world to help with the early stages of planning by creating a global knowledge network (Nonaka and Takeuchi, 1995).Environmental events that necessitate adjustive learning are those that have high unpredictability and low uncertainty. This type of environment indicates that information changes quickly and randomly. Organizations need to collect information rapidly about environmental events, and to distribute the information throughout the organization using new media and sensing units.Adjustive unlearning will be inhibited when there exists direct costs of change and cannibalization costs, which means change that brings success to a product but at the same time brings losses to a sister product, so it requires some kind of sacrifice.Operative unlearning Operative unlearning refers to the small-scale changes in beliefs and routines of an organization. This type of unlearning is related to incremental and continuous improvement and changes. This type of unlearning involves a limited departure from structural coherence. It helps organization maintain internal reliability and involves adjustments in systems, processes, or structures, but it does not involve fundamental change in strategy, core values, or corporate identity. Operative unlearning is most likely necessary during times of relative environmental stability and is likely to take place over extended periods of time.The environment, which stimulates operative learning, is low on uncertainty and is unpredictable. There is little information emerging and its direction is less unpredictable. In this sense, organizational change is a gradual and ongoing process which is aimed at fine tuning of the fit or match between the organizations' current strategy, structure, people and processes, and external environment.A typical example for an operative unlearning is the incremental product development activities in organizations. For instance, the HP DeskJet printer series, e.g. HP-800, developed from the adjustments to the production and commercialization processes, small-scale changes in production and design routines, and product improvement strategies to sustain the current market against the competitors - an on-going belief change efforts of management to adapt changes in competitive environment. As shown in Figure 1, this study:* denotes that unlearning can be operationalized as declarative and procedural memory eliminating - changing beliefs and routines in organizations based on the organizational change and memory literature; and* identifies different types of unlearning contingent upon the environmental conditions.Also, unlike that suggested by "common wisdom," which suggests that organizational unlearning is the opposite of organizational learning, Figure 1 shows that unlearning is an important sub-process of the organizational learning process and makes unique contributions to organizational learning literature (Akgun et al., 2003). Specifically, since, key contributions of unlearning in the organizational learning process have long been ignored in the literature, the legitimization of unlearning in the learning process was not forthcoming. Based on the organizational change and memory literature, we recognize that, unlearning:* catalyzes organizational learning process to foster a dynamic learning process;* provides a platform for shifting single-loop learning to double-loop learning; and* connects organizational learning and organizational change processes.As noted above, unlearning "catalyzes" the learning processes. The implication of the word, "catalyze" is that unlearning makes learning a dynamic process. As described by writers (Huber, 1991; Akgun et al., 2003), the organizational learning process encompasses continuous, reflexive and reciprocal cycle of activities that include memory, sensemaking, intelligence, thinking, emotions, improvisation, information acquisition, dissemination, and implementation. Since, it is a continuous process, no particular stage or sub-component can be taken as the starting point, therefore, missing or ignoring a component of the process halts the learning process in general (Akgun et al., 2003). In other words, all sub-components or processes work together rather than selectively. Unlearning facilitates a fluid and reflexive process of learning by changing beliefs and routines (Wijnhoven, 2001). For instance, organizational beliefs, which were created based on the positive feedback from a firm's success:* impose an interpretative inertia; and* facilitate inadequate reception of action signals.In addition, organizational routines, which indicate a motor, fixed or automatic action and behavior, pose an organizational inertia in changing and novel conditions. Routines reinforce the status quo (Nelson and Winter, 1982) and inhibit active seeking of alternatives. In this regard, unlearning makes the change and learning process unalterable, because old beliefs and routines were altered and replaced by a new set knowledge.In addition to the dynamic learning process, unlearning secures the learning levels. Specifically, unlearning is a means to shift from lower- (i.e. single-loop) to higher-level learning (i.e. double-loop). According to the organizational learning literature, single-loop learning or adaptive learning consists of detecting performance gaps and their elimination in line with standard operation procedures (Argyris and Schon, 1978). Here, an organization does not change its beliefs and routines. However, based on the environmental feedback, if an organization changes its beliefs and routines, one may say that double-loop learning has occurred. Here, unlearning provides a way for this transition, changing beliefs about cause-effect relations and operating procedures (Baker and Sinkula, 1999; Pawlowski, 2001). Wang and Pervaiz (2003), for instance, take a stronger stance and argue that organizational unlearning is needed to create quantum leaps.Finally, another critical role of unlearning is the interlocking of the organizational change and learning processes. There exists a common agreement that organizational change and learning are closely related concepts. For example, Pratt and Barnett (1997) have argued that learning is not a stand-alone concept; rather, it belongs to a family of change processes that encompasses unlearning and relearning (that is, adapting new mental models, procedures and routines). Indeed, based on contingency theory, the relation between an organization and its environment is the main concern of the theoretical perspectives of organizational change (Pawlowsky, 2001). At the same time, organizational learning requires changes in the relationships between organization and its environment (Hedberg and Wolff, 2001). Change involves some degree of learning, and learning provides a potential for change rather than any guarantee of its realization (Child and Heavens, 2001). Further, unlearning is one of the common dominators of change and learning process. For instance, Huy (2001) noted that beliefs and routines are the content of the change. Additionally, organizational learning concentrates on the process that brings such change about (Krebsbach-Gnath, 2001) and unlearning is one of the vital constructs of the learning process. In this perspective, unlearning acts as a linking pin between organizational learning and change processes. Unlearning is an important construct of organizational change and learning process, and thereby warrants an in-depth investigation. However, since the term "unlearning" was accompanied by the term "learning," such as "learning and unlearning," it was often seen as a demode construct or study area due to semantic usage of the terms and that it really is something different from learning, not simply its opposite. In this study, we further explored unlearning by reviewing organizational change and memory literature. However, our study is just one of a few attempts to discover the unlearning concept, see also Hedberg (1981), Nystrom and Starbuck (1984) and others. In future studies, we suggest that:1. Unlearning be investigated in the nomonological web of other organizational learning constructs. The recent research slightly mentions the niche of unlearning in the organizational learning literature (Akgun et al., 2003; Huber, 1991). For instance, according to Akgun et al. (2003), organizational learning is a process of the reciprocal relations among the information processing (i.e. memory, unlearning, sensemaking, etc.), intelligence and emotions. An empirical investigation of interwoven relations among the sub-components of learning would give a greater understanding of the process of learning at the organizational level.2. The types of unlearning can be investigated in detail using a longitudinal study. Specifically, the:* triggers or antecedents;* primary role of managers;* behavioral aspects of the practical (and unexpected) use; and* behavioral and operational consequences of each types of unlearning can be studied in case studies.3. Time processing orientation and agreement preferences should be investigated in unlearning process. Specifically, conflicts arose during changes in routines and beliefs, and resolution of those conflicts, majority and minority issues and power structures during the changes in beliefs and routines should be studied.4. How unlearning occurs should be investigated and empirically tested in groups, such as in new product development team. For instance, literature on new product development demonstrates that the groups that have strong memories are least able to deviate from previous routines during the new product development processes (Moorman and Miner, 1998). Therefore, eliminating group/team memory or unlearning, is critical for new product performance in turbulent environments/conditions and begs an empirical study. Also, empirically investigating unlearning in cross-functional new product development teams may illuminate researchers at the organizational level. Specifically, it has been asserted that new product development teams show the similar behavioral patterns as organizations do or represent microcosms of organizations based on the literature on the fractal geometry and images in system thinking (Capra, 1996). The basic philosophy of fractal images is that all characteristic patterns of the system are found repeatedly at the descending scales (Capra, 1996, p. 138). This is similar to holographic systems, which create processes where the whole can be encoded in all of the parts, so that "each and every part represents the whole" (Morgan, 1998, p. 71). In this sake, new product development team unlearning process empirically sheds light on the organizational unlearning process based on the functional isomorphism rather than homomorphic conceptualization of unlearning process, because NPD team members exist on more than one level simultaneously, act differently as units and influence each other across levels (Weiss, 1993, p. 278), and NPD process provides a more controllable, identifiable, and accurate unit definability (Meyers and Wilemon, 1986). Opens in a new window.Figure 1 Organizational unlearning Opens in a new window.Table I Types of organizational unlearning
|
- It is argued that unlearning is conceptualized as organizational memory eliminating, and is operationalized as changing beliefs and routines covariates in organizations. This is followed with a discussion of unlearning types, specifically, reinventive, formative, operative and adjustive, which are contingent on the environmental conditions. Finally, future research suggestions are proposed to leverage understanding on unlearning in the literature.
|
[SECTION: Findings] Rapid changes and unpredictable events occur in the business environment. These changes are, in part, the result of market growth or technology development, and they create turbulence that can destroy the existing competencies of an organization (Tushman and Anderson, 1986). In such a changing environment, organizations find that their previous strategies, core competencies (Prahalad and Hamel, 1990), beliefs, values, and cultures (Moorman and Miner, 1997; Sproull, 1981) are becoming less effective, or are no longer effective at all. These core competencies, which might have taken many years to develop and refine, can become core rigidities (Leonard-Barton, 1995), and can hinder an organization's ability to compete and win. Hedberg (1981) has noted that knowledge becomes obsolete under turbulent environments and must be renewed. Hedberg (1981) has defined this renewing activity as "unlearning," and has emphasized that an inability to unlearn is a critical weakness of many organizations. Hamel and Prahalad (1994, p. 71) have also noted, "Companies are going to have to unlearn a lot of their past - and also to forget it! The future will not be an extrapolation of the past."Unlearning is vital for organizations to learn how to survive and compete in the competitive landscape (Nystrom and Starbuck, 1984; Hedberg, 1981). Specifically, organizations make it more difficult to learn without first unlearning (Hedberg, 1981). Organizations have widely established and accepted beliefs and methods that persuade them to neglect important new technologies and markets, because they have a great emotional investment in old ways of working. Those established beliefs and methods create rules and competency traps that negatively affect the operations of organizations (Mezias et al., 2001). In particular, those procedures, methods and beliefs that inhibit the reception and evaluation of new market and technology information, and reduce the value of perceived new information. For instance, Day (1994, p. 24) has stated that: "The presumed correctness of past actions and interpretations is reinforced by repeated success, and ensuing complacency breeds rejection of information that conflicts with conventional wisdom." Also, people focus on information that supports their current beliefs and methods. However, incorrect beliefs and methods assimilate errors in judgment and actions because people alter the perception of reality to fit their beliefs and methods (Rousseau, 2001). Fixed beliefs lead to perception rigidity or inaccurate causal attributions, and these result in organizations becoming slower to recognize changes (Dickson, 1992). For instance, Starbuck (1996) points out that:People in organizations find it very difficult to deal effectively with information that conflicts with their current beliefs and methods. They do not know how to accommodate dissonant information and they find it difficult to change a few elements of their interdependent beliefs and methods.Even though unlearning is important, it is problematic due to difficulty of conceptualizing, operationalizing and testing it in the current scholarship. As Crossan et al. (1995) have pointed out, the "unlearning" concept has generated great interest, but it has received limited acceptance in the literature due to confusion regarding the terms "learning" and "unlearning" both in theory and in practice. In particular, confusion was enhanced when the term "unlearning" was used as the reverse of "learning." One of the reasons for this confusion is that most of the arguments on the unlearning has been conducted in the organizational learning literature (Akgun et al., 2003). However, the unlearning concept is not just bounded by or restricted to the organizational learning literature. Since, March's organizational adaptation theory (Cyert and March, 1963) and Lewin's (1951) organizational change theory, unlearning has made claim as the heart of the organizational change process (Walsh and Ungson, 1991). For example, scholars in the school of organizational change and memory pointed out that unlearning and memory are closely related concepts, because memory creates competency traps due to fixed and well-accepted routines and values in organizations (Levitt and March, 1988; Moorman and Miner, 1997). Accordingly, an understanding of unlearning can be leveraged, reducing the confusion between unlearning and learning by reviewing organizational change and memory literature.The purpose of this paper therefore is to shed light on the unlearning concept based on the organizational memory and change literature. Specifically, the research questions addressed in this study are:RQ1. How the unlearning concept can be conceptualized and operationalized for future empirical investigations.RQ2. What types of unlearning can be revealed based on the organizational memory and change literature to enhance the organizational learning and change management scholarship.In the sections that follow, this study:1. explains the concept of unlearning by reviewing different streams of research;2. argues that unlearning reflects two underlying components, changes in beliefs and routines; and then proposes that the combination of these components of unlearning creates distinct types of organizational unlearning based on environmental contingencies; and3. provides guidelines for future research. The concept of "unlearning" has attracted many researchers from diverse fields for the last 40 years with numerous studies coming from the individual learning and cognitive psychology literature. At the individual level, unlearning has been discussed as part of verbal learning psychology and individual cognition. In the verbal learning psychology literature, unlearning has primarily been viewed as deleting and replacing old stimuli by interpolated learning, which is also called interference theory (Postman et al., 1965). Scholars in this stream of research (Postman et al., 1965) have mentioned that unlearning is the gradual weakening of associations between stimuli and responses by retroactive inhibition.In addition to the verbal learning psychology, the cognitive science literature has viewed unlearning as changes in belief structure (Fiske and Taylor, 1984), mental model (Johnson-Laird, 1983), frame of reference (Shrivastava and Schneider, 1984), or cognitive maps (Walsh, 1988). Moreover, most of the studies attributable to unlearning in individual psychology and cognition have focused on memory loss (Hull, 1943), deterioration of the trace in memory (Koffka, 1935), intentional forgetting (Freud, 1943), and retroactive inhibition (Postman et al., 1965). Scholars have thus viewed unlearning as "memory eliminating" in a system (that is, the individual) (Greeno et al., 1971), and that this elimination helps bring about new learning, e.g. acceptance of novelty (new knowledge).Apart from studies at the individual level, unlearning has also been studied at the organizational level, especially in the scholarship on organizational change and memory (Lewin, 1951; Huber, 1991; Bartunek, 1988). Studies on unlearning at the organizational level reveal a similar, although not identical, pattern to that found in the scholarship on individual unlearning - that is, the elimination of memories (Walsh and Ungson, 1991). Scholars in the school of organizational memory have viewed organizational unlearning in various ways as:* changing acquisition processes and possible retrieval processes in the organizational memory (Walsh and Ungson, 1991);* disruption and re-creation of portions of the organization's memory (Anand et al., 1998);* purposely eliminating memories (Stein, 1995);* dissembling and disconfirming the causal connections in the organizational memory (Nicollini and Meznar, 1995); and* disintegrating the community's collective infrastructure of routines (Blackler et al., 1999).Thus, the common theme, to-date, for the unlearning, both in individual and organizational level studies, is:* eliminating memory via disconfirmation;* the disassembly of the connections and mechanisms of memory; and/or* changing how memory is manifested. Because unlearning has been conceptualized as memory eliminating, an investigation of how memory is formed and manifested could help in understanding and operationalizing unlearning in organizations. Moorman and Miner (1997), for instance, proposed that organizational memory is manifested in three forms:1. organizational beliefs that include knowledge, frame of references, models, values, and norms;2. formal and informal behavioral routines, procedures, and scripts that include standard operating procedural, managerial, and technical systems, capabilities, and information-sharing mechanisms; and3. the organization's physical artifacts including tools, programming, assembly-line layout, and features of products and product lines (such as product design, materials, packaging, and logos).Therefore, logically, unlearning must include eliminating beliefs, routines, and physical artifacts in organizations based on the insights garnered from the cognitive psychology, organizational memory and change literature. However, this definition is too broad for a parsimonious operationalization of unlearning, and thereby raises three concerns, namely ontological, construct validity, and pragmatistic. These three concerns clarify the operationalization of the unlearning concept in a systematic fashion. Specifically, these concerns:1. provide scrutiny and scientific explanations, which have the stature and explanatory significance assumed for unlearning;2. demonstrate what makes up the unlearning concept that are observed both in academia and practice; and3. explain variables or constructs of which the unlearning is a function by applying the scientific principles of system thinking, i.e. these concerns are mutually dependent.Ontological concern The term "eliminating" is often referred to as replacing, discarding, removing, and reducing in unlearning studies. Each verb has different meaning in english and encompasses a complex mental structure that is integrative and contains diverse reference points. For instance, when members of an organization replace their current belief with a new belief in the current organizational setting, they decide to use new belief, because it is expected to be more valid and trustworthy. However, the old belief still remains in the new organizational setting, but it is blocked or no longer considered. Unlike belief replacing, which has cognitive ramifications (Weick, 1979), discarding the current belief indicates that current belief is no longer available in the new organizational setting. The organizational actors decide to stop accepting current beliefs, and are unable to recall them from the resultant belief state represented by the new organizational setting. Nevertheless, the common theme of belief replacing, reducing, and discarding is that the stimulus or response assemblies are incongruent with the current organizational schema, and either an incremental or revolutionary schema change is needed. Indeed, belief replacing (i.e. another belief takes the place of the original) indicates an incremental schema change, whereas belief eliminating (the whole memory structure does not exist any more) shows a revolutionary schema change. In this sense, while adapting only one definition partially explains the unlearning concept, using different terms confounds the operationalization of unlearning. Therefore, for a gestalt understanding for the unlearning concept, all the terms should be combined into a task verb to define real activities or experiences of unlearning (Sandelands and Drazin, 1989). It is interesting to note that the ontological properties of unlearning - discarding, replacing and reducing, converge on the task verb of "changing" as noted by the research in artificial intelligence (AI) (Boutilier, 1998; Burns and Gomolinska, 2001) and social and cognitive psychology literature (Fiske and Taylor, 1991; Akgun et al., 2003)[1]. According to AI researchers, the change of an agent's belief state reflects a response to a change or changes in the world as well as what is perceived to be an incorrect or incomplete prior belief, leading to a new belief state (Boutilier, 1998). One might suggest that replacing, reducing or discarding of a belief is a semantic change of a belief state. Also, in the field of social psychology, Crocker et al. (1984, p. 217) state that, "The schematic approach to belief change is concerned with changes in the content and structure of social knowledge in response to information that is incongruent with the schema."The similar rationalization of discarding, replacing, and reducing for routines, and physical artifacts can be made in light of the production and operations management (Tranfield and Smith, 1998) and management literature (Pentland and Rueter, 1994). For instance, Tranfield and Smith (1998, p. 119) wrote that, "Reviewing, revising, redesigning and replacing routines either as automatic responses or effortful accomplishments lies at the heart of change management for routines." This is further supported by Mezias et al. (2001) where they stated that unlearning old rationales embedded in routines is essential to changing those routines. Accordingly, the ontological concern indicates that unlearning is the changing of beliefs, routines and physical artifacts. However, these three forms of unlearning (that is, a change in beliefs, routines, and physical artifacts) pose problems from the perspective of construct validity, leading to the second concern.Construct validity concern Since, unlearning is a cognitive process, changes in organizational artifacts, which are not cognitive, should be coupled with changes in beliefs and routines to facilitate unlearning. For instance, if a manufacturing group changes its beliefs regarding the operational methods and techniques for the production of a particular product, the group also needs to change the "machinery" in that process to remove the previous habits and knowledge. This suggests that a more comprehensive memory structure is needed to operationalize organizational unlearning. It is of interest to note that these three forms of memory structures (i.e. beliefs, routines, and physical artifacts) converge on what cognitive psychologists call "declarative and procedural memory" (Anderson, 1993). Declarative memory is related to "know-what" in organizations - that is, a memory of what needs to be done. This is a memory for facts, events, and propositions (Moorman and Miner, 1998). Moorman and Miner (1998) have indicated that a key characteristic of declarative memory is the variety of uses to which it can be put - which is similar to knowledge structure, frames of reference, mental models, values, and norms. In essence, these constitute beliefs, because beliefs explain the world in terms of cause-and-effect relationships, in interpretative schemata that map the experience of the world as mental models, and cognitive frames (Sproull, 1981; Allard-Poesi, 1998).Apart from declarative memory, another memory structure is procedural memory. Procedural memory is related to "know-how" - that is, how to do it or how things are done (Cohen and Bacdayan, 1994; Moorman and Miner, 1998). Procedural memory involves skills or routines, such as behavioral routines, procedures, tools, programming, and so on (Cohen and Bacdayan, 1994; Pentland and Rueter, 1994). Winter (1986, p. 165), for instance, has pointed out that the word "routine" is a term used generically and includes decision-making, techniques, skills, standard operational procedures, management practices, policies, strategies, information systems, information structures, programs, scripts, and organizational forms.A change in beliefs and routines as unlearning is also connected to the theories on organizational learning. Scholars (Levitt and March, 1988) have indicated that routines are a source of organizational learning. For instance, Hendry (1996, p. 636) wrote that "A way round this whole problem of learning/not learning may be through the notion of organizational routines. Routines offer a reference point to link different perspectives on organizational learning." This is particularly apropos because routines, the fabric of organizational life, shape how organizations define and solve problems and learn. Feldman (2000) also demonstrated organizational routines as a process of organizational learning. In addition to routines, organizational learning implies shared knowledge, perceptions, and beliefs. Specifically, since organizational beliefs filter environmental stimuli, they affect the way people perceive, interpret and evaluate events, how they connect the organizational capabilities with external environment, and govern "what is good and what is bad." As noted by Schein (1993), learning processes can be influenced by organizational beliefs. Unlike the beliefs and routines, which are the important part of the learning process, physical artifacts are the results of prior learning, as noted by Moorman and Miner (1997, p. 93), and are closely related to the organizational knowing rather than learning (Orlikowski, 2002). Kuh and Whitt (1988), for example, noted that physical artifacts surround people physically and provide them with the immediate sensory stimuli and memory. Accordingly, construct validity concern denotes that unlearning is the changing beliefs and routines. As demonstrated empirically, the components of unlearning should occur together for an effective unlearning (Starbuck, 1996), leading to the third concern.Pragmatistic The pragmatistic concern indicates that beliefs and routines should occur simultaneously. From a statistical point of view, changes in beliefs and routines are covariates. For instance, changing the organizational beliefs, such as anticipated market requirements, initiates unlearning. However, changes in manufacturing, distribution or logistics, should also be sanctioned by the organizations for effective organizational unlearning (Sinkula, 2002). In a real sense, changes in beliefs per se cause the deterioration of the routines and paralyze the organization, albeit temporarily, because the existing patterns of actions as indicated by routines are less likely to be congruent with the new set of beliefs (Sinkula, 2002). Similarly, substituting routines with new ones without changes in organizational beliefs also partially initiate organizational unlearning. However, since the initial filters of incoming external stimuli (existing organizational beliefs) remain, i.e. sensing and funneling the same set of stimuli to organizational members, there will be a dissonance between relevant information processing and actual behaviors.Based on the above discussion, "organizational unlearning can be operationalized as covariately changing declarative memories (beliefs) and procedural memories (routines) in organizations"[2]. Since, organizational unlearning is operationalized as changes in beliefs and routines, this operationalization is confused with organizational change as well. Indeed, organizational change and learning occurs through revision of organizational scripts, cognitive schema establishing behavior and routines, or behavioral recipes involving beliefs as indicated in action learning (Johnson, 1990), neo-institutional theory of organization (Greenwood and Hinings, 1996), social cognition (Akgun et al., 2003), and social psychology (Weick, 1979). Unlearning shares similar etiological properties with organizational change (Johnson, 1990). However, note that organizational change is a generic term and is a broad concept involving "analytical, educational, learning and political process as well as a process which combines rational, political and cultural elements" (Hendry, 1996). In essence, organizational change is an end state of a transformation process (Tsoukas and Chia, 2002). Nevertheless, unlearning in particular focuses on the memory eliminating. Organizational unlearning is a change in collective cognition and routines that coordinate organizational change process. Specifically, unlearning is a stage or catalyst (one that causes a process or an event to happen) in the change process to make it a dynamic process, as indicated by organizational change theories (Mezias et al., 2001). For instance, Lewin's (1951) model for change involves three steps:1. unfreezing, which is suspending the current structure and involves disconfirmation of expectations, learning anxiety, and provision of psychological safety;2. transition, which is changing the mental structure and involves cognitive restructuring, semantic redefinition, and new standards of judgment; and3. refreezing, which is adapting the new mental structure and involves creation of supportive social norms, and making change congruent with personality.The second stage of Lewin's model, transition, is indicative of the unlearning phenomenon. Gemmill and Smith (1985), also explained the elements of system changes as:* disequilibrium, which is the result of external and internal forces, such as crisis and chaos;* symmetry breaking, which refers to breaking down of existing patterns of interactions or system habits;* experimentation, which refers to creating different new forms or configuration to reformulate the system; and* reformulation, which is selecting a new configuration.In their model, the system breaking stage denotes the unlearning concept.For that reason, we purport that "unlearning is embedded in organizational change process. However, the aim of unlearning is not performance improvement per se; rather it is a catalyst for the change process." We argued that unlearning involves the combination of the changes in beliefs and routines and these two components of unlearning must exist in tandem for unlearning to occur effectively. Nevertheless, in practice, the magnitude of the changes in beliefs and routines may vary. For instance, an organization is likely to put a higher emphasis on changes in routines than on changes in beliefs. One of the reasons would be the environmental conditions, in particular, environmental turbulence (Schein, 1993; Hedberg, 1981; Nystrom and Starbuck, 1981). However, considering environmental turbulence as a unidimensional construct narrows our understanding of the dynamic unlearning process, requiring a multidimensional view of environmental turbulence. Environmental turbulence refers to the degree of change and unpredictability of an environment (Glazer and Weiss, 1993, p. 510). The degree of change involves more information, leading to multiple and conflicted interpretations about an organizational situation. This type of contingency requires more interpretation and a variety of schemas or belief structures. Unpredictability of environment is related to the time-sensitivity of the information. Information in a given period loses its value in subsequent periods. Since, different environmental conditions require organizations to use the appropriate unlearning process, we propose that environmental turbulence results in four types of unlearning process as shown in Table I. The labels used for unlearning are borrowed from Gnyawali and Stewart (2003). Reinventive unlearning occurs when organizations put high emphasis on both changes in beliefs and routines. Formative unlearning is when more emphasis is placed on the beliefs and less on the routines. High emphasis on the changes in routines and low on the changes in beliefs results in adjustive unlearning, and low emphasis on both changes in beliefs and routines results in operative unlearning.It is interesting to note that these types of unlearning make good connections with the literature on planned (discontinuous) change. The proposed types of unlearning provide a meeting point between theories of planned change and theories of continuous change as Weick and Quinn (1999) stated. For instance, Weick and Quinn (1999, pp. 381-2) pointed out that:Recent work suggests, ironically, that to understand organizational change one must first understand organizational inertia, its content, its tenacity, its interdependencies. Recent work also suggests that change is not an on-off phenomenon nor is its effectiveness contingent on the degree to which it is planned. Furthermore, the trajectory of change is more often spiral or open ended than linear. All of these insights are more likely to be kept in play if researchers focus on "changing" rather than "change." A shift in vocabulary from "change" to "changing" directs attention to actions of substituting one thing for another, of making one thing into another thing, or of attracting one thing to become other than it was.Indeed, the types of unlearning indicate that change is an on-going process, never ceasing, and provide a continuum between continuous change and planned (discontinuous) change. Also, the magnitude of changes in beliefs and routines, in each type of unlearning, directs one's attention to the actions that can substitute one thing for another.Reinventive unlearning Reinventive unlearning indicates the fundamental changes in the existing beliefs and routines of an organization. This type of unlearning is at the heart of radical organizational change and transformation. Reinventive unlearning provides a strategic reorientation (Greenwood and Hinings, 1996), and takes organizations outside their familiar domains (Starbuck, 1996), such as radical shift in business strategy. In particular, reinventive unlearning is needed if an organization is to compete in the next generation of products or in the context of new disruptive technologies (Christensen, 1997). For instance, in the computer disk drive industry, manufacturers of 8 inch disk drives invested in process innovation to such a degree that when 5 inch disk drives began to capture the market, these manufacturers were not able to transform their production and operation procedures and beliefs (Christensen, 1997).The situation that triggers the reinventive unlearning is the high-level environmental change and unpredictability, such as "hyper" turbulent environments as indicated by D'Aveni (1984). In this type of environment, there exists a vast amount of information and a concurrent number of resulting interpretations. Also, information enters and leaves in numerous directions over short periods of time, loosing its validity in an unpredictable way.Reinventive unlearning is difficult and risky, and the outcome of reinventive unlearning is far from certain. It is expected that reinventive unlearning will be challenged when there exists:* myopia or inability of the company to look into the future with clarity;* inadequate strategic vision or lack of clear commitment of top management to changes; and* departmental politics or resistance from departments.Also, the more complacent firms are to their competitive and institutional environments, the more difficult it is for them to achieve reinventive unlearning. Tripsas and Gavetti (2000), for instance, showed how Polaroid Corporation failed to respond to the ongoing shift from analog to digital imagining. They showed how managers' beliefs, following the Gillette "razor blade" business model, delayed the commercialization of a stand-alone digital camera product.Formative unlearning Formative unlearning refers to the creation of new belief structures in organizations, with incremental changes in organizational routines. Organizations create new shared schemas and use these schemas to reinterpret information. In this type of unlearning, organizations develop new sense-making capability. For instance, Royal Dutch/Shell benefited from unlearning during the 1973 oil crisis by changing its managers' conceptual frames of references - developing new sense-making capabilities. Choo (1998, p. 16) notes that:Shell's managers have been assuming that oil demand will continue to grow at rates higher than GNP in a calm political environment where oil supply was unproblematic, a set of norms that managers had taken for granted for some time. The planning scenarios forced them to challenge these norms and to think about a low-growth world where oil consumption was increasing more slowly than GNP, where oil producers were reaching the limits of their capabilities and were reluctant to raise output further because they were unable to absorb the additional revenues.The high level of environmental change, which has numerous information sources, and more predictable change of information over time, triggers formative unlearning. In this type of environment, organizations will tend modify their strategies and rearrange their personnel.However, formative unlearning will be challenged when there exists denial or refusal to accept new belief structures. For instance, in his seminal book, Mastering the Dynamics of Innovation, Utterback (1994) noted how beliefs regarding natural ice harvesters (know what) impeded a change to machine-made ice production, because producers believed that people would prefer ice from nature rather than ice made by machines. Actually, ice harvesters were employing machines for their production and operation, indicating incremental changes in organizational routines, such as applying steam-powered circular saws for cutting blocks from rivers and ponds, and installing mechanical conveyors to provide continuous hauling of ice cakes from ponds to ice houses.Adjustive unlearning Adjustive unlearning refers to incremental changes in the existing knowledge structure of the organization, whereas fundamental changes occur in the organizational routines. This type of unlearning is related to evolutionary innovation and change and promotes new product lines, and new strategic business units. This type of unlearning includes the engineering intervention and work processes redesigns to improve the speed and quality of production. For instance, in a well known article by Imai et al. (1988) in which they discussed Honda's product-development strategy, to choose between a modified version of their Civic model or a totally new design. Even though the Civic was a successful product in the market, Honda changed its product-development mentality and designed a completely new car. During the development stage of the new Accord model, Honda carried out a design contest to identify the best car for its customers, and asked engineers and marketers around the world to help with the early stages of planning by creating a global knowledge network (Nonaka and Takeuchi, 1995).Environmental events that necessitate adjustive learning are those that have high unpredictability and low uncertainty. This type of environment indicates that information changes quickly and randomly. Organizations need to collect information rapidly about environmental events, and to distribute the information throughout the organization using new media and sensing units.Adjustive unlearning will be inhibited when there exists direct costs of change and cannibalization costs, which means change that brings success to a product but at the same time brings losses to a sister product, so it requires some kind of sacrifice.Operative unlearning Operative unlearning refers to the small-scale changes in beliefs and routines of an organization. This type of unlearning is related to incremental and continuous improvement and changes. This type of unlearning involves a limited departure from structural coherence. It helps organization maintain internal reliability and involves adjustments in systems, processes, or structures, but it does not involve fundamental change in strategy, core values, or corporate identity. Operative unlearning is most likely necessary during times of relative environmental stability and is likely to take place over extended periods of time.The environment, which stimulates operative learning, is low on uncertainty and is unpredictable. There is little information emerging and its direction is less unpredictable. In this sense, organizational change is a gradual and ongoing process which is aimed at fine tuning of the fit or match between the organizations' current strategy, structure, people and processes, and external environment.A typical example for an operative unlearning is the incremental product development activities in organizations. For instance, the HP DeskJet printer series, e.g. HP-800, developed from the adjustments to the production and commercialization processes, small-scale changes in production and design routines, and product improvement strategies to sustain the current market against the competitors - an on-going belief change efforts of management to adapt changes in competitive environment. As shown in Figure 1, this study:* denotes that unlearning can be operationalized as declarative and procedural memory eliminating - changing beliefs and routines in organizations based on the organizational change and memory literature; and* identifies different types of unlearning contingent upon the environmental conditions.Also, unlike that suggested by "common wisdom," which suggests that organizational unlearning is the opposite of organizational learning, Figure 1 shows that unlearning is an important sub-process of the organizational learning process and makes unique contributions to organizational learning literature (Akgun et al., 2003). Specifically, since, key contributions of unlearning in the organizational learning process have long been ignored in the literature, the legitimization of unlearning in the learning process was not forthcoming. Based on the organizational change and memory literature, we recognize that, unlearning:* catalyzes organizational learning process to foster a dynamic learning process;* provides a platform for shifting single-loop learning to double-loop learning; and* connects organizational learning and organizational change processes.As noted above, unlearning "catalyzes" the learning processes. The implication of the word, "catalyze" is that unlearning makes learning a dynamic process. As described by writers (Huber, 1991; Akgun et al., 2003), the organizational learning process encompasses continuous, reflexive and reciprocal cycle of activities that include memory, sensemaking, intelligence, thinking, emotions, improvisation, information acquisition, dissemination, and implementation. Since, it is a continuous process, no particular stage or sub-component can be taken as the starting point, therefore, missing or ignoring a component of the process halts the learning process in general (Akgun et al., 2003). In other words, all sub-components or processes work together rather than selectively. Unlearning facilitates a fluid and reflexive process of learning by changing beliefs and routines (Wijnhoven, 2001). For instance, organizational beliefs, which were created based on the positive feedback from a firm's success:* impose an interpretative inertia; and* facilitate inadequate reception of action signals.In addition, organizational routines, which indicate a motor, fixed or automatic action and behavior, pose an organizational inertia in changing and novel conditions. Routines reinforce the status quo (Nelson and Winter, 1982) and inhibit active seeking of alternatives. In this regard, unlearning makes the change and learning process unalterable, because old beliefs and routines were altered and replaced by a new set knowledge.In addition to the dynamic learning process, unlearning secures the learning levels. Specifically, unlearning is a means to shift from lower- (i.e. single-loop) to higher-level learning (i.e. double-loop). According to the organizational learning literature, single-loop learning or adaptive learning consists of detecting performance gaps and their elimination in line with standard operation procedures (Argyris and Schon, 1978). Here, an organization does not change its beliefs and routines. However, based on the environmental feedback, if an organization changes its beliefs and routines, one may say that double-loop learning has occurred. Here, unlearning provides a way for this transition, changing beliefs about cause-effect relations and operating procedures (Baker and Sinkula, 1999; Pawlowski, 2001). Wang and Pervaiz (2003), for instance, take a stronger stance and argue that organizational unlearning is needed to create quantum leaps.Finally, another critical role of unlearning is the interlocking of the organizational change and learning processes. There exists a common agreement that organizational change and learning are closely related concepts. For example, Pratt and Barnett (1997) have argued that learning is not a stand-alone concept; rather, it belongs to a family of change processes that encompasses unlearning and relearning (that is, adapting new mental models, procedures and routines). Indeed, based on contingency theory, the relation between an organization and its environment is the main concern of the theoretical perspectives of organizational change (Pawlowsky, 2001). At the same time, organizational learning requires changes in the relationships between organization and its environment (Hedberg and Wolff, 2001). Change involves some degree of learning, and learning provides a potential for change rather than any guarantee of its realization (Child and Heavens, 2001). Further, unlearning is one of the common dominators of change and learning process. For instance, Huy (2001) noted that beliefs and routines are the content of the change. Additionally, organizational learning concentrates on the process that brings such change about (Krebsbach-Gnath, 2001) and unlearning is one of the vital constructs of the learning process. In this perspective, unlearning acts as a linking pin between organizational learning and change processes. Unlearning is an important construct of organizational change and learning process, and thereby warrants an in-depth investigation. However, since the term "unlearning" was accompanied by the term "learning," such as "learning and unlearning," it was often seen as a demode construct or study area due to semantic usage of the terms and that it really is something different from learning, not simply its opposite. In this study, we further explored unlearning by reviewing organizational change and memory literature. However, our study is just one of a few attempts to discover the unlearning concept, see also Hedberg (1981), Nystrom and Starbuck (1984) and others. In future studies, we suggest that:1. Unlearning be investigated in the nomonological web of other organizational learning constructs. The recent research slightly mentions the niche of unlearning in the organizational learning literature (Akgun et al., 2003; Huber, 1991). For instance, according to Akgun et al. (2003), organizational learning is a process of the reciprocal relations among the information processing (i.e. memory, unlearning, sensemaking, etc.), intelligence and emotions. An empirical investigation of interwoven relations among the sub-components of learning would give a greater understanding of the process of learning at the organizational level.2. The types of unlearning can be investigated in detail using a longitudinal study. Specifically, the:* triggers or antecedents;* primary role of managers;* behavioral aspects of the practical (and unexpected) use; and* behavioral and operational consequences of each types of unlearning can be studied in case studies.3. Time processing orientation and agreement preferences should be investigated in unlearning process. Specifically, conflicts arose during changes in routines and beliefs, and resolution of those conflicts, majority and minority issues and power structures during the changes in beliefs and routines should be studied.4. How unlearning occurs should be investigated and empirically tested in groups, such as in new product development team. For instance, literature on new product development demonstrates that the groups that have strong memories are least able to deviate from previous routines during the new product development processes (Moorman and Miner, 1998). Therefore, eliminating group/team memory or unlearning, is critical for new product performance in turbulent environments/conditions and begs an empirical study. Also, empirically investigating unlearning in cross-functional new product development teams may illuminate researchers at the organizational level. Specifically, it has been asserted that new product development teams show the similar behavioral patterns as organizations do or represent microcosms of organizations based on the literature on the fractal geometry and images in system thinking (Capra, 1996). The basic philosophy of fractal images is that all characteristic patterns of the system are found repeatedly at the descending scales (Capra, 1996, p. 138). This is similar to holographic systems, which create processes where the whole can be encoded in all of the parts, so that "each and every part represents the whole" (Morgan, 1998, p. 71). In this sake, new product development team unlearning process empirically sheds light on the organizational unlearning process based on the functional isomorphism rather than homomorphic conceptualization of unlearning process, because NPD team members exist on more than one level simultaneously, act differently as units and influence each other across levels (Weiss, 1993, p. 278), and NPD process provides a more controllable, identifiable, and accurate unit definability (Meyers and Wilemon, 1986). Opens in a new window.Figure 1 Organizational unlearning Opens in a new window.Table I Types of organizational unlearning
|
- Shows that organizations first need to unlearn established beliefs and methods which have created rules and competency traps, in order to be receptive to new market and technology information.
|
[SECTION: Value] Rapid changes and unpredictable events occur in the business environment. These changes are, in part, the result of market growth or technology development, and they create turbulence that can destroy the existing competencies of an organization (Tushman and Anderson, 1986). In such a changing environment, organizations find that their previous strategies, core competencies (Prahalad and Hamel, 1990), beliefs, values, and cultures (Moorman and Miner, 1997; Sproull, 1981) are becoming less effective, or are no longer effective at all. These core competencies, which might have taken many years to develop and refine, can become core rigidities (Leonard-Barton, 1995), and can hinder an organization's ability to compete and win. Hedberg (1981) has noted that knowledge becomes obsolete under turbulent environments and must be renewed. Hedberg (1981) has defined this renewing activity as "unlearning," and has emphasized that an inability to unlearn is a critical weakness of many organizations. Hamel and Prahalad (1994, p. 71) have also noted, "Companies are going to have to unlearn a lot of their past - and also to forget it! The future will not be an extrapolation of the past."Unlearning is vital for organizations to learn how to survive and compete in the competitive landscape (Nystrom and Starbuck, 1984; Hedberg, 1981). Specifically, organizations make it more difficult to learn without first unlearning (Hedberg, 1981). Organizations have widely established and accepted beliefs and methods that persuade them to neglect important new technologies and markets, because they have a great emotional investment in old ways of working. Those established beliefs and methods create rules and competency traps that negatively affect the operations of organizations (Mezias et al., 2001). In particular, those procedures, methods and beliefs that inhibit the reception and evaluation of new market and technology information, and reduce the value of perceived new information. For instance, Day (1994, p. 24) has stated that: "The presumed correctness of past actions and interpretations is reinforced by repeated success, and ensuing complacency breeds rejection of information that conflicts with conventional wisdom." Also, people focus on information that supports their current beliefs and methods. However, incorrect beliefs and methods assimilate errors in judgment and actions because people alter the perception of reality to fit their beliefs and methods (Rousseau, 2001). Fixed beliefs lead to perception rigidity or inaccurate causal attributions, and these result in organizations becoming slower to recognize changes (Dickson, 1992). For instance, Starbuck (1996) points out that:People in organizations find it very difficult to deal effectively with information that conflicts with their current beliefs and methods. They do not know how to accommodate dissonant information and they find it difficult to change a few elements of their interdependent beliefs and methods.Even though unlearning is important, it is problematic due to difficulty of conceptualizing, operationalizing and testing it in the current scholarship. As Crossan et al. (1995) have pointed out, the "unlearning" concept has generated great interest, but it has received limited acceptance in the literature due to confusion regarding the terms "learning" and "unlearning" both in theory and in practice. In particular, confusion was enhanced when the term "unlearning" was used as the reverse of "learning." One of the reasons for this confusion is that most of the arguments on the unlearning has been conducted in the organizational learning literature (Akgun et al., 2003). However, the unlearning concept is not just bounded by or restricted to the organizational learning literature. Since, March's organizational adaptation theory (Cyert and March, 1963) and Lewin's (1951) organizational change theory, unlearning has made claim as the heart of the organizational change process (Walsh and Ungson, 1991). For example, scholars in the school of organizational change and memory pointed out that unlearning and memory are closely related concepts, because memory creates competency traps due to fixed and well-accepted routines and values in organizations (Levitt and March, 1988; Moorman and Miner, 1997). Accordingly, an understanding of unlearning can be leveraged, reducing the confusion between unlearning and learning by reviewing organizational change and memory literature.The purpose of this paper therefore is to shed light on the unlearning concept based on the organizational memory and change literature. Specifically, the research questions addressed in this study are:RQ1. How the unlearning concept can be conceptualized and operationalized for future empirical investigations.RQ2. What types of unlearning can be revealed based on the organizational memory and change literature to enhance the organizational learning and change management scholarship.In the sections that follow, this study:1. explains the concept of unlearning by reviewing different streams of research;2. argues that unlearning reflects two underlying components, changes in beliefs and routines; and then proposes that the combination of these components of unlearning creates distinct types of organizational unlearning based on environmental contingencies; and3. provides guidelines for future research. The concept of "unlearning" has attracted many researchers from diverse fields for the last 40 years with numerous studies coming from the individual learning and cognitive psychology literature. At the individual level, unlearning has been discussed as part of verbal learning psychology and individual cognition. In the verbal learning psychology literature, unlearning has primarily been viewed as deleting and replacing old stimuli by interpolated learning, which is also called interference theory (Postman et al., 1965). Scholars in this stream of research (Postman et al., 1965) have mentioned that unlearning is the gradual weakening of associations between stimuli and responses by retroactive inhibition.In addition to the verbal learning psychology, the cognitive science literature has viewed unlearning as changes in belief structure (Fiske and Taylor, 1984), mental model (Johnson-Laird, 1983), frame of reference (Shrivastava and Schneider, 1984), or cognitive maps (Walsh, 1988). Moreover, most of the studies attributable to unlearning in individual psychology and cognition have focused on memory loss (Hull, 1943), deterioration of the trace in memory (Koffka, 1935), intentional forgetting (Freud, 1943), and retroactive inhibition (Postman et al., 1965). Scholars have thus viewed unlearning as "memory eliminating" in a system (that is, the individual) (Greeno et al., 1971), and that this elimination helps bring about new learning, e.g. acceptance of novelty (new knowledge).Apart from studies at the individual level, unlearning has also been studied at the organizational level, especially in the scholarship on organizational change and memory (Lewin, 1951; Huber, 1991; Bartunek, 1988). Studies on unlearning at the organizational level reveal a similar, although not identical, pattern to that found in the scholarship on individual unlearning - that is, the elimination of memories (Walsh and Ungson, 1991). Scholars in the school of organizational memory have viewed organizational unlearning in various ways as:* changing acquisition processes and possible retrieval processes in the organizational memory (Walsh and Ungson, 1991);* disruption and re-creation of portions of the organization's memory (Anand et al., 1998);* purposely eliminating memories (Stein, 1995);* dissembling and disconfirming the causal connections in the organizational memory (Nicollini and Meznar, 1995); and* disintegrating the community's collective infrastructure of routines (Blackler et al., 1999).Thus, the common theme, to-date, for the unlearning, both in individual and organizational level studies, is:* eliminating memory via disconfirmation;* the disassembly of the connections and mechanisms of memory; and/or* changing how memory is manifested. Because unlearning has been conceptualized as memory eliminating, an investigation of how memory is formed and manifested could help in understanding and operationalizing unlearning in organizations. Moorman and Miner (1997), for instance, proposed that organizational memory is manifested in three forms:1. organizational beliefs that include knowledge, frame of references, models, values, and norms;2. formal and informal behavioral routines, procedures, and scripts that include standard operating procedural, managerial, and technical systems, capabilities, and information-sharing mechanisms; and3. the organization's physical artifacts including tools, programming, assembly-line layout, and features of products and product lines (such as product design, materials, packaging, and logos).Therefore, logically, unlearning must include eliminating beliefs, routines, and physical artifacts in organizations based on the insights garnered from the cognitive psychology, organizational memory and change literature. However, this definition is too broad for a parsimonious operationalization of unlearning, and thereby raises three concerns, namely ontological, construct validity, and pragmatistic. These three concerns clarify the operationalization of the unlearning concept in a systematic fashion. Specifically, these concerns:1. provide scrutiny and scientific explanations, which have the stature and explanatory significance assumed for unlearning;2. demonstrate what makes up the unlearning concept that are observed both in academia and practice; and3. explain variables or constructs of which the unlearning is a function by applying the scientific principles of system thinking, i.e. these concerns are mutually dependent.Ontological concern The term "eliminating" is often referred to as replacing, discarding, removing, and reducing in unlearning studies. Each verb has different meaning in english and encompasses a complex mental structure that is integrative and contains diverse reference points. For instance, when members of an organization replace their current belief with a new belief in the current organizational setting, they decide to use new belief, because it is expected to be more valid and trustworthy. However, the old belief still remains in the new organizational setting, but it is blocked or no longer considered. Unlike belief replacing, which has cognitive ramifications (Weick, 1979), discarding the current belief indicates that current belief is no longer available in the new organizational setting. The organizational actors decide to stop accepting current beliefs, and are unable to recall them from the resultant belief state represented by the new organizational setting. Nevertheless, the common theme of belief replacing, reducing, and discarding is that the stimulus or response assemblies are incongruent with the current organizational schema, and either an incremental or revolutionary schema change is needed. Indeed, belief replacing (i.e. another belief takes the place of the original) indicates an incremental schema change, whereas belief eliminating (the whole memory structure does not exist any more) shows a revolutionary schema change. In this sense, while adapting only one definition partially explains the unlearning concept, using different terms confounds the operationalization of unlearning. Therefore, for a gestalt understanding for the unlearning concept, all the terms should be combined into a task verb to define real activities or experiences of unlearning (Sandelands and Drazin, 1989). It is interesting to note that the ontological properties of unlearning - discarding, replacing and reducing, converge on the task verb of "changing" as noted by the research in artificial intelligence (AI) (Boutilier, 1998; Burns and Gomolinska, 2001) and social and cognitive psychology literature (Fiske and Taylor, 1991; Akgun et al., 2003)[1]. According to AI researchers, the change of an agent's belief state reflects a response to a change or changes in the world as well as what is perceived to be an incorrect or incomplete prior belief, leading to a new belief state (Boutilier, 1998). One might suggest that replacing, reducing or discarding of a belief is a semantic change of a belief state. Also, in the field of social psychology, Crocker et al. (1984, p. 217) state that, "The schematic approach to belief change is concerned with changes in the content and structure of social knowledge in response to information that is incongruent with the schema."The similar rationalization of discarding, replacing, and reducing for routines, and physical artifacts can be made in light of the production and operations management (Tranfield and Smith, 1998) and management literature (Pentland and Rueter, 1994). For instance, Tranfield and Smith (1998, p. 119) wrote that, "Reviewing, revising, redesigning and replacing routines either as automatic responses or effortful accomplishments lies at the heart of change management for routines." This is further supported by Mezias et al. (2001) where they stated that unlearning old rationales embedded in routines is essential to changing those routines. Accordingly, the ontological concern indicates that unlearning is the changing of beliefs, routines and physical artifacts. However, these three forms of unlearning (that is, a change in beliefs, routines, and physical artifacts) pose problems from the perspective of construct validity, leading to the second concern.Construct validity concern Since, unlearning is a cognitive process, changes in organizational artifacts, which are not cognitive, should be coupled with changes in beliefs and routines to facilitate unlearning. For instance, if a manufacturing group changes its beliefs regarding the operational methods and techniques for the production of a particular product, the group also needs to change the "machinery" in that process to remove the previous habits and knowledge. This suggests that a more comprehensive memory structure is needed to operationalize organizational unlearning. It is of interest to note that these three forms of memory structures (i.e. beliefs, routines, and physical artifacts) converge on what cognitive psychologists call "declarative and procedural memory" (Anderson, 1993). Declarative memory is related to "know-what" in organizations - that is, a memory of what needs to be done. This is a memory for facts, events, and propositions (Moorman and Miner, 1998). Moorman and Miner (1998) have indicated that a key characteristic of declarative memory is the variety of uses to which it can be put - which is similar to knowledge structure, frames of reference, mental models, values, and norms. In essence, these constitute beliefs, because beliefs explain the world in terms of cause-and-effect relationships, in interpretative schemata that map the experience of the world as mental models, and cognitive frames (Sproull, 1981; Allard-Poesi, 1998).Apart from declarative memory, another memory structure is procedural memory. Procedural memory is related to "know-how" - that is, how to do it or how things are done (Cohen and Bacdayan, 1994; Moorman and Miner, 1998). Procedural memory involves skills or routines, such as behavioral routines, procedures, tools, programming, and so on (Cohen and Bacdayan, 1994; Pentland and Rueter, 1994). Winter (1986, p. 165), for instance, has pointed out that the word "routine" is a term used generically and includes decision-making, techniques, skills, standard operational procedures, management practices, policies, strategies, information systems, information structures, programs, scripts, and organizational forms.A change in beliefs and routines as unlearning is also connected to the theories on organizational learning. Scholars (Levitt and March, 1988) have indicated that routines are a source of organizational learning. For instance, Hendry (1996, p. 636) wrote that "A way round this whole problem of learning/not learning may be through the notion of organizational routines. Routines offer a reference point to link different perspectives on organizational learning." This is particularly apropos because routines, the fabric of organizational life, shape how organizations define and solve problems and learn. Feldman (2000) also demonstrated organizational routines as a process of organizational learning. In addition to routines, organizational learning implies shared knowledge, perceptions, and beliefs. Specifically, since organizational beliefs filter environmental stimuli, they affect the way people perceive, interpret and evaluate events, how they connect the organizational capabilities with external environment, and govern "what is good and what is bad." As noted by Schein (1993), learning processes can be influenced by organizational beliefs. Unlike the beliefs and routines, which are the important part of the learning process, physical artifacts are the results of prior learning, as noted by Moorman and Miner (1997, p. 93), and are closely related to the organizational knowing rather than learning (Orlikowski, 2002). Kuh and Whitt (1988), for example, noted that physical artifacts surround people physically and provide them with the immediate sensory stimuli and memory. Accordingly, construct validity concern denotes that unlearning is the changing beliefs and routines. As demonstrated empirically, the components of unlearning should occur together for an effective unlearning (Starbuck, 1996), leading to the third concern.Pragmatistic The pragmatistic concern indicates that beliefs and routines should occur simultaneously. From a statistical point of view, changes in beliefs and routines are covariates. For instance, changing the organizational beliefs, such as anticipated market requirements, initiates unlearning. However, changes in manufacturing, distribution or logistics, should also be sanctioned by the organizations for effective organizational unlearning (Sinkula, 2002). In a real sense, changes in beliefs per se cause the deterioration of the routines and paralyze the organization, albeit temporarily, because the existing patterns of actions as indicated by routines are less likely to be congruent with the new set of beliefs (Sinkula, 2002). Similarly, substituting routines with new ones without changes in organizational beliefs also partially initiate organizational unlearning. However, since the initial filters of incoming external stimuli (existing organizational beliefs) remain, i.e. sensing and funneling the same set of stimuli to organizational members, there will be a dissonance between relevant information processing and actual behaviors.Based on the above discussion, "organizational unlearning can be operationalized as covariately changing declarative memories (beliefs) and procedural memories (routines) in organizations"[2]. Since, organizational unlearning is operationalized as changes in beliefs and routines, this operationalization is confused with organizational change as well. Indeed, organizational change and learning occurs through revision of organizational scripts, cognitive schema establishing behavior and routines, or behavioral recipes involving beliefs as indicated in action learning (Johnson, 1990), neo-institutional theory of organization (Greenwood and Hinings, 1996), social cognition (Akgun et al., 2003), and social psychology (Weick, 1979). Unlearning shares similar etiological properties with organizational change (Johnson, 1990). However, note that organizational change is a generic term and is a broad concept involving "analytical, educational, learning and political process as well as a process which combines rational, political and cultural elements" (Hendry, 1996). In essence, organizational change is an end state of a transformation process (Tsoukas and Chia, 2002). Nevertheless, unlearning in particular focuses on the memory eliminating. Organizational unlearning is a change in collective cognition and routines that coordinate organizational change process. Specifically, unlearning is a stage or catalyst (one that causes a process or an event to happen) in the change process to make it a dynamic process, as indicated by organizational change theories (Mezias et al., 2001). For instance, Lewin's (1951) model for change involves three steps:1. unfreezing, which is suspending the current structure and involves disconfirmation of expectations, learning anxiety, and provision of psychological safety;2. transition, which is changing the mental structure and involves cognitive restructuring, semantic redefinition, and new standards of judgment; and3. refreezing, which is adapting the new mental structure and involves creation of supportive social norms, and making change congruent with personality.The second stage of Lewin's model, transition, is indicative of the unlearning phenomenon. Gemmill and Smith (1985), also explained the elements of system changes as:* disequilibrium, which is the result of external and internal forces, such as crisis and chaos;* symmetry breaking, which refers to breaking down of existing patterns of interactions or system habits;* experimentation, which refers to creating different new forms or configuration to reformulate the system; and* reformulation, which is selecting a new configuration.In their model, the system breaking stage denotes the unlearning concept.For that reason, we purport that "unlearning is embedded in organizational change process. However, the aim of unlearning is not performance improvement per se; rather it is a catalyst for the change process." We argued that unlearning involves the combination of the changes in beliefs and routines and these two components of unlearning must exist in tandem for unlearning to occur effectively. Nevertheless, in practice, the magnitude of the changes in beliefs and routines may vary. For instance, an organization is likely to put a higher emphasis on changes in routines than on changes in beliefs. One of the reasons would be the environmental conditions, in particular, environmental turbulence (Schein, 1993; Hedberg, 1981; Nystrom and Starbuck, 1981). However, considering environmental turbulence as a unidimensional construct narrows our understanding of the dynamic unlearning process, requiring a multidimensional view of environmental turbulence. Environmental turbulence refers to the degree of change and unpredictability of an environment (Glazer and Weiss, 1993, p. 510). The degree of change involves more information, leading to multiple and conflicted interpretations about an organizational situation. This type of contingency requires more interpretation and a variety of schemas or belief structures. Unpredictability of environment is related to the time-sensitivity of the information. Information in a given period loses its value in subsequent periods. Since, different environmental conditions require organizations to use the appropriate unlearning process, we propose that environmental turbulence results in four types of unlearning process as shown in Table I. The labels used for unlearning are borrowed from Gnyawali and Stewart (2003). Reinventive unlearning occurs when organizations put high emphasis on both changes in beliefs and routines. Formative unlearning is when more emphasis is placed on the beliefs and less on the routines. High emphasis on the changes in routines and low on the changes in beliefs results in adjustive unlearning, and low emphasis on both changes in beliefs and routines results in operative unlearning.It is interesting to note that these types of unlearning make good connections with the literature on planned (discontinuous) change. The proposed types of unlearning provide a meeting point between theories of planned change and theories of continuous change as Weick and Quinn (1999) stated. For instance, Weick and Quinn (1999, pp. 381-2) pointed out that:Recent work suggests, ironically, that to understand organizational change one must first understand organizational inertia, its content, its tenacity, its interdependencies. Recent work also suggests that change is not an on-off phenomenon nor is its effectiveness contingent on the degree to which it is planned. Furthermore, the trajectory of change is more often spiral or open ended than linear. All of these insights are more likely to be kept in play if researchers focus on "changing" rather than "change." A shift in vocabulary from "change" to "changing" directs attention to actions of substituting one thing for another, of making one thing into another thing, or of attracting one thing to become other than it was.Indeed, the types of unlearning indicate that change is an on-going process, never ceasing, and provide a continuum between continuous change and planned (discontinuous) change. Also, the magnitude of changes in beliefs and routines, in each type of unlearning, directs one's attention to the actions that can substitute one thing for another.Reinventive unlearning Reinventive unlearning indicates the fundamental changes in the existing beliefs and routines of an organization. This type of unlearning is at the heart of radical organizational change and transformation. Reinventive unlearning provides a strategic reorientation (Greenwood and Hinings, 1996), and takes organizations outside their familiar domains (Starbuck, 1996), such as radical shift in business strategy. In particular, reinventive unlearning is needed if an organization is to compete in the next generation of products or in the context of new disruptive technologies (Christensen, 1997). For instance, in the computer disk drive industry, manufacturers of 8 inch disk drives invested in process innovation to such a degree that when 5 inch disk drives began to capture the market, these manufacturers were not able to transform their production and operation procedures and beliefs (Christensen, 1997).The situation that triggers the reinventive unlearning is the high-level environmental change and unpredictability, such as "hyper" turbulent environments as indicated by D'Aveni (1984). In this type of environment, there exists a vast amount of information and a concurrent number of resulting interpretations. Also, information enters and leaves in numerous directions over short periods of time, loosing its validity in an unpredictable way.Reinventive unlearning is difficult and risky, and the outcome of reinventive unlearning is far from certain. It is expected that reinventive unlearning will be challenged when there exists:* myopia or inability of the company to look into the future with clarity;* inadequate strategic vision or lack of clear commitment of top management to changes; and* departmental politics or resistance from departments.Also, the more complacent firms are to their competitive and institutional environments, the more difficult it is for them to achieve reinventive unlearning. Tripsas and Gavetti (2000), for instance, showed how Polaroid Corporation failed to respond to the ongoing shift from analog to digital imagining. They showed how managers' beliefs, following the Gillette "razor blade" business model, delayed the commercialization of a stand-alone digital camera product.Formative unlearning Formative unlearning refers to the creation of new belief structures in organizations, with incremental changes in organizational routines. Organizations create new shared schemas and use these schemas to reinterpret information. In this type of unlearning, organizations develop new sense-making capability. For instance, Royal Dutch/Shell benefited from unlearning during the 1973 oil crisis by changing its managers' conceptual frames of references - developing new sense-making capabilities. Choo (1998, p. 16) notes that:Shell's managers have been assuming that oil demand will continue to grow at rates higher than GNP in a calm political environment where oil supply was unproblematic, a set of norms that managers had taken for granted for some time. The planning scenarios forced them to challenge these norms and to think about a low-growth world where oil consumption was increasing more slowly than GNP, where oil producers were reaching the limits of their capabilities and were reluctant to raise output further because they were unable to absorb the additional revenues.The high level of environmental change, which has numerous information sources, and more predictable change of information over time, triggers formative unlearning. In this type of environment, organizations will tend modify their strategies and rearrange their personnel.However, formative unlearning will be challenged when there exists denial or refusal to accept new belief structures. For instance, in his seminal book, Mastering the Dynamics of Innovation, Utterback (1994) noted how beliefs regarding natural ice harvesters (know what) impeded a change to machine-made ice production, because producers believed that people would prefer ice from nature rather than ice made by machines. Actually, ice harvesters were employing machines for their production and operation, indicating incremental changes in organizational routines, such as applying steam-powered circular saws for cutting blocks from rivers and ponds, and installing mechanical conveyors to provide continuous hauling of ice cakes from ponds to ice houses.Adjustive unlearning Adjustive unlearning refers to incremental changes in the existing knowledge structure of the organization, whereas fundamental changes occur in the organizational routines. This type of unlearning is related to evolutionary innovation and change and promotes new product lines, and new strategic business units. This type of unlearning includes the engineering intervention and work processes redesigns to improve the speed and quality of production. For instance, in a well known article by Imai et al. (1988) in which they discussed Honda's product-development strategy, to choose between a modified version of their Civic model or a totally new design. Even though the Civic was a successful product in the market, Honda changed its product-development mentality and designed a completely new car. During the development stage of the new Accord model, Honda carried out a design contest to identify the best car for its customers, and asked engineers and marketers around the world to help with the early stages of planning by creating a global knowledge network (Nonaka and Takeuchi, 1995).Environmental events that necessitate adjustive learning are those that have high unpredictability and low uncertainty. This type of environment indicates that information changes quickly and randomly. Organizations need to collect information rapidly about environmental events, and to distribute the information throughout the organization using new media and sensing units.Adjustive unlearning will be inhibited when there exists direct costs of change and cannibalization costs, which means change that brings success to a product but at the same time brings losses to a sister product, so it requires some kind of sacrifice.Operative unlearning Operative unlearning refers to the small-scale changes in beliefs and routines of an organization. This type of unlearning is related to incremental and continuous improvement and changes. This type of unlearning involves a limited departure from structural coherence. It helps organization maintain internal reliability and involves adjustments in systems, processes, or structures, but it does not involve fundamental change in strategy, core values, or corporate identity. Operative unlearning is most likely necessary during times of relative environmental stability and is likely to take place over extended periods of time.The environment, which stimulates operative learning, is low on uncertainty and is unpredictable. There is little information emerging and its direction is less unpredictable. In this sense, organizational change is a gradual and ongoing process which is aimed at fine tuning of the fit or match between the organizations' current strategy, structure, people and processes, and external environment.A typical example for an operative unlearning is the incremental product development activities in organizations. For instance, the HP DeskJet printer series, e.g. HP-800, developed from the adjustments to the production and commercialization processes, small-scale changes in production and design routines, and product improvement strategies to sustain the current market against the competitors - an on-going belief change efforts of management to adapt changes in competitive environment. As shown in Figure 1, this study:* denotes that unlearning can be operationalized as declarative and procedural memory eliminating - changing beliefs and routines in organizations based on the organizational change and memory literature; and* identifies different types of unlearning contingent upon the environmental conditions.Also, unlike that suggested by "common wisdom," which suggests that organizational unlearning is the opposite of organizational learning, Figure 1 shows that unlearning is an important sub-process of the organizational learning process and makes unique contributions to organizational learning literature (Akgun et al., 2003). Specifically, since, key contributions of unlearning in the organizational learning process have long been ignored in the literature, the legitimization of unlearning in the learning process was not forthcoming. Based on the organizational change and memory literature, we recognize that, unlearning:* catalyzes organizational learning process to foster a dynamic learning process;* provides a platform for shifting single-loop learning to double-loop learning; and* connects organizational learning and organizational change processes.As noted above, unlearning "catalyzes" the learning processes. The implication of the word, "catalyze" is that unlearning makes learning a dynamic process. As described by writers (Huber, 1991; Akgun et al., 2003), the organizational learning process encompasses continuous, reflexive and reciprocal cycle of activities that include memory, sensemaking, intelligence, thinking, emotions, improvisation, information acquisition, dissemination, and implementation. Since, it is a continuous process, no particular stage or sub-component can be taken as the starting point, therefore, missing or ignoring a component of the process halts the learning process in general (Akgun et al., 2003). In other words, all sub-components or processes work together rather than selectively. Unlearning facilitates a fluid and reflexive process of learning by changing beliefs and routines (Wijnhoven, 2001). For instance, organizational beliefs, which were created based on the positive feedback from a firm's success:* impose an interpretative inertia; and* facilitate inadequate reception of action signals.In addition, organizational routines, which indicate a motor, fixed or automatic action and behavior, pose an organizational inertia in changing and novel conditions. Routines reinforce the status quo (Nelson and Winter, 1982) and inhibit active seeking of alternatives. In this regard, unlearning makes the change and learning process unalterable, because old beliefs and routines were altered and replaced by a new set knowledge.In addition to the dynamic learning process, unlearning secures the learning levels. Specifically, unlearning is a means to shift from lower- (i.e. single-loop) to higher-level learning (i.e. double-loop). According to the organizational learning literature, single-loop learning or adaptive learning consists of detecting performance gaps and their elimination in line with standard operation procedures (Argyris and Schon, 1978). Here, an organization does not change its beliefs and routines. However, based on the environmental feedback, if an organization changes its beliefs and routines, one may say that double-loop learning has occurred. Here, unlearning provides a way for this transition, changing beliefs about cause-effect relations and operating procedures (Baker and Sinkula, 1999; Pawlowski, 2001). Wang and Pervaiz (2003), for instance, take a stronger stance and argue that organizational unlearning is needed to create quantum leaps.Finally, another critical role of unlearning is the interlocking of the organizational change and learning processes. There exists a common agreement that organizational change and learning are closely related concepts. For example, Pratt and Barnett (1997) have argued that learning is not a stand-alone concept; rather, it belongs to a family of change processes that encompasses unlearning and relearning (that is, adapting new mental models, procedures and routines). Indeed, based on contingency theory, the relation between an organization and its environment is the main concern of the theoretical perspectives of organizational change (Pawlowsky, 2001). At the same time, organizational learning requires changes in the relationships between organization and its environment (Hedberg and Wolff, 2001). Change involves some degree of learning, and learning provides a potential for change rather than any guarantee of its realization (Child and Heavens, 2001). Further, unlearning is one of the common dominators of change and learning process. For instance, Huy (2001) noted that beliefs and routines are the content of the change. Additionally, organizational learning concentrates on the process that brings such change about (Krebsbach-Gnath, 2001) and unlearning is one of the vital constructs of the learning process. In this perspective, unlearning acts as a linking pin between organizational learning and change processes. Unlearning is an important construct of organizational change and learning process, and thereby warrants an in-depth investigation. However, since the term "unlearning" was accompanied by the term "learning," such as "learning and unlearning," it was often seen as a demode construct or study area due to semantic usage of the terms and that it really is something different from learning, not simply its opposite. In this study, we further explored unlearning by reviewing organizational change and memory literature. However, our study is just one of a few attempts to discover the unlearning concept, see also Hedberg (1981), Nystrom and Starbuck (1984) and others. In future studies, we suggest that:1. Unlearning be investigated in the nomonological web of other organizational learning constructs. The recent research slightly mentions the niche of unlearning in the organizational learning literature (Akgun et al., 2003; Huber, 1991). For instance, according to Akgun et al. (2003), organizational learning is a process of the reciprocal relations among the information processing (i.e. memory, unlearning, sensemaking, etc.), intelligence and emotions. An empirical investigation of interwoven relations among the sub-components of learning would give a greater understanding of the process of learning at the organizational level.2. The types of unlearning can be investigated in detail using a longitudinal study. Specifically, the:* triggers or antecedents;* primary role of managers;* behavioral aspects of the practical (and unexpected) use; and* behavioral and operational consequences of each types of unlearning can be studied in case studies.3. Time processing orientation and agreement preferences should be investigated in unlearning process. Specifically, conflicts arose during changes in routines and beliefs, and resolution of those conflicts, majority and minority issues and power structures during the changes in beliefs and routines should be studied.4. How unlearning occurs should be investigated and empirically tested in groups, such as in new product development team. For instance, literature on new product development demonstrates that the groups that have strong memories are least able to deviate from previous routines during the new product development processes (Moorman and Miner, 1998). Therefore, eliminating group/team memory or unlearning, is critical for new product performance in turbulent environments/conditions and begs an empirical study. Also, empirically investigating unlearning in cross-functional new product development teams may illuminate researchers at the organizational level. Specifically, it has been asserted that new product development teams show the similar behavioral patterns as organizations do or represent microcosms of organizations based on the literature on the fractal geometry and images in system thinking (Capra, 1996). The basic philosophy of fractal images is that all characteristic patterns of the system are found repeatedly at the descending scales (Capra, 1996, p. 138). This is similar to holographic systems, which create processes where the whole can be encoded in all of the parts, so that "each and every part represents the whole" (Morgan, 1998, p. 71). In this sake, new product development team unlearning process empirically sheds light on the organizational unlearning process based on the functional isomorphism rather than homomorphic conceptualization of unlearning process, because NPD team members exist on more than one level simultaneously, act differently as units and influence each other across levels (Weiss, 1993, p. 278), and NPD process provides a more controllable, identifiable, and accurate unit definability (Meyers and Wilemon, 1986). Opens in a new window.Figure 1 Organizational unlearning Opens in a new window.Table I Types of organizational unlearning
|
- This paper is of value in shedding light on the unlearning concept based on the organizational memory and change literature.
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.