| { | |
| "original_study": { | |
| "claim": { | |
| "hypothesis": "Electoral fraud initially increases as violence rises but declines once violence becomes sufficiently intense, producing a curvilinear relationship between violence and fraud.", | |
| "hypothesis_location": "Theory section under “Violence and Fraud,” where Hypothesis 1 is explicitly stated predicting an inverted U-shaped relationship between violence and election fraud.", | |
| "statement": "The empirical results show that fraud is more likely in districts experiencing moderate levels of violence but becomes less likely in districts with very high levels of violence, consistent with an inverted U-shaped relationship between violence and fraud.", | |
| "statement_location": "Results section, Table 2, where the linear violence term is positive and significant and the squared violence term is negative and significant.", | |
| "study_type": "Observational" | |
| }, | |
| "data": { | |
| "source": "Polling-station level election returns from the Afghan Independent Election Commission combined with geocoded violence data from ISAF SIGACT reports and auxiliary datasets.", | |
| "wave_or_subset": "2009 Afghanistan presidential election; violence measured in a five-day window around election day and, in alternative models, a two-month pre-election window.", | |
| "sample_size": "Approximately 389 districts for fraud measures; up to 398 districts for violence and control variables.", | |
| "unit_of_analysis": "District.", | |
| "access_details": "Election results were publicly released by the Afghan Independent Election Commission; violence data originate from military and media-based event datasets.", | |
| "notes": "Fraud is measured using a digit-based forensic test (last-digit test) aggregated to the district level and validated using data from a post-election recount of ballot boxes." | |
| }, | |
| "method": { | |
| "description": "The study tests whether the relationship between violence and election fraud is nonlinear by regressing district-level fraud indicators on measures of violence and their squared terms, while controlling for development and geographic factors.", | |
| "steps": [ | |
| "Aggregate polling-station election results to the district level.", | |
| "Apply the last-digit forensic test to generate a binary indicator of fraud for each district.", | |
| "Measure insurgent violence using counts of attacks per 1,000 population around election day.", | |
| "Include a squared violence term to capture potential nonlinearity.", | |
| "Add controls for closed polling stations, development, geography, and population characteristics.", | |
| "Estimate logit models for the binary fraud measure and OLS models for the recount-based fraud share, clustering standard errors at the regional command level." | |
| ], | |
| "models": "Logistic regression (for binary fraud indicator) and OLS regression (for recount-based fraud share) with linear and squared violence terms.", | |
| "outcome_variable": "Election fraud at the district level.", | |
| "independent_variables": "Violence per 1,000 population; squared violence term.", | |
| "control_variables": "Number of planned polling stations closed; electrification; per-capita expenditure; distance from Kabul; average elevation.", | |
| "tools_software": "not stated" | |
| }, | |
| "results": { | |
| "summary": "The regression results support a curvilinear association between violence and fraud: fraud increases with low to moderate violence but decreases as violence becomes very high.", | |
| "numerical_results": [ | |
| { | |
| "outcome_name": "Election fraud (last-digit test)", | |
| "value": "8.477", | |
| "unit": "log-odds (logit coefficient)", | |
| "effect_size": "logit regression coefficient for violence (linear term)", | |
| "confidence_interval": { | |
| "lower": "not stated", | |
| "upper": "not stated", | |
| "level": "not stated" | |
| }, | |
| "p_value": "< 0.1", | |
| "statistical_significance": 1, | |
| "direction": "positive" | |
| }, | |
| { | |
| "outcome_name": "Election fraud (last-digit test)", | |
| "value": "-13.748", | |
| "unit": "log-odds (logit coefficient)", | |
| "effect_size": "logit regression coefficient for squared violence term", | |
| "confidence_interval": { | |
| "lower": "not stated", | |
| "upper": "not stated", | |
| "level": "not stated" | |
| }, | |
| "p_value": "< 0.01", | |
| "statistical_significance": 1, | |
| "direction": "negative" | |
| } | |
| ] | |
| }, | |
| "metadata": { | |
| "original_paper_id": "10.1017/S0007123412000191", | |
| "original_paper_title": "Violence and Election Fraud: Evidence from Afghanistan", | |
| "original_paper_code": "not stated", | |
| "original_paper_data": "http://dvn.iq.harvard.edu/dvn/dv/nilsw" | |
| } | |
| } | |
| } |