Hugging Face's logo

Datasets:
Workopia
/
us-warn-layoffs

Tasks:
Tabular Classification
Time Series Forecasting
Modalities:
Tabular
Text
Formats:
parquet
Languages:
English
Size:
10K - 100K
Tags:
labor-economics
layoffs
WARN
employment
united-states
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
1
us-warn-layoffs / README.md
Shuang007's picture
Shuang007
Upload README.md with huggingface_hub
a71f391 verified
preview code
|
Raw
History Blame Contribute Delete
3.87 kB
metadata 
license: cc-by-4.0
language:
  - en
pretty_name: US WARN Layoffs Dataset (Workopia)
tags:
  - labor-economics
  - layoffs
  - WARN
  - employment
  - united-states
size_categories:
  - 10K<n<100K
task_categories:
  - tabular-classification
  - time-series-forecasting
configs:
  - config_name: default
    data_files: data/layoffs_us_warn.parquet

US WARN Layoffs Dataset

A cleaned, event-level dataset of United States layoff and plant-closure notices filed under the WARN Act (Worker Adjustment and Retraining Notification), plus a small set of layoff disclosures from SEC 8-K filings. Compiled and maintained by Workopia.

WARN notices are public records that employers are legally required to file with state workforce agencies ahead of mass layoffs or plant closings. This dataset normalizes those filings across states into a single, analysis-ready table.

At a glance

Events 18,468
Total employees affected 2,306,915
Notice-date range 2004-01-12 → 2026-06-10
Geographic coverage United States, 35 states
Primary source State WARN portals (CA, TX, NJ, IL, WA, MD, …)
Secondary source SEC 8-K Item 2.05 disclosures (140 events)
Granularity One row per filed notice
PII None — company / facility level only

Density note. Coverage is sparse before ~2015 and dense from 2020 onward (COVID-era spike, then a 2025–2026 rise). Treat pre-2015 years as a thin historical tail, not a complete record. The dataset reflects WARN's universe (large layoffs above state thresholds, plant closings) — it is not a census of all U.S. job losses.

Schema

Column Type Description
event_id string Stable unique id for the notice
company string Company name as filed (may include facility detail)
company_normalized string Cleaned company name
employees int Workers affected (present for 94% of rows)
kind string Layoff / Closure / Unknown
permanence string Permanent / Temporary / Unknown
notice_date date Date the notice was filed
effective_date date Date layoffs take effect
country string Always United States
state string US state (2-letter)
county string County
city string City
address string Facility address (public WARN record)
naics2 string 2-digit NAICS sector (present for ~8% only)
sic string SIC code where available
industry string Industry label (present for ~51%)
ticker string Stock ticker where mappable (~0.75% only)
cik string SEC CIK where available
source_url string Source filing / portal URL
source_type string e.g. WARN_CA, WARN_TX, SEC_8K
sec_item_codes string SEC 8-K item code(s) for 8-K-sourced rows
notice_year int Year of notice_date (convenience)

Known limitations

  • U.S.-only. Non-U.S. layoffs are not included in this release.
  • Industry & ticker are sparse (naics2 ~8%, industry ~51%, ticker ~0.75%). Sector- and equity-level analysis requires additional enrichment.
  • WARN is not exhaustive — thresholds, exemptions, and state-by-state filing differences mean smaller layoffs and some states are underrepresented.

Suggested uses

  • Tracking U.S. layoff trends over time (national and by state)
  • Labor-market nowcasting and event studies
  • Training/evaluation data for entity normalization and employment NLP

Citation 

Workopia (2026). US WARN Layoffs Dataset. https://workopia.io

License

Released under CC BY 4.0. Underlying WARN notices are public records; this distribution is the cleaned/normalized derivative compiled by Workopia.