Datasetpublic

teosv1/us-warn-act-layoff-notices

US WARN Act Layoff & Closure Notices — normalized snapshot (2026-07-29) 5,660 US WARN Act layoff and closure notices, normalized to one schema across 5 states: TX (2,358), CA (1,632), IL (1,369), NC (236), NY (65). Covering 528,129 affected workers and 4,635 distinct employers, with effective dates from 2010-10-30 to 2028-02-01. This is a frozen snapshot Taken 2026-07-29. It does not update. WARN filings arrive continuously, so this file begins going stale the day… See the full description on the dataset page: https://huggingface.co/datasets/teosv1/us-warn-act-layoff-notices.

sourceHugging Facecc-by-4.0updated 2mo agoView on Hugging Face

1likes20downloads

Dataset Card

US WARN Act Layoff & Closure Notices — normalized snapshot (2026-07-29)

5,660 US WARN Act layoff and closure notices, normalized to one schema across 5 states: TX (2,358), CA (1,632), IL (1,369), NC (236), NY (65). Covering 528,129 affected workers and 4,635 distinct employers, with effective dates from 2010-10-30 to 2028-02-01.

This is a frozen snapshot

Taken 2026-07-29. It does not update. WARN filings arrive continuously, so this file begins going stale the day it is published — that is intentional and it is stated plainly rather than buried.

[Get the live API](https://buy.stripe.com/7sY5kweawcc1gqw0P7fEk0d) — 49.00 USD/month, self-serve, refreshed every 12 hours.

Why this exists

WARN Act notices are public records, published by individual state labor agencies. They are also published in five mutually incompatible ways: two spreadsheet dialects, born-digital PDFs with differing table geometry, and portals that disagree about what a date is. The work in this dataset is not access, it is normalization — one schema, typed fields, ISO dates, and a stable provenance key per source artifact.

Schema

Column	Type	Notes
`state`	string	USPS two-letter code
`company_name`	string	Employer as filed
`employees_impacted`	integer	Headcount from the notice; null when the filing omits it
`effective_date`	string	ISO date the layoff or closure takes effect
`notice_date`	string	ISO date the notice was filed — the lead-time signal
`location`	string	County-level location as filed
`layoff_type`	string	e.g. Closure Permanent, Layoff Temporary
`raw_source_file`	string	Stable provenance key for the source artifact
`parsed_timestamp`	string	When this row entered the corpus

Nulls are honest. Where a filing genuinely does not state a value, the field is null rather than zero or a guess — a stored NULL beats confident garbage.

—warn_notices.csv — 5,660 rows, header included
—warn_notices.jsonl — the same rows, one JSON object per line
—coverage.json — counts by state, date range, totals

Known data quality notes

—55 rows have no `effective_date`. The underlying filing did not state one. Stored null rather than guessed.
—2 rows carry an `effective_date` before the WARN Act existed (1930-03-30, 1930-03-31). These are parsing artifacts, not history — their notice_date values sit in the correct era. They are retained rather than silently dropped, and excluded from the headline date range above.
—Company names are as-filed and not entity-resolved. One employer may appear under several spellings.

Provenance and limitations

Derived from public state WARN Act filings. Coverage is the five states above; states absent here are absent because of acquisition barriers (WAF blocks, JS-rendered portals, unstable PDF geometry), not because their data is unimportant. No personal data is included; these are corporate filings.

Citation

TEOS Intelligence. US WARN Act Layoff & Closure Notices, normalized snapshot 2026-07-29.