US WARN Notices — 9 Largest States

Get every current WARN Act layoff and closure notice from the nine largest publishing states — California, Texas, New York, Florida, Illinois, Ohio, Pennsylvania, Michigan, Washington — in one clean, normalized dataset, pulled straight from each state's official government publication, with a source link on every record.

WARN notices are the earliest public layoff signal there is: US employers must file them with the state 60+ days before a mass layoff or closure, weeks before anything hits the news. Recruiters use them to reach affected workers at filing time; B2B sales teams use them as churn/budget-change signals; researchers get a clean multi-state layoff series with full provenance.

Why this actor

• The 9 biggest WARN states in one run — including the states other actors skip because their sources are genuinely hard: Florida's paged tables, Pennsylvania's accordion page, Michigan's search API, Washington's postback-paged grid, Ohio's constantly-moving CSV. Most WARN actors on the store cover 2–6 states, usually only the easy ones.
• Official sources only. Every record links back to the state publication it was parsed from (source_url), and to the underlying notice document where the state publishes one (record_url — FL/OH/MI attachments). Verify any row in one click. No third-party aggregators, no news scraping, no login, no personal data.
• A format change never becomes silent garbage. Every state's parser opens with a signature check on the source's own structure (exact CSV/ table headers, required JSON keys). If a state agency redesigns its page, that one state fails loudly and alone — reported in SOURCE_STATUS — while the other states still return data. You will never quietly get wrong columns.
• Honest metadata. Pennsylvania publishes only the month a notice was received, so PA records carry notice_date_precision: "month" instead of a fabricated day. Employee counts are the states' own published figures; where a state left the field blank, it is null, not guessed.

Input

Field	Type	Default	Notes
states	array	all 9	Any of CA TX NY FL IL OH PA MI WA
sinceDays	integer	365	Only notices dated within this window. 0 = everything each state currently publishes
minEmployees	integer	—	Keep notices affecting at least this many workers (notices with no published count are always kept)
companySearch	string	—	Case-insensitive substring match on employer name
maxRecords	integer	10000	Stop after this many records

{
  "states": ["TX", "OH", "MI"],
  "sinceDays": 90,
  "minEmployees": 100
}

Output

One dataset item per notice:

{
  "state": "CA",
  "state_agency": "California Employment Development Department (EDD)",
  "company": "Foundation for California Community Colleges",
  "location": "1102 Q Street, Suite 4800 Sacramento CA 95811; Sacramento County",
  "notice_date": "2026-09-02",
  "notice_date_precision": "day",
  "effective_date": "2026-10-30",
  "effective_date_raw": "",
  "employees_affected": 178,
  "action": "Layoff Permanent",
  "record_url": "",
  "source_url": "https://edd.ca.gov/en/jobs_and_training/Layoff_Services_WARN/",
  "fetched_at": "2026-09-05T17:26:31Z"
}

• notice_date — date the state received/posted the notice (YYYY-MM-DD).
• notice_date_precision — day, or month for PA (its listing groups by month received; such records carry the 1st of the month).
• effective_date — first layoff/closure date when it parses to a single clean day; effective_date_raw preserves the state's verbatim text when it doesn't (e.g. "02-06-26 thru 03-13-26").
• employees_affected — the state's own figure; null where the state published none.
• record_url — per-notice document link where the state publishes one (FL/OH attachments, MI PDFs), else empty.

The run also writes SOURCE_STATUS to the key-value store: per-state status (ok/failed), record/parse counts, timing, and a floor warning if an unfiltered run returned suspiciously few records from a source.

Per-state coverage (what each state actually publishes)

State	Source	Window published	Notes
CA	EDD WARN report workbook	current CA fiscal year (Jul 1 onward)	file replaced in place as notices arrive
TX	TWC dataset on data.texas.gov	multi-year history	official state open-data API
NY	NYSDOL WARN dashboard export	what the dashboard exports	one row per impacted site
FL	FloridaCommerce REACT	current + prior calendar year	per-notice attachment links
IL	DCEO / Illinois workNet	trailing 12 months, active notices	completed layoffs age out of the state's own query
OH	ODJFS notices CSV	current + prior year	CSV URL changes on update; re-discovered every run
PA	L&I WARN notices page	current + prior year	month precision (see above)
MI	LEO WARN notices	what the state's search returns	per-notice PDFs on older records
WA	ESD WARN search	multi-year grid, newest first	paged server-side; the actor walks the pages

Georgia is excluded, not missing by accident: GA currently publishes no machine-readable WARN listing at all (its legacy search app is retired and the new site has no WARN page — checked directly). We don't guess.

Honest limitations

• Coverage is these 9 states — the largest WARN publishers, not all 50. More states are added based on demand; ask in Issues.
• FL's listing has no location column (the location is inside the linked notice PDF) — FL location is empty and record_url carries the PDF.
• IL replicates the state app's own default query (active notices); completed layoffs age out. State-tracked layoffs without a WARN notice are deliberately excluded.
• Each state publishes a different history window (table above). sinceDays can only narrow what a state publishes, never extend it.
• A full unfiltered 9-state run takes ~2–4 minutes (Florida paginates server-side); a filtered or few-state run takes seconds.

Typical usages

• Weekly layoff feed: schedule the actor weekly with {"sinceDays": 8} and pipe the dataset to your CRM/sheet — a clean feed of every new notice across all 9 states.
• Recruiting in one state: {"states": ["OH"], "sinceDays": 90}.
• Account monitoring: {"companySearch": "amazon", "sinceDays": 730}.

Legality

WARN notices are public records that US states are required to publish under the federal WARN Act. This actor reads only official government publications: no logins, no paywalls, no personal data — records are at the employer/site level.