US layoff notices by year, 1988–2026

Every US employer over a certain size must file a WARN Act notice before a mass layoff or plant closing, and every state publishes those notices its own way. This is all 57,975 of them we can reach, from 46 state portals, normalized into one schema and re-derived every day — counted by year — and every one of them is free to download, back to 1988, with no paywalled years and no account.

57,975layoff notices on file1988–2026
0of them dated before 2024archive only — not in the free tier
6,449,610workers coveredas published by the states
46state portals in one schemarebuilt daily

Download the full 1988+ archive — free Get alerted when YOUR employers file — $49/year

Workers in WARN notices per year, 1988–2025388,599 workers19882025
Every complete year on file. 2020 is the peak: 1,962,468 workers across 15,629 notices. · Source: Official state labor department WARN listings, normalized · as of 2026-09-10
2020 is the year that dwarfs the rest. 15,629 WARN notices covering 1,962,468 workers were filed in 2020 alone — more workers than the 5 complete years since put together (2021–2025: 1,375,412). Every row behind that number is in the free download, in the same schema as this morning's notices.
WARN notices filed per year, 1988–20254,086 notices19882025
Notice counts rise partly because coverage widens: early years are the few states that publish a deep archive, recent years are all 46. The states column below makes that explicit. · Source: Official state labor department WARN listings, normalized · as of 2026-09-10

Year by year

Every year below is in the free files. “States filing” is the number of distinct states with at least one notice that year — it is the coverage caveat, stated rather than hidden.

YearNoticesWorkersStates filingWhere it lives
2027 (future-dated)24401free
2026 (year to date)2,754260,43446free
20254,086388,59945free
20243,689318,70741free
20233,570326,83342free
20221,883167,00337free
20211,894174,27038free
202015,6291,962,46838free
20192,801271,95636free
20182,471243,11635free
20172,276211,90735free
20162,259218,29333free
20151,969220,55931free
20141,175141,89526free
2013910111,28026free
20121,258117,48525free
20111,03987,75222free
201065577,80022free
20091,037136,22815free
2008742122,25113free
200747574,51813free
200646061,29510free
200545362,41010free
200448672,9299free
200346769,1399free
200246477,1177free
2001587103,6807free
200037662,0796free
199930845,6235free
199819562,0265free
199712416,2352free
199615526,0152free
199517027,0272free
199411517,7132free
199314320,9692free
199211721,4262free
199114327,0372free
199015422,9652free
19899814,4202free
198875,7112free
Doing research, or building a model, on US layoffs? Take the whole thing: every notice back to 1988 (57,975 rows, 46 states, CSV + JSON), rebuilt every morning rather than frozen at the day you downloaded it, CC BY 4.0. Download the archive — free → Need a name on your list watched daily instead? WARN Watch is $49/year.

FAQ

What is in the free dataset and what is in the archive?

The free CSV/JSON covers notices dated 1988-01-01 to today — complete, no delay, no login, CC BY 4.0. The archive is everything before that: 0 notices back to 1988, 0 workers, same schema, one payment.

Why is 2020 so much larger than every other year?

2020 is the COVID-19 shutdown. US employers filed 15,629 WARN notices covering 1,962,468 workers that year — roughly 7x the year before. It is the single most-requested slice of this dataset and it sits entirely in the archive, because the free tier starts at 1988.

Do early years mean fewer layoffs, or fewer states publishing?

Mostly the latter, and the table says so directly: the 'States filing' column shows how many distinct states appear in each year. Early years are the handful of agencies that publish a deep back catalogue. We never backfill or estimate a state that does not publish — a gap stays a gap.

Can I cite these yearly totals?

Yes. They are aggregates of public records, CC BY 4.0 — credit “WARN Feed” and link this page. The per-year numbers here are regenerated by the same daily pipeline that builds the dataset, so they match the files exactly. This build is dated 2026-09-10.

How current is the history?

The archive is re-derived on every daily refresh, not frozen at purchase: states amend headcounts and re-issue old notices, and those corrections flow through. What you buy is the current best reconstruction, not a snapshot someone uploaded once.