APProjects/us-layoffs-monthly-time-series-warn-act

US layoffs, month by month — 455 months of WARN notices, 1988-11 → 2026-09, rebuilt daily

Last rebuilt: 2026-09-16. One row per calendar month: how many US WARN Act layoff notices were filed, how many workers they named, and how many states contributed — as a regular series with every month present (zeros included), ready for pandas, a chart or a forecasting model. A second table gives the same series per state.

455	consecutive months, 1988-11 → 2026-09, no gaps
60,930	dated notices in the series (375 undated rows excluded, said so below)
4,769	notices in April 2020, the busiest month on record (40 states reporting)
1.2×	April 2020 alone, measured against the entire last 12 months combined (4,091 notices)
265	notices in August 2026, the last complete month; 27,025 workers named
6,576	state-month rows in the by_state table

Notices per month, notice-date basis, all reporting states, through August 2026 (the partial current month is left off the chart). The April 2020 spike is the pandemic; the flat left half is a handful of states' worth of history, not a quiet economy — see "Honest scope" before you compare decades.

This file is free forever — CC BY 4.0, no login, no API key, rebuilt daily.

If you track layoffs for specific employers or states rather than the national aggregate, WARN Watch runs your list against every daily refresh and gives you a private alert page, an RSS feed and an optional Slack/Discord/Teams webhook.

Try it free for 30 days — no card (up to 3 employers or one state) · then $49/year, one-off, no auto-renew, 14-day refund

Quickstart

from datasets import load_dataset
national = load_dataset("APProjects/us-layoffs-monthly-time-series-warn-act", split="train")
by_state = load_dataset("APProjects/us-layoffs-monthly-time-series-warn-act", "by_state", split="train")

import pandas as pd
base = "https://huggingface.co/datasets/APProjects/us-layoffs-monthly-time-series-warn-act/resolve/main/data/"
m = pd.read_csv(base + "monthly_series.csv", parse_dates=["month"]).set_index("month")

m.loc["2015":, "notices"].plot()                       # the modern, ~30-40 state era
m.loc["2015":, "notices"].rolling(12).mean()            # smoothed
m["workers_affected"].idxmax()                          # 2020-04
s = pd.read_csv(base + "monthly_by_state.csv")
s[s.state == "CA"].set_index("month")["notices"]        # one state's own series

Columns — monthly_series.csv (default config)

column	meaning
month	calendar month, YYYY-MM
notices	WARN notices dated in this month (notice date; effective date when the state publishes no notice date)
notices_with_headcount	how many of those publish an integer headcount
workers_affected	sum of headcounts over notices_with_headcount rows only
states_reporting	distinct states with at least one notice this month — read this before comparing years
dated_by_effective	rows in this month that were dated by effective date because no notice date exists
notices_effective_month	the same notices keyed by effective date (when the separations actually happen)
workers_effective_month	headcount sum on that effective-date basis

monthly_by_state.csv (by_state config): month, state (USPS code), notices, notices_with_headcount, workers_affected — only (month, state) pairs with activity; zero rows are implied.

Honest scope — read this before you cite it

• This is not a national count before roughly 2015. State portals differ wildly in how far back they publish: 2 states reach back to 1988, 42 have reported in the last twelve months, 48 are covered today. A rise from 1995 to 2025 is mostly coverage, not layoffs. Use states_reporting to restrict to comparable periods, or use the by_state table and pick states with long archives.
• Notice month ≠ layoff month. Employers must file 60 days ahead; the notice-date basis leads the effective-date basis by one to three months. Both are in the file; pick the one your question needs and say which.
• Dates. MI, PA and SC publish no notice date. Their rows are dated by effective date and counted in dated_by_effective (2,862 rows in total). 375 rows carry neither date and are excluded; 25 rows are notice-dated after the rebuild month and 352 have effective dates after it — those are held out of the series rather than plotted as the future.
• workers_affected undercounts. Only rows with an integer headcount are summed; notices_with_headcount says how many that was per month.
• The current month is partial and refills every day until it closes; the last complete month is August 2026. Late-arriving notices are added to their own month retroactively, so history can revise slightly between rebuilds.
• Coverage is 48 states, not 50. Not covered yet: AR, HI, ID, MA, MN, MO, ND, NH, NV, OH, WY. The notice-level mirror carries coverage.json with the authoritative per-state list and scrape stamps.
• Compiled from state workforce-agency portals. Independent project, not affiliated with any government agency; not legal, financial or employment advice.

Where this comes from

Derived on every daily refresh from the notice-level dataset APProjects/us-warn-act-layoffs-notices-daily (61,330 notices, 48 states, back to 1988). Companion rollups from the same morning's run: by employer. Full site with per-state and per-month pages, search and RSS: approjects-warn-act-notices.static.hf.space.

A one-off scrape of WARN data starts rotting the week it is posted — states amend headcounts, re-issue notices and drop rows. Compare the rebuild date at the top of this card with the "last modified" date on any other US layoffs series before you pick one.

Getting told when the next notice lands

Everything above is free, CC BY 4.0, no account. The one paid thing this project sells is the watching: WARN Watch — up to 500 employer terms plus whole states, matched on every daily rebuild for a year, with a private alert page, RSS and an optional Slack / Discord / Teams webhook ($49/year). Try it first for nothing: free 30-day watch, no card, nothing renews.

Corrections: open an issue — they ship the same day.

Cite as: "WARN Feed — US layoffs monthly time series (WARN Act), rebuilt 2026-09-16, huggingface.co/datasets/APProjects/us-layoffs-monthly-time-series-warn-act".

A layoff record you can audit, not just download

This dataset is one cut of a single daily rebuild: 61,330 US WARN Act layoff notices from 48 state agencies, 1988 to today, one schema, no login, no delay, CC BY 4.0. Snapshot as of 2026-09-16; the files above are rebuilt every day, so the live count is the truth.

Several projects publish a current WARN scrape and two of them carry more rows than we do. None of them publish what the records used to say:

• 617 observed changes to already-published notices, logged daily since 2026-08-31. data/revisions.csv records every field that differed between two consecutive daily builds — employee counts, effective dates, notice types, company names — with the old value, the new value and the date we saw it. We publish the observation and not the cause: a change is equally explained by the agency amending the notice or by our own parser improving, and we do not guess which (see data/revisions.README.txt). A scrape that starts tomorrow cannot backfill any of it; it only exists if someone was watching.
• 6,799 notices whose state agency page no longer lists them. Agencies take notices down. We keep them, flagged as archive-only, so a count you ran last year still reconciles.
• Point-in-time employer identity. The ticker crosswalk resolves a filer to the company as it existed at the time of the notice — Kmart, Sears Holdings, Symantec — not to whatever is on today's ticker file.

If you have to defend a number to an editor, a referee or a compliance reviewer, that provenance layer is the part you cannot rebuild yourself. How to cite this dataset →

Look something up right now — free, no signup, nothing to install. Check any employer or state against the last 180 days → It runs in your browser against these same files.

Building something with it? The same files are a free HTTP API — JSON and CSV, no key, no signup, access-control-allow-origin: * so fetch() works from a browser: endpoints, schema and curl examples →

Need one industry only? The same filings, cut by an auditable employer-name rule (each row keeps the rule that fired): tech companies · hospitals & healthcare · retail store closings · restaurants & hotels · factory & plant closings · banks, insurance & finance · warehouses, trucking & logistics · all 20 sectors.

Or have it watch a list for you. Coming back to look is the part a CSV cannot do. WARN Watch — $49 for a year, one payment, nothing auto-renews, 14-day refund, no login: up to 500 employer names plus whole states, matched on every daily refresh, delivered to a private alert page + calendar (.ics) + RSS + an optional Slack / Discord / Teams webhook. Every alert carries that employer's whole filing history from the archive, which a keyword rule on an RSS feed cannot see. There is no built-in email — we do not claim one.

• See a real alert page before paying · what you get
• Try it free for 30 days, no card · Buy — $49/year

Not deciding today? Join the update list → — one email when a new dataset or tier is published; nothing promotional. A state added or a column renamed ships in the daily release instead, no address needed. The list is shared across APProjects datasets, holds an email address only, is run by Gumroad, and any message unsubscribes you. Rather give no address at all? Watch the repo's releases — GitHub notifies you on every daily republish, and a new state or changed field is in those notes the day it lands.

Reaching a human. WARN Feed is published by APProjects, an automated data publisher — that is stated plainly rather than dressed up. Corrections, coverage gaps, schema questions and refund requests all go here and are read: open an issue. Payments are handled by Gumroad as merchant of record, so an invoice can carry your company name.

Source, scrapers and methodology · the 48-state site

_{Card views are counted anonymously: one 1×1 image on a public CDN, no cookies, no script, no personal data. The count is public: jsDelivr stats.}