Blog
What WARN Data Does — and Doesn't — Tell You About H-1B Employers
September 3, 2026
WARN notices can confirm that an employer announced a mass layoff, but they can't tell you whether it happened, who was affected, or whether any H-1B workers were involved. Here is exactly where the line sits.
Can WARN data tell you whether an H-1B employer laid off its visa workers? No — and it is worth being precise about why. A WARN notice can confirm that a covered employer formally announced a planned mass layoff or closure to a state agency, on a certain date, at a certain site, with the employer's own estimated headcount. It cannot confirm that the layoff happened as announced, cannot identify who was affected, and never records visa status. This is the fifth and final piece in our series on reading H-1B disclosure data honestly, and it is the one the others lean on: whenever we connect a layoff to a sponsor, these are the rules that constrain what the connection means.
What WARN can actually tell you
The Worker Adjustment and Retraining Notification (WARN) Act requires larger employers to give advance notice of certain mass layoffs and plant closures. When they do, the notice becomes a public record held by a state labor agency. From that record you get four reliable facts: the employer name as the company wrote it, the date the notice was filed, the location or worksite, and the employer's estimated number of affected positions at the time of filing.
That is a genuine, primary-source signal. It tells you a company was planning a workforce reduction big enough to trigger a legal notice — not a rumor, not a press leak, but a filing the employer made itself. For the layoffs we track, that is the anchor everything else hangs on.
What WARN cannot tell you
The gaps matter more than the facts, because they are where misreadings happen.
It can't tell you the layoff happened as announced. A WARN notice is an announcement, not a receipt. Notices get revised, delayed, or rescinded entirely. A company can file for 400 positions and cut 250, or restructure and cut none. The headcount is an estimate at filing time, not a final count.
It can't tell you who was affected. WARN never identifies individual workers, and it certainly never flags visa status. There is no field for "H-1B," no occupation breakdown, nothing that links a notice to any named employee. Any claim that a specific layoff hit H-1B workers is going beyond what the record contains.
It can't see most of the country. Only some states publish usable WARN data, and the coverage is uneven. We track six — California, Texas, New Jersey, Illinois, Massachusetts, and Virginia. The other 44 are effectively dark for our purposes.
It can't see smaller employers. WARN thresholds exclude companies below roughly 100 employees and layoffs below event-size cutoffs. A small sponsor can shed its entire staff without ever generating a notice.
How big is the blind spot, honestly
Big enough that "national picture" is the wrong phrase for it. The six tracked states account for roughly 45% of all LCA filings — about 4.12 million of 9.12 million. That is a large, meaningful window onto where H-1B work concentrates, and it is worth taking seriously. It is not a census.
Within even that window, the numbers have soft edges. Of the 1,038 sponsor-linked notices we hold, 186 carry no headcount at all — and 175 of those are in Illinois. Where headcount is missing, any worker total we report is a floor, not a full count. New Jersey compounds this by publishing dates at month granularity rather than exact days, so a New Jersey notice tells you the month, not the morning.
None of this makes the data useless. It makes it a lower bound with known holes — which is a fine thing to work with, as long as you say so.
The matching problem, and why we treat it as a feature
Here is the quiet hard part. A WARN notice names an employer as free text, typed by whoever filed it. An H-1B sponsor in the disclosure data is a separate, differently-spelled entity. Joining the two is not a lookup; it is a judgment call, and a wrong join invents a fact that never existed.
So we set a deliberately strict bar. A notice is linked to a sponsor only when the normalized employer name matches exactly and unambiguously, or when a human operator confirms it. Fuzzy candidates go to manual review — they are never auto-published on a guess.
We know the strict rule earns its keep because we tested it. A hand audit of 816 early automatic matches found 10 wrong ones — a 98.4% hit rate. Every single error came from one class: stray parentheticals like "TRIGO (Tesla)," where the parenthetical is the client, not the employer filing the notice. We fixed the matching rules to strip that pattern, re-ran the audit to 100%, and turned that audit into a regression gate that re-runs on every monthly refresh. If a future data update reintroduces the bug, the build catches it before anything publishes.
We spell this out not to boast about tidy engineering, but because "we linked this layoff to this sponsor" is exactly the kind of claim that should come with its evidence attached. Our full methodology documents the normalization and review steps in more detail.
The guardrails that apply to everything
A few principles run through this whole series, and they apply here doubled:
- A notice is an announcement, not a completed layoff. Read every WARN figure as a plan filed, subject to change.
- There is no causal link between LCA filings and layoffs — in either direction. A sponsor with many filings and a nearby WARN notice has not been shown to have laid off the people it sponsored, or anyone else.
- LCA is not a visa. A Labor Condition Application is a wage-and-conditions attestation an employer files with the Department of Labor; it is a step that precedes a petition, not proof anyone received or holds H-1B status. The glossary walks through these terms one at a time.
It is worth stacking the related caveats from the rest of the series in one place, because they compound: a DOL certification rate is not a USCIS approval rate; a withdrawn application is not a denied one; the offered wage on a filing is not take-home pay; and FY2026 figures are partial, reflecting roughly two reported quarters. Each of these is a place where a fast reading produces a wrong conclusion.
How to read a WARN-to-sponsor connection
Put it together and the responsible reading of anything on our layoffs pages is narrow but real:
A covered employer, in one of six tracked states, filed a public notice announcing a planned workforce reduction, on or around a given date, at a given site, of an estimated size — and that employer's normalized name matched an H-1B sponsor exactly, or a person confirmed it did.
That sentence is defensible in every clause. What it does not license is any claim about who lost a job, whether they held a visa, whether the cut occurred, or what happened in the 44 states we cannot see. When you see a layoff sitting next to a sponsor profile, that is the boundary of the fact — and knowing the boundary is what makes the data trustworthy in the first place. For the full scope of what we do and do not assert, see our disclaimer.
- WARN Act
- layoffs
- H-1B
- data limitations
- methodology