Open algorithm

How the displacement estimate works

The formula, unit, and limits behind XH1B's modeled displacement total. It counts worker equivalents across records, not unique people or worker-years.

Exposure method v1.2

What does the total count?

One worker equivalent is the model's base contribution for one position requested in a certified H-1B filing. The total is not a count of unique people or worker-years. A wage offer 20% below the BLS occupation median adds 0.2 equivalents for that position.

The total adds records across the available filing history. It does not multiply positions by the length of a visa or job. Renewals, amendments, and repeated filings can represent the same person more than once. The records do not identify every worker, so the model cannot count unique people.

The calculation

Total = sum of [requested positions × (1 + wage-gap share)] + matched WARN layoff counts

Each filing is compared with the latest national BLS occupational median wage for its canonical SOC code. The gap is the shortfall below that median, as a share of the median: max(0, (BLS median − wage offer) / BLS median). Filings with no BLS match or no positive wage offer contribute no wage gap.

A filing with no recorded worker count counts as one worker.

A worked example

Ten requested positions at 20% below the BLS occupation median contribute 12 worker equivalents: 10 base equivalents plus 2 from the wage-gap term. A matched WARN notice reporting 5 layoffs adds 5, for a total of 17. The result does not mean that 17 different people lost their jobs.

Which filings count?

  • Visa class H-1B only.
  • Case status starts with Certified (including certified-then-withdrawn).
  • Rows carrying a data-quality flag are excluded. This covers flagged H-1B filings, and WARN notices whose layoff count is flagged suspect.

Which records are added?

For employer and geography totals, the model adds reported layoffs from matched WARN events. The event matcher uses same-employer H-1B submission dates from three months before through six months after the notice date, with the effective date as fallback. Matching includes all decision outcomes and worksites. It is separate from the certified filing cohort used for the LCA term. A WARN notice with a suspect layoff count or an unproved location association matches nothing until review repairs it. Complete source observations stay stored. A match does not prove replacement.

  • Employers: LCA term plus matched WARN layoffs.
  • Persons (attorneys and points of contact): LCA term only.
  • Law firms: LCA term only, per firm name.
  • States, counties, and ZIP codes: LCA term plus matched WARN layoffs for the area.

Totals and components round to three decimals; the average wage-gap percentage rounds to one decimal.

What the estimate cannot establish

  • The model does not observe an actual H-1B hire and does not prove replacement.
  • Terms are additive across filings and notices, so the same person can be represented more than once.
  • A WARN count is a filed-notice count, not a verified layoff count.
  • Geography totals exclude flagged WARN counts. The math expander shows the as-published total beside the quarantined one.
  • The wage-gap term is a wage-depression exposure, not a wage-law finding.

The version labels a traceable calculation. It is not a claim that each requested position became a hire, that each hire displaced a worker, or that a matched notice proves replacement.

Public source records and site analysis. We do not attest to accuracy. Read the full disclaimer.

Dataset guides and site methods

  1. [1] Dataset documentation · DOL OFLC. LCA public disclosure record layout (FY2026 Q3). Pages 1–4: application, DOL outcome and dates, requested positions, employer, representatives, worksites and pay unitsThis layout covers determinations issued October 1, 2025 through June 30, 2026. It documents employer-provided application fields and DOL decisions, not USCIS decisions, payroll verification, legal bills or worker-linked moves. Older layouts can differ.Checked 2026-10-17.
  2. [2] Dataset documentation · BLS. OEWS technical notes (May 2025). Scope of the survey; occupational coding; concepts and estimation methodologyDocuments occupation and geographic wage estimates, not an individual worker’s pay. Use the estimate year and geography stated by the site; older releases can differ.Checked 2026-10-17.
  3. [3] XH1B method or registry · XH1B. WARN Notices (mass layoffs): source registry; Calculation and Matching Methods; Reading Public Records and Derived Indicators. Source registry entry: warn; Filing units, annualization, weighting, wage comparisons, and inference limits; Reading records: units, source coverage, wage gaps, entity links, and inference limitsXH1B dataset description: One row per source_notice_id. The definition of a mass layoff and the layoff threshold vary by state. The threshold is typically 50 or more employees.; These are the site's review methods, not validated predictions of wrongdoing. Inspect underlying records and alternative explanations.