An Entity-Resolved Crosswalk Between U.S. Department of Labor H-1B Labor Condition Applications and USCIS Petition Outcomes, FY2019-FY2026
Year: 2026
Type: preprint
Abstract: The United States publishes H-1B sponsorship activity in two unlinked federal datasets: the Department of Labor’s Labor Condition Application (LCA) disclosure files, which record an employer’s certified intent to sponsor, and the USCIS H-1B Employer Data Hub, which records adjudicated petition outcomes. Neither carries an identifier usable to join them, and the commercial tools that job seekers rely on treat LCA presence alone as evidence that an employer sponsors. That inference has never been validated. We release an entity-resolved employer-level crosswalk between the two sources covering FY2019–FY2026, built from 3,871,927 certified H-1B labor condition applications and 164,070 distinct employers. Our central methodological contribution is a machine-verified reference standard: we establish empirically that the Hub’s undocumented Tax ID field is the last four digits of the petitioner’s Employer Identification Number (98.63% agreement, n = 40,049, 95% CI [98.52%, 98.74%]), and that DOL began publishing full EMPLOYER_FEIN in FY2024 Q4. Together these permit evaluation of name-based linkage against agency-assigned identifiers rather than hand-labelled judgement. Held out from the EIN feature, the configuration that applies to the 5.5 fiscal years in which DOL publishes no FEIN, linkage achieves F1 = 0.9446 (FY2025) and F1 = 0.9473 (FY2026) at the same operating point. We additionally document three previously unreported defects in the source data, including that DOL’s quarterly releases are inconsistently cumulative across and within fiscal years, such that both obvious loading strategies silently corrupt the panel by up to 67%. As a demonstration, we report that 25.06% of employers holding a certified labor condition application have no corroborating USCIS approval.
Author: Sarpong, Felix Kwasi
Institution: Texas Tech University
Language: English
Topic: Data Quality and Management
Field: Decision Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7203643752
Cites:
Cited by:
Related to: