A follow-on to our FLAG contract-history report: that piece traced who built the software that processes H-1B applications. This one asks what the software is actually required to check — and finds the answer written directly into federal regulation. DOL's review of a Labor Condition Application is limited by law to completeness and "obvious inaccuracies," not whether the attestation is true. A report GAO sent to Congress in 2011 called the result "cursory." Eleven years of DOL's own disclosure data show exactly what that produces: a denial rate under 1%.
Three independent sources, cross-checked against each other: (1) the current regulatory text at 20 CFR §655.740, which
defines exactly what DOL's Certifying Officer is permitted to review before approving a Labor Condition Application;
(2) GAO-11-26, "H-1B Visa Program: Reforms Are Needed to Minimize the Risks and Costs of Current Program"
(January 2011), a report GAO delivered directly to congressional committees at Congress's own request; and (3) our
locally mirrored copy of DOL's own LCA and PERM disclosure data files, FY2015 through FY2026 year-to-date for LCA and
FY2015 through FY2025 for PERM (PERM's FY2026 file isn't published yet). Case-status rows recorded as NULL
(a known data-quality artifact in a handful of source files — see our disclosure-data quality report)
were excluded from both the numerator and denominator rather than assumed either way.
Before certifying a Labor Condition Application, the regulation gives DOL's Certifying Officer two things to look for — and nothing else:
That's the whole test: is the form filled out, and is anything on its face obviously wrong. The regulation does not authorize DOL to investigate whether the employer's attestation — that it will pay the prevailing wage, that hiring the foreign worker won't adversely affect similarly employed U.S. workers, that no strike or lockout is underway — is actually true. That determination is left to after-the-fact enforcement (complaint-driven investigations by DOL's Wage and Hour Division), not to the certification step itself. Contrast this with PERM, the green-card labor certification program run by the same office: PERM requires the employer to conduct and document actual recruitment of U.S. workers — placing job orders, running print ads, interviewing applicants, and filing a recruitment report — before DOL will certify anything. LCA has no equivalent requirement. That difference in what the two programs are legally required to verify is the entire subject of this report.
Congress didn't need to speculate about the effect of the attestation-only design — it asked GAO to check, and GAO reported back in a document sent directly "to congressional committees":
GAO's report went to Congress with a specific recommendation: that lawmakers "re-examine key H-1B program provisions and make appropriate changes as needed." Homeland Security pushed back on two of GAO's recommendations. Labor's response to the report, as GAO recorded it, was silence: "Labor did not respond to our recommendations." Fifteen years later, 20 CFR §655.740's completeness-and-obvious-inaccuracy standard is unchanged.
An attestation-only review with no substantive fact-check should produce almost no denials — because there's almost nothing left to deny an application for, short of a blank field or a wage typo. That's exactly what DOL's own disclosure files show for every year on record locally:
PERM sits in the same OFLC office, processed through the same FLAG system, and covers a comparable volume of cases — but its regulation requires a real labor-market test: documented recruitment, not just a signed attestation. Its denial rate has run 3 to 12 times higher than LCA's every single year since 2015:
| Fiscal Year | LCA Filed | LCA Denied | LCA Denial Rate | PERM Filed | PERM Denied | PERM Denial Rate | Gap |
|---|---|---|---|---|---|---|---|
| 2015 | 618,804 | 10,983 | 1.78% | 89,299 | 5,999 | 6.72% | 3.8× |
| 2016 | 647,852 | 9,220 | 1.42% | 126,143 | 5,560 | 4.41% | 3.1× |
| 2017 | 624,650 | 8,480 | 1.36% | 97,603 | 6,413 | 6.57% | 4.8× |
| 2018 | 654,360 | 8,627 | 1.32% | 119,776 | 6,255 | 5.22% | 4.0× |
| 2019* | 664,616 | 5,893 | 0.89% | 102,655 | 5,535 | 5.39% | 6.1× |
| 2020 | 577,334 | 3,983 | 0.69% | 94,019 | 4,213 | 4.48% | 6.5× |
| 2021 | 826,305 | 4,369 | 0.53% | 108,264 | 4,141 | 3.82% | 7.2× |
| 2022 | 626,084 | 3,096 | 0.49% | 104,600 | 4,573 | 4.37% | 8.8× |
| 2023 | 644,607 | 3,016 | 0.47% | 116,427 | 6,364 | 5.47% | 11.7× |
| 2024 | 890,368 | 7,313 | 0.82% | 92,258 | 4,892 | 5.30% | 6.5× |
| 2025 | 827,189 | 4,559 | 0.55% | 147,056† | 2,615† | 1.78%† | 3.2׆ |
| 2026 (YTD) | 293,507 | 1,377 | 0.47% | — | — | — | — |
| 2015–2026 total | 7,895,676 | 70,916 | 0.90% | 1,198,100 | 56,560 | 4.72% | 5.3× |
* FY2019 LCA figures exclude 383,931 NULL-status rows. † FY2025 PERM figures likely incomplete (see chart footnote); excluded from consideration when describing the "typical" gap, which runs 3.1×–11.7× across the eleven complete years 2015–2024.
This investigation is produced independently — one person, 23 years of data work, living on $1,419/month Social Security. If this work matters to you, a small donation makes a real difference.
♥ Donate via Square More Reports