The H-1B worker pool shrank by at least 124,553, but public data cannot say why

On July 24, the First Circuit denied the federal government's motion to keep the agencies' $100,000 H-1BH-1BA temporary US immigration classification that allows an employer to hire a foreign worker for a specialty occupation that usually requires at least a bachelor's degree. payment policy in force while an appeal proceeds. The court said the government had not made the required strong showing that it was likely to overturn the district court's conclusion that the agencies exceeded their statutory authority. Source: https://www.ca1.uscourts.gov/sites/ca1/files/opnfiles/26-1699O-01A.pdf

That order is important but narrow. It leaves the district court's June 8 vacaturvacaturA court remedy that sets aside a government rule or action. of the implementing policy operative for now. It does not finally decide the government's appeal, and it does not erase the fact that employers faced the payment policy when they decided whether to register workers in March. Sources: https://www.ca1.uscourts.gov/sites/ca1/files/opnfiles/26-1699O-01A.pdf and https://oag.ca.gov/system/files/attachments/press-docs/H1B%20Order.pdf

The fresh court order therefore sharpens a different question. The policy was present during the registrationregistrationThe short electronic entry an employer submits before it may be selected to file a complete cap-subject H-1B petition. decision, but can the large decline in registrations tell us how much the payment changed employer behavior? The answer is no. The decline contains one strong finding about the number of prospective workers and almost no causal information about why those workers disappeared from the pool.

USCISUSCISUS Citizenship and Immigration Services, the federal agency that administers H-1B registrations and petitions. reported that properly submitted registrations fell from 343,981 for FY2026 to 211,600 for FY2027. Subtracting the second count from the first gives 132,381 fewer registrations. Dividing that decline by 343,981 gives 38.5%, matching the agency's reported percentage. Source: https://www.linkedin.com/posts/uscis_this-years-h-1b-season-is-wrapping-up-and-activity-7463328105806618624-oR21

Registrations are not automatically the same as people because more than one employer can register the same worker. Think of a registration as a labeled ticket: each ticket names one worker, but the same worker can appear on several tickets. That distinction caused a major interpretive problem in earlier cap seasons. USCIS later moved to beneficiary-centered selectionbeneficiary-centered selectionA system designed to give each prospective worker one selection opportunity regardless of how many employers registered that person. so that multiple employers would not give one worker multiple independent chances. Sources: https://www.uscis.gov/working-in-the-united-states/temporary-workers/h-1b-specialty-occupations/h-1b-electronic-registration-process and https://content.govdelivery.com/accounts/USDHSCIS/bulletins/3999fb3

The FY2026 table nevertheless gives a solid starting point: 336,153 eligible registrations were for beneficiaries with no other eligible registration. Each of those registrations necessarily represents a different person, so FY2026 contained at least 336,153 prospective workers. The true count was higher because the remaining registrations included additional people who had multiple registrations. Source: https://www.uscis.gov/working-in-the-united-states/temporary-workers/h-1b-specialty-occupations/h-1b-electronic-registration-process

For FY2027, USCIS has released 211,600 properly submitted registrations but not a distinct-worker count. The number of people cannot exceed the number of registrations because one registration names one beneficiary. The most conservative comparison is therefore the smallest confirmed FY2026 worker count, 336,153, against the largest possible FY2027 worker count, 211,600.

The result is 124,553. Even if every FY2027 registration named a different person and the FY2026 multiple-registration category added no one to the earlier floor, the prospective-worker pool still fell by at least 124,553 people. Relative to the conservative FY2026 floor, that is a decline of at least 37.1%.

USCIS's FY2026 table assigns 7,828 eligible registrations to beneficiaries with multiple eligible registrations. That entire category equals only 5.9% of the subsequent 132,381-registration decline. Source: https://www.uscis.gov/working-in-the-united-states/temporary-workers/h-1b-specialty-occupations/h-1b-electronic-registration-process

Even 5.9% is intentionally too generous as an explanation based on duplicate or extra registrations. The 7,828 category includes the first legitimate registration for every multiply registered worker, not just registrations beyond the first. Removing the whole category pretends that every one of those workers and every one of their registrations vanished solely because registration multiplicity changed.

The conservative worker decline of 124,553 is 94.1% as large as the total registration decline. That comparison does not assign causes. It establishes something narrower: the vast majority of the numerical drop cannot be dismissed as the same number of workers generating fewer entries.

The FY2025 season produced a superficially similar fall in registrations but a very different worker trend. Eligible registrations declined from 758,994 in FY2024 to 470,342 in FY2025, which is 38.0% when calculated from the published counts. Approximate unique-beneficiarybeneficiaryThe prospective foreign worker named in a registration or petition. counts moved only from 446,000 to 442,000, a decline of about 0.9%. Source: https://content.govdelivery.com/accounts/USDHSCIS/bulletins/3999fb3

That earlier registration collapse occurred as beneficiary-centered selection removed the advantage of accumulating several registrations for one person. The number of tickets plunged while the approximate number of people barely moved. USCIS attributed much of that change to its integrity measures, but its published figures were descriptive rather than a controlled causal study. Source: https://content.govdelivery.com/accounts/USDHSCIS/bulletins/3999fb3

FY2027 is different. The lower-bound calculation shows that people, not merely repeated entries, account for almost all of the latest numerical decline. What it still does not reveal is why employers registered fewer people.

The September 19, 2025 proclamation imposed a $100,000 payment condition on certain H-1B petitions involving workers outside the United States and later entry, while permitting national-interest exceptions. USCIS told employers before registration that the proclamation did not directly change the electronic registration process, but that a selected employer might need to pay the additional amount before filing a petitionpetitionThe detailed application an employer files with USCIS after a worker has a valid selection.. Sources: https://www.whitehouse.gov/presidential-actions/2025/09/restriction-on-entry-of-certain-nonimmigrant-workers/ and https://content.govdelivery.com/accounts/USDHSCIS/bulletins/406da02

The distinction between direct procedure and anticipated cost matters. An employer did not pay $100,000 to submit a registration, but a firm expecting to face the payment after selection could rationally decide that registration was no longer worthwhile. That is a possible behavioral channel, not an observed effect in the public data.

At the same time, DHSDHSThe Department of Homeland Security, the cabinet department that includes USCIS. replaced equal-probability selection with a wage-weighted process. A Level I worker received one entry in the selection pool, Level II received two, Level III received three and Level IV received four. The final rule took effect on February 27, 2026, before registration opened on March 4. Sources: https://www.govinfo.gov/content/pkg/FR-2025-12-29/pdf/2025-23853.pdf and https://content.govdelivery.com/accounts/USDHSCIS/bulletins/406da02

DHS's simple model put selection probabilities at 15.29% for Level I, 30.58% for Level II, 45.87% for Level III and 61.16% for Level IV. Those figures are projections, not FY2027 results, but they demonstrate that the rule changed the expected return from registering at different wage levels. Source: https://www.govinfo.gov/content/pkg/FR-2025-12-29/pdf/2025-23853.pdf

A lower-wage employer could therefore register fewer workers because the odds were worse even if the $100,000 policy had no effect. A higher-wage employer could respond differently. Changes in hiring needs, financing conditions, recruiting strategies, expected legal outcomes and the availability of workers already in the United States could also alter registrations. The single nationwide total contains no way to separate those possibilities.

The payment policy did not place every prospective worker in the same position. In FY2024, USCIS approved 141,205 initial-employment H-1B petitions. Almost 46% requested consular or port-of-entry notification, while approximately 54% requested a change to H-1B status for someone already in the United States. Source: https://www.uscis.gov/sites/default/files/document/reports/ola_signed_h1b_characteristics_congressional_report_FY24.pdf

Those historical percentages cannot be applied directly to FY2027. They describe approved petitions from an earlier year, not registrations submitted under the new policies. A change-of-status request can also become a consular-notification case if USCIS cannot approve the requested status change.

The useful lesson is simply that exposure varied. A credible estimate of the payment's effect would need to compare registration changes among employers and workers likely to face the payment with changes among otherwise similar groups that were less exposed, while also accounting for wage level and the redesigned lottery.

USCIS's public FY2027 figure is one before-and-after comparison for the entire country. There is no published breakdown linking each registration to intended petition route, payment liability, national-interest exception, wage levelwage levelOne of four categories that compares an offered wage with wages for the same occupation and area, from Level I to Level IV., occupation, employer, work location or eventual petition outcome. The agency's social post reported several selection characteristics, but not the linked records required for causal attributioncausal attributionDetermining how much of an observed change was produced by one cause rather than merely occurring at the same time.. Source: https://www.linkedin.com/posts/uscis_this-years-h-1b-season-is-wrapping-up-and-activity-7463328105806618624-oR21

This is the statistical equivalent of seeing that traffic fell after a city simultaneously introduced a toll, closed two lanes and entered a recession. The traffic count proves that fewer cars arrived. It does not reveal how many were deterred by the toll.

A difference-in-differencesdifference-in-differencesA method that compares how an exposed group changed with how a similar unexposed group changed over the same period. study could help if researchers had suitable data. It would compare the change for a group likely to face the payment with the contemporaneous change for a similar group unlikely to face it. The design would still require evidence that the groups had been moving similarly before the policy and careful controls for wage-weighted selectionweighted selectionA lottery in which entries receive different weights, with the FY2027 system giving greater weight to higher wage levels..

• A unique, privacy-protected identifier linking registrations, selections and petitions across cap seasons.
• The intended change-of-status or consular route recorded at registration and the route ultimately requested.
• Whether the $100,000 payment was required, paid, waived or avoided because the case changed route.
• The wage level, offered wage, occupation, location, employer and US advanced-degree status for each registration.
• Comparable records from several seasons before the policy changes so pre-existing trends can be tested.
• Petition filing, approval, denial, withdrawal, visa issuance and admission outcomes.

Without those fields, any claim that the payment caused a specified number or percentage of the 124,553-worker minimum decline would be invented precision.

On July 17, USCIS said it had received enough petitions to reach the 65,000 regular cap and the 20,000 advanced-degree exemptionadvanced-degree exemptionThe allocation of up to 20,000 additional H-1B places for workers with qualifying master's or higher degrees from US institutions. for FY2027. That announcement concerns petition receipts sufficient to meet the statutory allocations. It does not say that every petition was approved, that every approved worker obtained a visa or that every worker entered the United States. Source: https://www.uscis.gov/newsroom/alerts/uscis-reaches-fiscal-year-2027-h-1b-cap

DHS itself modeled more petition receipts than the approximately 85,000 statutory places because some approved workers may not seek a visa or admission, may not receive a visa or may not be admitted. Selection is therefore permission to file a petition, not the final immigration outcome. Source: https://www.govinfo.gov/content/pkg/FR-2025-12-29/pdf/2025-23853.pdf

The cap announcement does establish that the smaller pool still produced enough filed petitions for USCIS to stop making additional selections. It does not undo the measured contraction in prospective workers and does not identify its cause.

The public record proves that the FY2027 pool contained at least 124,553 fewer prospective workers. It does not prove how many were deterred by the $100,000 payment policy.

That is a meaningful finding. It rules out the easiest alternative explanation that registrations merely fell because the same workers generated fewer entries. But the remaining causal question is unresolved because the payment policy, wage-weighted lottery and other employer decisions changed together.

USCIS may eventually publish the distinct FY2027 beneficiary count and more detailed registration and petition data. Until then, the defensible conclusion is a bound on the size of the contraction, not a verdict on what caused it.