Weighted H-1B Selection

The lottery that allocated capped H-1B visas now gives applicants between one and four entries according to the wage they are offered.

AI evaluation · not yet reviewed by a human

This evaluation was produced and sourced by an AI model; a human review is still pending. Figures and conclusions may still change. The review log is at the foot of the page.How review works →

Every year 85,000 H-1B visas are available to employers not exempt from the cap, and for two decades they were drawn at random from far more registrations than places. Since February 2026 each registration receives one entry for a job paying at the lowest of the Labour Department's four prevailing wage levels and four entries at the highest. The cap itself did not change: the same number of people enter, selected differently. The first year under the new rule drew 211,600 registrations against 343,981 a year earlier, and the share of selections at the lowest wage level fell to 17.7 percent. This evaluation looks four years ahead from the first weighted draw in March 2026.

Balance

Better for the future · 0.64 previous scale

Balance on the previous scale. The Bilanz 2.0 simulation is not yet available for this evaluation. The category comes from the share of the debate on the pro side (r).

For 34 · 64 % Against 19 · 36 %

Size class: small Scale of this evaluation: Normalised Impact — unitless, calibrated to this topic. For comparison: one point here is worth roughly 50 million euro per year. The cap is unchanged, so nothing here turns on how many people enter the country. Every figure in this evaluation is the difference between two sets of 85,000 people, which is why all of them are small. How we score →

Arguments for

Arguments against

6 arguments evaluated · Scoring v1.3 Δ absolute +15

Arguments — For

3 arguments

The visa goes where it is worth most

17of 100

A lottery hands a scarce permit to whoever draws it. Weighting by wage hands it to whoever will pay most for it, which is a rough measure of where the worker will produce most. Rough is the right word, and it is the subject of the first argument against.

Value 6 · OutputImpact 6.4Plausibility 4.5

▸ Show reasoning & sources ▾ Hide reasoning & sources

The stream is output: goods and services produced that would not otherwise exist, because a scarce work permit sits with the employer who values it most rather than with whoever won a draw. That belongs to the class this site uses for economic systems and prosperity. The wage itself is not the value — it is a payment from an employer to a worker, and both sides of it are inside the economy — so what is counted is the surplus above the wage, which is what the employer keeps and what competition passes to customers. Nothing about the worker's own welfare enters here. The value sits in the middle-upper part of the scale, at the level this site uses for economic output.

Allocating a fixed number of permits to those who bid most for them raises total output as long as the bid tracks the value produced. The bid here is the offered salary, and the shift in the mix is worth about 1.62 billion euro a year in additional pay across the 85,000 selections. What matters for output is not that pay but the margin above it: what the employer keeps after wages, which is what made the job worth filling. Twenty percent of the wage is used for that margin, in a range from 5 to 40 percent, giving 324 million euro a year. The upper end of the range would be justified if the specialist roles at issue are genuinely hard to fill; the lower end if employers are largely bidding against each other for the same people. Nothing here assumes any additional worker enters the country, because none does. The Impact is small and of the same order as the tax gain, which is what one expects when a fixed quota is reshuffled rather than enlarged.

▸ Show calculation ▾ Hide calculation

	Additional pay of the people selected [2][3]	85,000 × 19,000 euro	1,615 million euro
×	Margin the employer keeps above the wage Setting, range 5 to 40 percent: what made the job worth filling; the low half of the range is used because firms winning permits in the old lottery largely substituted rather than expanded [4]	20 %	320 million euro
÷	Normalised Impact scale of this evaluation	50 million euro a point	6.4

Score 6.4 Impact × 6 Value × 4.5 Plausibility ÷ 10 = 17 of 100

The claim that a higher bid signals higher value is a proposition about markets, not a measured finding, and the market in question is unusual. The counterfactual is the same 85,000 permits drawn at random, and no study compares output under the two allocations, because the weighted rule is nine months old. What can be said is that the H-1B lottery has been used as a natural experiment before: comparing firms that won permits against those that did not, researchers found the winners did not expand employment or patenting much, which suggests they substituted the new worker for someone they would otherwise have hired ^[4]. If that holds generally, the surplus from any single permit is smaller than the wage suggests, which is why the margin used here sits at the lower end. The confounder that would matter — that high-paying firms differ from low-paying ones in ways that also raise output — is not removed by anything, since no design separates the two. Reverse causation does not arise. The Plausibility is below the middle: the reasoning is sound in outline, no measurement supports the size, and the one clean study on adjacent ground points to a smaller effect than the wage implies.

evidence basis: Mechanism · P ceiling 6 identification: Mechanistic · rung ceiling 6 band: Chain closed, unevidenced · P 4–5

Counterfactual: the same 85,000 permits drawn at random. Design: mechanistic — the chain (higher bid → higher marginal product → more output) is named but no design separates bid from firm characteristics. Confounder: high-paying firms differing from low-paying ones in ways that also raise output; unaddressed. Direction: no reverse causation, the rule preceded the draw. Ceiling: mechanistic 6.0 binds. Band: chain closed but unevidenced — each link is named and the substitution counter-mechanism from the lottery study is answered by setting the margin at the low end of its range; only the measurement is missing.

Nothing measured argues against the direction; the lottery study showing that winning firms substituted rather than expanded is answered by taking the low end of the margin range [4]. Read back: reshuffling the permits toward higher bidders raises output by roughly the amount assumed here about as often as it does not.

Open: Firm-level output and employment for FY2027 winners against FY2026 winners, matched on size and sector, would put a number on the margin and could carry P to 6.

Higher earners pay more into the public purse

14of 100

The same 85,000 visas now go to people earning more. A higher salary means more income and payroll tax and no more use of public services. What the visa holder loses in take-home terms weighs less than what the budget gains, because they are high earners.

Value 5 · Public financesImpact 5.2Plausibility 5.5

▸ Show reasoning & sources ▾ Hide reasoning & sources

The stream is tax revenue, priced at the middle of the scale like any other euro. What makes it a gain rather than a wash is where it comes from and where it goes: a euro taken from someone in the top tenth of American earners and spent by a government serving everyone is worth more in the second place than in the first. That difference in weight is counted in the Impact, not here. The services these workers use do not rise with their salary, so no offsetting cost enters. Whether the money funds defence or Medicaid makes no difference to the weight. The value is the middle of the scale, the level this site uses for public money whatever it is spent on.

The cap is fixed, so the question is what the 85,000 selected people earn under one rule against the other. Under the lottery the wage levels were drawn in proportion to how they were registered; under the weighting the lowest level fell to 17.7 percent of selections and the share holding an American advanced degree rose from 57 to 71.5 percent ^[2]. That is roughly a one-level shift in the mix. Against a median compensation of about 110,000 dollars for a cap-subject worker starting employment, one level is worth about 20 percent, or 19,000 euro a year at 1.16 dollars to the euro ^[3]. Across 85,000 people that is 1.62 billion euro of additional earnings, in a range from 0.8 to 2.4 billion. About 32 percent of it reaches federal, state and payroll tax, which is 517 million euro. The money moves from people in the top tenth of American earners, where this site counts a euro at half its worth at median income, to a government budget, where it counts at one: half of it is a gain, 259 million euro. The Impact is small because the cap is unchanged — only the mix of who fills it moves, and only the tax on the difference is counted.

▸ Show calculation ▾ Hide calculation

	Visas allocated under the cap each year [1]		85,000 visas
×	Additional pay of the people selected Setting, range 9,000 to 28,000 euro: the share of selections at the lowest wage level fell to 17.7 percent, about a one-level shift, worth roughly 20 percent of a 110,000 dollar median; converted at 1 euro = 1.16 dollars [2][3]	19,000 euro a year each	1,615 million euro
×	Share reaching federal, state and payroll tax Setting, range 25 to 38 percent: the combined marginal rate at these incomes; the services these workers use do not rise with salary	32 %	517 million euro
×	Difference in what a euro is worth the money moves from the top tenth of American earners, where this site counts a euro at 0.5, to a public budget, where it counts at 1.0	1.0 minus 0.5	259 million euro
÷	Normalised Impact scale of this evaluation	50 million euro a point	5.2

Score 5.2 Impact × 5 Value × 5.5 Plausibility ÷ 10 = 14 of 100

The rule is arithmetic applied to a draw, and its first outcome has already been observed rather than predicted. The counterfactual is the FY2026 cap season under the old lottery, run on the same population of employers a year earlier, and the difference in the wage-level composition of selections between the two years is published ^[2]. The confounder is that other things changed at the same time — a 215 dollar registration fee and, for part of the year, a proposed 100,000 dollar entry fee that a court later struck down — and both would depress registrations from lower-paying employers independently of the weighting. That is why the wage shift used here is the composition of selections rather than the fall in registrations. Reverse causation does not arise: the rule was written before the draw. What is estimated rather than observed is the size of the wage gap between adjacent levels and the share of it that reaches tax, and both are stated as ranges. The Plausibility is at the upper end of what a projection can carry: the composition change is measured, and only its translation into salary and tax is estimated.

evidence basis: Projection · P ceiling 6 identification: Definitional · no rung ceiling

Counterfactual: the FY2026 cap season under the old random lottery, on the same employer population [2]. Design: definitional — the number of entries per registration follows from the rule, and the resulting composition of selections was published rather than modelled. Confounder: the 215 dollar registration fee and the proposed 100,000 dollar entry fee, both of which independently deter lower-paying employers; handled by taking the composition of selections rather than the fall in registrations. Direction: no reverse causation, the rule preceded the draw. Ceiling: projektion 6.0 binds because the wage gap between levels and the tax share are estimated; definitional carries no ceiling of its own.

Fewer entry-level posts filled from abroad

2.5of 100

The share of selections at the lowest prevailing wage level fell from about half to 17.7 percent. Those are the posts an American graduate would most plausibly have competed for. Roughly thirteen thousand of them a year now go to a domestic applicant instead.

Value 5 · Household budgetsImpact 1.4Plausibility 3.5

▸ Show reasoning & sources ▾ Hide reasoning & sources

The stream is earnings in the hands of American workers who take a post they would not otherwise have got, priced at the middle of the scale like any other money. What is counted is not their whole salary — the job exists either way and someone fills it — but the difference between that post and the next best one available to them. The employer is no worse off in a way that is counted here, because the position is filled at the same wage. That the workers concerned are early in their careers and earning below the median raises the weight the euro carries, which is in the Impact rather than here. The value is the middle of the scale, because the stream is a difference in earnings and the incomes it lands in are priced in the Impact.

Before the change, roughly half of cap selections were at the two lowest prevailing wage levels; afterwards, 17.7 percent were at the lowest ^[2]. Taking a fifteen percentage point shift out of the entry-level band gives about 12,750 posts a year that a lower-paid applicant no longer fills. Whether an American takes each of those posts is the open part: the employer may raise the wage and hire the same kind of candidate, or may not fill it at all. Three in ten is used here, in a range from one to six in ten, giving 3,825 people. What each gains is not the salary but the gap between this post and their next best option, about 15,000 euro a year. Those are early-career earners in the second fifth of American incomes, where this site counts a euro at one and a half. The result is 69 million euro a year. The Impact is the smallest figure in this debate, because a shift within a quota of 85,000 is a small event in a labour market of 168 million.

▸ Show calculation ▾ Hide calculation

	Entry-level selections that no longer occur the share of selections at the lowest wage level fell from about half of the two lowest bands to 17.7 percent [2]	85,000 × 15 percentage points	12,750 posts a year
×	Posts an American applicant takes instead Setting, range 10 to 60 percent: the employer may raise the wage and hire the same kind of candidate, or leave the post unfilled [4]	30 %	3,825 people
×	Gap between this post and their next best option	15,000 euro a year	57 million euro
×	Weight of a euro at these incomes early-career earners in the second fifth of American incomes, where this site counts a euro at 1.5, mixed with some above it	1.2	69 million euro
÷	Normalised Impact scale of this evaluation	50 million euro a point	1.4

Score 1.4 Impact × 5 Value × 3.5 Plausibility ÷ 10 = 2.5 of 100

Whether restricting one source of labour raises employment for domestic workers in the same occupation is among the most contested questions in labour economics, and the general answer is that displacement is smaller than intuition suggests. The counterfactual here is the same 85,000 posts filled under the old draw. The nearest clean evidence comes from the H-1B lottery itself, where firms that won permits did not expand overall employment, which implies they were substituting between candidates — the mechanism this argument relies on ^[4]. That is support for the direction but not for the size, and it comes from firm-level data rather than from any measurement of who fills the post instead. The confounder that would matter — that high-wage and low-wage employers hire from different pools entirely, so no substitution happens at all — is exactly what has not been checked. Reverse causation does not arise. The Plausibility is below the middle because the link that carries the whole quantity, the share of posts an American actually takes, has never been measured.

evidence basis: Mechanism · P ceiling 6 identification: Mechanistic · rung ceiling 6 band: Chain open · P 3–3.5

Counterfactual: the same posts filled under the old random draw. Design: mechanistic — chain named (fewer entry-level selections → post filled domestically → earnings gain), with the lottery study supporting substitution at firm level [4]. Confounder: high-wage and low-wage employers hiring from separate pools, so no substitution occurs; unchecked. Direction: no reverse causation. Ceiling: mechanistic 6.0 binds. Band: chain open, because the share of posts taken domestically has no carrier and the separate-pools counter-mechanism is unanswered.

The chain is named but the link carrying the quantity — how many of the freed posts an American actually takes — is unmeasured, and the possibility that the two wage bands draw on entirely separate pools is unanswered. Read back: about a third of the time, roughly three in ten of the freed entry-level posts go to a domestic applicant.

Open: Job-posting and hiring data for the affected occupations, comparing FY2027 against FY2026 cap seasons, would show whether the freed posts were filled domestically and could carry P to 6.

Arguments — Against

3 arguments

Graduates the country trained and does not keep

9of 100

American universities award several hundred thousand advanced degrees to international students each year, and the capped H-1B is the main way any of them stay. A new graduate is typically offered a first- or second-level wage. The first year's data cut against this argument rather than for it.

Value 6 · OutputImpact 5Plausibility 3

▸ Show reasoning & sources ▾ Hide reasoning & sources

The stream is output that leaves the country with the person who produces it: someone educated at an American university who works somewhere else instead. It is priced at the level this site uses for economic output, the same as the two arguments it argues against. What is counted is the contribution the person would have made here, not the cost of educating them, which is already spent whether they stay or go. Their own welfare is not counted, because it goes with them. The value sits in the middle-upper part of the scale, at the level this site uses for economic output.

The mechanism is straightforward: a new graduate is offered a first- or second-level prevailing wage almost by definition, since the levels track experience, so weighting by wage should push them out. Ten thousand graduates a year not retained is used here, in a range from zero to 30,000, and each would have contributed about 25,000 euro a year in output above what they consume in public services. That gives 250 million euro a year. The number is set well below what the mechanism alone would suggest, and the reason is the first year's data. The share of selections held by people with an American advanced degree did not fall under the new rule; it rose from 57 to 71.5 percent ^[2]. Whatever the weighting does to new graduates, it did not push them out in aggregate, and the argument survives only for the subset in lower-paying fields. The Impact is stated at a fifth of what the mechanism alone would give, because the one year of data available runs against the mechanism.

▸ Show calculation ▾ Hide calculation

	Graduates of American universities not retained Setting, range 0 to 30,000: set well below what the mechanism alone would give, because the advanced-degree share of selections rose rather than fell [2]		10,000 people a year
×	Output each would have added above the services they use Setting, range 15,000 to 40,000 euro	25,000 euro a year	250 million euro
÷	Normalised Impact scale of this evaluation	50 million euro a point	5

Score 5 Impact × 6 Value × 3 Plausibility ÷ 10 = 9 of 100

This argument has been tested by events and came off badly. The counterfactual is the FY2026 cap season under the old lottery; the comparison is the FY2027 season under the new rule, on the same population of employers a year apart ^[2]. Under the weighting, holders of American advanced degrees took a larger share of selections, not a smaller one — the opposite of what the argument predicts. Two things keep it alive rather than dismissed. The measured share covers all advanced degrees together, so it can rise while graduates in lower-paying fields such as public health or education are pushed out, and one cap season is a short series that also contains the effects of a registration fee and a proposed entry fee. The confounder is therefore real in both directions and unresolved. Reverse causation does not arise. The Plausibility is low: the aggregate finding contradicts the argument, and only a named subset of it is still supported.

evidence basis: Precedent · P ceiling 6 identification: Mechanistic · rung ceiling 6 band: Partial aspect supported · P 2.5–3

Counterfactual: the FY2026 cap season under the old lottery [2]. Design: mechanistic — the chain is named and one year of before-and-after data exists, but no design separates the weighting from the fee changes in the same season. Confounder: the 215 dollar registration fee and the proposed 100,000 dollar entry fee, both affecting the same year; unresolved. Direction: no reverse causation. Ceiling: mechanistic 6.0 binds below the praezedenz ceiling of 8.5. Band: partial aspect supported — the lowest wage level falling supports the mechanism while the advanced-degree share rising contradicts it, so only a named subset survives.

Supporting finding: the share of selections at the lowest wage level fell sharply, which is where new graduates sit. Contradicting finding: the share holding an American advanced degree rose from 57 to 71.5 percent, the opposite of what the argument predicts [2]. Only the subset in lower-paying fields survives. Read back: for roughly three graduates in ten of the kind this argument describes, the weighting is what keeps them out.

Open: A breakdown of selections by degree field rather than by degree level would show whether graduates in lower-paying disciplines were displaced, and could raise this argument or remove it.

A wage is a poor measure of what a job is worth

7.6of 100

The rule ranks a specialist by what an employer will pay. A physical therapist in a rural hospital, a mathematics teacher in a district that cannot fill the post, and a research associate on a grant all sit at the bottom of the wage table. What they contribute is not what they are paid.

Value 6 · OutputImpact 3.6Plausibility 3.5

▸ Show reasoning & sources ▾ Hide reasoning & sources

The stream is the same one the argument on allocation counts, running the other way: output and public value that does not happen because the permit went elsewhere. It carries the same weight, because it is the same kind of good — what a person produces for others. Where the value is public rather than commercial, as with a school or a rural clinic, it is still output and is still priced here rather than at the higher weight the site reserves for health or education as such, because what is lost is a marginal post rather than a service. The people who would have been served are not identified individually, so no separate stream is opened for them. The value sits in the middle-upper part of the scale, at the level this site uses for economic output.

About 406,000 H-1B petitions were approved in 2025, six in ten of them in computer occupations ^[3]. The cap-subject minority working in schools, therapy, social services and grant-funded research is small — many such employers are exempt from the cap entirely — but not nothing: about 6,000 selections a year, in a range from 3,000 to 12,000. Those posts sit at the bottom of the prevailing wage table in their areas, so under the weighting they now draw one entry against four. What is lost is the gap between what such a post pays and what it is worth: a mathematics teacher in a district that has advertised for two years, a therapist in a county with none. Thirty thousand euro a year is used for that gap, in a range from 10,000 to 60,000 — the number is a judgment, not a measurement, and it is the whole content of the argument. The result is 180 million euro a year. The Impact is roughly half the allocation gain it argues against, which is the honest shape of the objection: the wage is a decent proxy most of the time and a bad one in a minority of cases.

▸ Show calculation ▾ Hide calculation

	Cap-subject selections in schools, therapy, social services and grant-funded research Setting, range 3,000 to 12,000: assembled from occupation shares; most such employers are exempt from the cap entirely [3]		6,000 posts a year
×	Gap between what the post pays and what it is worth Setting, range 10,000 to 60,000 euro: a mathematics teacher in a district that has advertised for two years, a therapist in a county with none	30,000 euro a year	180 million euro
÷	Normalised Impact scale of this evaluation	50 million euro a point	3.6

Score 3.6 Impact × 6 Value × 3.5 Plausibility ÷ 10 = 7.6 of 100

That market wages understate social value in teaching, care and public research is not seriously contested; what has no measurement here is how much of the H-1B cap those occupations actually occupy and how large the gap is. The counterfactual is the old random draw, under which such a post had the same one-in-four odds as any other. No source counts cap-subject selections by whether the employer is a school district or a rural hospital, and the 6,000 used here is assembled from occupation shares rather than read off a table. The counter-argument is strong and only partly answered: most of these employers — universities, non-profit research institutions, affiliated hospitals — are exempt from the cap altogether and are untouched by the rule, which is why the figure is small. What is not answered is whether the remaining cap-subject share is 6,000 or a third of that. Reverse causation does not arise. The Plausibility is below the middle because the count of affected posts and the size of the gap are both constructions, and the exemption counter-argument removes an unknown part of the claim.

evidence basis: Mechanism · P ceiling 6 identification: Mechanistic · rung ceiling 6 band: Chain open · P 3–3.5

Counterfactual: the old random draw, under which a low-wage post had the same odds as any other. Design: mechanistic — chain named (low prevailing wage → one entry instead of four → post unfilled → public value lost); no source carries either the count or the gap. Confounder: cap exemption for universities and non-profit research, which removes an unknown share of the claim and is only partly answered. Direction: no reverse causation. Ceiling: mechanistic 6.0 binds. Band: chain open, because both load-bearing numbers are unmeasured and the exemption counter-mechanism is unresolved.

The chain is named but neither the count of affected posts nor the gap between pay and worth has a source, and the cap exemption for universities and non-profit research is only partly answered. Read back: about a third of the time, roughly 6,000 socially valuable low-wage posts a year are displaced by the weighting.

Open: A breakdown of cap-subject selections by employer type would settle the count within one cap season and could carry P to 6, or shrink the argument to a fraction of its size.

A small employer's odds fall further than a large one's

2.1of 100

Under a lottery a firm with one critical hire had the same chance per registration as a firm with a hundred. Weighting by wage changes that, because a start-up or a regional employer cannot match what a large technology firm pays for the same skill. Registrations fell by 38.5 percent in the first year.

Value 6 · OutputImpact 1Plausibility 3.5

▸ Show reasoning & sources ▾ Hide reasoning & sources

The stream is output that does not happen because a smaller employer cannot fill a specialist post. It is priced at the level this site uses for economic output, the same as the arguments it offsets. Nothing here treats a small firm as worth more than a large one for its own sake; what is counted is the work that goes undone, not who was supposed to do it. Whether the firm was going to succeed is not knowable and is not assumed. The value sits in the middle-upper part of the scale, at the level this site uses for economic output.

Registrations fell 38.5 percent in the first weighted season, from 343,981 to 211,600 ^[2]. Most of that is employers who calculated that a first- or second-level offer had become a poor bet. Behind those registrations sit perhaps 20,000 genuine hiring needs at smaller and regional employers, in a range from 8,000 to 40,000. Their odds did not vanish, though — under the old lottery a registration had about a one-in-four chance, and under the weighting a low-level registration still has roughly one in eight. Only the difference counts: thirteen percentage points across 20,000 needs is about 2,600 posts a year. Each unfilled specialist post costs a small firm something like 20,000 euro of output in a year, giving 52 million euro. The Impact is the smallest figure in this debate, because the change is a shift in odds rather than an exclusion.

▸ Show calculation ▾ Hide calculation

	Genuine hiring needs behind the withdrawn registrations Setting, range 8,000 to 40,000: registrations fell from 343,981 to 211,600, and speculative filings were a large share of the old total [2]		20,000 posts a year
×	Fall in their odds of selection about one in four under the old draw against roughly one in eight for a low-level registration under the weighting	13 percentage points	2,600 posts a year
×	Output foregone per unfilled post Setting, range 10,000 to 40,000 euro	20,000 euro a year	52 million euro
÷	Normalised Impact scale of this evaluation	50 million euro a point	1

Score 1 Impact × 6 Value × 3.5 Plausibility ÷ 10 = 2.1 of 100

The direction is not in doubt — a rule that rewards higher offers disadvantages employers who cannot make them — and the 38.5 percent fall in registrations is measured rather than inferred ^[2]. The counterfactual is the FY2026 season under the old lottery. What the measurement cannot separate is how much of the fall is the weighting and how much is the 215 dollar registration fee and the proposed 100,000 dollar entry fee that hung over the same season before a court struck it down; that confounder is unresolved and it works in the argument's favour, which is why the figure is set low. Nothing measures how many of the withdrawn registrations represented a real hiring need rather than a speculative filing, and speculative filings were widely believed to be a large share of the old total. Reverse causation does not arise. The Plausibility is below the middle: the fall in registrations is measured but cannot be attributed, and the share of it that was real demand is unknown.

evidence basis: Precedent · P ceiling 6 identification: Mechanistic · rung ceiling 6 band: Chain open · P 3–3.5

Counterfactual: the FY2026 cap season under the old lottery [2]. Design: mechanistic — the fall in registrations is observed, but nothing separates the weighting from the two fee changes in the same season, and nothing measures how many registrations were real demand. Confounder: the 215 dollar registration fee and the proposed 100,000 dollar entry fee; unresolved, and working in the argument's favour. Direction: no reverse causation. Ceiling: mechanistic 6.0 binds below the praezedenz ceiling of 8.5. Band: chain open, because the share of the fall attributable to the rule and the share representing real demand are both unchecked.

The fall in registrations is measured but not attributable, and the share of withdrawn registrations that were real hiring needs rather than speculative filings is unknown. Read back: about a third of the time, roughly 2,600 posts a year go unfilled at smaller employers because of the weighting.

Open: Registration counts by employer size for the FY2028 season, once the fee litigation has settled, would separate the weighting from the fees and could carry P to 6.

Summary

This rule does not change how many people come; it changes which 85,000 of them do, so every figure in the debate is small and none of them turns on immigration levels. The case for it is that a higher salary carries more tax and signals a job worth filling, and both of those are real if modest. The case against is that a salary is a rough measure of what a job is worth, and it is roughest exactly where the public interest is strongest — a teacher, a rural therapist, a research associate. What settles the balance in the rule's favour is not the strength of that reasoning but the first year's data: the share of selections held by graduates of American universities rose rather than fell, which is the opposite of what the strongest objection predicted. One cap season is a short series, and it also contains two separate fee changes, so that finding could yet reverse.

Outlook — effect over time

Better for the future · 0.64 previous scale

Sources

U.S. Citizenship and Immigration Services: FY 2027 H-1B Cap Initial Registration Period Opens on March 4. uscis.gov
American Immigration Lawyers Association: Featured Issue: FY2027 H-1B Cap Season. aila.org
U.S. Citizenship and Immigration Services: Characteristics of H-1B Specialty Occupation Workers, Fiscal Year 2025. uscis.gov
Doran, Gelber and Isen, Journal of Political Economy 130(10): The Effects of High-Skilled Immigration Policy on Firms: Evidence from Visa Lotteries. journals.uchicago.edu

Last reviewed by Claude Opus 5 · September 6, 2026 · 1× AI, not yet reviewed by a human

September 6, 2026AI reviewClaude Opus 5First evaluation
Created for the English side: six arguments on a reallocation of a fixed cap, with the first weighted cap season used as the observed outcome.

Evaluations are produced with AI support and reviewed on a schedule for new developments; human passes are marked separately.How we review →