💬 Discussion
Two Greek thinkers compare professional tools, work permit cards and weighted lottery vessels.

Weighted H-1B Selection

For two decades the capped H-1B work visas were drawn by lottery from far more registrations than places. The Department of Homeland Security replaced the draw with one weighted by the offered wage, which immigration lawyers, universities and business groups opposed. The rule took effect on 27 February 2026, and the first weighted draw took place in March 2026 [1].

  • 211,600 registrations in the first weighted season, against 343,981 a year earlier [2]
  • 17.8 percent of selections went to jobs at the lowest wage level in that season [2]
  • 71.5 percent of selections went to holders of an American advanced degree, up from 57 percent [2]

The proposal

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

The debate

The administration argues that employers gamed the random draw to bring in foreign workers below American wages, and that weighting by wage protects American workers. Critics answer that the wage levels measure experience rather than skill, that graduates and schools lose out, and that the law requires selection in filing order. Congress did not vote on the change. The department issued the final rule on 23 December 2025, after public comments on its proposal of September 2025.

  • “The existing random selection process of H-1B registrations was exploited and abused by U.S. employers who were primarily seeking to import foreign workers at lower wages than they would pay American workers.”

    argues for: Matthew Tragesser, U.S. Citizenship and Immigration Services Press release, 23 December 2025

  • “A proposed rule would significantly change H-1B selection by heavily weighting the process in favor of individuals in positions that require more experience and against early-career professionals, such as recent international students.”

    argues against: National Foundation for American Policy, National Foundation for American Policy Policy brief, September 2025

  • “Under the proposed rule, many schools would also be unable to hire teachers on H-1B petitions in middle school or high school, as around 90% are paid at Level I or Level II.”

    argues against: National Foundation for American Policy, National Foundation for American Policy Policy brief, September 2025

The analysis

Weighted H-1B Selection

81% of the weight is for, 19% against.

An argument's weight combines three questions: how big the effect is, how certain it is and how much the good at stake matters. The larger the slice, the more the argument weighs.

For

Against

Weighting the H-1B lottery by wage is a good measure on the current analysis, and a small one: it changes which 85,000 people come, not how many.

Very likely better

Very likely better: Under the assumptions used, the proposal brings more than it costs in 98 of 100 model runs.

 Overall effect: medium

How strongly the proposal works overall, in five steps from very small to very large, comparable across all debates. The ring is always full, even when the overall effect is small; this step says how large it is.

Like the department, the analysis favors the weighting, though for a reason the department barely mentions. The department speaks of protecting American workers, while the result rests on output: the same visas now go where employers bid most for them. The entry-level posts freed for American graduates, its headline claim, weigh little. The point critics pressed hardest, that graduates of American universities lose out, ran against the first year's data, when their share of selections rose. Their objection that a wage misprices schools and small employers holds for few posts, and one season shaped by a fee on petitions is a short test.

In detail

All arguments

Arguments for3 arguments

The visa goes where it is worth most

23points

A lottery hands a scarce permit to whoever draws it. Weighting by wage level hands it to whoever pays most for it against the going rate for that occupation and place, 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 8.6Plausibility 4.5
▸ Show reasoning & sources ▾ Hide reasoning & sources

Value

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. Where higher pay reflects more productive work, the whole of that extra output counts, whoever ends up with it — the worker as pay, the government as tax, the employer as margin. Where it reflects only a higher position on the wage table, nothing more is produced and nothing is counted here. The value sits in the middle-upper part of the scale, at the level this site uses for economic output.

Impact

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 measured against the four wage levels published for the occupation and the place — not an absolute salary ranking across occupations. Against a median compensation of about 103,000 dollars for workers starting H-1B employment in 2025 [3], one level is worth about 20 percent, or 20,600 dollars a year, in a range from 9,900 to 30,000 dollars; the range is wide at the bottom because part of the shift comes from a smaller pool and from employers choosing the occupation code or work site that lifts the level without lifting the pay. Across 85,000 people that is about 1.75 billion dollars of additional pay. Half of that is taken to reflect more productive work, in a range from a quarter to three quarters; the rest is a higher position on the wage table without more output, which the argument on public finances deals with. On the productive half the employer keeps a margin above the wage, which is what made the job worth filling: twenty percent is used, in a range from 5 to 40 percent. That gives about 1.05 billion dollars a year for each cohort, and 2.0 billion averaged across the four years, with 1.9 cohorts in the workforce at once. The upper end 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 the largest in this debate, because a better match between permit and job raises what the same 85,000 people produce.

▸ Show calculation ▾ Hide calculation
Additional pay of the people selected range 8,500 to 26,000 euro-equivalent each (about 9,900 to 30,000 dollars) [2][3] 85,000 × 17,800 euro-equivalent (about 20,600 dollars each, at 1.16 dollars to the euro) 1,510 million euro-equivalent
× Part that reflects more productive work Setting, range 25 to 75 percent: the rest is a higher position on the wage table without more output 50 % 755 million euro-equivalent
× Plus the margin the employer keeps above the wage Setting, range 1.05 to 1.4: 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] 1.2 906 million euro-equivalent
× Cohorts in the workforce at once, averaged over the four years each year's 85,000 selections start work from October at the earliest and stay for at least three years; across the four years from the first weighted draw in March 2026 that averages about 1.9 cohorts in the workforce at once 1.9 1,721 million euro-equivalent
÷ Normalised Impact scale of this evaluation 200 million euro-equivalent a point 8.6
Score 8.6 Impact × 6 Value × 4.5 Plausibility ÷ 10 = 23 of 100

Plausibility

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: Mechanism · plausibility at most 6 Evidence of effect: Mechanistic · at most 6 Placement: Chain closed, unevidenced · plausibility 4–5

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

3.2points

The same 85,000 visas now go to people earning more. Where that higher pay is a matter of where the job sits on the wage table rather than of what the person produces, the extra tax on it moves money from high earners to the public budget.

Value 5 · Public financesImpact 1.1Plausibility 5.5
▸ Show reasoning & sources ▾ Hide reasoning & sources

Value

The stream is tax revenue, priced at the middle of the scale like any other dollar. What makes it a gain rather than a wash is where it comes from and where it goes: a dollar taken from someone in the top fifth 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.

Impact

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 about 17.8 percent of selections [2][5]. That is roughly a one-level shift in the mix. Against a median compensation of about 103,000 dollars for workers starting H-1B employment in 2025 [3], one level is worth about 20 percent, or 20,600 dollars a year, in a range from 9,900 to 30,000 dollars; the range is wide at the bottom because part of the shift comes from a smaller pool and from employers choosing the occupation code or work site that lifts the level without lifting the pay. Across 85,000 people that is about 1.75 billion dollars of additional pay. Half of it is taken to reflect more productive work, which the argument on allocation counts in full, tax included; the other half, in a range from a quarter to three quarters, is pay that is higher on the table without more being produced. About 32 percent of that half reaches federal, state and payroll tax, about 280 million dollars. The money moves from households in the top fifth of American incomes, where this site counts a dollar at half its worth at median income, to a government budget, where it counts at one: half of it is a gain. Averaged over the four years, with 1.9 cohorts in the workforce at once, that is a Normalised Impact of 1.15 on the scale of this evaluation. The Impact is small because only the tax on the part of the higher pay that reflects no more output is a gain of this kind; the rest belongs to the argument on allocation.

▸ Show calculation ▾ Hide calculation
Visas allocated under the cap each year [1] 85,000 visas
× Additional pay of the people selected Setting, range 8,500 to 26,000 euro-equivalent (about 9,900 to 30,000 dollars): about a one-level shift in the wage mix, worth roughly 20 percent of the 103,000 dollar median for workers starting H-1B employment; converted at 1.16 dollars to the euro; the lower end allows for a smaller pool and for levels chosen without higher pay [2][3] 17,800 euro-equivalent a year each (about 20,600 dollars) 1,510 million euro-equivalent
× Part that is higher on the table without more output Setting, range 25 to 75 percent: the other part reflects more productive work and is counted in the argument on allocation 50 % 755 million euro-equivalent
× Share reaching federal, state and payroll tax Setting, range 25 to 38 percent: the combined marginal rate at these incomes 32 % 242 million euro-equivalent
× Difference in what a dollar is worth the money moves from households in the top fifth of American incomes, where this site counts a dollar at 0.5, to a public budget, where it counts at 1.0 1.0 minus 0.5 121 million euro-equivalent
× Cohorts in the workforce at once, averaged over the four years each year's 85,000 selections start work from October at the earliest and stay for at least three years; across the four years from the first weighted draw in March 2026 that averages about 1.9 cohorts in the workforce at once 1.9 230 million euro-equivalent
÷ Normalised Impact scale of this evaluation 200 million euro-equivalent a point 1.15
Score 1.15 Impact × 5 Value × 5.5 Plausibility ÷ 10 = 3.2 of 100

Plausibility

The rule is arithmetic applied to a draw, but what it does to salaries runs through choices it does not control: which employers register, and at which wage level they offer. Its first outcome has 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 design is a before-and-after comparison on two different pools of registrations, not a controlled one. The confounder is that something else changed at the same time — a 100,000 dollar fee on new petitions, imposed by presidential proclamation in September 2025 and in force during the March 2026 registration, before a federal court struck it down in June 2026; the 215 dollar registration fee is not a confounder, because it already applied to the year compared against. That fee deterred 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, how much of it reflects more output, and the share that reaches tax, and all three 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: Projection · plausibility at most 6 Evidence of effect: Mechanistic · at most 6

Fewer entry-level posts filled from abroad

0.91points

On the department's own model, the weighting cuts selections at the lowest prevailing wage level by about 10,000 a year. Those are the posts an American graduate would most plausibly have competed for, and roughly 3,000 of them are modelled to go to a domestic applicant instead.

Value 5 · Household budgetsImpact 0.5Plausibility 3.5
▸ Show reasoning & sources ▾ Hide reasoning & sources

Value

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 dollar 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.

Impact

The department's own analysis of the rule, run on the same pool of registrations under both methods, puts the selections at the lowest prevailing wage level at about 23,800 under the lottery and 13,700 under the weighting for 85,000 visas: about 10,100 fewer a year [6]. The first weighted season fits that: 17.8 percent of selections were at the lowest level [5], against a share of about 28 percent of registrations in earlier years [6]. A range from 8,000 to 12,750 is used. 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, may move the work abroad, or may not fill it at all. Three in ten is used here, in a range from one to six in ten, giving about 3,000 people. What each gains is not the salary but the gap between this post and their next best option, about 17,000 dollars a year. Those are early-career earners mostly in the second fifth of American incomes, where this site counts a dollar at one and a half, blended to 1.2 here because some of them sit above that band. The result is a Normalised Impact of 0.52 on the scale of this evaluation, averaged across the 1.9 cohorts of the four-year horizon. 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 about 160 million.

▸ Show calculation ▾ Hide calculation
Entry-level selections that no longer occur Setting, range 8,000 to 12,750: the department's model of the rule on the same pool of registrations; the first weighted season had 17.8 percent of selections at the lowest level [5][6] 23,830 under the lottery against 13,731 under the weighting 10,100 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, move the work abroad, or leave the post unfilled [4][8] 30 % 3,030 people
× Gap between this post and their next best option 15,000 euro-equivalent a year (about 17,000 dollars) 45.5 million euro-equivalent
× Weight of a dollar at these incomes early-career earners in the second fifth of American incomes, where this site counts a dollar at 1.5, mixed with some above it 1.2 54.5 million euro-equivalent
× Cohorts in the workforce at once, averaged over the four years each year's 85,000 selections start work from October at the earliest and stay for at least three years; across the four years from the first weighted draw in March 2026 that averages about 1.9 cohorts in the workforce at once 1.9 104 million euro-equivalent
÷ Normalised Impact scale of this evaluation 200 million euro-equivalent a point 0.52
Score 0.52 Impact × 5 Value × 3.5 Plausibility ÷ 10 = 0.9 of 100

Plausibility

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]. Firms that lose permits also hire abroad: among American multinationals, about 0.4 workers abroad for each permit lost [8], which is one more reason a freed post need not go to an American. 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: Mechanism · plausibility at most 6 Evidence of effect: Mechanistic · at most 6 Placement: Chain open · plausibility 3–3.5

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 against3 arguments

Graduates the country trained and does not keep

4.3points

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 2.4Plausibility 3
▸ Show reasoning & sources ▾ Hide reasoning & sources

Value

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. What the visa is worth to the person who holds it is not counted here; it is the same open question for both sets of 85,000 people and belongs to a separate stream this evaluation has not priced. The value sits in the middle-upper part of the scale, at the level this site uses for economic output.

Impact

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 29,000 dollars a year in output above what they consume in public services. That gives about 290 million dollars a year for each cohort, and about 550 million averaged across the four years, with 1.9 cohorts missing at once. 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], and the share of registrants with at least an American master's degree rose from 45 to 66 percent [5]. 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-equivalent (about 17,000 to 46,000 dollars) 25,000 euro-equivalent a year (about 29,000 dollars) 250 million euro-equivalent
× Cohorts in the workforce at once, averaged over the four years each year's 85,000 selections start work from October at the earliest and stay for at least three years; across the four years from the first weighted draw in March 2026 that averages about 1.9 cohorts in the workforce at once 1.9 475 million euro-equivalent
÷ Normalised Impact scale of this evaluation 200 million euro-equivalent a point 2.38
Score 2.38 Impact × 6 Value × 3 Plausibility ÷ 10 = 4.3 of 100

Plausibility

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 effect of the 100,000 dollar fee on new petitions, in force during the March 2026 registration until a court struck it down in June 2026. 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: Precedent · plausibility at most 6 Evidence of effect: Mechanistic · at most 6 Placement: Partial aspect supported · plausibility 2.5–3

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

1.2points

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 offers for the same occupation in the same place, and so sits at a lower level of the same wage table. Registrations fell by 38.5 percent in the first year.

Value 6 · OutputImpact 0.6Plausibility 3.5
▸ Show reasoning & sources ▾ Hide reasoning & sources

Value

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.

Impact

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. Compared on the same pool, as the department's own analysis does, a first-level registration's chance falls from 29.6 to 15.3 percent [6]. Only the difference counts: 14.3 percentage points, in a range from 10 to 17, across 20,000 needs is about 2,860 posts a year. Each unfilled specialist post costs a small firm something like 23,000 dollars of output in a year, giving about 66 million dollars for each year's cohort and about 130 million averaged across the 1.9 cohorts of the horizon. 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 about 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 Setting, range 10 to 17 points: on the same pool a first-level registration's chance falls from 29.6 to 15.3 percent [6] 14.3 percentage points 2,860 posts a year
× Output foregone per unfilled post Setting, range 10,000 to 40,000 euro-equivalent (about 12,000 to 46,000 dollars) 20,000 euro-equivalent a year (about 23,000 dollars) 57.2 million euro-equivalent
× Cohorts in the workforce at once, averaged over the four years each year's 85,000 selections start work from October at the earliest and stay for at least three years; across the four years from the first weighted draw in March 2026 that averages about 1.9 cohorts in the workforce at once 1.9 109 million euro-equivalent
÷ Normalised Impact scale of this evaluation 200 million euro-equivalent a point 0.55
Score 0.55 Impact × 6 Value × 3.5 Plausibility ÷ 10 = 1.2 of 100

Plausibility

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 100,000 dollar fee on new petitions, imposed by proclamation in September 2025 and in force during the March 2026 registration before a court struck it down in June 2026; 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. Larger firms that lose permits can move part of the work abroad, about 0.4 hires abroad per permit lost among American multinationals [8]; with the cap fixed that is no loss to the United States against the lottery, and small employers rarely have that option. 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: Precedent · plausibility at most 6 Evidence of effect: Mechanistic · at most 6 Placement: Chain open · plausibility 3–3.5

Open: Registration counts by employer size for the FY2028 season, once the litigation over the 100,000 dollar fee has settled, would separate the weighting from the fees and could carry P to 6.

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

0.9points

The rule ranks a post by where its offered pay sits in the wage table for that occupation and place. A school district, a rural hospital or a grant-funded laboratory usually hires at the first or second of the four levels, and at the first level such a post now has about half its former chance. What those posts contribute is not what they are paid.

Value 6 · OutputImpact 0.4Plausibility 3.5
▸ Show reasoning & sources ▾ Hide reasoning & sources

Value

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.

Impact

About 406,000 H-1B petitions were approved in 2025, most 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. The weighting compares the offered pay with the four levels published for that occupation in that place, not with what other occupations pay, so a teacher paid at the third level of the teachers' table draws three entries like anyone else; the argument holds for the posts these employers offer at the first level of their own table. On the department's own model, run on the same pool, a first-level registration's chance falls from 29.6 to 15.3 percent, while a second-level one rises slightly, from 29.6 to 30.6 percent [6]. With these employers hiring more often at the first level than others do, about a quarter of their selections is taken as lost, in a range from a tenth to about half — 1,500 a year. 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. About 35,000 dollars a year is used for that gap, in a range from 12,000 to 70,000 — the number is a judgment, not a measurement, and it is the whole content of the argument. The result is about 52 million dollars a year for each cohort, and about 100 million averaged across the 1.9 cohorts of the horizon. The Impact is a twentieth of 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
× Share of them lost under the weighting Setting, range 10 to 48 percent: on the same pool a first-level registration's chance falls from 29.6 to 15.3 percent, a second-level one rises slightly to 30.6 percent; these employers hire more often at the first level [6] 25 % 1,500 posts a year
× Gap between what the post pays and what it is worth Setting, range 10,000 to 60,000 euro-equivalent (about 12,000 to 70,000 dollars): a mathematics teacher in a district that has advertised for two years, a therapist in a county with none 30,000 euro-equivalent a year (about 35,000 dollars) 45 million euro-equivalent
× Cohorts in the workforce at once, averaged over the four years each year's 85,000 selections start work from October at the earliest and stay for at least three years; across the four years from the first weighted draw in March 2026 that averages about 1.9 cohorts in the workforce at once 1.9 85.5 million euro-equivalent
÷ Normalised Impact scale of this evaluation 200 million euro-equivalent a point 0.43
Score 0.43 Impact × 6 Value × 3.5 Plausibility ÷ 10 = 0.9 of 100

Plausibility

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: Mechanism · plausibility at most 6 Evidence of effect: Mechanistic · at most 6 Placement: Chain open · plausibility 3–3.5

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.

How robust the result is

Under the assumptions used, the proposal brings more than it costs in 98 of 100 model runs.

gains more than it costscosts more than it gains

Each dot stands for 200 of the 20,000 model runs.

Outside the calculation the result hangs on how much of the higher pay of those now selected reflects more productive work. The analysis counts half; the other half enters only as tax moved from high earners to the public budget. The cap is unchanged, so nothing here turns on how many people enter the country; every figure is the difference between two sets of 85,000 people. A proposed fee of 103,265 dollars on every cap-subject petition would reshape the pool more than the weighting does and is not part of this comparison. On these settings the rule lies ahead in 98 of 100 runs.

Show the distribution of the model runs

20,000 simulated futures

Very likely better · 98 %

In 98 of 100 simulated futures the proposal brings more than it costs.

No effect is known exactly; every argument has a range. The calculation draws a possible size for each argument from its range 20,000 times and counts how often the benefits exceed the costs.

Net effect in points per future. Left of zero the drawbacks outweigh the benefits, right of zero the benefits win. Eight in ten futures fall between +8.1 and +37 points.
Scale and assessment

Scale of this evaluation: Normalised Impact, unitless and calibrated to this topic. For comparison, one point here corresponds to about 200 million euro-equivalent a year.

Size class:medium

Effect medium (+4.2) · 81 % of the effect is positive

Areas show individual argument weights. The positive share of the effect nets out money transfers between the two sides; the probability of a positive net effect is determined separately.

How we score →

Review and sources

Last reviewed by Claude Opus 5.5 · October 29, 2026 · 8× AI, 1× human
  1. October 29, 2026AI reviewClaude Opus 5.5record updated

    Page texts for the new page layout (draft G) with quotes from primary sources; numbers unchanged.

  2. October 21, 2026AI reviewClaude Fable 5.1record updated

    Money amounts in argument texts stated in dollars per style-en.md; scores, chains and balance unchanged

  3. October 20, 2026AI reviewClaude Fable 5.1record updated

    Summary and scale note rewritten to the style of 20 Oct 2026; numbers unchanged

  4. October 16, 2026AI reviewClaude Opus 5.5record updated

    Texts updated to Bilanz 2.2

  5. October 15, 2026Assessed by a human, implemented by AIClaude Opus 5.5re-scored

    Review of 15 October, second pass with Julian's decisions: H-1B: wage gain counted half as more output, cohorts 1.9, odds compared on the same pool, sources replaced by the proclamation and the rule's own analysis; offshoring checked and noted; simulated balance 0.954, much better.

  6. October 15, 2026AI reviewClaude Opus 5.5record updated

    Review of 15 October: pro-1 no longer definitional, fee history corrected (the 215 dollar fee already applied in FY2026; the 100,000 dollar fee was a proclamation, since struck down), pro-3 now about 3,800 modelled domestic hires, ranges on all six arguments; simulated balance 0.723, better as before.

  7. September 14, 2026AI reviewClaude Fable 5.1re-scored

    Gesamtprüfung 08.09.: Gewichtung als OEWS-Stufe je Beruf/Ort korrekt beschrieben; alle sechs Ketten auf 2,5 Kohorten im Vierjahreshorizont (Bestand statt Fluss); con-1 Chancen halbiert statt Totalausfall (i 0,18 → 0,225); k 20 → 10; r 0,64 → 0,69 (besser).

  8. 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 →

Show all 9 sourcesHide sources
  1. U.S. Citizenship and Immigration Services: FY 2027 H-1B Cap Initial Registration Period Opens on March 4. uscis.gov
  2. SHRM (Society for Human Resource Management): USCIS Reaches FY2027 H-1B Visa Cap (preliminary agency figures from May 2026). shrm.org
  3. U.S. Citizenship and Immigration Services: Characteristics of H-1B Specialty Occupation Workers, Fiscal Year 2025. uscis.gov
  4. 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
  5. The White House: Restriction on Entry of Certain Nonimmigrant Workers (proclamation of 18 September 2026). whitehouse.gov
  6. Department of Homeland Security, Federal Register 2025-23853: Weighted Selection Process for Registrants and Petitioners Seeking To File Cap-Subject H-1B Petitions (final rule and regulatory impact analysis). federalregister.gov
  7. Department of Homeland Security, Federal Register 2026-17324: Fee for Certain H-1B Petitions (proposed rule). federalregister.gov
  8. Glennon, Management Science 70(2): How Do Restrictions on High-Skilled Immigration Affect Offshoring? Evidence from the H-1B Program. nber.org
  9. Forbes (Stuart Anderson): Judge Strikes Down Latest Version Of $100,000 H-1B Immigration Fee. forbes.com

More debates in this area

Very likely worse · 0 %
Size: large
The Detention Buildout

Roughly double the immigration detention estate and the removal machinery that feeds it, using the funds appropriated in 2025.
All debates in Immigration and Border Enforcement →