25 Million in Crisis, and We’re Blaming the Wrong People

As McKinsey’s Aaron De Smet and Tim Koller put it, “Misframing a problem can lead to solving the wrong one.”[1] And right now, America is asking, “How do we stop foreign workers from taking jobs?” when we should be asking, “Why are we producing millions more graduates in fields the economy doesn’t need, and not producing those the economy wants?”

Think you can frame problems effectively? Would you make better decisions than your elected officials? Let’s find out.

“Man always believes more readily that which he prefers.”

Francis Bacon, Novum Organum (1620), Book I, Aphorism 49, Joseph Devey’s edition

You’ve just been elected to the U.S. Senate. Congratulations. Now here’s your first crisis:

A constituent, a physics graduate, sharp and hard-working, can’t find a job. Their parents are furious. They spent four years and invested $150,000 in that degree. Then another email arrives. And another. Engineers, computer scientists, mathematicians, all struggling. Your inbox is flooding with stories of American graduates losing out.

You check the Federal Reserve data. In the New York Fed’s release at the time, physics graduates faced 7.8 percent unemployment, second-highest among all majors.[2]

Your staff pulls more numbers: foreign workers on H-1B and OPT visas hold roughly 1 million jobs in America.[3][4] The connection seems obvious.

What would you do?

If you’re thinking “restrict those visa programs and protect American workers,” you’re not alone. That’s precisely what was being proposed in late 2025.[5] It’s intuitive. It’s visible. It shows you’re taking action.

There’s just one problem: about 24 million college graduates in the U.S. are out of work or in jobs that don’t require their degree.[6][7] Even if you eliminated every single visa holder tomorrow and perfectly replaced each one with an American graduate (economically impossible, as we’ll see), you’d help only about 4 percent of them.

Where do the other 23 million go?

Still confident in your decision? Or are you starting to suspect there’s something deeper going on, something that easy answers won’t fix?

This isn’t a story about one senator making one mistake. It’s about how smart people, facing enormous pressure to act and limited time to analyze, end up solving the wrong problem while the real crisis goes unaddressed. It’s about the forces that lead even brilliant decision-makers astray.

And it’s about why the most challenging problems demand that we resist our instincts and ask better questions.

Why Smart People Solve Wrong Problems: The Forces That Lead Even Brilliant Leaders Astray

In November 2025, Senator Eric Schmitt pointed to physics graduates’ high unemployment rate as evidence that foreign workers are taking American jobs.[5] He’s not wrong to be concerned. American graduates genuinely are struggling. But he’s looking at the wrong cause.

The senator isn’t acting in bad faith. He graduated cum laude, earned a law degree, served as Missouri’s Attorney General, and now represents millions of constituents.[8] He’s an intelligent person facing what all leaders face: a genuinely challenging situation, public pressure to do something, a complex system most experts don’t fully understand, and limited time to become an expert himself.

That’s the human condition of leadership. It’s also precisely why we should demand better analysis.

Four Systematic Forces Push Leaders Toward Wrong Solutions

Several powerful forces push even intelligent leaders toward wrong conclusions:

• The Story-Over-Data Trap: A vivid narrative, such as “my constituent lost a job to an H-1B worker,” feels more real than a large and abstract statistical pattern you haven’t seen visualized. We evolved to learn from stories. But policy should be built on facts, analysis, and statistical patterns, not anecdotes.

• The Visibility Bias: Foreign workers are visible and easy to target. The broken university system that steers about 137,000 students a year into psychology (the BLS projects only about 12,900 psychologist openings a year, most requiring a graduate degree, and about 48 percent of recent psychology graduates work in jobs that typically do not require a degree)[9][2] is invisible and diffuse. Politicians get credit for visible action, not for fixing invisible systems.

• Ideology Over Analysis: For some politicians, opposing immigration is a core principle. For others, defending free markets is foundational. Both camps have investments that make dispassionate analysis difficult. When your worldview already suggests visa programs are bad (or good), confirming evidence attracts attention while disconfirming evidence gets dismissed.

• The Analyst Gap: In my view, most senators have limited analytical infrastructure for deep policy work. Even a good senator may have only a few staffers who genuinely know visa programs or higher-ed economics. They’re competing for attention with a hundred other issues. There is not enough capacity to conduct the rigorous root-cause analysis this kind of problem requires.

The result? Smart people point at physics unemployment without understanding what’s actually happening in physics labor markets. They propose solutions that help about 4 percent of affected graduates while potentially undermining American competitiveness in a global talent war.

This isn’t unique to this senator, his followers, or this issue. In my experience, it’s common in critical decision-making.

I’d Argue H-1B and OPT Aren’t Broken. They’re Rigorous, Adjudicated Processes With Real Oversight

Before we can evaluate whether H-1B and OPT programs are “broken,” we need to understand what they actually are. Most of the public debate I’ve followed skips this part.

H-1B and OPT are not unstructured, unregulated channels. They are adjudicated legal processes with multiple layers of government review, specific evidentiary requirements, and denial mechanisms. A government official reviews every single approval in accordance with statutory criteria. Understanding this process and its rigor is foundational to intelligent policy discussion.

H-1B Applicants Face Three Separate Gates, And Two-Thirds Are Rejected Before Review Even Begins

To understand how selective the H-1B process actually is, walk through what an employer must accomplish:

H-1B Registration for the Lottery

Before diving into the full H-1B petition process, there’s a crucial preliminary step for cap-subject petitions: online registration.[10] This electronic system, implemented by USCIS, streamlines the initial selection process.

• Employer creates a myUSCIS account (if they don’t already have one)
• During the registration period, the employer submits an electronic registration for each prospective H-1B worker
• USCIS conducts a lottery to select registrations for further consideration
• Selected registrants are notified and invited to continue the process and file a full H-1B petition.

Labor Condition Application (LCA) with the Department of Labor

Before even filing an H-1B petition with USCIS, an employer must file a Labor Condition Application (Form ETA-9035) with the Department of Labor.[11] DOL doesn’t vet the merits, but the employer must make specific legal attestations that can be enforced later, and USCIS separately decides whether the job qualifies:[11][12]

• The job qualifies as a “specialty occupation”, defined as requiring a bachelor’s degree in a specific field, or equivalent specialized knowledge.
• The employer will pay the higher of: (a) the prevailing wage for that occupation in that geographic area, or (b) the actual wage paid to similarly employed U.S. workers.
• Hiring the H-1B worker will not adversely affect the working conditions of U.S. workers similarly employed.

DOL reviews the LCA within seven working days, and only for completeness and obvious errors or inaccuracies. The attestation then becomes enforceable. If violated, the employer can face penalties, including back wages owed to the underpaid H-1B worker, civil money penalties, and debarment from the program.[11] So this gate is not a merits review. It works by binding the employer to promises the government can enforce after the fact.

Lottery Selection

Even before USCIS evaluates a single petition, most applications must survive a numerical cap:[12][13]

• Regular cap: 65,000 visas
• Advanced degree exemption (Master’s degree or higher from a U.S. institution): 20,000 additional visas
• Universities, research institutions, and specific nonprofits: exempt from cap

For fiscal year 2026, USCIS received 358,737 electronic registrations. USCIS conducts preliminary checks to remove duplicates, invalid passports or travel documents, or failed payments. Eligible registrations were 343,981 (about 339,000 unique beneficiaries), of which 120,141 were selected to apply for approximately 85,000 available cap slots.[13] USCIS selected about 35 percent of eligible registrations, and 33 percent of all registrations. USCIS selects more registrations than there are slots, likely because not every selected employer goes on to file.

That bears emphasis: two out of three would-be H-1B hires were rejected at the lottery stage, before any petition was even fully prepared. Not because of fraud or abuse, but because demand dramatically exceeds legal supply.

This gate is incredibly tight.

USCIS Adjudication of Form I-129

Only after surviving the lottery does the complete petition go to USCIS for detailed review. This is where the most rigorous gate sits.

USCIS adjudicators must evaluate two distinct legal questions, each with its own evidentiary burden:[14]

USCIS Officers Demand Detailed Evidence That Jobs Require Specialized Degrees. It’s Not Rubber-Stamping

The occupation must require a bachelor’s degree in a specific field as a minimum entry requirement. USCIS’s Adjudicator’s Field Manual, parts of which USCIS has since replaced with its Policy Manual, points officers to references like the Occupational Outlook Handbook but cautions that “job titles themselves are often meaningless.” Instead, it says, “it is necessary to consider all the facts surrounding the petition”: the worker’s education and experience, the nature of the employer’s business, industry practice, and salary.

The standard is specific: a degree in a directly related field must normally be required for the occupation, for parallel positions in the employer’s industry, or by the employer itself, or the duties must be so specialized that such a degree is normally needed.[12]

This is real scrutiny. It’s the opposite of rubber-stamping.

Does this particular individual actually possess the required qualifications?

Even if the job qualifies, does the candidate have the proper credentials? USCIS examines:

• Educational transcripts and degree certificates
• Work experience documentation
• Whether experience is directly related to the specialty occupation
• Whether the candidate’s background demonstrates the specific knowledge the job requires

An employer cannot simply claim, “We need this person,” and get approval. They must document that this specific person has the particular qualifications for this specific role.

Approval Rates Tell the Story

After all gates, what percentage of H-1B petitions actually get approved?

According to USCIS data for fiscal 2026, the roughly 85,000 cap slots equal only about 24 percent of registrations.[13] Denial rates for initial-employment H-1B petitions were 2.5 percent in FY 2024 and 2.8 percent in FY 2025.[15] That sounds low, but I think the denominator matters: by the time a petition reaches USCIS for full adjudication, many weak applications have likely been withdrawn or never filed by employers who knew they wouldn’t survive scrutiny.

The more instructive number is the Requests for Evidence (RFE) rate, the percentage of petitions USCIS asks employers to supplement or clarify. By one tracker’s count it ran near 20 percent by FY 2023.[16] USCIS’s own report puts it at 8 percent of all H-1B petitions completed in FY 2024 and 13 percent of those for initial employment, nearly 19,000 new petitions in a single year.[17] That means USCIS is actively questioning, scrutinizing, and demanding better documentation on roughly one in eight new cases.

This is active government oversight, not a rubber stamp.

OPT Requires Full-Time Enrollment, Formal School Approval, and Government Adjudication, Not a Loophole

Optional Practical Training requires students to maintain a valid F-1 status, be enrolled full-time for at least one academic year, get a formal recommendation from their school, and file detailed applications with immigration officials.[18]

For STEM OPT extensions (which allow up to 36 months instead of 12), the requirements tighten further: the job must be directly related to the STEM degree field, the employer must use E-Verify (an automated work authorization system), and the employer must provide a formal training plan detailing how the position develops the student’s knowledge in the STEM field.

Again, this is an adjudicated government process, not an honor system or a free-for-all.

Real Abuse Exists, but I’d Argue It Clusters in Body-Shop Firms, Not the Majority of H-1B Use

This is important: acknowledging that H-1B and OPT are rigorous processes doesn’t mean there’s no abuse. In my view, it means abuse is concentrated in a subset of employers rather than spread across the whole program.

Documented abuses cluster in specific segments of the economy:

• Outsourcing and Body-Shopping Firms: Some labor contractors use H-1B explicitly as a cost-arbitrage strategy. They hire foreign workers on H-1B, place them as contractors at client companies, pay them less than the law requires, and profit as middlemen while undercutting the wages U.S. engineers would command. This is wage theft disguised as visa sponsorship. In one case, Ron Hira and Daniel Costa of the Economic Policy Institute estimated that a single outsourcing firm, HCL, appears to have underpaid its H-1B workers by at least $95 million a year. It’s real, it’s documented, and it deserves enforcement.[19]

• Wage Violations: Some employers attest that they will pay the required wage but actually pay their H-1B workers substantially less than U.S. workers in the same roles, creating a de facto two-tier wage system. Again, real and documented.[19]

• Misclassification: I suspect some employers also claim jobs are specialty occupations when they’re not, or claim qualifications the candidate doesn’t possess, betting that adjudicators won’t catch the gap or USCIS will lack resources to vet thoroughly.

• The Economic Policy Institute, a labor-focused research organization, calls the abuse in the HCL case “the proverbial tip of the iceberg” that “points to widespread, systemic H-1B abuse,” and says underpaying H-1B workers is likely an industrywide practice among outsourcing firms. The evidence is compelling enough to justify enforcement and tighter oversight.[19]

But here’s the critical nuance: In my view, most H-1B approvals are not in these abuse-prone sectors. The Indian-based outsourcing firms critics focus on are a shrinking slice. In FY 2025, the top seven Indian-based companies had 4,573 new H-1B petitions approved, about 4 percent of the 114,806 approvals for initial employment and a 70 percent drop from FY 2015.[15] Universities also sponsor H-1B workers, and in my view their institutional oversight makes systematic wage violations less likely.

So, as I read the evidence, we have a situation where:

• The dominant use of H-1B aligns with statutory intent (attracting or retaining specialized talent in genuine specialty occupations).
• A meaningful subset of uses, primarily in certain outsourcing and staffing models, reflects concentrated, repeated misuse.
• The adjudication process itself is rigorous, but enforcement and penalty mechanisms could be tighter.

What we have is “some abuse occurs in certain sectors and deserves targeted enforcement,” not “the program is fundamentally broken.”

It’s also not the same as “foreign workers are taking American jobs from the general college graduate population.”

The Senator’s Dilemma, And Why Our Leadership Needs Better Support

Let’s return to that senator and have genuine empathy for the position he’s in.

He represents millions of constituents. Some have lost jobs or struggled with underemployment. Some are recent graduates with expensive degrees and disappointing employment prospects. Some are parents terrified that their children’s college investment won’t pay off. These are real people with real problems, and a good senator feels the weight of that.

He also knows that his job depends partly on being seen as doing something. The public wants action on problems they perceive as urgent. If a senator says, “This is actually a complex systemic issue in higher education that will take 10-15 years to reform,” voters think he’s doing nothing. If he proposes restricting H-1B visas, he’s shown leadership, even if the policy doesn’t solve the underlying problem.

He’s also constrained by what economists and policy experts tell him. If the dominant narrative among economists and immigration analysts is “H-1B and OPT are terrible for American workers,” he feels justified in acting on that narrative. If he’s not an expert in visa programs himself (and few senators are), he’s dependent on advisors, hearing testimony, and prevailing expert consensus.

Here’s where the system breaks down: the expert consensus on H-1B and OPT is genuinely contested, but the political framing treats it as settled.

Some economists, such as John Bound, Gaurav Khanna and Nicolas Morales, find that H-1B visas held down wages and employment for U.S. computer scientists, even as they raised Americans’ overall welfare. Others, including Anna Maria Mayda, Giovanni Peri and their co-authors, find that when the H-1B cap tightened, employment of similar American workers at for-profit firms did not change, which suggests H-1B workers rarely replace Americans. Some companies may also use H-1B to avoid training and developing American talent. Others, and I’m one of them, think American universities aren’t producing enough specialists in certain domains, which can make H-1B a necessity rather than a choice.[20]

A responsible senator faced with contested expert opinion should do several things:

First: Commission deeper analysis. Before proposing a significant policy change, have your staff (or the Congressional Research Service) produce a genuinely comprehensive assessment. What’s the actual scale of harm from H-1B/OPT? Where does it concentrate? What are the counterfactuals? What happens if we eliminate these programs?

Second: Demand root-cause analysis, not symptom-level thinking. Don’t ask, “Are foreign workers hurting American graduates?” Ask “what’s causing 24 million college graduates to be unemployed or underemployed, and what role do foreign visa holders play in that?” The first question leads to visa restrictions. The second might lead to very different conclusions.

Third: Consider unintended consequences. America does not have a monopoly on talent, and in my view it does not produce enough advanced skills to compete. We need to weigh the benefits and costs of every policy. If we eliminate H-1B and OPT:

• What happens to American universities’ ability to attract international students and faculty? (Many come expecting potential pathways to work authorization after graduation.)
• What happens to American companies’ competitiveness in global talent markets?
• What happens to specific sectors (like biotech, quantum computing, and advanced manufacturing) that depend on accessing specialized talent?
• Do the economic benefits to America outweigh the costs, or the opposite?
• Are there better ways to address abuse than abolishing the entire mechanism?

Fourth: Push back against ideological constraints. For senators predisposed to oppose immigration, resisting H-1B restrictions is uncomfortable. For senators predisposed to support immigration, acknowledging program abuses is uncomfortable. But serious solutions require confronting discomfort. It requires being willing to say things your political base doesn’t want to hear.

The senator we’re discussing would be doing the American public a service if he said something like this:

“I’m genuinely concerned about American college graduates struggling to find work. But the evidence suggests the problem is primarily structural. We’re producing far more graduates in low-demand fields than the labor market needs, while simultaneously failing to steer enough students into high-demand fields. Foreign workers on H-1B and OPT account for roughly 1 million jobs in an economy of 160 million workers, and while economists still disagree, some of the evidence suggests they’re concentrated in roles that complement rather than directly replace American workers. There are abuses in the visa programs, particularly in certain outsourcing sectors, and those need enforcement. But abolishing the programs wholesale would hurt America’s long-term competitiveness without solving the real problem: a higher-education system with broken incentives. Let me propose a comprehensive approach that addresses both visa program abuses and the structural education and employment issues affecting 24 million college graduates in the U.S.”

That’s harder politically. It requires resisting easy answers. It means defending the value of necessary foreign workers while still taking American concerns seriously. It requires admitting the problem is partly domestic and that legislation alone won’t fix it without long-term structural reform.

But that’s what serious policy leadership looks like.

Visa Workers Account for 1 Million Jobs, The Crisis Affects 24 Million Graduates

Before any other analysis, we need to establish a baseline scale. Because scale determines everything about how seriously we should take a policy lever.

About 24 Million College Graduates Face Unemployment or Underemployment

According to the Bureau of Labor Statistics Current Population Survey for September 2025, the United States has 163.65 million employed workers.[21] Of these, 65.59 million are college graduates age 25 and older (bachelor’s degree or higher). Another 1.92 million college graduates are unemployed, for a college-educated labor force of 67.51 million.[6]

Of that 67.51 million:

• 1.92 million are actively unemployed and searching for work.[6]
• About 22.1 million are underemployed, working in jobs that don’t require a college degree (33.7 percent of the 65.59 million who are employed, using the New York Fed’s underemployment rate for all college graduates).[7]

Combined, about 24.0 million college graduates are facing employment challenges. The underemployment share is not new. The New York Fed’s underemployment rate for all college graduates was about 34 percent in 1990 as well, so this is a long-running, structural problem, not a sudden shock.[7]

Think about that number. About 24 million people invested years and often tens of thousands of dollars in degrees, and today they are without work or in jobs that don’t need them. That’s the population affected that we need to help.

Even Eliminating Every Visa Worker Would Address Only 4% of Affected Graduates

H-1B and OPT programs together account for approximately 1.024 million foreign workers in the U.S. labor market:[3][4]

• H-1B: ~730,000 workers[3]
• OPT: ~294,000 workers[4]

Now do the math:

1.02 million foreign workers: 24.0 million struggling graduates = about 1 to 23.5

Even if you could eliminate every single visa position tomorrow and perfectly replace each one with an American graduate (economically impossible, as we’ll see), you’d address only about 4.3 percent of affected graduates.

Where do the other 23 million go?

This is not rhetorical. This is the fundamental policy question Washington should be pursuing.

Even Narrowing to STEM Fields, Visa Workers Hold About 4% of College-Educated STEM Jobs, About 10% in Science and Engineering

The strongest counterargument to the scale analysis is: “H-1B and OPT aren’t evenly distributed. They concentrate in STEM fields where the domestic supply is tightest. Maybe the impact is bigger there?”

It’s a fair point. Let’s test it.

STEM Workforce: According to NSF data, approximately 17.55 million people in the U.S. with bachelor’s degrees or higher work in STEM fields. Against the 163.65 million employed in September 2025, that’s roughly 11 percent (the NSF count is from 2021).[22] That count includes health jobs such as nursing and medicine, which the visa counts below leave out. In science and engineering jobs alone, NSF counts about 7.3 million workers with a bachelor’s degree or higher.[22]

STEM Unemployment/Underemployment: In the New York Fed’s release at the time, unemployment for recent STEM graduates (ages 22-27) varied dramatically:[2]

• Physics: 7.8% unemployment
• Computer Science: 6.1% unemployment
• Mathematics: 3.7% unemployment
• Civil Engineering: 1.0% unemployment

By my estimate, weighting the New York Fed’s rates by the number of degrees in each STEM major, recent graduate unemployment across STEM fields is roughly 3.8 percent, about the same as the 3.6 percent for all recent college graduates in the same New York Fed table. Where STEM graduates clearly do better, on the same weighting, is underemployment, working in roles that don’t require a degree (approximately 28.9% for STEM fields versus 38.3% for all recent graduates in that table).[23]

It’s tempting to apply these rates to the whole STEM workforce and count the affected. That doesn’t work. The 17.55 million are people of every age who are already employed in STEM, so unemployment and underemployment rates measured on 22-to-27-year-old graduates can’t be stretched across them. The comparison we can defend is simpler: how big is the visa workforce next to the college-educated STEM workforce it supposedly displaces?

STEM Visa Concentration: H-1B and OPT workers are heavily concentrated in STEM:

• H-1B in STEM fields: ~575,000 (~79% of H-1B petitions approved in FY 2024, applied to the 730,000 estimate)[24]
• OPT in STEM fields: at least ~144,000 (49% of those authorized for OPT in 2024 majored in computer science or engineering, applied to the 294,000 OPT total)[25]
• Combined in STEM: at least ~719,000

Now the STEM-only comparison:

At least 719,000 visa workers in STEM to 17.55 million college-educated STEM workers = about 1 to 24

Even narrowing to STEM, visa workers hold about 4 percent of the jobs held by college-educated STEM workers, and about 10 percent of college-educated science and engineering jobs, where they concentrate. That doesn’t prove they harm no one. It does set the scale: even if every one of those jobs went to an American graduate, it would open about 4 percent of the college-educated STEM jobs in the country, or about 1 in 10 science and engineering jobs.

If OPT workers in STEM fields are keeping Americans from STEM jobs, then eliminating even the 144,000 in computer science and engineering frees up under 1 percent of college-educated STEM jobs, or about 2 percent of science and engineering jobs! Why does anyone think this is the source of our problem?

But here’s where it gets fascinating.

Even when we examine categories where H-1B and OPT could be crowding out domestic workers, the New York Fed’s release at the time didn’t show the unemployment spike that argument would predict.

The CS Paradox: 3.55x More Visa Workers Than Affected Graduates, Yet Unemployment Was 6.1%, Not Far Higher

Zoom in further to just computer science and IT, the epicenter of the visa debate:

CS Recent Graduates Affected:

• Unemployed (at 6.1% unemployment rate): ~42,000[2]
• Underemployed (at 16.5% underemployment rate): ~115,000[2]
• Total: ~157,000

These counts are my estimate. They apply the New York Fed rates to roughly 700,000 recent CS graduates. NCES counts about 550,000 CS bachelor’s degrees across the six classes from 2016-17 to 2021-22,[22] so this base, and the affected count, may run high. If the base is smaller, the ratio below only grows.

Visa Workers in IT/CS:

• H-1B in computer occupations: ~467,000 (63.9% of H-1B petitions approved in FY 2024, applied to the 730,000 estimate)[17]
• OPT in computer science: ~91,000 (31% of those authorized for OPT in 2024 majored in computer science, applied to the 294,000 OPT total)[26]
• Combined in IT/CS: ~558,000

There are 3.55 times more visa workers in IT/CS than there are affected recent CS graduates combined. The two counts are different groups. The visa count covers workers of every age and experience level, while the graduate count covers only graduates ages 22 to 27, so read the ratio as a sense of scale, not a head-to-head match.

This creates a genuine paradox: if visa workers so dramatically outnumber the affected CS graduate population and are indeed crowding out Americans, why isn’t CS unemployment far higher? At 6.1 percent it sits above the STEM average, but in the New York Fed’s release at the time it was not even the peak among STEM majors. Physics and computer engineering were higher. (Update, September 2026: the New York Fed’s current release shows computer science at 7.0 percent, physics at 6.6 percent, with computer engineering at 7.8 percent. Computer science is now above physics, which reverses part of this contrast.)[2]

I think the answer to this paradox reveals what’s actually happening. And visa competition, I’d argue, is only a small part of it.

Domestic Oversupply and AI Automation, More Than Visa Competition, Likely Explain CS Graduate Unemployment

The computer science paradox, that visa workers vastly outnumber affected graduates, yet unemployment is elevated rather than extreme, points, in my view, to structural problems that visa policy alone can’t fix.

CS Degree Production Grew 138% Since 2001, While the Regular H-1B Cap Has Held at 65,000 Since 2004

Computer science degree production has exploded over the past 20 years. In 2001-02, the USA conferred 68,290 Bachelor’s, Master’s, and Doctor’s degrees. By 2021-22 that number grew to 162,631.[22]

Between 2001 and 2021, CS degree production increased by 138 percent, to about 2.4 times its 2001 level. Part of that growth is in graduate degrees, where international students make up 71 percent of full-time graduate students in computer and information sciences, according to the National Foundation for American Policy.[15] That is why I lean on the bachelor’s numbers below.

Universities, seeing student demand and tuition revenue, expanded CS programs dramatically. Schools from elite research universities to regional colleges now offer computer science degrees, and I doubt all of them have the faculty expertise, industry connections, or placement success to justify the investment.

Berkeley computer science professor Hany Farid describes the change from where he sits. “Our students typically had five internship offers throughout their first four years of college. They would graduate with exceedingly high salaries, multiple offers. They had the run of the place. That is not happening today. They’re happy to get one job offer.”[27] Farid doesn’t pin it on one cause. He calls it “a confluence of many things,” with AI among them.

This shift from scarcity to surplus came while the regular H-1B cap held at 65,000. OPT did grow, roughly doubling from about 147,000 students in 2015-16 to about 294,000 in 2024-25,[4] so visa policy is part of the picture. Those OPT counts cover every field of study, and in 2024 about 31 percent of those authorized for OPT majored in computer science.[26] But the bigger force, as I read these numbers, is degree production at American colleges, because each year’s output adds a whole new class to the job market. Bachelor’s degrees in CS alone more than doubled between 2001-02 and 2021-22, from about 50,400 to about 108,500, so each year’s class is now about 58,000 graduates larger.[22] In my view, that output outran the entry-level jobs available to absorb it.

AI Tools Are Squeezing the Entry-Level Developer Roles New Graduates Traditionally Filled

I’d say the jobs entry-level CS graduates traditionally filled are among the roles most exposed to AI automation.

GitHub Copilot writes basic code. ChatGPT debugs programs. Claude handles API documentation. A company that once needed 10 junior developers to build a feature might now get by with two mid-level developers armed with AI tools.

Bureau of Labor Statistics projections show 10.1 percent growth in computer and mathematical occupations through 2034, and BLS points to demand for roles like data scientists and information security analysts.[28] (Update, September 2026: the 2025-35 projections show 7.3 percent growth through 2035.[28]) The projections don’t say how much of that growth will be entry-level.

Andrew Ng, the Stanford professor who founded Google Brain, recently broke down what he’s observing in the talent markets. And his insights reveal a crucial aspect of how AI is reshaping hiring.[29]

Ng describes a clear hierarchy of engineering talent in the AI era:

The Bottom Tier: Computer science graduates who never learned AI. These are people entering the job market without AI skills, and Ng suggests some lack even basics like cloud computing.

Ng is blunt: this is “the tier that is in trouble.” He says university curricula haven’t kept pace, and he asks listeners to imagine a CS graduate who has never heard of cloud computing, essentially a graduate without the foundational tools of modern work.

This happened while the regular H-1B cap held steady. The bigger story looks less like entry-level jobs going to foreign workers and more like entry-level jobs thinning out.

First-Time H-1B Computer Workers, Including Master’s and PhD Holders, Earn a $101K Median, About 32% More Than the $76K Projected Average for New Computer Sciences Grads

Here’s a critical point that gets lost in political rhetoric: most H-1B visa workers don’t compete for entry-level positions.

The median H-1B salary in computer occupations is approximately $125,000 across all approved petitions, most of them for workers already in the program. For first-time H-1B petitions (initial employment), many of them for recent graduates moving from student visas, the median is about $101,000.[30]

The BLS puts the 2025 median wage of all US workers in computer and mathematical occupations at $109,280.[31] Median H-1B is about 14 percent higher.

The National Association of Colleges and Employers (NACE) projects an average starting salary of $76,251 for Class of 2025 computer sciences majors, a group that also includes information sciences and software applications majors.[32] The first-time H-1B median is about 32 percent higher, though it includes workers with master’s and doctoral degrees, and the all-petition median is about 64 percent higher. That’s not a slight difference.

Why? Many H-1B workers fill specific, specialized roles, project-based work, and mid- to senior-level positions. In my experience, many of these roles require several years of experience, which recent graduates don’t have.

In my view, outside the body shops described above, the economics point the same way: most companies don’t sponsor expensive H-1B visas, with legal costs, paperwork, uncertainty, and timing delays, if they could hire fresh American graduates more cheaply and easily. For them, the economics only make sense if the visa worker fills a genuinely different market need.

Those companies sponsor H-1B visas when they genuinely can’t find domestic workers with specific expertise, when they’re retaining someone who proved valuable during OPT, when they need specialists in emerging technologies, or when they’re competing internationally for talent.

This market segmentation is crucial because it means most H-1B workers are not directly competing for the entry-level positions where recent graduates face unemployment.

The Senator’s Key Evidence Rests on a Few Hundred Survey Responses

Remember that senator’s physics statistic? This example shows how one vivid number can stand in for analysis in policy thinking.

The New York Fed figure, which at the time showed physics graduates with 7.8 percent unemployment, the second-highest of all majors, comes from the American Community Survey (ACS). The ACS is massive and valuable, but the public data files the New York Fed uses are a 1 percent sample of the U.S. population.[33]

For large populations, 1 percent sampling provides excellent statistical reliability. For tiny subgroups like physics majors, the estimates get much less precise.

Physics Degrees Are 0.43% of All Bachelors, Too Small for Precise ACS Estimates

Physics awards approximately 8,600 bachelor’s degrees a year (8,618 in the class of 2022, according to the American Institute of Physics).[34] Out of about 2 million bachelor’s degrees awarded in the U.S., that’s just 0.43 percent.[35]

The New York Fed’s rate comes from a single year of ACS data covering graduates ages 22 to 27, roughly six graduating classes. At a 1 percent sampling rate, that means several hundred physics graduates in the survey.[33][35]

That is a real sample, but a small one for a number that moves the national debate. With a few hundred people, a measured 7.8 percent could plausibly sit two or three points higher or lower.

That’s not a canyon. It’s not a rounding error either.

Year-to-year fluctuations of 1–2 percentage points can come from sampling variance alone, so on their own they tell us little about actual labor market changes.

A senator making major policy based on this statistic is making policy on a number that is less precise than it looks, and, as the AIP data below shows, describes a long-standing feature of physics rather than a new crisis.

Higher-Quality AIP Data Shows Physics Unemployment Rises and Falls With the Economy

The American Institute of Physics (AIP) surveys new physics bachelors directly, with far larger samples: 3,939 responses across the classes of 2021 and 2022, about 2,000 a year. On their chart of the classes of 1996 through 2022, the share of new physics bachelors who are unemployed a year after graduating ranges from about 2 to 8 percent, and about 4 to 6 percent in most years.[36] That share counts everyone, including the nearly half who go on to graduate school. Among those actually in the job market, the latest figure works out to about 7 percent, close to the New York Fed’s number. So the New York Fed isn’t wrong. What AIP adds is the long view.

Most critically, the peaks line up with recessions: about 7 percent after the post-9/11 downturn, 7 percent in 2008, and 8 percent in the pandemic year of 2020. In between, the share settles back to about 4 to 6 percent.

If H-1B and OPT were primary drivers of physics unemployment, the peaks should follow visa expansions. Two of them line up: the temporary increase of the H-1B cap to 195,000, three times the 65,000 base, in fiscal 2001 through 2003,[3] and the new STEM extension of OPT in 2008.[26] Both overlapped recessions, so the chart can’t separate cause from timing. The telling stretch comes later. OPT grew by about half between 2015-16 and 2019-20, from about 147,000 to about 224,000,[4] and outside the pandemic year the unemployed share stayed around 4 to 5 percent.

That pattern looks less like foreign worker competition than like the business cycle plus features specific to physics labor markets.

Physics Unemployment Reflects Field-Specific Dynamics, Nearly Half Pursue Grad School, Many Choose Non-Physics Careers

Physics has unique labor market characteristics that may help explain persistently elevated unemployment:

Graduate School Self-Selection: Nearly half (46 percent) of physics bachelors go straight to graduate study.[36] My guess is that this draws many of the strongest students into graduate programs, which can leave a less competitive pool in the immediate job market. And because physics PhD programs take years, the timing can create cycles of gluts and shortages.

Field Leakage: Physics bachelor’s degree holders who go on to employment rather than continue to graduate school are not predominantly entering careers in physics. They choose a variety of careers, including engineering, software, other STEM fields, teaching, and non-STEM work.[36]

The New York Fed counts a graduate as “underemployed” only when the job typically doesn’t require a degree, so a physics graduate in software or engineering isn’t counted that way. Many of those moves are deliberate career choices that use physics training, not labor market failures.

Small Absolute Numbers Mean Minimal Policy Impact: Even if you eliminated all physics unemployment through visa restrictions (impossible given structural factors), you’d affect roughly 350 people a year, about 2,100 across the roughly six graduating classes the New York Fed counts, still less than 0.01 percent of the 24 million college graduates affected nationwide.

Using physics unemployment to justify broad visa restrictions is policy analysis by anecdote. It’s like diagnosing a heart attack based on a hangnail. The symptom is visible, but it’s fundamentally unrelated to the actual patient’s condition.

And it exemplifies a broader problem: in my experience, the loudest voices in policy rarely have the time to dig down to root causes, and when we rush to adopt solutions that address symptoms, we end up breaking other, far more important goals.

China Graduated 4.4 Times as Many STEM Students in 2020, and Visa Programs Help America Keep Its Edge

Here’s what I think the visa debate gives too little weight to: the strategic context in which these programs operate.

The U.S. is in an intensifying, long-term competition with China and other nations for advanced talent in critical fields.

China Produced 3.6 Million STEM Grads in 2020. America Produced 820,000

In 2020, China graduated approximately 3.6 million STEM students. India graduated approximately 2.6 million. The United States graduated approximately 820,000.[37]

That year, China produced about 4.4 times as many STEM graduates as America.

Quality matters. American universities remain globally prestigious. I’d argue our research infrastructure is still unmatched. And by my arithmetic, dividing those counts by each country’s population, America turns out STEM graduates at about the same rate as China, roughly 2.5 per 1,000 people. But when you’re competing in quantum computing, artificial intelligence, hypersonics, biotechnology, and other domains where national security depends on technical leadership, sheer numbers start to matter enormously.

America’s Competitive Edge Is Attracting and Retaining World-Class Talent

In my view, the biggest advantage America has kept in this competition is our ability to attract and retain world-class scientists, engineers, and entrepreneurs from around the globe.[38] Please read: America’s Talent Crisis: The War America Is Losing (And Doesn’t Even Know It’s Fighting)

Many of the best students from China, India, Europe, Latin America, and Africa want to study in American universities, and many of them want to stay and contribute to American innovation. H-1B and OPT programs are mechanisms for converting that educational advantage into a sustained competitive advantage by keeping talented people here after we’ve trained them.[38]

When we treat H-1B and OPT primarily as political punching bags rather than strategic talent-retention tools, we risk undermining one of the few clear advantages America still holds over competitors with vastly larger populations.

Restricting Visas Sends American-Trained Talent Back to China, India, and Europe, With Their Skills

When we make it harder for international students to work in America after graduation or create a hostile atmosphere suggesting they’re unwelcome, many return home. And they take their American training with them.

That physics PhD who studied quantum computing at MIT? If they can’t work in America, they work in Shanghai, advancing Chinese quantum capabilities.

That computer science graduate who trained in AI/ML at Carnegie Mellon? If they face hostile rhetoric about “stealing American jobs,” they return to Seoul or Singapore, building their countries’ AI sectors.

I think the notion that restricting these visas “protects American jobs” misses how much these programs help America keep its technological lead against far larger competitors determined to overtake us.[38]

It’s the policy equivalent of poisoning our own well to keep others from drinking from it.

America Faces Critical STEM Shortages While 24 Million Graduates Struggle, Partly Because We Train for the Wrong Fields

Here’s the brutal irony that exposes the incomplete framing of the visa debate:

The National Science Board of NSF says “[t]he United States is facing an accelerating science, technology, engineering, and math (STEM) talent crisis that increasingly puts our economic and national security at risk.”[39]

While we have 24 million college graduates struggling with unemployment or underemployment, we simultaneously face critical skills shortages in areas essential to national security and economic competitiveness.

These aren’t jobs that visa workers are stealing. In my view, many of these jobs are hard to fill because we don’t have enough Americans with the proper training, and we’re not training enough Americans for these fields.

Why?

Because of the exact structural dysfunction we’ve been discussing. We are producing significant numbers of degrees in low-demand skills.

The result is a terrible misallocation of human capital:

• Millions of college-educated Americans are struggling to find appropriate employment
• Critical jobs are hard to fill
• America risks losing competitive ground in fields essential to national security and economic leadership

This is not a problem that visa restrictions solve. This requires fixing how we guide students into fields of study.

And yet, political energy goes into restricting the number of needed foreign workers rather than reforming higher education incentives.

Restricting Visas Sacrifices Talent Retention and Competitive Advantage, For 4% of the Problem

When we look at the whole landscape, the structural dysfunction in higher education behind much of the 24 million underemployed or unemployed graduates, the critical unfilled positions in national security-critical fields, and the competition for talent from China and India, restricting H-1B and OPT visas represents a specific strategic choice with foreseeable consequences.

Eliminating Visa Programs Would Leave Critical Jobs Unfilled and American Innovation Offshored

Domestic Consequences:

• Fewer international students come to American universities (that’s a genuine advantage we lose: their full-cost fees or exceptional talent).
• Many of the top ones, after graduating with American training, return home rather than face visa restrictions.
• This reduces the efficiency of our educational investment. We’ve trained them, but don’t retain their talent.
• Over time, international enrollment declines as students realize they can’t stay post-graduation. American universities lose revenue and prestige.

Competitive Consequences:

• American companies lose flexibility to bring in or retain specialized talent for a competitive advantage.
• Critical positions in AI, quantum, advanced manufacturing, and cybersecurity remain unfilled.
• American innovation slows in fields where we compete directly with China and other nations.
• American companies increasingly offshore their R&D and strategic functions because they can’t access talent domestically.

Employment Consequences for American Graduates:

• The 24 million underemployed/unemployed college graduates are not hired en masse to fill the cybersecurity, AI/ML, and advanced manufacturing roles.
• Why? Because they don’t have the proper training, and restricting visa workers doesn’t create that training retroactively.
• The jobs remain unfilled or go offshore.
• The college employment crisis continues unchecked because the actual problem, structural dysfunction in higher education, remains unfixed.

55% of Billion-Dollar Startups and 65% of Top AI Companies Have Immigrant Founders

Here are some facts worth considering, from research by the National Foundation for American Policy (NFAP):

• “55% of America’s startup companies valued at $1 billion or more have at least one immigrant founder.”[15]
• Immigrants “have founded or cofounded nearly two-thirds (65% or 28 of 43) of the top AI companies in the United States.”[15]
• “70% of full-time graduate students in fields related to artificial intelligence are international students.”[15]
• In my view, if we get rid of these visas, we run the risk of missing the next Elon Musk, Satya Nadella, or Sundar Pichai. There are many more…

By focusing political capital on visa restrictions rather than structural higher-ed reform, we sacrifice:

1. Talent retention: The ability to keep exceptional international talent in America after we’ve educated them.
2. Competitive advantage in critical fields: The flexibility to bring in or retain specialists when domestic supply is insufficient.
3. University prestige and revenue: International students and faculty are crucial to top-tier institutions’ research and educational missions.
4. Innovation in emerging fields: Where first-mover advantage matters (AI, quantum, biotech), restricted talent access directly impacts innovation speed and quality.

All in exchange for addressing approximately 4 percent of the college employment crisis.

Solving the Crisis Requires Visa Program Fixes, Not Elimination, Plus Comprehensive Higher-Ed Reform

If the goal is genuinely to improve outcomes for American college graduates while maintaining America’s competitive position in global talent competition, the policy approach should be:

Target Documented Abuse Through Wage-Based Selection and Enforcement, Keep Talent Retention Mechanisms

Target documented abuses without eliminating talent retention mechanisms: shift H-1B allocation from lottery to wage-based selection, prioritizing higher salaries. (Update, September 2026: in December 2025 DHS adopted a weighted selection rule that favors registrations at higher wage levels when a lottery is needed.[40] A presidential proclamation that took effect on September 21, 2025 also requires a $100,000 payment with certain H-1B petitions, and a September 2026 proclamation extended it to September 21, 2027.[41]) Strengthen prevailing wage enforcement with meaningful penalties. Increase Department of Labor auditing resources. Restrict body-shop and staffing-firm usage operating primarily for cost arbitrage. And fast-track advanced degree holders from U.S. institutions in critical shortage fields.

For OPT: maintain STEM extensions with tighter direct-employment oversight, require training plans demonstrating actual skill development, tighten restrictions on staffing firms, and expand for critical shortage fields.

These reforms address real abuse without eliminating talent retention.

Require Outcome Transparency, Gainful Employment Standards, and Strategic Workforce Development

Remember psychology? For 2024 to 2034, the BLS projected about 12,900 psychologist openings a year, most requiring a graduate degree, while we produce about 137,000 bachelor’s degrees in psychology a year, plus about 38,000 graduate degrees (2020-21 count).[9][42] Even the graduate degrees alone outnumber those openings about three to one. In the New York Fed’s current release (updated February 2026), about 48 percent of recent psychology graduates work in jobs that don’t require a degree.[2] They will have given up four or more years of earnings, paid out-of-pocket costs to attend college, and graduate with a degree that is not necessary for the job they end up filling. In my experience, that job can start at a lower salary than a peer who went to work straight out of high school.

This crisis is not isolated. I believe it’s the challenge causing the lion’s share of our problem.

Fixing this is harder politically, but infinitely more important:

Transparency on Outcomes: Mandate that universities publish detailed employment-outcome data by major, including underemployment rates, earnings distributions, debt-service ratios, and net present value calculations. Make poor outcomes visible and create reputational pressure.

Gainful Employment Standards: Require programs to demonstrate that graduates earn enough to repay student debt in a reasonable time. Programs consistently failing this test lose access to federal student aid, creating institutional incentive to (a) control costs, (b) improve placement, and (c) limit enrollment in oversupplied fields.

Strategic Workforce Development: Subsidize education in critical shortage fields, cybersecurity, AI/ML, advanced manufacturing, biotech, quantum, and semiconductor engineering. Provide free or heavily subsidized tuition, accelerated loan forgiveness, and employer tax credits for hiring graduates in these fields.

No Major Policy Vote Without Comprehensive Analysis of Scale, Root Causes, and Unintended Consequences

Here’s where I’m most sympathetic but also most demanding of our senators:

Require comprehensive analysis before major policy votes: Before supporting visa restrictions, demand staff produce a detailed analysis of the actual impact scale (not anecdotes), a root-cause analysis of the problem being solved, an analysis of unintended consequences, a comparison of different policy levers and their impacts, and evidence for proposed solution effectiveness.

Create incentive for honest discussion: Build political space where senators can acknowledge complexity without losing credibility. “This is more complicated than initially thought” shouldn’t be a political weakness.

Demand iterative evaluation: Major policy changes should include mechanisms for evaluation. If we restrict H-1B visas and graduate unemployment doesn’t improve, we should reverse course without losing face.

Restricting Visas Won’t Solve the College Employment Crisis. It Requires Harder Work: Higher-Ed Reform

Restricting H-1B and OPT visas will not solve the college employment crisis.

If we restrict them to zero, the 24 million unemployed or underemployed graduates won’t automatically get hired for positions currently filled by visa workers. Why? Most H-1B workers and recent graduates occupy different market segments. H-1B workers mostly do specialized technical roles. Recent graduates seek entry-level work. OPT workers, who are recent graduates themselves, and first-time H-1B workers coming off student visas, are the exception.

Critical positions in emerging fields will remain unfilled or go offshore. The real problem, structural dysfunction producing graduates in oversupplied fields, remains unchanged.

Visa reforms can address abuses and improve program integrity. But they’re not solutions to the main crisis. That requires harder work: higher-education reform, better workforce planning, and skills-based assessment.

Smart Leaders Deserve Better Analytical Support, And Political Space to Change Their Minds

That senator I mentioned at the beginning? He’s not a villain. He’s an intelligent person trying to address a real problem: millions of Americans struggling with college employment outcomes. His instinct to “do something” is correct.

His error wasn’t in caring. It was in analyzing too quickly, letting visible politics override rigorous root-cause thinking, and focusing on a symptom (visa workers) that explains only a tiny fraction of the actual problem.

I think this is an American tendency: we’re drawn to visible, concrete scapegoats. They’re easier to address than diffuse structural dysfunction. They generate political credit for “taking action.”

But when we make major policy decisions based on incomplete analysis, we end up solving the wrong problems while breaking other vital things.

The college employment crisis is real and urgent. So is America’s need to maintain a competitive advantage in talent competition against rivals with far larger populations. These aren’t necessarily in conflict, but they are if we waste political capital and policy creativity on the wrong levers.

Our leaders deserve better analytical support. They should have research staff and expert advisors saying: “Yes, this is a real problem. No, visa programs aren’t the root cause. Here’s what the data actually shows. Here’s what would actually solve this. Here are the unintended consequences if we go down that path instead.”

And they should have cover to say publicly: “I care deeply about this problem, but the solution isn’t what I initially thought. Here’s what the data shows we should do.”

That’s not a weakness. That’s integrity. It’s also what good governance requires.

The question is whether our political system, with its pressure for visible action, ideological constraints, and limited analytical infrastructure, can support that kind of thinking.

The stakes are high enough that we have to try.

This Isn’t Just Policy. Every Organization Faces the Same Wrong-Problem Risk Without Dissent Culture

This essay has been mostly about policy and politicians. But if you think your CEO, VP, or manager is exempt from wrong framing and going down rabbit holes, you are likely ignoring reality.

At McKinsey and Company, we had an obligation to dissent. That means that everyone had to assist in redirecting the conversation to the facts and what was important. That simple cultural expectation reduced framing error. Still, it occurred frequently.

When leadership is hierarchical, without room for diverse points of view and opportunities for dissent, you will find significant opportunities for misframing and the resultant misallocation of focus and resources!

Reference Sources

1. De Smet, Aaron, and Tim Koller. “Bias Busters: When the question—not the answer—is the mistake.” McKinsey Quarterly, McKinsey & Company, August 22, 2025. https://www.mckinsey.com/capabilities/strategy-and-corporate-finance/our-insights/bias-busters-when-the-question-not-the-answer-is-the-mistake
2. Federal Reserve Bank of New York. “The Labor Market for Recent College Graduates.” Interactive data tool showing unemployment rates by major. Accessed November 2025. The release at the time was updated February 20, 2025, with 2023 data. The current release, updated February 4, 2026, is the source of the psychology figure and the September 2026 update. https://www.newyorkfed.org/research/college-labor-market
3. FWD.us. “H-1B Visa Program.” Policy brief on legal immigration and workforce needs. Accessed November 2025. https://www.fwd.us/news/h1b-visa-program/
4. Institute of International Education (IIE). “Open Doors 2025 International Student Census Tables.” Annual data on international student enrollment. November 2025. https://opendoorsdata.org/wp-content/uploads/2025/11/OD25_Intl-Student-Census_Tables.xlsx
5. Schmitt, Eric (@Eric_Schmitt). “The Optional Practical Training program is a cheap foreign-labor program…” and related posts on immigration policy. X (formerly Twitter), November 2025. https://x.com/Eric_Schmitt/status/1989741493595308461 and https://x.com/Eric_Schmitt/status/1993035234812838338 and https://x.com/Eric_Schmitt/status/1989741410137108558
6. U.S. Bureau of Labor Statistics. “Table A-4. Employment status of the civilian population 25 years and over by educational attainment.” Current Population Survey, September 2025. https://www.bls.gov/news.release/archives/empsit_11202025.htm
7. Federal Reserve Bank of New York. “The Labor Market for Recent College Graduates: Underemployment Rates.” Data showing 33.7% underemployment for all college graduates (the June 2026 reading), the rate used above. In the current release (updated August 6, 2026), the rate is 33.8% for June 2025 and 34.2% for September 2025, so the 22.1 million count leans low. https://www.newyorkfed.org/research/college-labor-market#–:explore:underemployment
8. U.S. Senate. “About Senator Eric Schmitt.” Official biography and committee assignments. Accessed November 2025. https://www.schmitt.senate.gov/about/
9. Shivamber, Leon. “Warning: This Degree Could Be Hazardous To Your Wealth!” Analysis of the college wealth premium. CollegeROI.org, September 3, 2025. https://collegeroi.org/warning-this-degree-could-be-hazardous-to-your-wealth and U.S. Bureau of Labor Statistics, “Psychologists,” Occupational Outlook Handbook, 2024-34 projections (archived copy), https://web.archive.org/web/20260819060440/https://www.bls.gov/ooh/life-physical-and-social-science/psychologists.htm
10. U.S. Citizenship and Immigration Services. “H-1B Electronic Registration Process.” Overview of the registration requirement for cap-subject petitions. Accessed November 2025. https://www.uscis.gov/working-in-the-united-states/temporary-workers/h-1b-specialty-occupations/h-1b-electronic-registration-process
11. U.S. Department of Labor. “Form ETA-9035: Labor Condition Application for H-1B Filings.” LCA requirements and attestations. Accessed November 2025. https://www.dol.gov/agencies/eta/foreign-labor/programs/h-1b and https://flag.dol.gov/sites/default/files/2019-09/ETA_Form_9035.pdf and 20 CFR 655.810, “What remedies may be ordered if violations are found?” https://www.ecfr.gov/current/title-20/chapter-V/part-655/subpart-I/section-655.810
12. U.S. Citizenship and Immigration Services. “H-1B Specialty Occupations.” Defines specialty occupation requirements and adjudication standards. Accessed November 2025. https://www.uscis.gov/working-in-the-united-states/h-1b-specialty-occupations
13. U.S. Citizenship and Immigration Services. “H-1B Electronic Registration Process.” Overview of the registration requirement for cap-subject petitions. Accessed November 2025. https://www.uscis.gov/working-in-the-united-states/temporary-workers/h-1b-specialty-occupations/h-1b-electronic-registration-process
14. U.S. Citizenship and Immigration Services. “Adjudicator’s Field Manual, Chapter 31: Petitions for Temporary Workers (H Classifications).” Section 31.3 partially superseded as of June 17, 2020. USCIS directs readers to its Policy Manual for current policy. Accessed November 2025. https://www.uscis.gov/sites/default/files/document/policy-manual-afm/afm31-external.pdf
15. National Foundation for American Policy (NFAP). “H-1B Petitions and Denial Rates for FY 2025.” Policy brief on H-1B adjudication trends. November 2025. https://nfap.com/wp-content/uploads/2025/11/H-1B-Petitions-and-Denial-Rates-For-FY-2025.NFAP-Policy-Brief.2025.pdf
16. DavidsonMorris. “H1B Data 2026: Trends, Employers & Approval Rates.” Overview of cap usage and processing trends. Accessed November 2025. https://www.davidsonmorris.com/h1b-data/
17. U.S. Citizenship and Immigration Services. “Characteristics of H-1B Specialty Occupation Workers: Fiscal Year 2024 Annual Report to Congress.” October 2025. https://www.uscis.gov/sites/default/files/document/reports/ola_signed_h1b_characteristics_congressional_report_FY24.pdf
18. U.S. Department of Homeland Security. “OPT Process at a Glance” and “Optional Practical Training (OPT) for F-1 Students.” SEVIS Help Hub and USCIS guidelines. Accessed November 2025. https://studyinthestates.dhs.gov/sevis-help-hub/student-records/fm-student-employment/f-1-optional-practical-training-opt and https://www.uscis.gov/working-in-the-united-states/students-and-exchange-visitors/optional-practical-training-opt-for-f-1-students#:~:text=Pre%2Dcompletion%20OPT:%20You%20may,per%20week)%20or%20full%20time. and “Form I-983 Overview.” https://studyinthestates.dhs.gov/stem-opt-hub/additional-resources/form-i-983-overview
19. Hira, Ron and Costa, Daniel. “New Evidence of Widespread Wage Theft in the H-1B Visa Program.” Economic Policy Institute, December 9, 2021. https://www.epi.org/publication/new-evidence-widespread-wage-theft-in-the-h-1b-program/
20. John Bound, Gaurav Khanna, and Nicolas Morales, “Understanding the Economic Impact of the H-1B Program on the U.S.,” NBER Working Paper 23153 (2017), https://doi.org/10.3386/w23153; Anna Maria Mayda, Francesc Ortega, Giovanni Peri, Kevin Shih, and Chad Sparber, “The Effect of the H-1B Quota on Employment and Selection of Foreign-Born Labor,” NBER Working Paper 23902 (2017), https://doi.org/10.3386/w23902.
21. U.S. Bureau of Labor Statistics. “The Employment Situation.” Monthly economic news release, September 2025. https://www.bls.gov/news.release/archives/empsit_11202025.htm
22. National Science Foundation. “The STEM Labor Force: Scientists, Engineers, and Skilled Technical Workers (NSB-2024-5).” Science and Engineering Indicators 2024. https://ncses.nsf.gov/pubs/nsb20245/ and supplemental Tables SLBR-2 and SLBR-23, https://ncses.nsf.gov/pubs/nsb20245/assets/supplemental-tables/nsb20245-supplemental-tables.pdf and National Center for Education Statistics, Digest of Education Statistics, Table 325.35, “Degrees in computer and information sciences conferred by postsecondary institutions, by level of degree and sex of student: Academic years 1964-65 through 2021-22” (the computer science degree counts), https://nces.ed.gov/programs/digest/d23/tables/dt23_325.35.asp
23. Federal Reserve Bank of New York. “The Labor Market for Recent College Graduates.” Unemployment and underemployment rates by major from American Community Survey data (2023). Accessed November 2025. https://www.newyorkfed.org/research/college-labor-market Weights were derived from NCES bachelor’s degree data (2021-22, Digest Table 322.10, https://nces.ed.gov/programs/digest/d23/tables/dt23_322.10.asp), which reports about 436,000 STEM bachelor’s degrees in 2021-22. Individual major breakdowns were estimated using NCES aggregate categories and subfield data from ASEE and College Factual. The NY Fed unemployment and underemployment rates are based on 2023 American Community Survey data for early-career graduates ages 22-27
24. U.S. Citizenship and Immigration Services. “Characteristics of H-1B Specialty Occupation Workers: Fiscal Year 2024 Annual Report to Congress.” October 2025. https://www.uscis.gov/sites/default/files/document/reports/ola_signed_h1b_characteristics_congressional_report_FY24.pdf
25. NAFSA: Association of International Educators. “CRS Focus Report: Optional Practical Training (OPT) for Foreign Students.” Regulatory information brief. Accessed November 2025. https://www.nafsa.org/regulatory-information/crs-focus-report-optional-practical-training-opt-foreign-students-united
26. NAFSA: Association of International Educators. “CRS Focus Report: Optional Practical Training (OPT) for Foreign Students.” Regulatory information brief. Accessed November 2025. https://www.nafsa.org/regulatory-information/crs-focus-report-optional-practical-training-opt-foreign-students-united
27. Griffiths, Brent D. “Leading computer science professor says ‘everybody’ is struggling to get jobs: ‘Something is happening in the industry’.” Business Insider, September 28, 2025. https://www.businessinsider.com/computer-science-students-job-search-ai-hany-farid-2025-9
28. U.S. Bureau of Labor Statistics. “Employment Projections: 2024-2034.” News release on occupational growth and AI automation. August 2025. https://www.bls.gov/news.release/archives/ecopro_08282025.htm The September 2026 update uses the 2025-35 projections: U.S. Bureau of Labor Statistics, “Occupational projections and worker characteristics,” Table 1.2 (also note 31), https://www.bls.gov/emp/tables/occupational-projections-and-characteristics.htm
29. Lee, Chong Ming. “Andrew Ng lays out a hierarchy of engineering talent — including who’s in trouble and who he refuses to hire.” Business Insider, November 18, 2025. https://www.businessinsider.com/andrew-ng-talent-engineer-ai-hire-college-graduates-computer-science-2025-11
30. U.S. Citizenship and Immigration Services. “Characteristics of H-1B Specialty Occupation Workers: Fiscal Year 2024 Annual Report to Congress.” October 2025. https://www.uscis.gov/sites/default/files/document/reports/ola_signed_h1b_characteristics_congressional_report_FY24.pdf
31. U.S. Bureau of Labor Statistics. “Occupational projections and worker characteristics.” Employment Projections program, Table 1.2 (2025-35 projections, 2025 median wages). Accessed September 2026. https://www.bls.gov/emp/tables/occupational-projections-and-characteristics.htm
32. National Association of Colleges and Employers (NACE). “Class of 2025 Salary Projections Mixed.” Projections for Class of 2025 bachelor’s graduates. February 10, 2025. https://www.naceweb.org/job-market/compensation/class-of-2025-salary-projections-mixed
33. U.S. Census Bureau. “American Community Survey: Sample Size and Data Quality.” Methodology and variance documentation. Accessed November 2025. https://www.census.gov/acs/www/methodology/sample-size-and-data-quality/ and https://nces.ed.gov/ipeds/Search?query=bachelors%20degree%20conferred&query2=bachelors%20degree%20conferred&resultType=all&page=1&sortBy=relevance&overlayDigestTableId=202328 and IPUMS USA, “Descriptions of IPUMS Samples” (ACS 2023 1-year sample: a 1-in-100 national random sample), https://usa.ipums.org/usa/sampdesc.shtml
34. American Institute of Physics. “Roster of Physics Departments with Enrollment and Degree Data, 2022.” https://www.aip.org/statistics/roster-of-physics-departments-with-enrollment-and-degree-data-2022
35. National Center for Education Statistics (NCES). “Bachelor’s degrees conferred by postsecondary institutions, by field of study.” Digest of Education Statistics, Table 322.10 (2021-22). https://nces.ed.gov/programs/digest/d23/tables/dt23_322.10.asp
36. American Institute of Physics (AIP). “Physics Bachelors Initial Employment.” Statistical Research Center data (1996-2022). Accessed November 2025. https://aip.brightspotcdn.com/f6/e4/2a07108a4456aad9e6483afb26eb/0037-employment-booklet-2021and2022-bach-i1.pdf
37. Center for Security and Emerging Technology (CSET). “The Global Distribution of STEM Graduates: Which Countries Lead the Way?” Georgetown University, Accessed November 2025. https://cset.georgetown.edu/article/the-global-distribution-of-stem-graduates-which-countries-lead-the-way/
38. Shivamber, Leon. “America’s Talent Crisis: The War America Is Losing (And Doesn’t Even Know It’s Fighting).” Shivamber.com, November 4, 2025. https://shivamber.com/americas-talent-crisis-the-war-america-is-losing-and-doesnt-even-know-its-fighting/
39. National Science Board. “Talent Is the Treasure.” National Science Foundation, 2024. https://nsf-gov-resources.nsf.gov/files/nsb-talent-treasure-2024.pdf
40. U.S. Department of Homeland Security. “Weighted Selection Process for Registrants and Petitioners Seeking to File Cap-Subject H-1B Petitions.” Final rule, 90 Fed. Reg. 60864, December 29, 2025. https://www.federalregister.gov/documents/2025/12/29/2025-23853/weighted-selection-process-for-registrants-and-petitioners-seeking-to-file-cap-subject-h-1b
41. The White House. “Restriction on Entry of Certain Nonimmigrant Workers.” Presidential proclamation, Federal Register, September 23, 2026. https://www.federalregister.gov/documents/2026/09/23/2026-19554/restriction-on-entry-of-certain-nonimmigrant-workers
42. National Center for Education Statistics. “Degrees in psychology conferred by postsecondary institutions, by level of degree and sex of student.” Digest of Education Statistics, Table 325.80. https://nces.ed.gov/programs/digest/d22/tables/dt22_325.80.asp