H-1B Visas and the American Economy

This post is part of a living literature review on migration. For more information on the project, see here.

---

This was not entirely the post I was planning on writing this week,¹ but: H-1B visas have recently been In The News.

The incoming Trump administration seems to be signaling that it is at least somewhat sympathetic to the “tech right” with the appointment of Sriram Krishnan to an advisory position on AI. Far-right activists were not thrilled with this choice, as Krishnan has advocated to make it easier for foreign workers to come to the US, particularly if they work in tech.

Thus sparked a “civil war” among Trump supporters. I think it would be more accurate to describe this as a large number of people shouting at each other on the platform formerly known as Twitter.com, but nonetheless: there has been extensive discussion of skilled migration generally and H-1B visas in particular.

Elon Musk is vocally in favor of expanding the program. Steve Bannon is opposed. Trump says he’s a big believer in H-1B visas, though he moved to restrict them during his first term. Bernie Sanders has also joined the fray, saying Musk just wants cheap immigrant labor.

So it seemed worthwhile spending some time summarizing what we do - and do not - know about H-1B visas and their effects on the American labor market.

What is an H-1B visa?

The short version

1. A H-1B visa is the US’ primary “professional worker” visa.² It is issued to people with at least a college degree³ in a specialty occupation, for a specific job; the visa requires a job offer in hand. H-1B visas are relatively common, compared to other types of work-based visas; if someone is working on a visa in the US, it is likely an H-1B.

2. However, H-1B visa holders are relatively rare in the US population as a whole. There are about 600,000 workers currently on a H-1B (or at least, there were in 2019). Since there are 167 million members of the US labor force, this means about 0.4% of the US workforce is on an H-1B visa. Most people in the US probably don’t know anyone on an H-1B,⁴ and most firms don’t employ any workers on one.⁵

3. Jobs that offer an H-1B visa must pay at least $60,000, or the prevailing wage for that job in their area, whichever is higher.⁶ This means that by construction, a person on a H-1B visa makes more than the average American.⁷

4. H-1B visas are not permanent. They expire, and no one goes directly from an H-1B visa to US citizenship. If you don’t have sponsorship for green card status (not part of H-1B status), you cannot stay in the US indefinitely on a H-1B visa.⁸

Wait, I thought we were talking about people who move to the US to stay here.

Well, sort of. Like most skilled worker programs, a H-1B visa is issued for a particular job opening. It is considered to be a non-immigrant visa, because you can’t stay or renew indefinitely.⁹

If your maximum six year period of H-1B status is going to expire, you can either 1) depart the US for one year, and then reapply for a new H-1B, or 2) one’s employer can petition for an individual to be classified as an immigrant that would allow the individual, when a green card number becomes available, to file for an adjustment of status that grants permanent residency.¹⁰ (Note that they would, at that point, no longer have or need H-1B status, as they would be a permanent resident.)

The likelihood of option 1 or option 2 varies significantly by your employer. Some firms eventually apply for employment-based green cards for nearly all their H-1B visa holders (for instance, Qualcomm and Microsoft), while others send nearly all of their workers home after the visa is up (for instance, Tata Consultancy Services and Wipro).¹¹

Indeed, this is one of the complications of looking at the H-1B visa program as a whole; different actors and firms use the H-1B programs in different ways. Some firms treat it as a way to bring in highly skilled immigrants with skills they could not find elsewhere; others treat it as a guest worker visa. This is one of the reasons evaluating the whole H-1B program is difficult; it is true that in some cases, it is (as per Elon Musk) to bring in top engineering talent; others (as per Bernie Sanders) really do use it as an offshoring method. There is no one, single way in which H-1B recipients fit into the US economy.

OK, but I’ve heard about a H-1B lottery. Why do people talk about a visa lottery if this is for a specific job opening?

If you are one of the Americans who knows anyone on an H-1B, you will have heard people discuss the H-1B lottery. And, yes, H-1B visas are determined partially by lottery.

Like most US visa types, there is a limited number of H-1B visas that can be issued in a single year. Since 2004, there have been 65,000 H-1B visas issued to any skilled worker, with an additional 20,000 reserved for workers with U.S. advanced degrees, an additional 5,400 reserved for Singaporean nationals, and an additional 1,400 reserved for Chilean nationals.¹² (If, by this point, you are thinking that US immigration law is needlessly complicated: congratulations! You are correct!)

However, there are far more than 91,800 foreign nationals who receive job offers that are conditional on receipt of a H-1B visa per year. Indeed, there are about 450,000 such people every year. Thus, there is a lottery to assign these 91,800 visas to 450,000 people. Or rather, several lotteries, as there is both a “regular” lottery and an “U.S. advanced degree” lottery. Because who ends up being selected is random, the lottery serves to assign H-1Bs to some firms and not others by pure chance. Several of the papers I discuss below will use this to determine what the effect of having an additional H-1B worker is.

There is one last caveat to how many H-1B visas are issued: some categories of employers – nonprofit research organizations, universities - are not subject to the cap. This means despite the H-1B “cap” being 91,800 visas/year, 386,000 H-1B visas were issued in FY 2023.

OK, enough background. What are the impacts of H-1B labor on the American economy?

So: who’s more right? Is Elon Musk right, and H-1B workers provide useful skills that aren’t available in the American labor market? Are they advancing the American economy? Or is Bernie Sanders right, and firms use H-1B workers as cheap foreign labor?¹³

The Labor Market

Many Americans want to know if H-1B visa holders are “taking American jobs”. I want to be a bit more precise than this. I want to know if H-1B visa holders complement American workers (perhaps by providing useful skills they do not have, allowing for more output or increasing overall output) or substitutes (thus, decreasing American employment or wages). Most likely, they are some combination of the two - but how much of each?

Native-Born Employment And Wages At Firms That Hire H-1Bs

Positive Supply Shocks

We begin by trying to answer if firms that successfully employed H-1B labor employed fewer natives as a result. The below papers use getting an additional H-1B visa (or many H-1B visas) to examine this question.

Doran, Gelber and Isen 2022 and Bound, Braga, Golden and Khanna 2015 both say that yes, native employment decreases as a result of H-1B labor.

Doran, Gelber and Isen 2022 finds that an additional H-1B visa crowds out 1.5 other workers at the firm and Bound, Braga, Golden and Khanna 2015 finds that the H-1B program reduced native-born employment in computer science in 2004 by 7-13.6% and wages for native-born computer scientists by 2.6%-5.1% relative to a counterfactual where immigration stayed at 1994 rates.

However, Mahajan, Morales, Shih and Chen 2024 finds the opposite - that firms that hire an H-1B worker end up hiring another 0.83 additional employees. These employees tend to be non-college-educated workers, perhaps providing complementary skills to those of H-1B visa holders (who are, by definition, college educated).

Of these three papers, I like the design of Doran, Gelber and Isen 2022 the best, because it uses the universe of firms with H-1B applications (albeit only in two years). However, it has possible external validity problems, as the identification strategy uses variation between outcomes of firms that applied the day the cap was reached. These firms differed somewhat from the universe of firms, which might mean that their results don’t apply broadly.

Still, I have more issues with the other two papers; Mahajan, Morales, Shih and Chen 2024 looks at only a single year and a subset of firms, and does not have additional data to determine how much the included firms might differ from firms not included in their data. Bound, Braga, Golden and Khanna 2015 uses a model rather than using the natural experiment.

Negative Supply Shocks (or “Great! Let’s fire all the H-1Bs and get back those jobs!”)

OK, so if a marginal H-1B visa recipient leads to 1.5 fewer Americans being hired by a firm, perhaps we should have fewer H-1B visas

The US government has, in fact, already sort of tried this by capping the number of H-1B visas. For the first few years of the current iteration of the H-1B visa program, starting in FY1992, the cap was not reached but since FY1997 H-1B numbers have been used before the end of the allotted fiscal year except for three years when Congress temporarily raised the cap (FY2001-2003). However, in 2007 when filings were made for FY2008, the cap was met before the start of the fiscal year and the government introduced the visa lottery.

Since a lottery is, by definition, random, some areas received fewer denied H-1B visas and some regions received more. Did the regions where more H-1B visas were denied hire more computer workers?

According to Peri, Shih and Sparber 2015, no, they did not. In fact, native-born computer based employment may have even fallen when firms had more H-1B visas denied, suggesting complementarities between native-born and immigrant computer workers.

So: hiring more immigrants - possibly bad for native-born employment. Not hiring immigrants? Also possibly bad for native-born employment. Unsatisfying as this conclusion is: the evidence base here is just kind of messy; I don’t really know that I can say for certain how firms respond to getting or being denied H-1B visas.

Native-Born Employment In The Economy As A Whole

There are also complex interactions between firms within an economy; a decrease in employment at a particular firm doesn’t necessarily mean a decrease in employment across the economy, if there are new needs as a result of either migration or productivity.

As an example, let’s say that H-1B workers are much more productive than Americans at designing some new invention. The firm that makes invention X is going to probably hire more H-1B workers and fewer Americans. But if invention X needs input Y, you might also see an expansion in firms that produce input Y - and some of the displaced Americans might work there.

There is also a simpler example of how the economy might adjust to more H-1B workers even if native-born employees are displaced from a firm. H-1B workers are, in fact, people; they will need to consume goods and services. They will need accountants and barbers and childcare and lawyers. Indeed, since H-1B visa holders are relatively high-income in comparison to the average American (see the fiscal impact section below), they probably consume more goods and services than the average American.¹⁴

However, this is, unfortunately, a really hard question to study, though, and the evidence is very limited.

Peri, Shih, and Sparber 2015 (a different paper, same authors) finds that in the long-run, an increased number of H-1B visa holders in an area leads to higher wages for college graduates.

However, this is a shift-share paper, and I’ve previously discussed why I don’t love shift-share designs in migration. Still: it’s the evidence we have, and it is weak evidence that the US economy does adjust over time - and that H-1B visas are good for the US.

Welfare and Output

Furthermore, productive firms are good for the US economy. Bound, Braga, Golden and Khanna 2015 estimates that increased immigration of IT workers in the 1990s resulted in the prices of IT goods and services being 1.9% lower than they otherwise would have been.

This may sound small, but IT is a significant portion of the US economy. In 2024, the US government alone spent about $100B on IT investments, and thus, immigration in the 1990s has saved US taxpayers perhaps $2B a year.

Furthermore, it also increased output in information technology by 2.5% relative to a world where immigration did not expand in the 1990s. There are times I am not sure more information technology is good - generally, whenever I am using a Microsoft product - but overall, an increase in information technology seems probably beneficial.

These immigrants produced new things, and that made goods and services more common and also cheaper. Seems good.

Innovation

The H-1B visa is often pitched as a way to increase innovation, as it is used for skilled workers, and is used particularly extensively in tech. Elon Musk says he wants to use H-1Bs to hire the top 0.1% of engineering talent with H-1Bs - that should increase innovation. But does it really?

The empirical evidence suggests that, yes, H-1B visas are good for American innovation.¹⁵ Hunt 2011 finds that immigrants who enter on skilled work visas are much more likely to produce patents and scientific papers than the native-born. This seems like strong evidence that H-1B visas do increase innovation; if you add more of a population that overproduces innovation, you end up with a higher rate of innovation.

Furthermore, Chen, Hshieh and Zhang 2021 and Dimmock, Huang, and Weisbenner 2019 find that winning the H-1B lottery significantly increases patenting rates and financial performance of startups.¹⁶

However, this conclusion is somewhat tempered by Doran, Gelber and Isen 2022. This paper looks at the universe of firms that use H-1B labor and finds that firms that receive an additional H-1B worker do not increase patenting rates (though this is not a tight null). However, given the lack of statistical significance of their patenting results, I don’t want to lean on this result that hard - and thus, still believe that H-1B visas increase innovation in the US.

Fiscal Impact

Another common worry about migrants is their impact on public finances. Do immigrants use more in public resources than they pay in taxes?

H-1B visa recipients are perhaps a best case for those worried about the effects of migration on the public purse. H-1B holders are relatively young (median age of 33), high earning (average income $108,000 in 2021) and have relatively few dependents (there are far fewer dependents of people on H-1B visas than H-1B visa recipients). All of these traits mean they are likely to pay more in taxes than they receive in benefits.

Furthermore, because it is a non-immigrant visa, and requires employment, H-1B recipients cannot be unemployed and in the country for more than 60 days. Furthermore, the minimum salary threshold of $60,000¹⁷ means that H-1B visa holders and their families¹⁸ would be unlikely to qualify for much income-based government assistance.

I have been unable to find a calculation of net fiscal impact of H-1B visa holders in particular, but the net fiscal impact of immigrants with a bachelor’s degree or higher who arrive between ages 25 and 44 is highly positive. Such an immigrant is expected to pay about $500,000 more in taxes than they receive in benefits in net present value terms.¹⁹

High Skilled Immigration More Broadly

There is a further caveat to all discussion of H-1B visas. H-1B visas are a gateway for other types of skilled immigration. The other large category of employment-based skilled immigration, EB visas, largely do not go to people living in other countries at the time of application. They go to workers already in the US, largely on H-1B visas.

This means that innovation and economic growth in the US due to immigrants who have green cards are also, kind of, due to H-1B visas. Indeed, H-1B visas tend to be early career; other types of visa tend to be more common later in one’s career; H-1B visa holders turn into other types of skilled immigration.

I haven’t focused on the broader literature on the effects of skilled migration on the US here, since I wanted to address the literature on H-1Bs specifically, but skilled migrants (on all visa types) contribute to the US in many ways. Immigrants are more likely to be inventors than the native-born, and the patents they produce are more innovative. Immigrants are more likely to be entrepreneurs, and foreign masters students increase the number of US startups. 38% of US Nobel Prize winners are immigrants - even though only 15% of the US population are immigrants.²⁰

I’ll discuss this literature more fully in a future post, but if we think that H-1Bs are a “gateway drug” to more skilled migrants, this also seems positive.

Takeaways

So: is Elon right, or is Bernie? Are H-1B visas good or bad for the American economy?

My read of the evidence is as follows:

• Do firms hire H-1B workers instead of Americans in the short term? I genuinely do not know; the state of the evidence is murkier than I would like.

• Does that mean that American workers suffer and have permanently lower wages? I don’t think so; one paper finds increased immigrant employment in STEM has increased American wages and employment. However, this conclusion is based on a single paper that I don’t love, and my level of certainty here is low. If another paper found otherwise, I’d be willing to believe that result.

• Did you write 3000 words telling me that you don’t know how H-1B visas have affected American job outcomes? Unfortunately yes.

• Do H-1B visas increase American innovation? Yes, this seems likely.

• Do H-1B recipients pay more in taxes than they receive in benefits? This I actually am sure about - yes.

• Is any of this evidence going to stop people arguing on the website formerly known as Twitter.com? I doubt it.

---

Many, many thanks to Amy Nice and Alex Nowrasteh for fact-checking this; any remaining mistakes are my own. Thank you to Nathan Barnard, Phoebe Arslanagić-Little, and Pradyumna Prasad for their comments; all remaining incoherence is due to who I am as a person.

1. Immigration and crime in Europe awaits.

2. The US government approved a total of 386,000 H-1B petitions visas in FY 2023 (including both initial and continuation of H-1B status for both employers subject to numerical limits and those, like universities, exempt from caps). In 2022, the US issued 270,284 EB-category green cards. The EB permanent residency program is actually smaller than it looks from those numbers, though, because dependents of the person sponsored are counted in the EB total, but not the H-1B visa numbers. Dependents of a EB permanent resident also get an EB green card; dependents of H-1B visa holders are issued H-4 visas. So only 121,628 workers received EB green cards in 2022, about a third of the number that received H-1B visas. Moreover, the EB-category green card cap was artificially inflated because of post-Covid results from the green card formula that doubled the EB allocation in FY2022 based on numbers intended for family immigration but left unused when American consulates were not completing family-sponsored green card applications overseas in FY2021. In a typical year, there are 120,000 green cards awarded solely based on skills, experience, and education (the EB-1, EB-2, and EB-3 classifications) of which about 55,000 are the principal workers being sponsored. The H-1B program is thus far larger than the EB program.

3. With the exception of fashion models, who are not required to have a degree.

4. Since H-1Bs tend to be concentrated in particular industries, such as academia, some people (including many people reading this post) probably know quite a few.

5. Since the average US workplace has 24 employees, the significant majority of firms in the US have no employees on a H-1B (though some workplaces employ thousands).

6. For instance, if you hire someone on a H-1B to be a software developer in San Francisco, you must pay a minimum of $130,250.

7. Though likely not forever, because this is hard-coded into the law, and nominal wages do rise over time.

8. H-1Bs are unusual in being dual-intent - most non-immigrant visas will not allow you to pursue options for permanent residency while on them; you are required to go home when it’s up. It is allowable on an H-1B to actively seek to stay permanently through a green card.

9. Though see previous footnote.

10. The ease of transitioning from a H-1B to an employment-based green card also depends significantly on the immigrant’s country of origin. If you are from Canada, it will probably take about two years. If you are from India, it could take 12 or more.

11. For more data, see table 9.1 here.

12. Weirdly, the Chilean and Singaporean quotas do not allow for dual intent.

13. OK, this we can rule out, because even outsourcing firms are required to pay the prevailing wage. The average H-1B migrant makes $108,000.

14. Yes, many immigrants do send home significant portions of their income as remittances. This does reduce the amount of consumption that occurs in the US, but 1) I am not sure if it reduces the amount of consumption to less than the average American, given the average H-1B visa holder makes considerably more than the average American, and 2) this is America. We’re really good at getting you to spend your money if you have it.

15. Note that I do not examine Hunt and Gauthier-Loiselle 2010 because 1) it includes all skilled immigrants, so includes visa types other than H-1B, and the panel data they use is for 1940-2000. Since the H-1B visa was only created in 1990, 80% of their data cannot be about H-1B recipients.

16. There is even some evidence that this productivity can spill over to the native-born through ethnic networks. Kerr and Lincoln 2010 finds that cities that have more H-1B recipients see more patents produced by native-born co-ethnics - that is, that H-1B recipients from India and China write patents with Chinese-Americans and Indian-Americans and increase overall patent rates in the cities and firms where there are lots of H-1B visa holders.

However, I don’t think this paper provides particularly strong evidence. One might imagine areas or firms where there are a lot of H-1B visa holders are different to other areas or firms (for instance, an area might have a lot of tech companies); other properties of the city - besides the H-1B rate - might drive the high rate of co-ethnic patent production.

17. This is below the median household income in the United States - a H-1B recipients’ dependents (generally) cannot work and thus the H-1B income will (generally) be the same as the household income - but well over the federal poverty line for any household of under 9 individuals. Some US benefits are available at up to 200% of the federal poverty line - for instance, food stamps in some states - but H-1B visa holders are unlikely to qualify for these either.

18. H-1B workers can bring their spouse and children under the age of 21 into the US; those dependents are allowed to study but not work (unless they apply for and receive an employment-based visa of their own, the person on a H-1B is in the process of getting a green card, or some other exceptions I won’t get into).

19. There are also fees paid by the employer to receive authorization to hire an H-1B professional. These are about $500M a year, distributed to the Department of Labor and NSF. While this is a small amount in comparison to the taxes paid in by H-1B visa holders, it’s also not nothing.

20. There’s a reason tech people tend to be pro-skilled-immigration; it is difficult to work in tech and not work with some incredible skilled immigrants.