How Much of U.S. Innovation Comes From High-Skilled Immigrants?
How much of America's innovation comes from immigrants? Bernstein et al. (2026) answer this question by linking U.S. patents granted from 1990 to 2016 to address and Social Security records for about 650,000 inventors. Immigrants who arrived as adults make up 16% of U.S. inventors but produced 23% of patents and 25% of the economic value those patents created. They also make their colleagues more productive: when an immigrant inventor dies early, co-inventors’ patenting falls about 16%, compared with 9.5% when a U.S.-born inventor dies. Counting these spillovers, the authors attribute about 32% of U.S. innovation to immigrants. The study does not show how innovation would change if immigration levels changed.
Why This Article Was Selected for The Policy Scientist
Economic growth in advanced economies depends heavily on new ideas, and the United States has long drawn a disproportionate share of its scientists and engineers from abroad. High-skilled immigration policy is now contested: recent restrictions on student and H-1B visas, including a $100,000 fee on new H-1B petitions announced in 2025, sharpen the question of what the country gains or loses by admitting fewer skilled workers. Earlier work, notably Hunt and Gauthier-Loiselle (2010) and Kerr and Lincoln (2010), established that immigrants patent at high rates; this article adds administrative data covering nearly all inventors and, more importantly, a credible estimate of how immigrants affect the productivity of the people they work with. Bernstein has published extensively on the economics of innovation. The spillover estimates rest on a difference-in-differences design using premature deaths, a strong causal strategy, while the headline 32% figure is a model-based accounting exercise rather than a policy counterfactual. The article is forthcoming in the American Economic Review, the flagship journal of the American Economic Association and among the most selective outlets in economics.
Full Citation and Link to Article
Bernstein, S., Diamond, R., Jiranaphawiboon, A., McQuade, T., & Pousada, B. (2026). The contribution of high-skilled immigrants to innovation in the United States. American Economic Review. Forthcoming.
Central Research Question
The study asks how much of U.S. innovation since 1990 can be attributed to immigrants, counting both the patents they produce themselves and the indirect contribution they make by raising the productivity of their collaborators. Along the way, it examines how immigrant and U.S.-born inventors differ in productivity over their careers, whether those differences reflect where immigrants live and which technologies they work in, how immigrants connect U.S. research to foreign knowledge, and whether immigrant inventors become more integrated with U.S.-born colleagues over time.
Previous Literature
Hunt and Gauthier-Loiselle (2010), using 2003 survey data, found that immigrants made up 14% of inventors but authored 24% of U.S. patents, and that a 1 percentage-point increase in the immigrant share of college graduates raised patents per capita by 9–18%. Kerr (2008b) inferred inventor ethnicity from names and showed rising Chinese and Indian contributions to U.S. patenting in the 1990s, though that approach cannot separate foreign-born from U.S.-born inventors. Kerr and Lincoln (2010) found that expansions of the H-1B program increased patenting in exposed firms and cities, while Doran, Gelber, and Isen (2014), using H-1B visa lotteries, found only modest and statistically insignificant effects on firm-level innovation. Studies of large immigration shocks point in different directions on crowd-out: Borjas and Doran (2012) found that Soviet mathematicians displaced American scholars, whereas Moser, Voena, and Waldinger (2014) found that German Jewish émigrés increased U.S. chemistry patenting by 71%. Methodologically, the study builds on Jaravel, Petkova, and Bell (2018), who used inventor deaths to measure how much co-inventors depend on one another.
Data
The study links all U.S. patents granted from 1990 to 2016 to the Infutor database, a commercial file containing the address histories of more than 300 million adults along with names, birth years, and the first five digits of Social Security numbers. Because the first five digits reveal roughly when a number was issued, and nearly all U.S.-born people receive one in childhood, anyone first assigned a number after age 20 is classified as an immigrant. About 70% of patent-inventor records were matched, roughly 90% of those covered by Infutor, producing a panel of about 650,000 inventors with usable birth and Social Security information. Innovation is measured by patents, citations adjusted for technology and year, "top patents" in the top 10% of citations, and the stock-market value of patents, imputed for private firms. The method misses immigrants who arrived before age 20; in 2014, that group made up 39% of all immigrants but only 19% of immigrant PhDs. For the spillover analysis, the authors used the Social Security Death Master File to identify 6,043 inventors who died between ages 25 and 60, yielding 3,947 matched groups and 15,470 surviving co-inventors.
Methods
The descriptive analysis divides each patent's credit equally among its inventors and compares immigrant and U.S.-born output, then uses regressions with birth-cohort, county-by-year, and technology-by-year fixed effects to test whether sorting explains the gap. The causal analysis is a difference-in-differences event study. Co-inventors of inventors who died prematurely are compared with co-inventors of "placebo" inventors who did not die but were matched on immigrant status, age, cumulative patents, citations, year, and number of collaborators. The key assumption is that the timing of an early death is unrelated to trends in collaborators' productivity, which the authors support by showing no decline before the death. Results hold using the Callaway and Sant'Anna (2021) estimator. Weaknesses remain: a death may disrupt a firm or project in ways beyond lost knowledge, and the design cannot detect whether other inventors fill the gap. A structural team-production model then separates team-specific effects from transferable human capital and produces the aggregate decomposition, which the authors describe as accounting rather than a counterfactual.
Findings/Size Effects
Immigrants make up 16% of inventors but produced about 23% of patents, a 43% overrepresentation relative to their share of inventors. Their work is not lower in quality: they account for 24% of citation-adjusted output, about 25% of top patents, and 25% of the economic value of patents. Their share is roughly 25% in computers and communications, drugs and medical, electronics, and chemicals, but about 15% in mechanical and other traditional fields. Both groups' productivity peaks in the late 30s, when immigrants produce more than 50% more patents per year and nearly twice the citation-adjusted output. Immigrants produce 0.093 more patents per year with only year controls; adding county-by-year and technology-by-year fixed effects reduces this to 0.063 (significant at the 1% level), and controlling for team size and peer quality reduces it to 0.052. Regional sorting accounts for about 32% of the gap; in 2005, 13.2% of immigrant inventors lived in Santa Clara County, compared with 4.4% of U.S.-born inventors. Immigrants collaborate with a foreign inventor on 16% of their patents, versus 9% for U.S.-born inventors.
The spillover estimates are the core contribution. Across all deaths, surviving co-inventors file 0.090 fewer patents per year, a 10.6% decline relative to their expected output. When the deceased inventor was an immigrant, the decline is 0.183 patents per year, or 16.4%; when U.S.-born, it is 0.077, or 9.5%. Put differently, for every 100 co-inventors, an immigrant's death costs about 18 patents per year, compared with about 8 for a U.S.-born inventor's death. The gap is larger for citation-adjusted output (22.1% versus 12.1%) and top patents (26.7% versus 11.1%), and present for economic value (16.6% versus 11.9%). Controlling for ten characteristics of the deceased inventor, collaborators, and relationship leaves the gap essentially unchanged. The structural model finds each living immigrant collaborator raises an inventor's general human capital about twice as much as a U.S.-born collaborator (0.693 versus 0.310). In the decomposition, immigrants directly account for 15% of U.S. innovation and indirectly for 17% through U.S.-born collaborators, totaling 32%; about 22% of U.S.-born inventors' output is attributable to immigrant spillovers.
Conclusion
The study shows that immigrants who arrived as adults contribute to U.S. innovation well beyond their numbers, producing more patents of equal or higher quality and making their collaborators more productive, with the spillover effect estimated through a credible quasi-experimental design. It also shows that collaboration runs both ways, with cross-group spillovers accounting for about 25% of U.S. innovation. The study does not establish how total innovation would change under different immigration policies, because it cannot measure whether immigrants crowd out other inventors, and the 32% figure depends on the structure of the model. Its findings also exclude immigrants who arrived as children and second-generation inventors. Future research could examine differences by country of origin, measure crowd-out directly, and test whether similar patterns hold in other countries.
Comments