The H-1B, the U.S. high-skilled work visa, raised wages in non-STEM jobs, a working paper finds

Twenty years of exposure to the H-1B visa, the U.S. program for sponsoring high-skilled foreign workers, raised incomes for native-born Americans and earlier immigrants, but mostly in occupations the visa does not directly target, a new National Bureau of Economic Research working paper finds.

The paper, [Long-Run Effects of H-1B Immigration on the U.S. Economy](https://www.nber.org/papers/w35560) by Ran Abramitzky, Leah Boustan, Elif Gulek, and Jens Hainmueller, exploits the 1999–2003 statutory cap expansion as a natural experiment. Industries that recruit heavily from the H-1B pool absorbed a measurable shock when the annual visa ceiling was temporarily lifted. By comparing those exposed industries with similar industries that were not, the authors estimate what two extra decades of H-1B presence actually did to wages, firms, and innovation.

Income gains were concentrated in non-STEM occupations, not in the science, technology, engineering, and math roles the visa is designed to fill. Wages rose in construction, retail, and food services in affected industries. That pattern is hard to square with a story in which the visa simply adds skilled workers to the high-skill labor market and stops there.

The authors read that finding as a clue and use it to test two competing mechanisms. A labor-supply shock, in which more H-1B workers just join the existing pool, should produce localized wage effects and few spillovers elsewhere. A productivity shock, in which H-1B workers make existing teams measurably more productive, should show up in the firms that buy from the affected industries, because those buyers get better inputs and the gains propagate forward through the supply chain.

Downstream industries, those that purchase from H-1B-exposed sectors, gained. Upstream suppliers did not. That asymmetry is the paper's most informative result, because pure labor-supply substitution would not generate directional spillovers of this shape, and the directional pattern is exactly what a productivity story predicts.

A second result backs the same interpretation. The authors find no detectable effect of H-1B exposure on patenting. That null is informative on its own. If the program were driving breakthrough innovation, the patent record should show it. The absence of a patenting bump is consistent with productivity gains coming from better task execution inside existing teams, the kind of marginal improvement that lifts wages and propagates through buyers but does not register in patent counts.

The reading complicates the active H-1B debate, which usually centers on labor-market substitution: how many native workers are displaced, and at what wage. The paper's evidence treats the question as a propagation problem: the visa appears to change how work gets done across a supply chain, not just how many people are competing for a given job.

Three limits apply. The identification rests on a single 1999–2003 cap expansion, and extrapolating from that shock to today's labor market is the authors' own stretch. The paper is a working paper, not yet peer-reviewed, and the productivity interpretation is the authors' reading of the spillover pattern, not an established fact. Wage gains in retail and construction are real, but they are also consistent with localized demand effects from clusters of higher-paid H-1B workers, an alternative the data cannot fully rule out.

The narrower claim the evidence supports: long-run H-1B exposure does not appear to have crowded native workers out of the high-skill labor market, and the income gains landed in places the cap was never designed to reach. Whether that pattern survives reanalysis, the 2017 and 2024 policy changes, and a tighter AI-adjacent labor market is the open question, and the next paper's job.