低技能移民限制对美国企业和工人的影响：来自随机抽签的证据

Michael A. Clemensand Ethan G. Lewis.July 2026. © American Economic Association. This paper has been published inAmericanEconomic Journal: Applied Economics. Reprinted with permission. ABSTRACT. US firms hiring foreign workers in low-skill nonfarm jobs facea binding quota on the H-2B visa, allocated in part through arandomized lottery. We evaluate the quota’s marginal impactusing the lottery, a novel firm survey, and a preanalysis plan. Firmsexogenously employing more H-2B workers in low-skill jobs increaseproduction (elasticity 0.20–0.22), investment (1.5–2.1), and profits(0.15). The elasticity of substitution between H-2B and US workersis very low (0.8–2.0). Thus the effect on US employment is zeroor positive overall, and positive in rural areas. Forensic analysissuggests similarly low substitutability of black-market labor. Michael A. Clemensis an economistand professor atthe Johns HopkinsUniversity Schoolof Governmentand Policy andnonresident seniorfellow at thePeterson Institutefor InternationalEconomics. Ethan G. Lewisis professor ofeconomics atDartmouth College. JEL codes:F22, J61, D22. Keywords:low-skill immigration, US firms, randomized lottery, H-2Bvisa, US employment, low-skill jobs, elasticity of substitution. Note:Firm survey approved by the Dartmouth College Committee forthe Protection of Human Subjects #STUDY00032360. Pre-analysis planirreversibly registered before data collection, athttps://osf.io/zdyun;AEA RCT Registry:https://www.socialscienceregistry.org/trials/16896,data and code available at Clemens and Lewis (2025). We benefitedfrom interactions with Suresh Naidu, Thomas Chaney, Melanie Morten,Ran Abramitzky, Olivier Blanchard, Jeff Kling, Dean Yang, Leah Boustan,Chris Walters, Joan Llull, William Collins, Paolo Falco, Chad Sparber,Jeremy Weinstein, Todd Schoellman, Nicolas Morales, Anna Maria Mayda,Joan Monras, Giovanni Peri, Muly San, Parag Mahajan, Sharat Ganapati,Quy-Toan Do, Marta Prato, Britta Glennon, Jonathan Dingel, Nels Lind,Stefano Carattini, Federico Mandelman, Hyunju Lee, Justin Sandefur,Greg Auclair, and seminar participants at the NBER Summer Institute, theNBER Conference on Immigrants in the US Economy, Stanford UniversityDepartment of Economics, the CESifo Venice Summer Institute, theVanderbilt University Department of Economics, University of DelawareDepartment of Economics, the Federal Reserve Bank of Richmond, theBank of Canada Workshop on Macroeconomic Implications of Migration,the Federal Reserve Bank of Atlanta, the Congressional Budget Office,George Mason University, the Peterson Institute for International Economics,and the University of Pittsburgh. US Citizenship and Immigration Servicesand the US Department of Labor provided public data on the certificationlottery. The firm survey was distributed by the National Association ofLandscape Professionals, the Outdoor Amusement Business Association,the Seasonal Employment Alliance, and the American Seafood JobsAlliance. We acknowledge support from Open Philanthropy and we thankReva Resstack for research assistance. Any views expressed herein arethose of the authors alone and do not represent any organization. 1Introduction The effect of immigration on U.S. workers remains controversial. It hinges crucially on the de-gree to which foreign workers complement or substitute for U.S. workers in production. Theliterature generally finds that immigration for high-skill work—requiring higher education andspecialized knowledge—broadly complements native labor. There is less consensus about immi-gration for low-skill work (Dustmann et al. 2016a; Blau et al. 2017, 267; Blau and Hunt 2019, 174;Edo et al. 2020). Despite the great economic and political importance of restrictions on low-skillimmigration, estimates of its effect range widely depending on the assumptions used to approx-imate causal identification (Card 1990; Borjas 2003; Ottaviano and Peri 2012; Dustmann et al.2016b). Here we study the economic effects of low-skill immigration using a novel, large-scale policyexperiment in the United States: nationwide, firm-level, natural randomization of restrictionson the employment of immigrants in low-skill jobs. The United States has one principal workvisa for low-skill labor in the nonfarm economy—the H-2B visa. U.S. employers’ access to thatvisa is limited by a quota and allocated in part via a randomized lottery conducted by the federalgovernment. This exogenous variation allows unusually transparent, policy-relevant estimatesof how low-skill immigration restrictions affect U.S. firms in the short run. We first pre-registered the hypothesis tests, predicted treatment effects, and subgroup hetero-geneity tests implied by a simple model of a monopolistically-competitive firm facing a poten-tially imperfectly-competitive input market for foreign and U.S. workers.We then collectedsurvey data on 472 firms comprising both winners and losers of the H-2B visa lotteries for mid-2021 and mid-2022.This allows pre-specifi