The effect of low-skill immigration restrictions on U.S. firms and workers: Evidence from a randomized lottery

How to cite

Clemens, Michael; Lewis, Ethan. 2025. "The effect of low-skill immigration restrictions on U.S. firms and workers: Evidence from a randomized lottery." AEA RCT Registry. AEARCTR-0016896. https://doi.org/10.1257/rct.16896-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Michael Clemens* , Ethan Lewis

* Primary investigator

Registry ID
AEARCTR-0016896
DOI
https://doi.org/10.1257/rct.16896-1.0
First registered
Last updated
Status
Completed
Trial dates
2021-10-21 – 2023-04-25
Countries
United States of America (Nationwide)
JEL codes
F22, J61, D22
Registry record
http://www.socialscienceregistry.org/trials/16896

Abstract

The U.S. Department of Labor runs a lottery to allocate temporary work visas (called H-2B visas) for jobs that do not require a college education and are outside of agriculture. Each year, employers request far more visas than the legal cap allows. To make the visa allocation fair, the Department of Labor randomly assigns petitions to processing groups labeled A, B, C, and so on. Petitions in the earliest group (“A”) are almost guaranteed to be processed before the cap is reached, while later groups have a much smaller chance. Because the assignment to these groups is random, it creates a natural experiment: some firms are randomly given the ability to hire more foreign workers, while otherwise similar firms are not.

This setup allows the present study to compare the outcomes of firms who “won” the lottery to those who “lost,” as assessed in a novel firm survey, isolating the causal effect of restricted or expanded access to low-skill immigrant labor. It is not a designed randomized controlled trial, but rather studies the result of an existing, naturally randomized administrative process within the US government. The core outcomes and subgroup analysis were fixed by a pre-analysis plan irreversibly posted before the firm survey began. The pre-analysis plan is available at: https://osf.io/zdyun

The initial results from analysis of the firm survey are available in the NBER Working Paper based on this study: http://doi.org/10.3386/w30589

Intervention

This study did not design or carry out an intervention. For its own administrative reasons, the US Department of Labor randomizes the order in which petitions for Foreign Labor Certification for US firms to hire workers on the H-2B visa are processed. That is, the randomized trial consists of natural randomization by a US government agency, a process that arose with no intervention by the investigators in this study. The investigators collected survey data on firms that entered this naturally randomized lottery.

Experimental design

The Dept. of Labor's randomized intervention was to assign a letter (A, B, C, etc.), chosen by lottery, to each firm's petition for hiring workers under this visa. It then processed those petitions in the order given by the letter: A petitions first, then B, then C, and so on. This had the knock-on effect of limiting access to the visas for firms that did not receive letter A in the lottery, because all petitioning firms were competing for a fixed quota of visas that rapidly ran out.

Design details

Randomization method
Randomization was carried out by the Department of Labor, in an internal process. It is carried out on a computer, but the agency does not specify the precise method that it uses.
Randomization unit
Randomization is at the level of petitions, not firms.
Clusters
We could not plan for a specific sample size, because the sample size depended heavily on the response rate. In the event, for the January 2021 lottery the universe was 5,376 petitions and the sample represented 477 petitions. For the January 2022 lottery the universe was 7,877 petitions and the sample represented 357 petitions.
Observations
The regressions are carried out at the firm level. Combining the data on firms in a pooled sample from the 2021 and 2022 lotteries, 472 firms were the basis for the empirical results.
Arms
We use the natural randomization to construct two different instrumental variables, only one of which lends itself to the dichotomous 'treatment/control' suggested by this question. For that instrument, the fraction of 'winners' is 0.314, implying a 'treatment' sample of 0.314 * 472 = 148 firms, and thus a 'control' sample of 324 firms.
Primary outcomes
The principal outcomes, as specified in the Pre-Analysis Plan, were firm-level revenue and firm-level hiring of low-skill US workers.
Secondary outcomes
The secondary outcomes, as specified in the Pre-Analysis Plan, were firm-level investment and firm-level year-on-year percentage change in profits.

Papers from this trial

  1. Clemens, Michael A., and Ethan G. Lewis. The Effect of Low-Skill Immigration Restrictions on US Firms and Workers: Evidence from a Randomized Lottery. No. w30589. National Bureau of Economic Research, 2024 (revision of 2022). http://doi.org/10.3386/w30589

Cited by

No citation data available for this registration yet.

Keywords