When ICE ramps up, U.S.-born workers don’t fill the gap, study finds

A central argument for intensified immigration enforcement has long been economic: remove unauthorized workers, and jobs will open up for Americans born in the United States. Employers across agriculture, construction, and other immigrant-heavy sectors have reported difficulty finding workers in 2025, while news coverage has suggested U.S.-born workers are not stepping in to fill the vacancies. Whether that pattern holds up at the national level, and what it means for paychecks, has been an open empirical question.

A new NBER working paper offers the first national, causal analysis of how the second Trump administration’s immigration enforcement surge has reshaped labor market outcomes. The short answer from the authors: in places where Immigration and Customs Enforcement (ICE) activity spiked, likely undocumented immigrants who remained in the country worked less, and U.S.-born men without a college degree saw their employment in affected sectors decline rather than rise.

The question behind the research

Elizabeth Cox and Chloe N. East, both economists at the University of Colorado Boulder, set out to test whether heightened enforcement has actually expanded opportunity for U.S.-born workers, as proponents argue. Past research on enforcement episodes in the 1930s and 2010s has generally found that when unauthorized workers leave or withdraw from the workforce, U.S.-born workers don’t simply replace them. But Trump-era enforcement in 2025 has been larger and, according to the authors, more indiscriminate than prior campaigns, making it unclear whether past patterns would hold.

Science of Money

Sign up for our free weekly newsletter for the latest insights.

Name

Email

The authors also wanted to measure what economists call “chilling effects”: situations where unauthorized workers who remain in the country reduce their work activity out of fear that going to a job will expose them to ICE. A Pew survey cited in the paper found that 43% of immigrants in the U.S. in summer 2025 worried that they or someone they knew could be deported.

How they designed the study

Enforcement intensified everywhere in 2025, but not evenly. Cox and East used this geographic variation to build a natural comparison. Drawing on data from the Deportation Data Project, which tracks ICE Enforcement and Removal Operations arrests, they identified “treated” areas as those that experienced a sudden surge in arrests between January and October 2025, defined as roughly 50 additional arrests per 100,000 non-citizens in a single month, or about a doubling of the arrest rate. “Control” areas saw more gradual increases. They worked with 58 geographic units: 48 states plus sub-state divisions in California, New York, and Texas.

For labor market outcomes, they used the monthly Current Population Survey (CPS) from January 2024 through November 2025. Because the CPS doesn’t record immigration status, the authors followed a standard practice of identifying “likely undocumented” individuals as foreign-born people ages 20 to 64 with at most a high school education working in sectors where unauthorized workers are over-represented. Their primary sector list, based on both undocumented concentration and male-dominated workforces (since over 90% of ICE arrestees were male), included agriculture, construction, manufacturing, and wholesale.

The statistical approach, known as an event study with a difference-in-differences design, compares outcomes in treated versus control areas before and after the arrest surge, while accounting for seasonal patterns, national trends, and persistent differences between areas. The authors also ran a placebo test using 2016–2017 data around Trump’s first inauguration to check whether the same areas would show similar patterns absent the Trump 2.0 enforcement surge. They did not.

One important design choice: the study does not try to count how many immigrants left treated areas. The CPS has known measurement problems for immigrant population estimates, and survey nonresponse has shifted in ways that can distort counts. Instead, the authors focused on what happened to labor market outcomes among people who remained and were surveyed.

What the data showed for immigrants

First, the classification worked as intended. In treated areas, daily arrests rose by about 200 per month compared with control areas, a 114% increase over the pre-period mean. Roughly 85% of that increase came from arrests of men.

Among likely undocumented immigrants who remained in the U.S. and worked in the four main affected sectors, employment fell by 3.4 percentage points, or about 4%, relative to workers in control areas. For men specifically, employment dropped 4.6 percentage points (5%) and hours worked fell by 2 hours per week (5%). Effects for women were small and not statistically significant. The declines grew over time, which the authors suggest is consistent with fear spreading through communities as news of ICE activity circulated.

Wages and weekly earnings among likely undocumented workers who remained employed did not shift consistently in either direction. In other words, there was no clear sign that employers were bidding up pay to hold on to immigrant workers.

What happened to U.S.-born workers

Here the findings push back against the job-opening argument. Across all U.S.-born workers in the four affected sectors, the authors found no increase in employment or hours. Their estimates can rule out an employment gain larger than 0.1 percentage points, a tight bound.

When the authors broke the U.S.-born sample down further, they found something else: a small but statistically meaningful decline in employment among U.S.-born men with a high school education or less working in affected sectors, on the order of 1.3% relative to the baseline. There was no corresponding effect for U.S.-born women or for U.S.-born workers with more education. Wages for U.S.-born workers also did not rise.

Looking sector by sector, the authors found that the industries where likely undocumented men lost the most work, such as construction, were also the industries where less-educated U.S.-born men lost work. The authors interpret this pattern as consistent with a complementary relationship between the two groups: unauthorized workers tend to cluster in particular occupations within a sector (about 40% of likely undocumented men in construction work as laborers, compared with roughly 20% of less-educated U.S.-born men), and losing them appears to reduce overall labor demand in the sector rather than shift it toward native-born workers.

Putting numbers on it

Using American Community Survey data to translate percentages into counts, the authors estimate that in the average treated area, about 7,574 fewer likely undocumented men were working in affected sectors. Given the cumulative increase of roughly 1,200 arrests of men in those areas, this implies about six likely undocumented men stopped working for each arrest, a substantially larger chilling effect than the 2.3-to-1 ratio the authors found for the first Obama administration in earlier research. They attribute the larger ratio to the broader and less targeted nature of current enforcement. For every six likely undocumented men who stopped working, the authors estimate roughly one less-educated U.S.-born man also stopped working in the same sectors.

Caveats worth noting

The study captures only part of the total picture. Because the authors set aside questions about how many immigrants physically left the U.S., their employment estimates reflect only the chilling effect on those who remained. The true impact on total immigrant labor supply is almost certainly larger.

The results also depend on the “likely undocumented” proxy, which uses demographics to approximate status that isn’t directly recorded. The authors show their results hold when they narrow the sample to non-citizens or to Hispanic workers. Survey nonresponse could also be rising among the most affected groups, which the authors note would tend to make their estimates conservative rather than inflated.

The authors note that their findings speak directly to the policy argument that enforcement creates openings for U.S.-born workers. In their data, from January 2024 through November 2025, that reallocation does not appear. As they put it, the results are “consistent with employers reducing labor demand overall, including for jobs more often taken by U.S.-born workers.”