A new analysis of US federal immigration enforcement shows arrests rose under both Trump presidencies, but the share of those arrested who had actual criminal convictions fell sharply. The study, published in the Proceedings of the National Academy of Sciences, argues there is a disconnect between the public narrative of targeting criminals and what happens on the ground.
US Immigration and Customs Enforcement, known as ICE, is the federal agency responsible for enforcing immigration laws within the interior of the United States. During the second Trump term, officials announced an expansive enforcement push, repeatedly citing public safety and the need to remove individuals who commit crimes as the primary motivation.
There has been little real-time analysis of how ICE conducts interior arrests during periods of aggressive expansion. The research, led by Chloe N. East of the University of Colorado, Boulder, and Caitlin Patler of the University of California, Berkeley, aimed to map out exactly who is being targeted, where arrests occur and what methods are used.
To understand these patterns, the researchers analysed administrative records from the Deportation Data Project covering more than 1.5 million ICE arrests made between October 2015 and March 2026. Because enforcement levels dipped to a ten-year low during the Biden administration due to the COVID-19 pandemic and changes at the border, the team focused on two windows around the two Trump inaugurations: roughly December 2015 to March 2018, and December 2024 to March 2026.
The team looked at the total number of arrests, the geographic region where each arrest took place, and whether the arrested person had a criminal conviction of any kind. They also categorized the arrests by the specific method ICE used, identifying Law Enforcement Agency arrests and Community Arrests.
Law Enforcement Agency arrests occur when ICE takes custody of someone already held in a local or state jail or prison, while Community Arrests involve ICE officers apprehending people directly in their neighbourhoods, homes or workplaces. This distinction allowed the researchers to see whether shifts in who was arrested were driven by changes in how arrests were carried out.
The analysis showed arrests increased massively under both administrations, but the surge was especially pronounced in the second term. In the year before the first inauguration, ICE averaged 304 arrests per day. This rose to 436 per day in the first 14 months of the first term, a 43 per cent rise.
Before the second term, arrests were at a similar baseline of 304 per day. But between January 2025 and March 2026, the average jumped to 952 arrests per day, a 214 per cent rise on the previous year.
As the volume of arrests grew, the share of arrestees with a criminal conviction declined. During the second term, the conviction rate among those arrested fell to 34 per cent, near the lowest point observed in the ten-and-a-half-year study period.
There was also a dramatic shift in how ICE operated on the ground during the second term. Community Arrests rose from 17 per cent of all apprehensions in the year before the second term to 47 per cent between January 2025 and March 2026, a substantial tilt towards direct neighbourhood enforcement that did not occur during the first term.
Researchers tested whether the change in tactic explained much of the fall in conviction rates. They found that the shift in method accounted for roughly a quarter of the decline in the second term.
The remaining three-quarters of the drop occurred because ICE was arresting fewer people with criminal records within both categories. Specifically, the conviction rate fell by 10 percentage points during Community Arrests and 16 percentage points during Law Enforcement Agency arrests, indicating a broad shift in who was being targeted beyond the change in tactics.
The geographic pattern varied across ICE’s 25 regional jurisdictions. In the second term, El Paso, Miami, New Orleans and San Antonio saw spikes in jail-based arrests, while Boston, Chicago, Newark, Seattle and all jurisdictions in California and New York saw spikes in Community Arrests. The researchers note this aligns with the administration’s stated aim of targeting sanctuary jurisdictions, though the decline in the share of arrestees with criminal convictions was nearly universal across the country despite these regional differences.
As with all research, several caveats apply. The data rely on ICE’s administrative records, which capture only individuals formally processed into the system. The analysis focuses on interior enforcement, excluding border apprehensions. The researchers also cannot determine why the conviction rate fell specifically within jail-based arrests; possible explanations include shifts in ICE prioritisation or changes in local policing practices, warranting further study.
The study, “ICE arrests, 2015–2026: Variation in targeting, method, and geography,” is authored by Chloe N. East, Caitlin Patler and Elizabeth Cox.
