ICE arrests, 2015–2026: Variation in targeting, method, and geography Authors: Chloe N. East, Caitlin Patler, Elizabeth Cox Published: 2026 ABSTRACT We analyze and decompose administrative data on all 1.6 million US Immigration and Customs Enforcement (ICE) arrests from October 2015 to March 2026. Our results reveal that the reality of immigration enforcement diverges sharply from the public narrative that ICE arrests are necessary to protect public safety by removing people who commit crimes: Although arrests spiked at the outset of both Trump presidencies, the share of arrested individuals with criminal convictions fell significantly, with especially marked declines in 2025. A shift in ICE tactics partly explains this pattern, but even conditional on tactic, the share with a criminal conviction declined as arrests rose. Moreover, we find substantial geographic heterogeneity in ICE methods, despite nearly universal declines in the criminal conviction rate. Our findings provide important evidence for policymakers, the general public, and researchers studying this period of immigration policy. The second Trump administration has launched the most expansive immigration enforcement effort in recent history, citing public safety as a central justification (1). This rationale sits in tension with a large, interdisciplinary literature finding that heightened immigration enforcement does not meaningfully impact crime (2–5). Throughout 2025, reports documented striking increases in Immigration and Customs Enforcement (ICE) arrests and plummeting rates of arrests of individuals with criminal convictions (6, 7), yet few studies have examined the mechanisms driving these patterns, or how they may vary over time and across geography. Using data on all 1.6 million administrative ICE arrests from October 2015 through March 2026, this Brief Report analyzes the scale, methods, and geographic scope of ICE arrests during the second Trump term, as compared to the prior decade, including the first Trump term. Given the public focus on immigration enforcement and public safety, we decompose changes in the rate of arrests involving individuals with criminal convictions into two components: shifts in ICE’s methods and changes in enforcement prioritization within methods. We do this for the entire country, as well as across ICE’s 25 Areas of Responsibility. We provide systematic, mechanism-focused analyses of ICE arrests under the second Trump term. As ICE activity and funding continue to expand (8), real-time documentation of these patterns is essential. By documenting variation in where, how, and for whom interior enforcement has expanded, our analyses have direct implications for the communities subject to arrest, for democratic accountability over executive enforcement authority, and for emerging research on the mechanisms and consequences of contemporary immigration policy. Materials and Methods We use administrative data from the Deportation Data Project on the universe of Immigration and Customs Enforcement (ICE) Enforcement and Removal Operations (ERO) administrative arrests from October 1, 2015, to March 10, 2026 (n=1,565,651), the most recent date for which data are currently available. Data are deidentified prior to release. We analyze the entire period and then zoom in on the periods immediately before and after each Trump inauguration: December 3, 2015–March 10, 2018 (n=294,717), for Term 1 and December 3, 2024–March 10, 2026 (n=520,805), for Term 2 (414 d before and after each inauguration). Our data contain information about the location and method of each arrest, as well as whether the arrested individual had a criminal conviction of any kind. We categorize method as criminal Law Enforcement Agency collaborative arrests (LEA), Community Arrests (CA), and Other arrests.* LEA arrests occur when ICE takes custody of an individual already in criminal LEA custody, while CA are arrests of individuals made directly by ICE and not already in custody. We examine national trends, as well as variation across ICE’s 25 ICE Areas of Responsibility (AORs). We conduct a decomposition analysis to examine how much of the total change in the criminal conviction rate before and after each inauguration was due to changes in methods (the “Method Effect”), or to changes in targeting within methods (the “Within-Method Effect”). SI Appendix provides further detail on our data, measures, and analytic strategy. Results Scale of Arrests and Criminal Conviction Rate. The Top panel of Fig. 1 plots national trends in total daily ICE arrests per 100,000 noncitizens measured in the American Community Survey (black solid line) and the criminal conviction rate (green dashed line) among those arrested, from October 2015 to March 2026.† Total arrests by criminal conviction status and method, 2015–2026. Notes: Uses ICE arrest data from the Deportation Data Project covering October 2015 to March 2026. Vertical dashed lines show inauguration dates. Before proceeding with our discussion of Trump Terms 1 and 2, we note the uniqueness of the Biden Administration. When Biden assumed office, arrests were at a ten-year low, reflecting substantial disruptions to ICE’s operations due to the COVID-19 pandemic (9). However, starting in spring of 2021, arrests began to rise, with a stark decline in the criminal conviction rate–changes that reflect ICE’s new, temporary role aiding Customs and Border Protection in processing asylum-seekers at the Southern U.S. border, lasting through FY2023 (10, 11). Given these irregularities, we focus our discussion on Terms 1 and 2. The scale of arrests in Term 2 is unprecedented in the past decade, and as compared to Term 1. In the year before the Term 1 inauguration, ICE averaged 304 arrests per day; this rose to 436 per day in the first 14 mo of Term 1, a 43% increase. The same pre-inauguration baseline (304 arrests per day) preceded Term 2, but arrests surged to an average of 952 per day between January 2025 and March 2026—a 214% increase. The criminal conviction rate has correspondingly fallen to 34% in Term 2, near its lowest point in the 10.5-y period we analyze. A consistent pattern emerges over the full period: Increases in arrests correspond to decreases in the criminal conviction rate. We estimate that each one-unit increase in arrests per 100,000 noncitizens is associated with a 19 percentage point (“pp”) decline in the criminal conviction rate over the whole period. However, this relationship is weaker in Term 2 (13 pp) compared to the prior 10 y, suggesting the increase in scale does not fully explain the drop in the criminal conviction rate in Term 2, which our decomposition analysis addresses further below. The Role of Arrest Methods. The Lower panel of Fig. 1 plots ICE arrests by method, revealing a striking shift in the method composition in Term 2. In particular, CA increased from 17% in the year prior to Term 2, to 47% from January 20, 2025 to March 10, 2026; no shift of such magnitude occurred during Term 1. While the percent with a criminal conviction is generally lower under the CA method compared to LEA arrests, the share of individuals with convictions dropped dramatically within both methods during Term 2. Our decomposition reveals that the percent of ICE arrests of individuals with a criminal conviction fell by 14.2 pp during Term 1 and 20.6 pp during Term 2. While shifts in arrest method explained only 0.2 pp of the decline in Term 1, they became more substantial in Term 2, explaining 5.1 pp. Thus, virtually all of the Term 1 change (14.0 of 14.2 pp) and about three-quarters of the larger Term 2 change (15.5 of 20.6 pp) are attributable to within-method change. Indeed, the conviction rate fell 10 pp within CA and 16 pp within LEA arrests in Term 2.‡ Geographic Heterogeneity. The final component of our analysis examines whether these patterns vary across geography. Fig. 2 plots arrests by method and term for each AOR and reveals meaningful variation over time and AOR in how the Term 2 ICE arrest surge has unfolded.§ During Term 1, increases in arrests were driven primarily by sustained growth in LEA arrests (pink line). During Term 2, several AORs experienced steep spikes in LEA arrests (red line), including those headquartered in El Paso, Miami, New Orleans, and San Antonio. In contrast, a distinct set of AORs experienced sharp spikes in CA in Term 2 (dark blue line), including those headquartered in Boston, Chicago, Newark, Seattle, St. Paul, and Washington DC, as well as all AORs in California and New York, consistent with the administration’s stated targeting of so-called sanctuary jurisdictions (12). This geographic variation in the use of CA was not nearly as pronounced in Term 1 (light blue line). Arrests per 100,000 noncitizens by AOR and method, relative to inauguration. Notes: Uses ICE ERO administrative arrest data from 2016, 2017, 2024, 2025, and 2026. Vertical red lines show inauguration dates. To obtain a rate arrests are divided by the number of noncitizens in the AOR (from the 2015 ACS for Term 1 and the 2023 ACS for Term 2), and multiplied by 100,000. To align the ICE classifications, the San Antonio AOR includes the Houston and San Antonio AORs in 2016–2017 and Harlingen, Houston, and San Antonio AORs in 2024–2026. Despite AOR-level variation in enforcement strategy, nearly all AORs experienced declines in the percent of arrested individuals with criminal convictions. We examine this in Fig. 3, which decomposes the change in conviction rate by AOR: Hollow dots show the total percentage point change, with Method and Within-Method portions in yellow and light blue for Term 1 and orange and dark blue for Term 2. Term 2 changes ranged from a +5 pp increase to a −45 pp decline, a much wider spread than Term 1’s −6 to −25 pp. Moreover, within-method change accounts for the majority of the decline in roughly three-quarters of AORs in Term 2, consistent with the national finding. The largest Term 2 within-method declines include San Diego (−34 pp), Washington DC (−28 pp), and Philadelphia (−27 pp). Decomposition of change in criminal conviction rate by AOR, Terms 1 and 2. Notes: Uses ICE ERO administrative arrest data from 2024, 2025, and 2026. Reports the percent of the pre–post inauguration change in percent with a conviction explained by method or within-method conviction rate. To align the ICE classifications, the San Antonio AOR includes the Houston and San Antonio AORs in 2016–2017 and Harlingen, Houston, and San Antonio AORs in 2024–2026. Method-of-arrest shifts also varied far more across AORs in Term 2 (ranging from −39 to +9 pp) than in Term 1 (ranging from −7 to +1 pp). These cross-AOR differences track the LEA-vs.-CA enforcement patterns documented in Fig. 2: AORs with sharp CA spikes show pronounced Method-of-arrest contributions (San Francisco −17 pp, Washington DC −12 pp, Boston −11 pp, and Los Angeles −10 pp), consistent with a shift toward arrest methods that historically prioritize fewer people with criminal convictions.¶ The geographic analysis thus reinforces two of our central findings; despite substantial AOR-level heterogeneity in ICE methods in Term 2, the share of arrested individuals with a criminal conviction nearly universally declined across the country due to within-method declines. Discussion Our results show a fundamental mismatch between the justifications for intensified immigration enforcement and the empirical patterns of that enforcement’s implementation. The majority of those arrested by ICE in the second Trump administration had no criminal conviction at arrest. This has been produced by three mutually reinforcing forces: Arrests have scaled to historically high levels, creating a mechanical downward pressure on the conviction rate; the composition of arrests has shifted toward CA, which yield lower conviction rates; and within nearly every method and AOR, ICE is arresting a population with lower criminal conviction rates. The Method Effect accounts for roughly 25% of the post-second-inauguration decline, and the dominant force, accounting for 75%, is a broad change in who is arrested, independent of method. While the geographic impact in Term 2 is more heterogeneous than Term 1, falling conviction rates due mostly to within-method changes is nearly universal in Term 2. Our findings provide direct evidence that intensified immigration enforcement increasingly contradicts its stated public-safety rationale. Supplementary Material Appendix 01 (PDF)