What percentage of ice arrests are violent criminals
Executive summary
Public statements from DHS and ICE insist a large majority of their arrests are of “criminal illegal aliens,” with officials citing a 70% figure that counts people with convictions or pending charges [1] [2], but independent analyses using ICE booking and detention data find that a small single‑digit share of those booked had violent convictions — roughly 5–8 percent — while a plurality or majority had no criminal conviction on record [3] [4] [5].
1. Why the headline numbers diverge: convictions vs. charges vs. encounters
DHS/ICE public messaging foregrounds a 70% statistic framed as arrests of people with convictions or pending criminal charges, a line repeatedly used in agency press releases and by spokespeople [1] [6] [2], while leaked or compiled ICE booking data analyzed by outside researchers (Cato Institute, TRAC, Deportation Data Project) show that by late 2025–early 2026 roughly 65–74 percent of people booked into ICE custody had no criminal conviction, and only about 5–8 percent carried violent convictions [3] [4] [7]. The difference arises from counting pending charges as “criminal,” using local jail holds and referrals to inflate interior arrests, and pooling encounters from Border Patrol and ICE that have distinct legal contexts [3] [8] [9].
2. What the best available numbers say about violent convictions
Independent analyses of ICE booking data indicate that people with violent convictions are a small minority of ICE detainees: the Cato Institute’s review found about 5 percent had violent convictions in one recent fiscal snapshot and elsewhere summarized as 8 percent in similar datasets depending on definitions and cutoffs [3] [5]. Reporting by TRAC and other data aggregators supports the broader finding that most people in ICE custody lacked felony violent convictions and that many convictions recorded were for lower‑level offenses or traffic violations [4] [10].
3. How official counts can be accurate yet misleading
ICE’s 70% claim can be factually true if the metric counts pending criminal charges alongside convictions and includes people wanted overseas or subject to separate criminal warrants; DHS spokespeople explicitly note the figure combines convictions and pending charges and excludes some categories like foreigners wanted for violent crimes abroad [2]. Agency press releases highlight high‑profile arrests of murderers, rapists and traffickers to reinforce that framing, but those case vignettes are selective and do not by themselves establish the overall percentage of violent offenders in the broader arrest pool [11] [12].
4. The policy and political incentives behind competing narratives
The Department of Homeland Security’s repeated “worst of the worst” framing serves recruitment and policy messaging goals and is amplified across DHS press releases [1] [6] [12], while researchers and advocacy groups spotlight the rise in arrests of non‑convicted migrants to argue the administration’s enforcement is sweeping and often targets noncriminal immigration cases [3] [8]. Both sides use slices of ICE’s complex operational data to support policy arguments; independent datasets (Deportation Data Project, TRAC, leaked ICE bookings) help journalists and researchers reconcile the differences but also reveal limits in public transparency [9] [4].
5. Limits of the reporting and what remains uncertain
Public and leaked datasets differ in scope and timing: ICE’s public releases, press vignettes, and internal booking files cover different populations (border encounters, interior arrests, local jail transfers), and many data snapshots are through mid‑October or other cutoffs, meaning year‑to‑date percentages shift month to month [9] [4]. The available reporting does not permit a single, definitive multiplier of “violent percentage” that applies to every ICE action type; instead, the best current evidence from independent analyses places the share of detainees with violent convictions in the low single digits (about 5–8 percent), while the agency’s broader 70% figure reflects a different denominator that includes pending charges and non‑conviction encounters [3] [2] [5].