73.6% of ICE Detainees Have No Conviction

Introduction

"The worst of the worst." DHS uses that phrase in the July 28 press release and the January 20 first-year recap, crediting deportations with driving the 2025 crime drop. Deporting those people, the administration says, is what made America safer in 2025. Then Cato Institute fellow David Bier got a hold of nonpublic ICE booking data: 73% of people booked into ICE custody in fiscal year 2026 had no criminal conviction at all, and only 5% had a violent one.

What "Worst of the Worst" Actually Means

The "no conviction" majority shows up in every independent dataset that's looked.

TRAC at Syracuse University reported that as of mid-November 2025, ICE was holding 65,135 people. Of those, 47,964 had no criminal convictions. That's 73.6%. By April 4, 2026, TRAC's data showed 70.8% of detainees had no conviction and no pending charges, meaning even the generous "involved with the criminal system somehow" definition didn't apply to seven in ten people locked up. The Deportation Data Project at UC Berkeley/UCLA, working from FOIA-obtained individual-level records, found that arrests of people with no criminal record increased 8.7x versus the Biden baseline. Street arrests went up 11x. Release rates for people without convictions dropped from 35% to 7%.

So how does DHS get to "70-75% criminal" in its press releases? By counting pending charges as if they're convictions. FactCheck.org broke down the methodology: about 36.5% of ICE arrestees have any conviction at all (including misdemeanors and traffic charges), plus another 29.8% with pending charges. Add those two together, round up, and you get the 70% number on the press release. The administration's version of the statistic puts a person fighting a dismissed disorderly conduct charge into the same bucket as a convicted murderer, and a lot of those pending charges end up dismissed.

Under a conviction-only standard, only about 7% of people booked by ICE have serious or violent convictions. Aaron Chalfin, the University of Pennsylvania criminologist, reviewed ICE's own data and landed on a similar number: about 13% had any prior conviction for violent or property crime. His phrase for the population being detained: "a population at lower risk of serious offending than those typically arrested by local police."

The Decline That Predates the Policy

Crime did fall in 2025. Council on Criminal Justice's Year-End 2025 Update found homicide down 21% from 2024, robbery down 23%, motor vehicle theft down 27%, carjacking down 43%. The U.S. murder rate is projected to land near 4.0 per 100,000, possibly the lowest in 125 years. DHS pointed at that report on July 28, 2025 and claimed credit for the cities CCJ analyzed.

Buried in the same document DHS cited is this line: "Any assertive claims about the influence of specific policy interventions, such as National Guard deployments and increased immigration enforcement … should be supported by robust research designs." The authors attribute the decline to "a complex tangle of broad social and technological changes and direct policy interventions," and explicitly do not credit deportations. DHS pulled the favorable headline figure and ignored the methodological warning sitting one paragraph away.

Then there's the timeline. The FBI's August 5, 2025 release showed national homicide fell 14.9% in 2024, covering the calendar year before Trump's second term began. Crime analyst Jeff Asher told The Trace that the decline "started in earnest in 2023, more than a year before Trump took office." Homicides peaked at 26,031 in 2021 and had already fallen to 20,162 by 2024. Cities the administration cited as enforcement wins, including D.C., Baltimore, Detroit, and St. Louis, were already trending toward 2014-era lows in 2024, before the deportation surge. To accept the administration's causal claim, you'd have to believe that a policy launched in January 2025 retroactively caused declines that started two full years before it existed.

The CCJ caveat is the cleanest example of how a public claim gets manufactured from careful research. CCJ's researchers flagged the limits of their own findings inside the report itself. DHS took the headline figure and attached it to a policy claim the paper's own authors said required robust research designs to support.

The Closest Study We Have, and What It Found

There's no peer-reviewed analysis of the 2025 deportation surge yet (that takes two to three years). But there is a peer-reviewed study of the program most directly analogous to it: Secure Communities, the Obama-era interior enforcement push that ramped up ICE detentions during 2008–2013. NBER Working Paper 32109 by Felipe Gonçalves, Elisa Jácome, and Emily Weisburst ran a difference-in-differences analysis using the National Crime Victimization Survey, which captures roughly 240,000 respondents annually and includes both reported and unreported crimes.

Here's what the paper found. Secure Communities increased ICE detentions by 54%. Total reported crime didn't budge. But Hispanic victimization rose by 0.15 percentage points monthly against a 0.9-point baseline, a 16% relative increase, while Hispanic crime reporting fell by 9 percentage points, a 30% decline from a 33-point base. The reported number stayed flat because more crime was happening against Hispanic residents while less of it was making it to police. The paper's estimate, on page 3, is that "Secure Communities resulted in 1.3 million additional crimes against Hispanics in the two years following program activation." Predominantly property crime, per the working paper itself, with smaller and less precisely estimated increases in violent crime.

That number is an econometric projection rather than a counted total of incidents, and the paper describes it as a credible estimate from a well-identified design. Two earlier peer-reviewed studies came at the same program from different angles. Treyger, Chalfin, and Loeffler (2014) in Criminology and Public Policy analyzed monthly crime rates in 335 U.S. cities and 3.6 million city-months of data from 2000 to 2011, specifically designed to isolate the effect of Secure Communities activation on crime rates. They found "no statistically discernible effects of activation on any category of crime under analysis" — not a small effect, an undetectable one. Miles and Cox (2014) in the Journal of Law and Economics used a different identification strategy, comparing counties that received Secure Communities before and after a staggered federal rollout across 47 states. Same conclusion: "The Secure Communities program has had no observable effect on crime rates."

Who Benefits

GEO Group and CoreCivic donated a combined $2.779 million to Trump's 2024 campaign, inaugural committee, and related entities, according to FEC records compiled by CREW. GEO put in $1.963 million; CoreCivic, $816,600. Each cut a $500,000 check to the inaugural committee. GEO Group CEO George Zoley personally bought 250,000 shares of his own company's stock in August 2024 at $12.28 each, a roughly $3 million position taken before the election. The day after Trump won, GEO Group stock rose 41% and CoreCivic rose 29%, and Zoley's pre-election position roughly doubled in value overnight.

What makes the bet pay off is detention bed demand, and that's exactly what mass interior enforcement produces. The "One Big, Beautiful Bill Act" allocated $45 billion for ICE detention contracting. According to OpenSecrets, GEO Group and CoreCivic together took $2.753 billion in ICE contract obligations in 2025 ($2.1 billion to GEO, $653.5 million to CoreCivic). GEO Group reported a record $254 million in 2025 profit, roughly 700% above 2024's $31.9 million. CoreCivic reported $116.5 million, a 70% jump, with Q4 2025 ICE revenue more than doubling year-over-year.

So $2.779 million in donations turned into $2.753 billion in contract obligations, a roughly 1,000-to-1 return that produced the kind of profit growth a specific policy environment has to manufacture. And the people filling those beds are the same population the administration's framing implies it's not targeting: 5% with violent records, most with no conviction at all.

The Causal Claim Doesn't Trace

Crime in 2025 fell. The question is what caused it, and the available evidence says it wasn't the deportation campaign. The decline started in 2023. Most of the people detained had no criminal record. The peer-reviewed research on the closest analogous program found 1.3 million additional crimes against Hispanic communities, mostly property crime, driven by chilling effects on police cooperation. The two private prison companies that bankrolled the 2024 effort posted record profits filling beds with the population the administration's own rhetoric implies it isn't targeting.

The open question is whether the 2025 crime trend continues into 2026 once the comparison year stops being the pandemic spike, and whether the chilling effect documented in the Secure Communities literature shows up in next year's National Crime Victimization Survey data, which is the only place a 2025 reporting decline would actually surface. The administration is making a causal claim. The data to test it is already on the way.