What are the differences between ICE administrative arrests, custodial arrests and removals in official data sets?
This fact-check may be outdated. Consider refreshing it to get the most current information.
Executive summary
Administrative arrests are ICE’s recorded interior “apprehensions” — a broad, civil category that includes both at‑large arrests in the community and arrests that occur after local or federal custody transfers; custodial arrests are a subcategory of those administrative arrests that happen when local, state or federal jails transfer people to ICE; and removals are the separate outcome data that record deportations actually carried out by ICE, but removals are captured differently, sometimes incompletely, across ICE public files and third‑party reconstructions (Deportation Data Project) [1] [2] [3].
1. What “administrative arrests” mean in ICE datasets
ICE labels the events counted in its public “arrests” tables as administrative arrests — interior detentions of noncitizens alleged to be removable — and these data are produced by ICE’s Enforcement and Removal Operations (ERO) for public use, with ICE asserting data integrity but warning that counts can change until year‑end “lock” [4] [1]. Analysts stress that these records reflect administrative custody events, not criminal convictions, and they exclude some Border Patrol apprehensions or cases that never enter ICE detention unless linked elsewhere in DHS reporting [1] [5].
2. Custodial arrests: arrests that flow from the criminal justice pipeline
Custodial arrests are arrests coded as occurring in custodial settings — primarily local jails and other lock‑ups — where ICE takes custody of people already held by local, state or federal authorities; historically and in recent years most ICE arrests have been custodial, driven by partnerships and detainer or transfer practices that feed ICE from jail rosters [6] [7] [8]. Reporting changes and aggregation choices by ICE — for example, collapsing Criminal Alien Program entries under “Custodial Arrests” in mid‑2025 — have reduced the granularity available to outsiders, and duplicative rows or transfer‑bookings complicate counts, so researchers often treat custodial arrests as a combined “local jails or other lock‑ups” category and clean duplicates before analysis [8] [5].
3. At‑large vs custodial within administrative arrests (how they differ operationally and in the data)
At‑large arrests are administrative arrests made in the community (homes, workplaces, check‑ins), whereas custodial arrests happen in a controlled, detention setting after another agency has arrested someone; at‑large arrests are more resource‑intensive and riskier for ICE, so custodial arrests historically outpace at‑large arrests and are easier to scale — a distinction ICE and public analysts document across ERO reports [2] [9] [6]. Data tables often tag “arrest method” or “arresting agency,” but changes in reporting and omitted temporary‑facility bookings mean public tallies can undercount or misclassify some events [1] [10].
4. Removals: the separate outcome dataset and its limits
Removals in ICE and Deportation Data Project files are the records of deportations actually carried out by ICE, and researchers typically identify removals by matching depart dates in arrests or by final “Remove” release reasons in detentions data; however, removals can occur without an ICE arrest or ICE book‑in (for example, Border Patrol expedited removals), and ICE’s published removals files have known quality issues and gaps that third‑party projects caution about [3] [5]. The Government Accountability Office has flagged that ICE’s public reporting understates some detentions and that removals have fallen even as arrests rose in certain years, illustrating how these datasets measure different stages of enforcement (arrest → detention → removal) imperfectly [10].
5. Practical implications for interpreting official counts
When reading ICE’s “arrest” numbers, treat them as administrative events that bundle distinct operational pathways: custodial arrests largely reflect cooperation with jails and produce higher volumes, while at‑large arrests are fewer and operationally different; removals are downstream and require separate matching logic to count reliably, and analysts must account for reporting changes, duplicate records, and exclusions that ICE and watchdogs (Deportation Data Project, GAO) have documented [8] [5] [10]. Alternative viewpoints exist: ICE asserts its public tables are authoritative for long‑term use, while independent researchers and news outlets warn that policy changes and dataset quirks can distort trends if not corrected for [4] [1] [11].