How have ICE arrest and removal numbers changed year‑by‑year since 2003 and what counting rules affect those totals?
Executive summary
From available government and independent analyses, ICE interior removals fell substantially over the past decade while arrests and detentions have fluctuated with policy shifts: removals from the interior averaged roughly 155,000 per year in FY2009–16, dropped to about 81,000 in FY2017–20 and to roughly 38,000 for FY2021–24 even as border removals rose dramatically to 224,000 in FY2024 [1]. Arrest counts show variation—declining from 2019 through 2021 then rising in 2022—and different data collections and counting rules meaningfully change year‑to‑year totals [2] [3].
1. What the trends look like, at a glance
Interior removals (the people ICE removes after interior arrests or final orders executed in the interior) show a long‑term downward trajectory from the high years of the early 2010s to much lower totals after 2016, with Migration Policy Center reporting averages of ~155,000 in FY2009–16, ~81,000 in FY2017–20, and ~38,000 in FY2021–24 [1]. By contrast, removals tied to border encounters increased in recent years, reaching 224,000 in FY2024 as ICE shifted resources toward border processing [1]. Arrests and bookings present a more jagged picture: GAO’s analysis found arrests varied from 2019–2022, falling through 2021 then increasing in 2022 (from 133,541 in 2019 to 154,204 in 2022), illustrating how enforcement emphasis and operational tempo can change rapidly [2].
2. Why “year‑by‑year” numbers can be misleading: event vs. person counts
Federal reporting often counts immigration “events” not unique people, meaning an individual encountered, arrested, or detained multiple times in a reporting period can be counted repeatedly, a practice described in DHS’s monthly tables and OHSS guidance [4] [5]. The Office of Homeland Security Statistics explicitly notes the unit of measurement is immigration events and that people encountered more than once are counted multiple times, a rule that inflates annual tallies relative to unique‑person measures [4]. Independent analysts and projects (e.g., Deportation Data Project) have found duplicated rows and booking‑same‑day repeats in ICE’s administrative arrest tables, which required cleaning to avoid overcounts [6] [7].
3. Definitions matter: administrative arrests, removals, returns, and book‑outs
Different labels reflect different processes: Journalist’s Resource emphasizes that the ICE ERO administrative arrest dataset records administrative arrests—detentions by ICE in the interior—and should not be conflated with criminal arrests or enforcement by other ICE components [3]. OHSS and ICE distinguish “removals” (formal deportations) from “returns” or “repatriations,” and they separately track “book‑ins” and “book‑outs,” with the KHSM and OHSS explaining that these operational distinctions affect counts and trend interpretation [5] [4]. GAO found ICE’s public reports understated total detentions because some temporary facility bookings were excluded from published detention totals, reinforcing the point that methodological choices shift the headline numbers [2].
4. Data quality, FOIA releases, and independent reconstructions
Much of what researchers use comes from ICE’s own releases plus FOIA‑provided datasets; the Deportation Data Project hosts FOIA data and links that permit pathway tracing but warns that some datasets omitted removals/encounters due to errors [6]. Independent groups and journalists have had to de‑duplicate and reconcile records—Prison Policy and others reported overcount issues and regional concentrations of arrests, noting nearly half of some arrest flows originated from local jails [7]. GAO’s architecture review recommends ICE improve transparency about methodology and include all detentions in public reporting, a recommendation DHS did not concur with but which highlights unresolved data limitations [2].
5. Competing narratives and what the official sources say
ICE and DHS public statements frame recent operations as concentrated on criminal “worst of the worst” removals and tout publicity around criminal arrests, underscoring an enforcement narrative [8] [9]. Independent analyses and Migration Policy’s explainer provide the countervailing numeric story—declining interior removals over a decade with a large reallocation of removals to border processing—while watchdogs stress that counting rules, event‑versus‑person measures, facility exclusions, and data duplications all materially shape the year‑by‑year story [1] [6] [2].