What share of ICE detentions in 2025 were explicitly coded as 'reentry after deportation' in FOIA datasets?
Executive summary
The FOIA datasets released by ICE to the Deportation Data Project and reposted by researchers do include categorical fields that can identify “reentry after deportation” as a reason for arrest or removal, but none of the publicly cited reports or dataset release notes in the provided sources report a single, verified share of all 2025 ICE detentions explicitly coded that way; available documentation instead emphasizes variable definitions, missing tables, and shifting coding practices that prevent a confident, sourced percentage from being stated here [1] [2] [3]. Analysts who have used these FOIA releases note changes and gaps in the 2025 data that make simple headline statistics unreliable without re‑processing the raw tables [4] [2].
1. What the FOIA data include and how “reentry after deportation” appears in them
ICE’s FOIA production—now published and curated by the Deportation Data Project—contains multiple tables (encounters, arrests, detentions, detainers, removals) and documentation describing fields and codes that track charges and categories such as illegal reentry/reentry after deportation, meaning the variable needed to flag those cases exists in the released data schema [2] [1]. The Deportation Data Project and downstream users have made both raw and processed versions of the detention and arrest tables available for analysis, explicitly recommending that researchers cite the government FOIA production when using these fields [1] [5].
2. Why a single, sourced share for 2025 is not present in the reporting
Multiple reporting and research pieces built from the FOIA dumps describe aggregate trends, spikes in detentions, and shifts in the criminality composition of the detained population—but none of the pieces in the supplied set publish a definitive percentage of 2025 ICE detentions that are explicitly coded as “reentry after deportation.” The Deportation Data Project release notes and preprint warn that parts of the release are incomplete or inconsistent (for example, removals tables with errors and shifting coding practices in 2025), and researchers have repeatedly noted data gaps and coding changes that complicate simple tallies [3] [2] [4]. That caveat appears across the project’s documentation and in downstream analyses that rely on reprocessing the raw tables rather than accepting an agency summary statistic [1] [6].
3. What related analyses do show (and what they do not show)
Analyses drawing on the FOIA data have reported that a large share of the 2025 increase in ICE book‑ins came from people without criminal convictions and that immigration convictions (which can include illegal entry or reentry) figure prominently in the detained population; for instance, Cato’s takeaways from leaked datasets note that immigration convictions are as common among detainees as violent convictions in a recent slice of data—but that analysis does not translate into a precise national share coded specifically as reentry after deportation across all 2025 detentions in the official FOIA tables [7]. Vera and other research dashboards visualize detention‑stint level data derived from the FOIA releases, but their public summaries focus on counts and trends rather than publishing a single, FOIA‑sourced percentage for the reentry category [8].
4. The practical path to the precise figure and the obstacles
To produce a rigorously sourced share requires direct analysis of the relevant FOIA detentions and arrests tables (or the processed one‑row‑per‑stay detentions dataset) to count records where the charge/coding field equals the reentry/reentry‑after‑deportation code, then divide by total detentions for the chosen timeframe; the Deportation Data Project provides the raw and processed files needed to perform that computation but its release notes caution about missing identifiers, a removals table with errors, and coding shifts in mid‑2025 that can change denominators and counts if not handled carefully [5] [2] [3]. In short, the sources document the data and the methodological hurdles but do not themselves publish the final percentage for 2025, so declaring a specific share here would exceed what the provided reporting supports [1] [2].