Where can I find annual ICE detention datasets and how to analyze them by status?
Executive summary
For annual ICE detention datasets, primary public sources include ICE’s own Enforcement & Removal Operations statistics and the OHSS (DHS) Key Homeland Security Measures detentions pages; independent cleaned, person-level releases and analysis are available from Vera’s ICE Detention Trends (data through June 10, 2025) and university-backed Deportation Data Project releases based on FOIA responses (covers Sept 2023–June 10, 2025) [1] [2] [3] [4]. Analysts commonly compare “book‑ins” (initial admissions), “book‑outs” (final releases), daily population counts, and detention‑stint records to tabulate status by outcome (released, removed, transferred, held) — Vera and Deportation Data Project documentation explain how to reconstruct status and length‑of‑stay using detention stints and book‑in/book‑out fields [3] [5] [4].
1. Where to get the raw annual and person‑level data — official and independent repositories
ICE’s public statistics page publishes biweekly and annual tables for arrests, removals, detentions and alternatives to detention and is the official starting point; ICE notes data fluctuate until a fiscal year is “locked” and affirms the integrity of what it posts [1]. DHS OHSS hosts Key Homeland Security Measures detentions data derived from ICE administrative records in the Enforcement Integrated Database (EID) and describes corrections and taxonomy changes across releases — useful for official annual aggregates and definitions [2]. For person‑level and researcher‑friendly datasets, Vera’s ICE Detention Trends provides detention history for every person in ICE custody from Oct 1, 2008 through June 10, 2025 and publishes code and technical appendices [3] [6]. The Deportation Data Project (UCLA/UC Berkeley/CILP) posts ICE data obtained via FOIA, includes processed/cleaned versions and documentation, and specifically released stints and arrest/detention records spanning Sept 2023–June 10, 2025 [4] [7].
2. Key variables and definitions to analyze detention “by status”
Analysts should work from these core constructs: book‑in (initial detention admission), book‑out (final release or disposition), detention stint (one facility stay), and daily population counts. “Status” can mean current custody status (detained vs. released), release reason (bond, order of supervision, deportation), or final outcome (removal/deportation vs. release). Vera’s documentation defines book‑ins, book‑outs and how multiple stints form a detention history and warns that dataset boundaries and coding changed across Dataset I/II/III — changes that affect counts and facility mapping [5] [6]. Deportation Data Project documentation lists table and field descriptions for arrests, detainers, detentions, encounters and removals useful for constructing status flags [4].
3. Practical analysis steps and pitfalls journalists must watch
Start by choosing your unit: per‑stint (each facility stay) vs. per‑person (unique detainee history). Use book‑in and book‑out timestamps to compute length of stay and derive current status at any date; aggregate book‑outs by disposition codes to tabulate outcomes. Beware that ICE public reporting sometimes uses a subset of detentions for annual “initial book‑ins,” a practice GAO criticized because it excludes tens of thousands and lacks transparent methodology — GAO recommended ICE report all detentions and explain exclusions [8]. Vera and Deportation Data Project warn of dataset boundary effects (date cutoffs and re‑coded facility IDs) that can create artificial jumps around October 1 or dataset change dates [3] [6].
4. Reconciliation and validation: authoritative vs. FOIA-derived sources
Official ICE/OHSS figures are authoritative for executive branch reporting but may omit records or apply internal filters; DHS/OHSS highlights that its SSOR is constructed from ICE’s EID and undergoes validation [2] [1]. FOIA‑obtained, researcher‑cleaned datasets (Vera, Deportation Data Project, academic teams) provide granular person‑level stints and cleaning code, enabling independent reproductions and corrections — Vera documents dataset merges and adjustments researchers made [3] [9] [4]. Journalists should cite both: ICE for official totals, and Vera/Deportation Data Project for person‑level checks and alternative counts.
5. What competing narratives to expect and how data supports them
Advocates and critics will use the same datasets to make very different claims: civil‑liberties groups and NIJC cite large annual totals and long stays to argue for an expanded, punitive detention system (e.g., reports noting hundreds of thousands detained over a year) while ICE emphasizes targeted custody for removals and public‑safety cases [10] [1]. Independent analyses show many detainees have no criminal convictions in certain periods and localities — reporting based on FOIA and academic work underpins those findings [4] [7]. Use person‑level stint records to test such claims (conviction flags, disposition codes), and always note limitations in ICE’s reported methods [8].
6. Quick reading list and reproducible resources to begin analysis
Pull official tables from ICE’s statistics portal and OHSS KHSM detentions pages for annual aggregates [1] [2]. Download Vera’s dashboard data and technical appendix (person‑level stints, codebook) and the Deportation Data Project’s data + documentation for FOIA‑derived records and cleaning scripts [3] [4] [9]. Cross‑check totals and state/facility breakdowns, and document any dataset edits or cutoff dates that could explain discrepancies [6] [5].
Limitations: available sources do not mention analysis code for every potential “status” flag you might want; inspect the Vera and Deportation Data Project codebooks for field names and conversion rules before publishing [3] [4].