How has ICE’s public dashboard reporting differed from FOIA releases, and what corrections or coding changes has ICE acknowledged for 2025 data?
Executive summary
ICE’s publicly visible statistics and its FOIA-produced datasets have diverged in important ways: independent researchers obtained detailed, individual-level ICE records via FOIA that reveal missing or differently coded fields compared with what ICE’s public releases and dashboards show [1] [2]. Analysts documenting the 2025 period report both large gaps in location data and explicit shifts in how ICE categorized arrests — changes that have prompted data corrections, withheld tables (removals/encounters), and calls for corrected versions [3] [2] [1].
1. What the FOIA releases contain that the public dashboard often does not
Groups that litigated for ICE records obtained individual-level spreadsheets through FOIA that include granular arrest, detainer and detention rows and allowed independent reprocessing and analysis — releases the Deportation Data Project says reproduce ICE’s original files but also require downstream processing to be usable for researchers [1] [4]. Those FOIA-produced datasets have been the basis for reporting that documents trends by location and program type that are not easily visible on aggregated public dashboards, and in some releases the deportations/removals or encounter tables were omitted because of suspected errors, leaving gaps FOIA recipients and analysts have sought to clarify [2] [5].
2. Concrete discrepancies and missing elements flagged by researchers
The Deportation Data Project repeatedly noted that ICE’s FOIA data releases omitted a reliable removals/encounters table or truncated older removals because of “significant errors,” forcing researchers to rely on partial tables or older exported files [5] [2]. Independent analysts also found that ICE’s datasets contained a “large number of cases with incomplete location information,” which undermined place-based analysis and differs from the impression given by some public summaries that emphasize geographic patterns [3]. Those omissions and incompleteness have produced materially different tallies depending on which ICE file or dashboard one uses [2] [3].
3. Coding changes in 2025 that altered how arrests were categorized
Researchers documented a clear coding shift in mid-2025: ICE began collapsing arrests that had previously been reported under specific Criminal Alien Program categories into a broader “Custodial Arrests” category, obscuring whether cases stemmed from local jails, state prisons, or federal custody transfers — a change that analysts say makes longitudinal comparisons harder and that ICE’s released files reflect [3]. The Prison Policy Initiative’s analysis states that as of the last week of July 2025 ICE stopped reporting various local, state, and federal Criminal Alien Program arrest types separately and instead reported many of those cases under the generic label “Custodial Arrests” [3].
4. Data corrections, withheld tables, and ICE’s response as documented
Several FOIA-driven releases explicitly withheld or excluded tables — most notably removals and encounters — because the agency or recipients flagged potential data errors, and the Deportation Data Project has been seeking corrected versions of those omitted tables [2] [1]. Reporting and organizational summaries note that ICE provided updated files covering arrests, detainer requests, and detentions through mid‑October 2025, but cautioned users that some tables were incomplete or excluded while corrections are sought [1] [2]. Public-facing ICE FOIA guidance and weekly DHS/ICE FOIA reports document the administrative pathway for corrections and releases but do not, in the available snippets, lay out a comprehensive technical correction log for 2025 [6] [7].
5. Litigation, transparency context, and limits of the public record
Courts have forced ICE to produce machine-readable records with substituted identifiers in prior cases — for example, judges found ICE must provide unique anonymous identifiers rather than withhold underlying keys like A‑numbers — which shaped the form of FOIA datasets and enabled the type of cross-stage tracking researchers achieved [8] [9]. At the same time, watchdogs and newsrooms report slow FOIA responses, selective withholding and opaque redactions that complicate efforts to reconcile dashboard tallies with FOIA files, and litigation continues over withheld authorizations and records [10] [11]. Public sources show researchers documenting coding changes and missing fields but do not provide a definitive ICE-originated, itemized admission of every correction for 2025, and those limitations persist in the record available to the public [2] [3].