How do ICE and CBP define and report 'removals' differently, and where are those datasets published?
Executive summary
ICE and CBP use different operational roles and counting rules to report "removals": CBP counts actions at the border — including refusals of admission and certain returns — in its encounters and custody/transfer datasets, while ICE reports removals tied to its Enforcement and Removal Operations (ERO) casework and detention workflows, including some voluntary returns and departures [1] [2] [3]. Those differences mean the agencies produce separate datasets that can overlap but are not reconciled; DHS’s Office of Homeland Security Statistics (OHSS) publishes synthesized annual and monthly tables drawing from both agencies to help analysts compare records [1] [4].
1. How the agencies’ missions shape what “removal” means on paper
CBP’s remit is the border and ports of entry, so its published numbers emphasize encounters, inadmissibility determinations, and operational removals or returns effected at or near the border; its Custody and Transfer Statistics and CBP encounter tables track people processed by Border Patrol and Office of Field Operations, including those referred for credible fear screening [2]. ICE’s published statistics come from ERO and reflect interior enforcement, detention, and formal removals—ICE emphasizes detention, case adjudication and executing removal orders, and its removals dataset is tied to individuals ICE processed or detained [3] [5]. Those mission differences produce distinct upstream events: many CBP actions never become ICE removals because the person is processed, turned back, or otherwise not transferred to ICE custody [6] [7].
2. Technical differences in counting: encounters vs. removals vs. returns
Public datasets use varied categories: OHSS and CBP publish "encounters," "inadmissibles," "returns," "repatriations," and other border-focused actions, while ICE distinguishes arrests, detentions, removals, and returns/voluntary departures within ERO’s administrative framework [4] [2] [3]. ICE removals historically have included forced removals and certain voluntary returns/departures processed through ICE systems, but ICE’s methodology excludes many CBP-only actions unless ICE had operational control or accepted the case for removal—creating a methodological gap between the agencies’ totals [7] [6]. Third-party researchers note the hierarchy: encounters are the broadest class, then detentions and removals, and that detentions/removals often include people first apprehended by CBP who later enter ICE datasets [5].
3. Where to find the datasets and what they contain
ICE publishes Enforcement and Removal Operations statistics and data tables on its site with tables for arrests, detentions, and removals (ICE statistics page) and those files include ERO-centered removals and related fields [3]. CBP posts Custody and Transfer Statistics and encounter tables on the CBP website, detailing sector-level encounters, inadmissibility outcomes, repatriations and programmatic removals like MPP or ENV [2]. DHS’s OHSS offers consolidated annual flow reports and monthly tables that draw from both CBP and ICE case records and serve as the statistical system of record for many indicators; OHSS also preserves earlier reports when terminology changes [1] [4]. Independent projects such as the Deportation Data Project document and reprocess ICE and CBP releases and note gaps, timing issues, and overlaps [8] [6].
4. Why totals don’t add up, and the limits of public reconciliation
Analysts repeatedly warn that ICE and CBP removal counts likely overlap but are not joined at the person-level in public releases, so comparing or summing agency totals risks double-counting or missing CBP-only actions; efforts to estimate the overlap have proven difficult because the agencies’ datasets use different inclusion rules and release cadences [8] [7]. OHSS mitigates some of this by producing harmonized flow tables, but OHSS itself relies on agency inputs and documented terminology changes, and public reporting does not fully disclose person-level linkage across CBP and ICE records [1] [4]. Where sources are silent—such as the exact number of CBP removals never handed to ICE—reporting cannot precisely resolve the gap without internal linkage data that is not published [6].