ICE Enforcement and Removal Operations: Reading the Federal Dataset Behind Immigration Enforcement

ICE publishes a detailed annual accounting of its Enforcement and Removal Operations: how many people were arrested, detained, removed, and returned, broken down by nationality, criminality designation, removal method, and field office. The dataset is publicly available, consistently misread, and far more granular than the summary numbers that dominate political coverage of immigration enforcement.

The four terms and why they are not interchangeable

Coverage of immigration enforcement frequently conflates four legally and operationally distinct categories. Using them interchangeably produces claims that are technically false.

An arrest (or administrative arrest) occurs when ICE takes a person into custody on civil immigration grounds. An arrest does not result in removal — the person may be released on bond, placed on an order of supervision, or transferred to detention pending a hearing. ICE's annual ERO report reports arrests as a distinct figure from removals.

Detention refers to civil immigration custody, typically in ICE detention facilities or contract facilities operated by private prison companies or local jails under intergovernmental service agreements. A person can be detained and then released, detained and then removed, or detained for months or years while immigration proceedings are pending. The detained population at any given moment is different from the annual flow of detained individuals. ICE reports both the average daily population (ADP) — the average number of people held on any given day — and the total number of individuals booked into ICE custody during the fiscal year.

A removal is the compulsory expulsion of a noncitizen from the United States pursuant to an order of removal, deportation, or exclusion under the Immigration and Nationality Act. A removal carries immigration consequences: the person is generally barred from re-entering for a defined period (ten years, twenty years, or permanently, depending on the grounds) and re-entry after removal is a federal felony under 8 U.S.C. § 1326. Removals are the primary metric in ICE's annual ERO report and the figure most relevant to enforcement trend analysis.

A return is a departure that is not a removal — typically a voluntary departure where a person agrees to leave without a formal order of removal being entered. Returns do not carry the same re-entry bar as removals. Historically, the United States recorded millions of returns annually, largely Mexican nationals who were encountered at the border and informally returned without formal proceedings. As enforcement practices shifted toward formal processing, the return count declined sharply relative to the removal count. The two numbers together constitute total “deportations” in casual usage, but they have meaningfully different legal consequences for the individuals involved.

What the ERO dataset contains

ICE's annual Enforcement and Removal Operations report is published as a PDF at ice.gov/removal-statistics. The reports cover fiscal years (October 1 through September 30) and include summary tables with the following dimensions:

• Country of origin — the nationality of removed individuals. ICE reports this as country of citizenship. The report lists the top twenty or so nationalities by removal volume; the Transactional Records Access Clearinghouse (TRAC) at Syracuse University publishes more granular country-level breakdowns through its TRAC-ICE portal.
• Criminal vs. non-criminal designation — ICE categorizes removed individuals as “convicted criminal” (someone with a criminal conviction), “non-convicted criminal” (someone with pending charges but no conviction), or “non-criminal” (no criminal record). The criminal designation is based on prior convictions in the US criminal justice system, not on immigration violations themselves. Immigration violations are civil matters; overstaying a visa is not a criminal offense.
• Method of removal — voluntary return vs. order of removal. Some reports further break down the order-of-removal category by whether the order was issued in absentia, by an immigration judge following a hearing, or through expedited removal without a hearing.
• Fiscal year — ICE's fiscal year runs October 1 through September 30, meaning FY2024 covers October 2023 through September 2024. This is the same fiscal year used across the federal government.
• Field office — ICE's Enforcement and Removal Operations is organized into 25 field offices across the country. The field office breakdown is particularly useful for distinguishing interior enforcement (arrests made away from the border, typically in the interior of the country) from border enforcement (arrests tied to recent border crossings). This distinction has become politically significant because different administrations have defined their enforcement priorities partly in terms of this interior/border split.
• Immigration violation type — the INA violation cited as the basis for the immigration enforcement action. Common categories include entry without inspection (EWI), overstay, and status violations. This field is more granular in TRAC data than in the summary PDFs.

How to access the data

The primary public access points are:

ICE ERO annual report PDFs at ice.gov/removal-statistics. ICE publishes full ERO reports back to FY2003. The reports are PDFs with summary tables; researchers typically use Camelot or pdfplumber to extract tables programmatically. The PDF format changes between years, requiring per-year extraction logic.

TRAC-ICE at Syracuse University at trac.syr.edu/immigration/reports/iceenforcementandremoval/. TRAC, which has obtained ICE enforcement data through sustained FOIA litigation since the 1990s, provides more granular data than the summary PDFs — monthly breakdowns, field-office-level detail, and nationality breakdowns that go deeper than ICE's published top-twenty lists. TRAC provides an interactive query tool and offers data downloads; some granular data requires a subscription.

DHS Office of Immigration Statistics (OIS) at dhs.gov/immigration-statistics. OIS publishes the Yearbook of Immigration Statistics, which covers removals, returns, and other enforcement metrics in a consistent historical format going back to 1892. The OIS Yearbook is the best source for long-run trend analysis because it provides a methodologically consistent time series across decades and administrations.

Key trends in the removal time series

The removal count is the single number most cited in immigration enforcement debates, and its trajectory across administrations is more nuanced than either side of that debate typically acknowledges.

Removals peaked at 432,281 in FY2013 under the Obama administration, which prioritized enforcement against recent border arrivals and individuals with criminal convictions. The Obama administration was the first to explicitly articulate a removal priority hierarchy — prioritizing recent entrants and those with serious criminal histories — while simultaneously deprioritizing long-settled residents without criminal records. The result was the highest removal numbers in recorded history, concentrated among specific populations.

Under the Trump administration's first term, the explicit removal priority system was rescinded: all undocumented individuals became enforcement priorities in principle, regardless of criminal history or length of residence. Interior enforcement arrests increased significantly. However, total removal numbers did not surpass the Obama-era peak. Removals reached 226,119 in FY2020, lower than any Obama-era year, partly because of COVID-19 disruptions and partly because the border encounter composition shifted toward populations that were harder to remove quickly (families and asylum seekers rather than single adults).

The Biden administration substantially reduced removals in its first year (FY2021: 59,011 removals) through a combination of policy memos, court-ordered pause on enforcement, and a narrowed enforcement priority framework. Numbers recovered in subsequent years as border encounter volumes increased and the administration faced political pressure. By FY2023 removals reached 142,580.

The second Trump administration, beginning in January 2025, launched an immediately visible surge in interior enforcement operations, including highly publicized workplace raids and targeted enforcement in major metropolitan areas. Early FY2025 arrest and removal numbers tracked significantly above prior-year baselines.

Interior vs. border enforcement

The interior/border distinction is analytically significant because the two enforcement streams involve different populations, different legal processes, and different policy levers.

Border enforcement primarily targets individuals apprehended shortly after crossing the border without authorization. Under expedited removal authority — 8 U.S.C. § 1225(b)(1), enacted in 1996 and expanded over subsequent years — these individuals can be removed without a hearing before an immigration judge if they have been in the country for fewer than two years (the Biden-era extension; originally 14 days within 100 miles of the border). Expedited removal is fast: it can be completed in days. It accounts for a large share of total removals in high-encounter years. CBP, not ICE, executes most border removals; CBP's removal numbers are included in the DHS OIS Yearbook but are reported separately from ICE's ERO numbers.

Interior enforcement targets individuals who are already living in the United States — those who entered legally and overstayed, those who entered without inspection years ago, and those who were arrested by ICE based on a criminal justice encounter (through the 287(g) program or Secure Communities, which automatically checked fingerprints of arrested individuals against immigration databases). Interior removals are slower, more resource-intensive, and politically more visible because they involve people with established community ties.

The criminality composition of removals differs between interior and border streams. Border removals are predominantly non-criminal (recent arrivals have not had time to accumulate US criminal records). Interior removals skew more heavily toward individuals with criminal convictions, because interior enforcement has historically been justified politically as targeting “dangerous criminals.” The ERO data allows researchers to disaggregate these streams, though the interior/border field-office proxy is imperfect.

Criminal vs. non-criminal and what the designation means

ICE's “convicted criminal” category covers anyone with a criminal conviction in the United States, regardless of the severity of the offense. The category includes individuals convicted of homicide and individuals convicted of traffic violations. ICE has been criticized for reporting aggregate “criminal removals” without breaking down conviction types, which obscures whether enforcement is targeting violent offenders or people with minor misdemeanor convictions.

TRAC has done the most sustained work on this question, obtaining ICE data on the specific offense types associated with criminal removals. Their analyses have consistently found that the largest single offense category among “criminal” removals is immigration violations — primarily improper re-entry under 8 U.S.C. § 1326 — which is circular (the criminal record is itself an immigration enforcement outcome). When immigration offenses are excluded from the criminal category, the share of removals involving individuals with violent-offense convictions is considerably smaller than public statements by enforcement officials typically suggest.

The non-criminal removal share has varied significantly across administrations as priority frameworks changed. Under Obama, non-criminal removals were nominally deprioritized; in practice, the large volume of recent border crossers drove non-criminal removal numbers up. Under the first Trump administration, the explicit statement that all undocumented individuals were enforcement priorities drove interior non-criminal arrests. The percentage of removals involving individuals with no criminal record rose during periods of expanded enforcement priority.

Nationality composition

For most of ICE ERO's recorded history, Mexico dominated the nationality breakdown by a wide margin — at peak, Mexican nationals constituted more than 70% of all removals. This reflected both the scale of US-Mexico migration and geographic enforcement logic: CBP encounters at the southern border were predominantly Mexican, and Mexican nationals could be removed by bus or foot across the border without the logistical complexity of international repatriation flights.

The nationality composition began shifting substantially around 2012–2014 as migration from the Northern Triangle of Central America — Guatemala, Honduras, and El Salvador — increased sharply. The rise of family unit and unaccompanied minor migration from Central America changed enforcement dynamics: families could not simply be returned across the border because they presented asylum claims, and their removal required coordination with destination governments. Guatemala, Honduras, and El Salvador collectively rose to constitute a significant share of annual removals through the late 2010s.

Post-2020, two nationalities emerged as notable additions to the high-volume removal categories: Venezuela and China. Venezuelan migration surged as economic collapse and political repression under the Maduro government produced a refugee crisis. Venezuelan removals were logistically complicated for years because the United States lacked a working repatriation agreement with Venezuela's government; a partial agreement announced in 2023 opened a limited removal pathway. Chinese nationals arriving at the southern border in large numbers from 2022 onward — many traveling through the Darien Gap and Central America — represented a new high-volume nationality that created its own diplomatic and logistical challenges for removal.

Expedited removal authority

The expedited removal provision at 8 U.S.C. § 1225(b)(1) allows immigration officers to order removed any arriving noncitizen who is inadmissible on fraud or misrepresentation grounds, or any noncitizen who cannot demonstrate continuous presence for more than the statutory period (two years as of the Biden-era expansion; previously 14 days within 100 miles of the border in the original 1996 Illegal Immigration Reform and Immigrant Responsibility Act formulation).

Expedited removal does not require a hearing before an immigration judge. The officer makes the determination in the field. The only exception is for individuals who express a fear of persecution or torture: they must be referred for a credible fear screening with an asylum officer. If the officer finds credible fear, the case proceeds to the regular immigration court system. If not, expedited removal can proceed immediately.

The expansion of expedited removal authority has been among the most significant drivers of total removal numbers in high-encounter years, because it allows large volumes of border crossers to be processed and removed without the years-long immigration court proceedings that a removal order before an immigration judge would otherwise require. ICE's ERO report distinguishes between removals via expedited removal order and removals via immigration judge order, though this breakdown is not always prominently featured in summary statistics.

Sanctuary jurisdictions and ICE detainer data

An ICE detainer (Form I-247) is a request from ICE to a state or local law enforcement agency to hold a person who would otherwise be eligible for release from criminal custody, for up to 48 hours, so that ICE can assume custody. Detainers are voluntary: ICE has no legal authority to compel a jurisdiction to comply. Jurisdictions that have adopted policies limiting detainer compliance are colloquially called “sanctuary jurisdictions,” though no uniform legal definition of that term exists.

ICE publishes data on detainer issuance and compliance through its detainer statistics page. The data includes the number of detainers issued to each jurisdiction, the number honored, and the number declined. TRAC maintains a searchable database of detainer activity by jurisdiction going back to 2008 that allows comparison across counties, states, and time periods.

The detainer data is politically significant because it quantifies the enforcement gap created by non-compliant jurisdictions — a gap that is real in the data but whose policy implications are contested. Enforcement-side analyses point to cases where individuals with prior criminal records who were released rather than transferred to ICE went on to commit additional offenses. Civil liberties analyses point to evidence that detainer requests are frequently issued for individuals with no criminal history and that jurisdictions holding individuals beyond their criminal release date on the basis of a civil immigration detainer may be violating the Fourth Amendment.

Detention facilities and detention statistics

ICE maintains a network of detention facilities that has expanded and contracted with enforcement priorities. As of 2024, ICE used approximately 200 facilities, a mix of dedicated ICE-operated facilities, contract facilities operated by private prison companies including GEO Group and CoreCivic (formerly CCA), and county jails operating under intergovernmental service agreements.

ICE publishes the detention facility list at ice.gov/detention-facilities, including location, facility type, rated capacity, and ADP. The average daily population in immigration detention peaked at approximately 55,000 in FY2019, declined sharply during COVID-19, and has trended upward since FY2022 as encounter volumes increased.

The Detention Watch Network and the Government Accountability Office have published analyses of detention conditions, death rates in ICE custody, and medical care standards. ICE is required to report deaths in custody, and these reports are available through the ICE website and through FOIA requests. TRAC maintains a database of deaths in ICE custody derived from these reports.

How journalists use TRAC-ICE data

The TRAC-ICE portal has been the primary data source for investigative and beat reporting on immigration enforcement since the early 2000s, precisely because it provides granularity that the official ERO PDFs do not. Journalists use the portal to:

• Identify surges or declines in arrests at the field-office level before national aggregate numbers suggest them.
• Disaggregate the “criminal removals” figure by offense type, comparing official characterizations of enforcement priorities against actual conviction-type breakdowns.
• Track detainer compliance rates by county and state to quantify the practical enforcement impact of sanctuary policies.
• Identify nationality-specific enforcement shifts — for instance, tracking when Venezuela or China emerged as high-volume removal nationalities before the annual ERO report was published.
• Cross-reference ICE arrest data against EOIR case completion data to estimate the share of arrests that result in removal orders vs. cases that are closed, administratively terminated, or result in grants of relief.

The following Python script demonstrates loading TRAC-formatted export data and computing the criminality breakdown and nationality trends that are the workhorses of enforcement reporting:

import pandas as pd
import requests
from io import StringIO

# TRAC-ICE provides data exports; ICE PDFs require tabula-py or pdfplumber for extraction.
# This example works with a TRAC-formatted CSV export.

df = pd.read_csv('trac_ice_removals.csv', low_memory=False)
df.columns = [c.strip().upper() for c in df.columns]

# Key fields typically available from TRAC:
# FISCAL_YEAR, NATIONALITY, REMOVAL_TYPE (order vs. voluntary),
# CRIMINALITY (criminal vs. non-criminal), FIELD_OFFICE, COUNT

# Interior vs. border removals proxy: field office geography
# ICE field offices in border states: San Antonio, El Paso, Phoenix, San Diego, Miami, Atlanta
border_offices = {'SAN ANTONIO', 'EL PASO', 'PHOENIX', 'SAN DIEGO'}
df['is_border_office'] = df['FIELD_OFFICE'].str.upper().isin(border_offices)

# Removal trend by year and criminality
pivot = (
    df.groupby(['FISCAL_YEAR', 'CRIMINALITY'])['COUNT']
    .sum()
    .unstack(fill_value=0)
    .assign(total=lambda x: x.sum(axis=1))
    .assign(criminal_pct=lambda x: x.get('CRIMINAL', 0) / x['total'])
)

print("Removals by year and criminality designation:")
print(pivot.to_string())

# Nationality breakdown — top 10 countries by fiscal year
top_nat = (
    df.groupby(['FISCAL_YEAR', 'NATIONALITY'])['COUNT']
    .sum()
    .reset_index()
    .sort_values(['FISCAL_YEAR', 'COUNT'], ascending=[True, False])
    .groupby('FISCAL_YEAR')
    .head(10)
)

print("\nTop 10 nationalities by fiscal year:")
print(top_nat.to_string(index=False))

# Removal type split: order of removal vs. voluntary return
if 'REMOVAL_TYPE' in df.columns:
    type_split = (
        df.groupby(['FISCAL_YEAR', 'REMOVAL_TYPE'])['COUNT']
        .sum()
        .unstack(fill_value=0)
    )
    print("\nRemoval type split by year:")
    print(type_split.to_string())

A common analytical error when working with this data is treating the criminal/non-criminal designation as a stable, consistent field across years. ICE's coding practices for criminality designation have changed across administrations, and the threshold for what counts as “criminal” for reporting purposes has shifted. Pre-2017 and post-2017 criminal designation numbers are not perfectly comparable because the 2017 executive order broadened the enforcement priority framework in ways that altered how ICE categorized cases in administrative records.

What the dataset establishes and what it cannot answer

The ICE ERO dataset, taken at face value, is one of the more complete federal enforcement datasets available. It provides fiscal-year-by-fiscal-year trend data on the volume and composition of immigration enforcement actions, with enough dimensional granularity — nationality, criminality, field office, removal method — to support substantive analysis of how enforcement priorities have actually been implemented rather than merely announced.

What the dataset cannot answer: whether the individuals removed had meritorious claims for relief that were not adjudicated, whether enforcement is concentrated in ways that produce disparate racial or national-origin outcomes relative to the undocumented population at large, and whether the prioritization of “criminal” removals is capturing genuinely dangerous individuals or primarily people with low-level or immigration-related convictions. These questions require joining the ERO data against EOIR case outcome data, criminal history records, and demographic estimates of the undocumented population — data integrations that are possible with TRAC resources and careful methodology but that are not available from ICE's published numbers alone.

The enforcement numbers are real and large regardless of interpretive framework: the United States has removed between 59,000 and 432,000 people per year for the past two decades. The dataset makes clear that the composition, not just the volume, is where the policy choices live.