U.S. Deportations and “Self-Deportations” in the First 18 Months of the Second Trump Term

Executive Summary

During its second term, the Trump administration has focused unprecedented attention and resources on expanding the number of arrests and deportations within the United States and encouraging existing unauthorized immigrants to “self deport.” Yet the impacts of these efforts are difficult to measure due both to reduced immigration enforcement reporting by the Department of Homeland Security (DHS) and long-standing confusion about the different ways the department counts deportations.

This fact sheet uses publicly available DHS data1 to examine the number of deportations (a term that comprises removals and returns; see Box 1 for a description of key terms) carried out by U.S. Immigration and Customs Enforcement (ICE) and U.S. Customs and Border Protection (CBP) during the first 18 months of the second Trump term. In the absence of official DHS reporting of detailed enforcement data since January 2025, these are the first estimates of total deportations for the current Trump administration, aside from brief statements by administration officials to news organizations and on social media.

"In the absence of official DHS reporting of detailed enforcement data since January 2025, these are the first estimates of total deportations for the current Trump administration, aside from brief statements by administration officials to news organizations and on social media."

Drawing on its analysis of the data, MPI finds that between January 20, 2025, and July 31, 2026:

• DHS (ICE and CBP combined) completed about 980,000 deportations, a figure in line with the department’s public statements. On an annual basis, total deportations are up from the Biden and first Trump administrations but remain well below the levels routinely observed in the early 2000s.
• In a major break with the pre-COVID-19-pandemic period, a growing share of deportations have been returns rather than formal removals. Of the 980,000 deportations, about 170,000 were ICE grants of voluntary departure or voluntary return as the Trump administration has sought to bypass the backlogged immigration courts, which hear removal cases.
• Even with U.S.-Mexico border encounters at historic lows, 69 percent of deportations were of individuals intercepted at the border: about 340,000 (35 percent) were of migrants intercepted at the border and turned around by CBP, and another 340,000 (34 percent) were of people initially arrested at the border and deported by ICE after being processed for removal within the United States. The other 31 percent of deportations (300,000) were of immigrants initially arrested within the United States—in other words, individuals likely to have some years of residence in U.S. communities.
• Interior arrests have accounted for an increasing share of total DHS deportations over this time, however: 28 percent in January 20–December 31, 2025; 34 percent in the first six months of 2026; and 39 percent in July 2026. The number of deportations of immigrants arrested in the country’s interior has also increased, from 8,000 per month on average in the first three months of the administration to 19,000 per month on average between May and July 2026.
• Excluding border turnarounds by CBP, 61 percent of deportations by ICE were of people who had been in the United States at least two years—390,000 deportations in all—and 15 percent (90,000) were of people who had been in the country ten or more years.
• Of total deportations of immigrants with two or more years of U.S. residence, 71 percent (280,000) were of people who had never been convicted of a crime, and of those with ten or more years of U.S. residence, 61 percent (57,000) had no criminal conviction. Overall (including all ICE and CBP removals and returns), just 20 percent (190,000) of the 980,000 deportations carried out between January 20, 2025, and July 31, 2026, were of people who had been convicted of a crime. And just 31 percent of removals were of convicted criminals, compared to an average removal criminality rate of 44 percent in fiscal years 2009–21.

The fact sheet also uses publicly available survey data and a new Migration Policy Institute (MPI) methodology to estimate the number of total immigrant departures—that is, the sum of deportations and “self-deportations” (also called self-departures) from the United States. Neither publicly available data nor internal DHS data allow for precise counts of total immigrant departures, so this exercise aims to fill the gap. (While DHS knows the full scope of deportations, it cannot directly measure self-departures because the department does not collect data on most land exits at the U.S.-Mexico border.) To focus on the outflow of people who have lived in U.S. communities and left amid heightened attention to immigration, this MPI methodology excludes border turnarounds (which are near immediate), short-term visitors, and other immigrants in the country for less than one year.

MPI estimates that there were likely between 790,000 and 1.9 million total departures of immigrants with at least one year of U.S. residence (via deportation or self-departure) between January 20, 2025, and July 31, 2026. This estimate includes about 480,000 deportations of immigrants who had been in the United States at least one year and between 310,000 and 1.5 million self-departures of immigrants who had at least one year of U.S. residence. Based on data availability, MPI’s estimates include self-departures of both unauthorized and lawfully present immigrants.

DHS officials have claimed more than 3 million total departures—all of unauthorized immigrants—during the same period. This DHS figure would include about 1 million deportations (including border turnarounds and deportations of other immigrants present in the United States less than one year, which are excluded from MPI’s estimate) and about 2 million unauthorized immigrant self-departures—approximately double MPI’s estimate for self-departures of unauthorized and lawfully present immigrants combined. It should also be noted that while DHS has characterized the 3 million departures figure as being unauthorized immigrants, this count likely includes certain lawfully present immigrants and U.S. citizens who have departed through the department’s Project Homecoming initiative, which provides a free flight out of the country and an exit bonus of up to $2,600.

1. Introduction

President Donald Trump made immigration enforcement a centerpiece of his re-election campaign, including a promise to focus on “the worst of the worst.”2 Since taking office in January 2025, the administration has devoted unprecedented resources to the issue, securing $240 billion dollars from Congress for the Department of Homeland Security (DHS) through fiscal year (FY) 2029.3 U.S. Immigration and Customs Enforcement (ICE), which has primary responsibility for deportations, hired 12,000 officers in 2025, in a single year more than doubling its overall number of officers.4 Even those investments understate the commitment to immigration enforcement, as DHS and its components (ICE, U.S. Customs and Border Protection [CBP], and U.S. Citizenship and Immigration Services [USCIS]) have reassigned significant numbers of personnel away from refugee and asylum processing, criminal investigations, and visa processing missions to focus instead on immigrant arrests and deportations. Departments and agencies outside DHS, including the Federal Bureau of Investigation (FBI) and U.S. attorneys offices, also have had personnel repointed to immigration enforcement work.5

Yet amid this unprecedented attention and investment, official reporting on immigration enforcement has been cut back—in contrast to the first Trump term, which saw no interruption in reporting on ICE and CBP activities. The DHS Office of Homeland Security Statistics (OHSS) released its last detailed monthly enforcement report, which offers a look across ICE and CBP activities, in January 2025. And despite statutory requirements for annual reporting, the Trump administration has not released a single report on border security metrics or on DHS immigration enforcement actions for FY 2024 or 2025.

This fact sheet seeks to help fill in this gap in deportation reporting by using publicly available data to approximately replicate OHSS reporting where possible.6 The fact sheet provides both estimates of total deportations and estimates broken out by type of deportation, arrest location (border versus interior), which agency was responsible, immigrants’ time in the United States, and criminality. For each of these categories, trends since January 2025 are compared to historic patterns.

"This fact sheet seeks to help fill in this gap in deportation reporting by using publicly available data to approximately replicate OHSS reporting where possible."

This fact sheet also explains what DHS knows and does not know about total unauthorized immigrant departures from the United States, which represents the sum of deportations and unauthorized immigrant “self-departures.” While neither public nor internal DHS data allow a precise count of total unauthorized immigrant departures, the Migration Policy Institute (MPI) has used public data to develop a broad estimate of recent total departures. (See the fact sheet’s Appendix for methodological information on this and other aspects of the data analysis.)

2. Total Deportations

When DHS officials describe total deportation numbers, the data point being described is the sum of removals and returns, including a mix of migrants arrested at the border and those arrested within the United States, and of individuals deported by ICE or by CBP.

These various categories, along with changes between one administration and the next as to which kinds of deportations receive the most focus, have resulted in public confusion about DHS deportation data. In addition, ICE releases formal reports on deportations that it carries out far more often than DHS reports on total (ICE plus CBP) deportations, leading to additional confusion between these two counts. Total deportations are the most comprehensive measure of federal immigration enforcement, and OHSS’s Yearbook of Immigration Statistics includes trend data on total deportations dating back to 1892.7

MPI’s analysis of publicly available data finds that ICE and CBP together completed about 980,000 total deportations between January 20, 2025, and July 31, 2026, consistent with certain public statements by DHS officials.8

Total deportations have fluctuated within a band of 46,000 to 57,000 per month during the second Trump term (see Figure 1). These totals show a modestly increasing trend, with the monthly average for the latest three-month period (55,000 per month on average in May–July 2026) 9 percent higher than the 51,000 per month average in January–March 2025, at the start of the administration.

On an annualized basis, the Trump administration completed 620,000 total deportations in FY 2025 and was on track to complete 660,000 in FY 2026.9 These annual totals represent a 26 percent increase over the average for FY 2021–24 and a 42 percent increase over FY 2017–20, but they are well below the all-time high of 1.86 million deportations in FY 2000 (see Figure 2) as well as the administration-set target of carrying out 1 million deportations annually.10

3. Deportations by Type 

Total DHS deportations—that is, total compulsory departures from within the United States or from its borders and ports of entry—fall into a few broad categories that differ in terms of their legal basis, consequences, and process.

The most important distinction is between a removal and a return. Removals are designed to be the most consequential form of repatriation. Noncitizens who are removed generally are barred from obtaining a visa to re-enter the United States for at least five years, and potentially forever, depending on the type of removal. Being removed also makes individuals subject to enhanced criminal charges if they are re-encountered in the United States in the future. Immigrants who are returned are not subject to visa restrictions or enhanced criminal penalties.

Relatedly, the removal process includes more robust due-process protections than do returns. In general, immigrants have the right to contest a removal order with DHS or in immigration court, but they have no right to contest a return. Returns are “voluntary” in the sense that individuals typically agree to be returned (i.e., agree to forgo the opportunity to contest their removal in immigration court) in order to avoid the penalties associated with removal. In this sense, returns are similar to plea bargains in criminal proceedings. DHS considers a grant of return in lieu of a potential removal to be a form of discretion that benefits the immigrant, but it bears emphasis that individuals with a “grant of return” are required to depart.

Returns can be further disaggregated into enforcement returns and administrative returns.11 Enforcement returns involve migrants judged by DHS to have tried to enter the United States unlawfully, while administrative returns generally involve migrants who did not make such an effort. From FY 2018 to FY 2023, DHS counted certain withdrawals of application for admission at ports of entry as administrative returns if the department never considered removal proceedings. Since the start of FY 2024, administrative returns have been limited to crew members who lack a visa allowing them to debark while their ship is at port and who are required to remain on board their vessels. All administrative returns therefore involve people who never entered the United States.

MPI analysis finds that of the 980,000 total ICE and CBP deportations between January 20, 2025, and July 31, 2026, about 59 percent (580,000) were removals, 30 percent (300,000) were enforcement returns, and 11 percent (110,000) were administrative returns.12 On an annualized basis, the average of 380,000 DHS removals per year during the second Trump term falls between the averages for the first and the second Obama administrations (392,000 and 373,000, respectively) and well below the single-year record of more than 432,000 removals in FY 2013 under President Barack Obama (see Figure 3).

Figure 3 also reveals a notable pattern with respect to enforcement returns. Enforcement returns dropped sharply from FY 2009 (and total returns did so for earlier years) through the mid-2010s, a trend that reflects a systematic effort by CBP to increase the “consequences” that unauthorized crossers at the U.S.-Mexico border face by processing them for removal rather than allowing them to voluntarily return to Mexico.13 Returns remained at low levels through the COVID-19 pandemic years despite high numbers of border encounters because most people encountered at the border were expelled under Title 42 authority, which does not impose any sanctions beyond expulsion (and are not counted in these data). Returns climbed sharply when use of Title 42 ended in May 2023, at which point CBP began relying on enforcement returns to quickly process still-high numbers of encounters in FY 2023 and FY 2024.

Despite the sharp drop in border encounters since January 2025, enforcement returns remain surprisingly high, as Figure 4 shows, because the Trump administration has increasingly relied on returns to speed up deportation processing for people arrested far from the border. DHS was on track to complete more than 230,000 enforcement returns in FY 2026, fewer than the Biden administration in FY 2023 or FY 2024, but otherwise the most since FY 2011.

Notably, about 170,000 of the enforcement returns between January 20, 2025, and July 31, 2026, were grants of voluntary departure or voluntary return by ICE—an all-time high by a wide margin—as DHS seeks to bypass immigration court proceedings for people arrested and/or detained inside the United States.14

4. Deportations by Arrest Location and Deportation Agency

In addition to type of deportation, it can be useful to distinguish deportations that occur after a person is arrested within the United States versus at a U.S. border or port of entry, and which DHS agency is involved.

• Border arrests leading to deportation by CBP: Historically, most migrants arrested at the border were repatriated directly to Mexico by the U.S. Border Patrol, part of CBP, or to the country from which they traveled to the United States by the CBP Office of Field Operations (OFO), all without ever entering the country or without ever leaving CBP custody. The Border Patrol handles encounters between ports of entry, and OFO handles arrivals of “inadmissibles” (those arriving without prior authorization to enter) at ports of entry. People arrested at the border and repatriated by CBP are sometimes called “border turnarounds.”
• Border arrests leading to deportation by ICE: Migrants encountered at the border15 who are not deported directly by CBP are transferred to ICE for removal processing and possible detention. Individuals may be turned over to ICE because they are encountered between ports of entry and are not from a contiguous country, claim fear of return to their home country (i.e., seek asylum), have an outstanding arrest warrant, due to limits on CBP processing capacity, or for other reason. If they are deported, they count as ICE deportations.
• Interior arrests, which are those carried out in U.S. communities away from the border, that result in deportations are always processed by ICE and counted as ICE removals or returns.16

Out of the 980,000 deportations between January 20, 2025, and July 31, 2026, 35 percent (340,000) were immediate border turnarounds by CBP, 34 percent (340,000) were border arrests that led to deportation by ICE, and 31 percent (300,000) were ICE deportations resulting from interior arrests (whether carried out by it or another agency). With encounters between border ports of entry at 50-year lows during this period, most border arrests leading to deportation by CBP involve migrants turned away by OFO at ports of entry. Border arrests leading to deportation by ICE are a mix of new arrivals and people who were encountered at the border months or years ago but who have only recently been deported. If someone was arrested at the border years ago and was recently re-arrested by ICE in the U.S. interior, that would be considered an ICE deportation resulting from a border arrest.

Not surprisingly, given the drop in border arrivals, deportations resulting from border arrests have represented a declining share of total deportations over the course of the second Trump term. CBP deportations (border turnarounds) in the latest three-month period (19,000 per month on average in May–July 2026) were down 14 percent compared to the start of the administration (22,000 per month on average in January–March 2025). Similarly, ICE deportations of migrants arrested at the border were down 21 percent, from 21,000 per month to 17,000 per month on average in the same period.

Conversely, deportations of immigrants arrested in the interior were up 151 percent, from 8,000 per month on average at the beginning of the administration to 19,000 per month on average in the most recent three-month period (see Figure 5). Interior arrests have also accounted for a growing share of total DHS deportations, rising from 28 percent in January 20–December 31, 2025, to 34 percent in the first six months of 2026, and further to 39 percent in July 2026.

Despite this trend—and despite the extraordinary resources being devoted to arrests within the country’s interior—deportations from the interior in FY 2025 only reached an annualized rate of about 160,000, well below the average of 218,000 for FY 2009 through FY 2012. And ICE was on track for about 230,000 deportations resulting from interior arrests in FY 2026, slightly below the levels reached in FY 2008 and FY 2009 (235,000 and 238,000, respectively), as shown in Figure 6. In this earlier period, DHS was often able to more quickly remove arrested immigrants because more of the unauthorized immigrant population was from Mexico at the time, and Mexicans are less likely than other groups to present asylum or other protection claims as a defense against removal. In addition, the immigration courts had less of a case backlog then, allowing removal cases to move faster. The unauthorized immigrant population today is more demographically diverse,17 comprised of many groups seeking protection from threats at home, and the immigration courts have massive case backlogs,18 making the turnaround time between interior arrest and removal much longer under normal processing pathways (with the caveat that a growing number of interior arrests are bypassing these normal processing pathways through voluntary departure, as noted above).19

5. Deportations by Immigrants’ Time in the United States

In the first 18 months of the second Trump term, 61 percent of ICE deportations20—about 390,000 deportations—were of unauthorized immigrants who had lived in the United States for two or more years prior to their departure, compared to 39 percent (250,000) that were of newer arrivals.21 More than 90,000 deportations (15 percent of the ICE total) were of immigrants who had ten or more years of U.S. residence prior to being deported (see Figure 7).22

As Figure 7 illustrates, ICE deportations have focused increasingly on the settled and long-settled populations. The number of deportations of new arrivals fell 24 percent from an average of 17,000 per month in January–March 2025 to 13,000 per month on average in May–July 2026, while deportations of immigrants who had two to ten years of U.S. residence grew 77 percent from a monthly average of 10,000 to 18,000 and deportations of long-settled immigrants (ten or more years) grew 147 percent from 2,400 to 6,000 per month in the same period.

6. Deportations with and without Criminal Convictions

DHS public statements, which often refer to arrest and removal of “the worst of the worst,” frequently combine data on criminal convictions and criminal charges.23 But ICE data made available by the Deportation Data Project allow those categories to be disaggregated to understand how many arrests or removals are of actual convicted criminals. Removal of convicted criminals is also the relevant category for historic comparisons, since official DHS reporting has always been limited to that category and has not included those with a pending criminal charge.24

DHS completed about 190,000 deportations of convicted criminals between January 20, 2025, and July 31, 2026 (20 percent of the 980,000 total deportations during that period), including 180,000 removals of convicted criminals, which is the metric DHS has long tracked. Focusing on removals of people with criminal convictions, the Trump administration completed an annualized 120,000 such removals in FY 2025 and 120,000 were projected in FY 2026. These counts represent an increase over the five years since the start of the pandemic (FY 2020–24) but are otherwise the lowest since FY 2008. Likewise, removals of convicted criminals have accounted for 31 percent of all removals during the second Trump term, a rate well below the average of 44 percent for FY 2009–21 (see Figure 8).

Following DHS’s approach in reporting on this topic, the MPI analysis combines all types of criminal convictions into a single category. It is notable, however, that the most common crimes deportees have been convicted of are illegal entry/re-entry (23 percent of total criminal removals) and traffic offenses, including driving under the influence (DUI) convictions (21 percent).

When examined by years of U.S. residence, the deportations of each cohort during the period of January 20, 2025–July 31, 2026, were overwhelmingly of people without a U.S. criminal conviction: 74 percent of deportations of immigrants with two to ten years of U.S. residence were of people with no criminal record, as were 61 percent of deportations of immigrants with more than ten years of U.S. residence. Overall, the Trump administration has carried out about 280,000 deportations of individuals without a criminal conviction and with two-plus years in the United States, including about 57,000 deportations of noncriminals with at least ten years of U.S. residence (see Table 1).

7. Total Immigrant Departures

DHS has repeatedly stated publicly that more than 3 million unauthorized immigrants have departed the United States since January 2025,25 accounting for both deportations and “self-deportations.” This count cannot be verified using publicly available data—and indeed, DHS itself is unable to count total departures since no data are collected from most people who leave the country via the U.S.-Mexico land border.

This section summarizes what DHS does and does not know about total migrant departures. It also provides an MPI estimate of the likely range of total departures since January 2025 using publicly available data sources. MPI’s methodology focuses on measuring the outflow of people who have lived in U.S. communities and left amid a heightened focus on immigration, excluding border turnarounds (which are near immediate) and migrants in the country for less than one year, and in some cases as little as a day.

A. The DHS Estimate

The DHS approach to counting total migrant departures is to add four figures:

• Total deportations: As described earlier, total deportations include the sum of ICE and CBP removals and returns, including border turnarounds.
• Migrants known to have participated in Project Homecoming26: This group consists of two partially overlapping subpopulations:
  • people who register with the CBP Home app and then are observed departing the United States, and
  • people who are contacted by a Project Homecoming contractor that facilitates their outbound travel and/or initiates a departure stipend and who are observed leaving the United States.
• Other known self-departures: In general, DHS has the ability to observe departures of all travelers who leave the United States by air, sea, or across the northern land border, as well as a small share of southern land border departures.27 DHS checks these departures against datasets that include immigrants who are known to have been in the United States without lawful status. These datasets include nonimmigrant visa overstayers, immigrants on ICE’s non-detained docket (i.e., those being processed for removal), people with expired parole or Temporary Protected Status (TPS), and holders of Deferred Action for Childhood Arrivals (DACA) or another form of deferred action. DHS is also able to identify those departing with a foreign passport who have no record of lawfully entering the United States.28
• Unobserved self-departures: DHS assumes, but cannot estimate, an additional number of unauthorized migrants who depart the United States across the southwest land border without leaving a footprint in available data.

The department’s methodology for counting total unauthorized immigrant departures is limited in several important respects. The first three categories identified above overlap to varying degrees. Large numbers of Project Homecoming departures are also counted as ICE deportations since Project Homecoming contractors mainly recruit participants from ICE detention facilities.29 (Project Homecoming, which provides a free flight out of the country and exit bonus of up to $2,600, accounts for a large share of ICE voluntary departures.) Known self-departures of immigrants on ICE’s non-detained docket who are observed on flight and other manifests are also counted as ICE deportations.30

"The department’s methodology for counting total unauthorized immigrant departures is limited in several important respects."

It should also be noted that not all of these departures are unauthorized immigrants. The CBP Home app is explicitly offered to all immigrants without criminal convictions and does not subject participants to status checks, meaning these self-departures not only include some lawfully present immigrants but also that it is more than likely that some U.S. citizens are included, particularly in the case of U.S.-citizen children who depart with unauthorized immigrant parents.

Furthermore, “total departures” as defined by DHS include border turnarounds and other rapid deportations of people who have never resided in the United States as well as self-departures by nonimmigrant visa holders who overstay their period of admission by as little as a single day. While these departures meet a technical definition of unauthorized immigrant departure, they do not meaningfully affect the actual population of unauthorized immigrants in the United States.31

Finally, DHS has no concrete data—or even a model-based estimate—of how many migrants depart across the southwest land border, nor how many of them are unauthorized immigrants. While the department has released relatively little data on known self-departures, it is likely that unobserved self-departures presumably occurring via the U.S.-Mexico border make up the largest category of the department’s 3 million total departures claim.

B. MPI’s Estimate

DHS has released limited data on known self-departures and lacks a rigorous estimate of unobserved self-departures. To help fill the knowledge gap, MPI developed a standalone estimate leveraging the best available data. This analysis draws on Mexican Census data that measure the number of Mexicans returning to Mexico after living in the United States the previous year, and then utilizes MPI estimates of the U.S. unauthorized and legal immigrant populations to estimate what share of all immigrant departures from the United States are likely to be comprised of Mexican immigrants versus immigrants from other parts of the world (see the Appendix for methodological details).

Since this exercise focuses on estimating departures that are substantively meaningful because the individuals had spent some period of time in the country, MPI’s estimates are of immigrants who had at least one year of U.S. residence prior to their departure. This definition excludes short-term visitors and seasonal migrants who travel back and forth to the United States, as well as border turnarounds and other deportations that take place less than one year after a migrant’s U.S. arrival, all of which are included in DHS’s count. And while DHS has characterized its 3 million total departure number as being a count of unauthorized immigrants, if this count includes Project Homecoming departures, the number also includes certain lawfully present immigrants and U.S. citizens. The MPI total departures estimate does not distinguish between departures of unauthorized immigrants and lawfully present immigrants, other than to exclude naturalized U.S. citizens.

The resulting estimate describes the likely range of total departures based on available data. The exact number of estimated outflows depends on the specific assumptions made about the relationship between Mexican outflows and total outflows (which, in turn, is shaped by differences in the demographic characteristics of immigrants from different parts of the world as well as variation in economic, safety, and other conditions in home countries, and differences in travel distances between the United States and those countries). This limitation is mitigated by offering a wide range, rather than a single-point estimate, to capture a variety of plausible assumptions about these outflows. With these caveats, this exercise makes it possible to evaluate the plausible level of departures and also how total departures and self-departures appear to be trending over time (see Figure 9).32

Based on this exercise, MPI estimates that in the first 18 months of the second Trump term, between 790,000 and 1.9 million departures likely occurred. While these total departure figures only include immigrants with at least one year of U.S. residence, if border turnarounds and other deportations of immigrants present for less than one year were to be included, these additional 500,000 deportations would bring MPI’s total estimate to between 1.3 million and 2.4 million total departures for January 20, 2025, through July 31, 2026.

Looking specifically at the subset of total departures comprised of people who self-deport, MPI estimates that in the first 18 months of the second Trump term, there were between 310,000 and 1.5 million self-departures of immigrants with at least one year of U.S. residence.

MPI’s estimates confirm that total departures in this period were likely two to three times higher than at any point for which data are available back to 2014 and that self-departures were likely two to four times higher. However, these estimates also suggest that it is highly unlikely that 3 million immigrants have departed the United States via deportation or self-departure—much less 3 million unauthorized immigrants.33

8. Conclusion

During its second term, the Trump administration has placed immigration enforcement at the center of its governing agenda, even as it has sharply curtailed DHS data transparency and reporting. This fact sheet uses publicly available DHS data to approximately replicate DHS Office of Homeland Security Statistics reporting for key deportation statistics and uses U.S. Census and Mexican data to generate a broad estimate for likely total immigrant departures, including self-departures.

Drawing on a newly developed MPI methodology that is described in the Appendix,34 the fact sheet finds that DHS, through its ICE and CBP components, carried out about 980,000 deportations during the first 18 months of the current Trump term, including of large numbers of immigrants with two or more years of residence and sizeable majorities with no criminal conviction. The data also reveal that the ICE deportations total includes record high numbers of voluntary departures, consistent with the administration’s effort to circumvent the backlogged immigration court system.

Finally, MPI estimates that while the total number of immigrants departing the United States through deportation or “self-deportation” are much higher than at any previous time dating back at least to 2014, it is highly unlikely that 3 million immigrants have departed the country—much less 3 million unauthorized immigrants. 

Appendix. MPI Methodology

In this fact sheet, data for historic comparisons come from the Department of Homeland Security (DHS) Office of Homeland Security Statistics (OHSS) Yearbook of Immigration Statistics35 and from month-by-month OHSS data on immigration enforcement and legal processes.36

Estimates of deportations since January 2025 are derived from publicly available data from three sources:

• U.S. Border Patrol (USBP) apprehension data posted on the Freedom of Information Act (FOIA) reading room on the U.S. Customs and Border Protection (CBP) website;37
• CBP Office of Field Operations (OFO) determinations of inadmissibility data posted on the CBP website’s FOIA reading room;38 and
• U.S. Immigration and Customs Enforcement (ICE) removals and returns data (which ICE refers to simply as “removals”) obtained through FOIA and posted by the University of California’s Deportation Data Project.39

Each of these datasets includes person-level records of DHS enforcement actions with enough information about processing disposition and arrest conditions for Migration Policy Institute (MPI) researchers to approximately replicate OHSS’s methodology for producing official DHS enforcement data.

The researchers have confidence in these source datasets because they were produced by the same CBP and ICE reporting teams that provide data to OHSS, and they include most of the core variables used by OHSS (and by CBP and ICE) to generate official reports. In the case of CBP data, MPI researchers were also able to compare total encounter numbers against published CBP encounter data.

At the same time, some data made available through FOIA are of a slightly different vintage than the final monthly and yearly snapshots used by OHSS for official reporting. In general, ICE and CBP finalize data extracts for OHSS reporting at the end of each month and the end of each fiscal year. These updates typically result in increases of up to one percentage point over the course of a fiscal year between the total suggested by monthly data and the final, adjusted, annual total. This means that the FOIA-based data used in this fact sheet will not be exact matches for official DHS data when they are released. Additional inconsistencies between the data used and official reporting are noted below.

To estimate total DHS removals and returns, MPI researchers summed up estimated ICE, USBP, and OFO removals and returns, following the same basic methodology as OHSS. In general, any migrant removed or returned by DHS who interacts with ICE (i.e., is arrested, reprocessed, transported, detained, or moved out of the United States by ICE) at any point between being apprehended or arrested and being removed or returned is counted as an ICE removal/return. Any migrant who is apprehended between ports of entry and turned around at the border without ever leaving USBP custody is counted as a USBP removal/return. And any migrant who is found inadmissible at a port of entry and turned away without being permitted to enter the United States is counted as an OFO removal/return. Certain removals or returns are recorded as jointly executed by ICE and either USBP or OFO (e.g., a migrant who books out of USBP custody directly to an airport where ICE flies them to their removal or return destination without ever booking into ICE detention). OHSS reporting generally assigns to ICE removals or returns that are jointly claimed by CBP and ICE; MPI researchers followed that same approach in this fact sheet.

Data pulled directly from OHSS public reporting retain OHSS rounding rules (which typically means rounding to the nearest 10). For MPI estimates based on FOIA data, figures below 10,000 are rounded to the nearest 100; figures between 10,000 and 100,000 are rounded to the nearest 1,000; figures between 100,000 and 1,000,000 are rounded to the nearest 10,000; and figures above 1,000,000 are rounded to the nearest 100,000.

A. ICE Removal and Returns

To estimate ICE removals and returns, MPI researchers used the Deportation Data Project’s data file on ICE removals for fiscal year (FY) 2025 through August 6, 2026.40 This is essentially the same dataset used by OHSS for official reporting, with one exception noted below.

MPI unzipped the Deportation Data Project data and combined records with departure dates between January 20, 2025, and July 31, 2026, into a single file. Every record in the file is counted as an ICE removal or an ICE return. Following OHSS’s methodology, records with case statuses of withdrawal permitted, voluntary departure confirmed, or voluntary return witnessed were counted as ICE returns; records with case statuses of Title 50 expulsion,41 deported/removed deportability, excluded/deported/removed, or excluded/removed - inadmissibility were counted as ICE removals.

Notably, the ICE data released by the Deportation Data Project do not include a field for “case closure date,” which is the date field used by OHSS and ICE to assign calendar dates to removals and returns in official reporting. In this fact sheet, removal dates were assigned based on the “departure date” field. In general, departure date is a more relevant date field because it more accurately represents when a migrant actually leaves the United States. In most cases, departure and case closure dates are close in time, but in some cases ICE agents delay finalizing the paperwork to close a case until the end of the month or the end of the fiscal year.

The regular pattern of some departures occurring in different months than their cases are closed means that none of the monthly totals in this fact sheet will be exact matches for official reporting, but the differences will balance out over time. Similarly, year-end reconciliation typically results in a few thousand removals being executed (i.e., having a departure date) in one year and being officially counted (case closure date) the following year. These differences also balance out over time, and both ICE and OHSS generally ignore this problem in their official reporting.

The difference between reporting by departure date and reporting by case closure date will be much larger than usual when DHS releases official removal and return numbers for FY 2025, however, because DHS’s analysis of known migrant self-departures in FY 2025 identified a large number of migrants who self-departed in FY 2024. Any of those migrants who were on ICE’s case processing docket had their cases closed as a result of their departure, and ICE counted these case closures as FY 2025 removals or returns. These removals and returns are not counted as FY 2025 ICE deportations in this fact sheet because the fact sheet reports by departure date rather than case closure date; as such, this fact sheet estimates fewer ICE removals in that fiscal year than official DHS reporting will eventually reflect.

B. USBP Removals and Returns

To estimate USBP removals and returns, MPI researchers used USBP apprehension data from CBP’s FOIA reading room. The researchers combined monthly and quarterly apprehensions datasets for January 2025 through July 2026. (CBP posts some of these files as quarterly data.) The researchers confirmed that total apprehensions in the resulting dataset fall within well under one percentage point of total apprehensions in public USBP reporting based on the CBP public data portal in every month from February 2025 through July 2026, other than May 2026 (for which the FOIA dataset appears to be missing about 2,000 records). The total count of apprehensions in the consolidated dataset falls within 1.5 percentage points of CBP public reporting.

Since FY 2024, USBP data provided to OHSS include a “transfer to” field that has information on migrants transferred to a port of entry to be deported directly from USBP custody. That field is mostly redacted in the USBP FOIA data, so MPI researchers were unable to replicate OHSS’s current methodology with available data.

In the absence of “transfer to” data, the researchers approximately replicated OHSS’s previous methodology, which was to count migrants from countries that could be deported to Mexico or Canada, who had processing dispositions that did not require onward processing in the United States, and who had no record of booking into ICE custody. For its analysis, MPI was unable to link USBP apprehensions to ICE custody data, so MPI researchers focused exclusively on Mexicans and Canadians apprehended at the southwest and northern borders, respectively, with processing dispositions that could permit USBP to deport the migrants without further processing within the United States. Apprehensions with dispositions indicating expedited removal without a fear claim (those marked in the data as “ER” or “EREBC”), reinstatement without a fear claim (“Reinst”), or administrative deportation of an aggravated felon (“ADMDPT”) were counted as removals. Apprehensions with dispositions indicating voluntary departure (“V”) or voluntary return (“VD”) were counted as returns.

The MPI methodology likely excludes a number of non-Mexicans who were deported to Mexico, mainly including migrants from Cuba, Haiti, Nicaragua, or Venezuela who may be removed to Mexico based on an arrangement put in place by the Biden administration, as well as migrants from Guatemala, Honduras, and El Salvador who may be returned to Mexico under certain circumstances. This undercount is likely very small, however, since most of these deportations involve ICE at some point in the process and would be counted as ICE removals and since relatively small numbers of non-Mexicans have been encountered at the southwest border since January 2025.

The MPI methodology also overcounts certain Mexicans who have USBP processing dispositions that permit immediate deportation but who nonetheless are processed by both CBP and ICE, and therefore count as ICE deportations rather than USBP deportations (or in some cases may remain in the United States, still being processed by ICE). The inability to subtract these cases biases MPI’s estimate of USBP removals and returns upward—likely more than enough to offset the downward bias discussed above. As a result, MPI’s overall estimate of USBP removals and returns is likely slightly higher than the actual official number.

C. OFO Removals and Returns

To estimate OFO removals and returns, MPI researchers used monthly and quarterly OFO inadmissibility determinations data from CBP’s FOIA reading room. Public OFO data include a field labeled “duplicate subject indicator,” which MPI used to filter out duplicate records (only including cases for which this field was blank or coded “N”). Where multiple OFO datasets for the same time period were found on the CBP website, the researchers selected the file for which total encounters match OFO reporting in the CBP public data portal. Following this rule, the researchers confirmed that total inadmissibility determinations in the MPI dataset fell within 0.4 percent of total inadmissibility determinations based on the CBP public data portal.

In the case of OFO data, the MPI researchers were able to closely match the current OHSS methodology. First, any inadmissibility determination with a disposition of crewmember refused landing rights, detained on board, or related disposition was coded as an administrative return. Second, all remaining OFO inadmissibility determinations with “I216 custody transfer to” values of “returned to foreign” were coded as deportations. Within this group, inadmissibility determinations with dispositions of visa waiver program refusal, voluntary return, or withdrawal were coded as enforcement returns and those with dispositions of expedited removal (no fear claim), reinstatement of removal (no fear claim), or visa waiver program removal were coded as OFO removals.

All other OFO inadmissibility determinations refer to migrants either paroled into the United States or turned over to ICE or another law enforcement agency. None of these cases were counted as OFO deportations. (Some migrants who receive an OFO inadmissibility determination and are turned over to ICE or another law enforcement agency are eventually removed or returned by ICE, in which case they will show up in ICE removal/return data.)

D. Time Periods

Throughout this fact sheet, “annualized” data for FY 2025 are based exclusively on data for January 20–September 30, 2025 (i.e., the parts of FY 2025 that occurred under the Trump administration). The annualized FY 2025 numbers were calculated by multiplying the total for the 254 days from January 20, 2025, to September 30, 2025, by the ratio of 365/254. The projected annual total for FY 2026 is similarly based on the total count for the 304 days from October 1, 2025–July 31, 2026, multiplied by 365/304. For monthly trend data and comparisons between the first three months and most recent three months of the second Trump term, January 2025 data are based exclusively on January 20–31, 2025, with those 12 days multiplied by 31/12 to simulate a 31-day total.

E. Time in the United States

In this analysis, all CBP deportations (border turnarounds) are assumed to be of migrants who spend less than two years in the United States. (In fact, virtually all of these deportations occur within a few days up to a few weeks.) For ICE deportations, time in the United States is calculated by subtracting “entry date” from “depart date.” A total of 218,560 records were missing entry dates, including 209,687 deportations resulting from ICE arrests, 8,744 resulting from USBP arrests, and 129 from OFO arrests. MPI researchers randomly assigned entry dates in these cases to be proportional to observed dates broken down by arresting agency. To the extent that entry dates are more likely to be missing for people who entered longer ago, as seems likely to be the case, MPI’s methodology understates the length of time deportees have been in the United States, particularly in the case of deportations resulting from ICE arrests.

F. Criminality

Criminality among ICE removals and returns is based on the “case criminality” variable in the ICE removals dataset, using the code “convicted criminals.” Criminality among USBP removals and returns is based on the “criminal convictions” variable in the USBP encounters dataset, using the “Y” code to identify convicted criminals. No criminality data are available for OFO inadmissibility determinations or deportations. Data on the removal of convicted criminals were calculated by filtering deportations for removal dispositions as described above.

G. Estimated Total Departures

MPI’s overall approach to estimating total departures is to assume that the propensity of any given migrant to leave is a function of three factors, among others: 1) the proximity and ease of travel to home countries, as well as home-country conditions that are more or less welcoming to returning migrants; 2) immigrants’ legal status and time in the United States; and 3) negative U.S. economic conditions, the intensity and nature of immigration enforcement, and other “push factors” that encourage people to leave the United States.

To capture differences across countries of origin, the model used to derive the total departure estimates in this fact sheet divides the world into seven country groupings. The first group consists of Mexico, which is assumed to have the highest propensity for return migration (i.e., departures from the United States) as a function of its proximity to the United States, long history of circular migration, and relative political and economic stability. The other groups are: the three Northen Central American countries (El Salvador, Guatemala, and Honduras); Cuba, Haiti, Nicaragua, and Venezuela; the rest of Latin America and the Caribbean; Ukraine; Canada and the rest of Europe; and all of Africa and the rest of the Eastern Hemisphere. The model makes high- and low-end assumptions about the general propensity of migrants from each of these country groupings to depart the United States, expressed in each case relative to Mexico, which represents the model’s baseline. For example, the model assumes Canadians and Europeans are between 60 percent and 95 percent as likely as Mexicans to return home, all other things equal.42

Table A–1. Assumed Propensities of Migrants to Depart, by Country Grouping Relative to Mexican Migrants

Next, MPI researchers drew upon MPI’s unique methodology for assigning immigration status to noncitizens in data from the U.S. Census Bureau’s American Community Survey (ACS) to generate population estimates from each of these seven country groupings in six segments based on time in the United States (less than four years, four to ten years, or ten-plus years) and legal status (unauthorized versus lawfully present noncitizens). ACS data are not yet available for 2025 or 2026, so population data were held constant after 2024. The model assumes more recent arrivals and unauthorized immigrants are the most likely to depart and that the ten-plus-year population and lawfully present noncitizens are the least likely to depart, with the four-to-ten-year population in the middle. The researchers constructed three weighting schemes to place greater emphasis on immigration status or time in the United States versus weighting them more evenly.

Table A–2. Assumed Propensities of Migrants to Depart, by Immigration Status and Time in the United States

After that, the researchers used data from Mexico’s National Institute of Statistics and Geography (INEGI) National Survey of Occupation and Employment (ENOE)43 and the INEGI National Survey of Demographic Dynamics (ENADID)44 to estimate the number of Mexican-born returnees who lived in the United States in the year prior to each year of available data, defining years to match ACS data, which run from July 1 of one year to June 30 of the next.45 (Consistent ENOE and ENADID time series data are not available prior to 2014 or for 2015–17 or 2019–21.) The model then divides the number of Mexicans who recently returned from the United States, as reported by ENOE and ENADID, by the number of Mexican immigrants in the United States to calculate the strength of “push factors” in each year for which data were available. This exercise was done three separate times, accounting for the demographic profile of Mexican migrants in each year and the three different assumptions about how those demographics affect the propensity to depart the United States, as described above.

Finally, the model multiplies the three sets of migration push factors by each of the country groupings, accounting for each one’s demographic segments, and by the set of assumptions about each country grouping’s relative propensity to depart the United States. This step yielded a total of six departure estimates for each year for which ENOE and ENADID data are available. The model’s upper- and lower-bound estimates were defined by the highest and lowest predictions from each year.

To calculate self-departures, the MPI researchers subtracted estimated deportations of immigrants who had been in the country for at least one year from estimated total departures of immigrants who had been in the United States at least one year. To estimate deportations of migrants who had resided in the United States for at least one year, the researchers assumed that all deportations resulting from interior arrests and half of ICE deportations resulting from border arrests fell into this category, with the other half of ICE deportations of border arrests and all CBP deportations assumed to be of migrants in the United States less than one year prior to deportation.46