This report presents estimates of the number of unauthorized immigrants in the United States as of July 2023 and for earlier years. As noted earlier in this report, the estimates presented here supersede all previously published Pew Research Center estimates.
The principal survey data used to produce the estimates and modifications made to the underlying surveys is described below (Methodology B). Two other methodology sections describe the other parts of the report’s methodology. The first (Methodology A) describes the methods used to produce the estimates of the U.S. unauthorized immigrant population and its characteristics. The third section (Methodology C) provides details on how the estimates are rounded and the definitions of various concepts used in the report.
Overview
Estimates of the size and characteristics of the unauthorized immigrant population rely on official U.S. Census Bureau surveys – the American Community Survey (ACS) and the Current Population Survey (CPS). These surveys provide information on the total foreign-born population and detailed characteristics of immigrants. They, in turn, rely on official Census Bureau population estimates for survey weights and population numbers. This methodology chapter describes the role of the population estimates and how we have incorporated the bureau’s normal revisions to these estimates into our estimates of the unauthorized immigrant population.
The estimates of the unauthorized immigrant population for 2005-2019 and 2021-2023 use the ACS, while those for 1995-2003 use the Annual Social and Economic Supplement (ASEC) to the March CPS. The estimates for 1990, which cover only state totals and selected countries of birth, use the 1990 decennial census. The estimates for the unauthorized immigrant population that we produce from these surveys are comparable across all years. This allows for tracking trends over time in the size and characteristics of the unauthorized immigrant population.
However, to achieve this consistency, we made modifications to some of the surveys. Each year’s surveys, as released by the Census Bureau, agree with population figures they develop for that year. But these population estimates may not be consistent with previous years because of changes in the bureau’s methods for estimating population and because of breaks in the population series that occur when the results of a new decennial census are introduced into the population estimates.
This section of the report highlights the causes of specific discontinuities in the Census Bureau’s population estimates which lead to inconsistencies in the population figures from the surveys. It also describes the methods we have used at Pew Research Center to make modifications that allow for comparisons across years. The modifications bring the underlying bureau surveys in line with actual population changes over the more than three decades covered by our estimates of unauthorized immigrants. We focus in some detail on revisions to the 2022 and 2023 population estimates that affect recent ACS data.
(More details regarding the survey designs and sample sizes of the ACS and CPS are described below.)
Official population estimates and surveys
The Census Bureau regularly issues new estimates of the U.S. population as of July 1. Initial estimates for the nation and states are released in December of every non-decennial census year. The annual estimates are referred to by their “vintage,” or the year they are released. For example, the population estimates for July 1, 2024, are released in December 2024 and are called the Vintage 2024 population estimates. Each annual release also updates the previous vintage and includes population estimates for July 1 of every year since the most recent decennial census.
During the ensuing year, the bureau releases additional estimates with more demographic detail (age, sex, race/Hispanic origin) and more geographic detail (counties, cities, places). These estimates are also referred to as “postcensal” population estimates because they update the population from the last decennial census.
The postcensal estimates are used to weight the Census Bureau’s surveys. Each year’s ACS is consistent with the official population estimates released in the year of the survey. For example, the 2023 ACS is consistent with the Census Bureau’s Vintage 2023 population estimates released in December 2023. That is, the weights in the survey produce population figures from the survey that agree with the Vintage 2023 populations for states and smaller areas subdivided by age, sex, race and Hispanic origin.
Similarly, each year’s March CPS ASEC is weighted to the most recent Census Bureau population estimates, but in this case the weights are from the previous year’s vintage. So, for example, the weights for the March CPS ASEC for 2003 are based on the Census Bureau’s Vintage 2002 population estimates for states and age, sex, race and Hispanic origin groups.
When the Census Bureau updates their postcensal population estimates for the years since the previous decennial census, they almost never update the weights for previous surveys. In most years, the revisions to the time series of population data are fairly small, so any potential updates to the previous surveys would be negligible. However, when revisions to population estimates are large – as they were for Vintage 2024 – the impact on recent surveys can also be big. This was the case for the Vintage 2024 population estimates and the 2022 and 2023 ACS. Large revisions can result in major discontinuities and inconsistencies in estimates of the unauthorized immigrant population that are based on these surveys.
After each decennial census, the bureau assesses the quality of population estimates for the previous decade based on how closely the estimates series matched the census count. The bureau then issues a set of population estimates for the previous decade that are consistent with both the previous decennial census and the new one. These are called “intercensal” population estimates because they align with two censuses. The intercensal population estimates can sometimes result in very large discrepancies from the previous postcensal estimates. However, regardless of the magnitude of any disagreement between postcensal and intercensal population estimates, the Census Bureau has generally not updated its surveys and their weights to be consistent with the new intercensal population estimates.
To address these inconsistencies, Pew Research Center has developed updated survey weights that are used in our estimates of the U.S. unauthorized immigrant population. The Center has reweighted some of the bureau’s surveys to take into account the intercensal and revised postcensal population estimates. The unauthorized immigrant estimates for 1995 and 1998 are based on reweighted March CPS ASEC surveys consistent with the 1990-2000 intercensal population estimates. The unauthorized immigrant estimates for 2003 (CPS-based) and for 2005-2009 (ACS-based) are based on reweighted surveys consistent with the 2000-2010 intercensal population estimates. We have also modified the 2022 and 2023 ACS to take into account the large revisions to net international migration in the Vintage 2024 population estimates. (The revisions to the pre-2010 surveys are described in detail in the methodology of this previous report.)
Census Bureau’s revised estimates of net immigration, 2021-2024
When the Census Bureau issued new Vintage 2024 population estimates for 2020-2024, the estimates included a significantly higher estimate of net immigration after July 2021 than had been included in previous population estimates. These revised population estimates more accurately reflect the level of immigration to the U.S. for those years.
However, these revisions meant that our published 2022 estimates of the unauthorized immigrant population (based on the Vintage 2022-weighted ACS) were lower than what was implied by the new Census Bureau population estimates. Similarly, if we based our 2023 estimates of the unauthorized immigrant population on the already-released 2023 ACS (weighted to the Vintage 2023 estimates), these would be lower than what was implied by the Vintage 2024 population estimates of international migration.
For July 1, 2021, to June 30, 2022, the bureau’s Vintage 2024 estimate for net international migration was almost 1.7 million, an increase of 700,000 over the previous Vintage 2023 (and 2022) estimates for that year. These revised net immigration estimates were based on administrative data sources for immigrants admitted under humanitarian programs not previously part of the bureau’s population estimates. Virtually all of the additional immigrants added as a result of the Census Bureau’s adjustment are included in what we define as our unauthorized immigrant population. As a result of the bureau’s revised population estimates, we updated our estimate of the unauthorized immigrant population to 11.8 million in 2022, up from our original estimate of 11.0 million that used the bureau’s original ACS figures.
For the next year (July 1, 2022, to June 30, 2023), the bureau’s new net immigration estimate of 2.3 million was another 1.2 million higher than the previous published estimate. The total U.S. population for July 1, 2023, in the Vintage 2024 population estimates was 336.8 million. In contrast, the total population for the same year (as measured by the 2023 ACS) was 334.9 million. The difference of about 1.9 million was almost all attributable to immigration during 2021-2023 that the unadjusted 2023 ACS did not capture. In other words, nearly the entire revision was due to migrants that would be considered unauthorized.
Modifications to 2023 ACS
The Center’s approach to making the 2023 ACS consistent with the Census Bureau’s Vintage 2024 population estimates involves creating copies of certain existing households in the microdata (called “replicates”) and making adjustments to individual weights for other individual cases in the survey. The households and individuals selected for replication and adjustment depend on the respondents’ countries of birth and year of entry to the U.S. plus their state of residence. The first step in the process involves developing “targets” for the adjustment, or an estimate of how many people need to be added to ensure that the population totals for the immigrant population and states in the augmented 2023 ACS dataset are equal to the Vintage 2024 population estimates.
Adjustment targets
The Census Bureau’s revised methodology for the Vintage 2024 population estimates added certain humanitarian immigrants to the net international migration estimate for two years that the Census Bureau uses for its estimates: July 1, 2021-June 30, 2022 (designated here as CY21-22) and July 1, 2022-June 30, 2023 (or CY22-23). At the time Pew Research Center implemented these adjustments (January-February 2025), the information available was limited to the total population for each year and components of demographic change, particularly net immigration for census years, for the U.S. and states. The targets for the adjustments are: (1) the difference between net international migration for CY21-23 from Vintage 2024 minus Vintage 2023; and (2) any remaining difference in total population. The first represents additional immigrants to be added to the population, and the second component, which is very small, is change in the U.S.-born population.
The Census Bureau’s documentation does not provide further information on the countries of birth for the additional immigrants, nor their race/ethnicity. It does provide information on the type of immigrants added. So, we constructed estimates of the countries of origin for the additional immigrants from the original sources:
- Refugees: Refugee admissions from U.S. Refugee Admissions Program, collected by the State Department, Refugee Processing Center by state and country of birth.
- U.S. Border Patrol (USBP) releases: Reported number of migrants released at the southwest border from the Department of Homeland Security (DHS) Office of Homeland Security Statistics (OHSS) for major countries of birth.
- Office of Field Operations (OFO) paroles: Reported number of migrants paroled at the southwest border from DHS OHSS for major countries of birth.
- Other paroles: These include CHNV paroles from Cuba, Haiti, Nicaragua and Venezuela from DHS OHSS by country of birth, as well as Pew Research Center estimates of Ukrainians admitted under the Uniting for Ukraine (U4U) program and Afghanis admitted under the Operation Allies Welcome (OAW) program.
The total of the targets for major countries of birth and regions is adjusted to agree with the total first derived from the Census Bureau’s population estimates. Only the refugees in these estimates are lawful immigrants; all others are part of what we define as the unauthorized immigrant population.
Selection of cases for replication and adjustment
As a first step in making the necessary adjustments, we made preliminary status assignments to identify individual ACS respondents as lawful temporary migrants, refugees and potential unauthorized immigrants. For replication, we selected households including only recently arrived immigrants (post-2020) who were refugees or potential unauthorized immigrants. These households could also include U.S.-born children born after the parents’ arrival in the U.S. Not all of these households were replicated, as the selection was guided by the national targets by status and country/region of birth and the state targets.
Some additional recently arrived individuals and families in other households were also selected for a separate weight adjustment (i.e., not replication). The weight adjustments for these immigrants were limited to a factor in the range of 0.6 to 1.4 so as not to distort the relationship of the adjusted weights to other household members.
After the initial set of replications and adjustments, we checked the alignment of the newly weighted ACS respondents with the target totals for states and for countries and regions of birth. A final set of adjustments, generally small, was made. The resulting reweighted ACS data for immigrants added to the 2023 ACS agreed with the Census Bureau’s estimates of additional net international migration for CY21-23 in the Vintage 2024 population estimates.
Final adjustments
To align with the Vintage 2024 estimates totals for states in 2023, some relatively small adjustments were still required to the native population. In some states, adjustment required reducing the U.S.-born population slightly, but most of the final adjustments added more people. We selected individual respondents with relatively large ACS weights and made adjustments of less than 5% to individual weights.
Use of the supplemented 2023 ACS dataset
The result of the modification and adjustment is a 2023 ACS dataset with records for individuals and households that reproduces the state totals in the Census Bureau’s Vintage 2024 population estimates for 2023. The new dataset has an immigrant population that is larger than in the original 2023 ACS by exactly the additional immigrants added by the Census Bureau through their Vintage 2024 revisions for each state. We used this dataset to estimate the size of the unauthorized immigrant population included in the ACS following the methods described in Methodology A.
Next, we assigned immigrant status – lawful temporary resident, lawful permanent resident and unauthorized immigrant – to individual respondents using the methods described in Methodology A. Then, the estimates were adjusted for omissions from the ACS by adjusting the individual weights in the dataset. This final dataset is the “augmented 2023 ACS” and is the source for the 2023 estimates published in this report.
Modifications to 2022 ACS
The modifications to the 2022 ACS followed roughly the same procedures as for the 2023 ACS. Recently arrived immigrant households were selected for replication; individual immigrant respondents had their weights adjusted; and individual U.S.-born respondents had small adjustments to align with state targets for additional net international migration and the total population. The two main differences from 2023 were the calculation of population targets for adjustment and the nature of the dataset in which the adjustments were made.
The original 2022 ACS used weights based on the Vintage 2022 population estimates. So, the calculation of targets for revised state population totals involved comparing the Vintage 2024 population estimates for states in 2022 with the Vintage 2022 population estimates for states. Similarly, the additional immigration total needed was the difference of the Census Bureau’s revised estimate of net international migration for CY21-22 (only) minus the Vintage 2022 estimate for this figure, by state.
Because we had already constructed a modified dataset for the 2022 ACS with immigration statuses already assigned, we could select households for replication and individuals for weight adjustment that conformed to the statuses implied by the Census Bureau’s adjusted immigration levels. Specifically, most of the adjustment went to individuals and households assigned as unauthorized immigrants in our 2022 ACS dataset; the remainder went to individuals who were identified as refugees in 2022. After the adjusted data set was aligned with the target totals, we applied the adjustments for undercount following the same procedures we had used in developing our original estimates. The augmented dataset for the 2022 ACS was used to produce updated estimates for 2022 published in this report.
Design of the ACS and CPS-ASEC
The ACS is an ongoing survey conducted by the U.S. Census Bureau that represents the U.S. population as of July 1 of each year. The survey collects detailed information on a broad range of topics, including country of birth, year of immigration and citizenship – the information required for the residual estimates. The ACS has a continuous collection design with monthly samples of almost 300,000 households. The initial sample has been slightly more than 3.5 million addresses each year since 2012.9 Initial nonresponse addresses are then subsampled for further interviews. The final sample has been almost 2 million addresses since 2021 and exceeded 2.3 million addresses for several years in the 2010s.
The ACS began full-scale operation in 2005, covering only the U.S. household population. Since 2006 it has covered the entire U.S. population. ACS data is released by the Census Bureau in September and October for the previous year. For this report, public-use samples of individual survey records from the ACS are tabulated to provide the data used in the estimation process. The public-use file is a representative 1% sample of the entire U.S. (including more than 3 million individual records for each year since 2008 – 3.4 million in 2023). The public-use data employed in these estimates was obtained from IPUMS-USA.
The other survey data source used for residual estimates comes from March Supplements to the CPS. The CPS is a monthly survey currently of about 60,000 U.S. households conducted jointly by the U.S. Bureau of Labor Statistics and the Census Bureau. The March CPS Supplement has a greatly expanded questionnaire that collects detailed information on income sources, participation in a range of government programs and health insurance coverage. The March sample is also expanded over the regular monthly CPS by including additional households with children, Hispanics and specific race groups. The CPS was redesigned in 1994 and, for the first time, included the information required for the residual estimates (i.e., country of birth, date of immigration and citizenship). The CPS universe covers the U.S. civilian noninstitutional population. The March Supplement data from the CPS is released by the Census Bureau in September for the previous March. The public-use files for the CPS ASEC include all households that participated in the survey.
The CPS ASEC sample, while large compared with many surveys, is much smaller than the ACS. From 1995 through 2000, the sample was around 50,000 households. Beginning with 2001, the March Supplement sample was expanded significantly to more than 75,000 households through 2011. Before the ACS was fully implemented in 2005, the March CPS was the principal option for making estimates of the size of the unauthorized immigrant population.
Because of the much larger sample size in the ACS (more than 3.4 million sample cases in 2023, including more than 425,000 foreign-born cases) than in the March CPS (146,000 sample cases in 2023 with about 21,000 foreign-born cases), the ACS-based estimates should be considered more accurate than the CPS-based estimates. The ACS-based estimates are much more precise as well – meaning that the sample variability and margins of error are much smaller in ACS-based estimates.
