Why the Decline in the Foreign-Born in the Monthly Household Survey in 2025 Is Very Likely Real

Based on the monthly Current Population Survey (CPS) from January to July of this year the Center for Immigration Studies previously reported that the foreign-born or immigrant population (legal and illegal) declined by 2.2 million, with illegal immigrants accounting for nearly three-fourths of the falloff.¹ The August data shows a nearly identical number of immigrants as July, so the decline remains 2.2 million through August. As we noted in our prior report, because of stepped-up immigration enforcement it is possible the decline “was due, at least in part, to a greater reluctance by immigrants to participate in the survey or to identify as foreign-born”. However, based on multiple months of data, response rates to the CPS, continued willingness of survey participants to answer immigration-related questions, slowing job growth reported by employers, anecdotal evidence, and other data, we find no reason to doubt that the decline in the foreign-born is real.

Among our findings:

• The decrease in the foreign-born might be dismissed as a statistical anomaly if only one month showed a much smaller foreign-born population compared to January. But, in fact, multiple months of CPS data show a large decline in the foreign-born after January.
• The Bureau of Labor Statistics reports no sudden increase in the share of households unwilling or unable to participate in the CPS. In fact, in August, survey response rates jumped but there was no increase in immigrants in the survey.
• There has been no increase in the share of respondents to the CPS who were unable or unwilling to answer the specific immigration-related questions in the survey, such as country of birth or citizenship.
• Although the government’s establishment survey of employers does not identify the foreign-born or U.S.-born, it does show a dramatic slowdown in the overall number of employees businesses report having added in recent months.
• The Job Opening and Labor Force Turnover Survey (JOLTS) data shows an increase in people leaving jobs in accommodation and food services, and construction — two industries with large concentrations of immigrants.
• Despite a methodology that is biased toward showing growth in the foreign-born in 2025, the CPS still shows a significant decline so far this year.
• The huge number of stories in major media outlets since the start of this year about immigrants leaving voluntarily, or thinking about doing so, further buttresses our finding that the foreign-born population has declined.

Reasons to Believe the Decline Is Real

Multiple Months of Data Show Trends. As Figure 1 shows, the foreign-born population grew dramatically by 8.3 million from January 2021 to January 2025. This growth represents clear evidence that the CPS was, to a significant extent, able to capture the impact of the border crisis and the sudden surge in immigration, much of it illegal.² The large falloff in the immigrant population over the first seven months of this year is an indication that, like the large increase in the foreign-born over the prior four years, the CPS is again capturing a real and sudden change in migration patterns. The decrease in the foreign-born since January could be more easily dismissed as a statistical fluke or a problem with the data if only one month showed a smaller foreign-born population compared to January of this year. Instead, the immigrant population has been lower in every month since January.

No Sudden Decline in Response Rate. Response rates to the CPS, as reported by the BLS, are found in Figure 2. According to the BLS, a non-response “is the failure to obtain the minimum required information from an eligible housing unit in the sample during the survey collection period”. This happens when “respondents are unable or unwilling to participate, interviewers are unable to locate addresses or respondents, or when other barriers exist to completing the interview”. If immigrants are here but refusing to take part in the survey, then it might show up as lower response rates. As Figure 2 shows, there has been a long-term decline in response rates to the survey.³ However, response rates between January and July of this year are only slightly lower than the long-term trend as shown in Figure 2.⁴

Response rates improved markedly in August, matching a level not seen since October 2024. The response rate generally improves some each August, but not as much as the increase this year. At present, it is unclear if there is some explanation for this jump beyond just the normal fluctuations in the data. But what we can say is that the large increase in August 2025 did not result in a sudden increase in the size of the immigrant population, which remained virtually unchanged from July to August.⁵ This is an indication that a steady decline in response rates from January to July this year, which is consistent with the prior trend, does not seem to be linked to the decline in the foreign-born.

Furthermore, as Figure 3 demonstrates, the response rate has seen the same gradual decline for years even as the CPS recorded very different trends in the size of the immigrant population — large, measured increases in the foreign-born through January 2025, and decreases since then. There is simply no clear indication that overall response rates necessarily drive increases or decreases in the size of the foreign-born in the survey.

Rotation Groups Show No Sudden Shift. The CPS is divided into eight roughly equal-sized rotation groups. Each month, one rotation group leaves and another joins the survey. Members of a group are in the survey for four months, leave for eight months, and then rejoin for four more months. Participants in the CPS know they will be taking part for multiple months and one reason the survey is structured this way is that it should improve the accuracy of responses by creating a relationship with the Census Bureau, which conducts the survey for the Bureau of Labor Statistics. Figure 4 shows the change in the number of foreign-born individuals (unweighted) in each rotation group compared to January 2025 for each month.⁶ Figure 5 shows the same thing using weighted data. The number of immigrants in each group is not that large and by themselves rotation groups are not designed to be fully representative of the U.S. population. As a result, it is not surprising that Figures 4 and 5 show variation in the unweighted and weighted foreign-born sample in each rotation group. But the decline is generally spread across all of the rotation groups since January.

Those in the first rotation group are joining the survey for the first time each month and might be expected to be the most reluctant to take part in the survey since they have no history with the Census Bureau and are being asked each survey question for the first time, including those related to immigration. The first rotation group is shown in red in Figures 4 and 5, which is also referred to as the incoming rotation group. There is no indication of an obvious problem or sudden change in the number of foreign-born respondents captured in the incoming rotation group — weighted or unweighted. Figure 4 shows a modest decline in the number of immigrants in rotation group 1 (unweighted) from January to August, while Figure 5 shows a slight increase in group 1 over the same time period. But overall there is no indication that the government is struggling to capture foreign-born individuals among those joining the survey for the first time each month.

Willingness to Answer Immigration Questions Unchanged. When data is missing in the CPS, the Census Bureau “imputes” a value, which means a response is filled in based on complex methodology. This includes editing values that are implausible. The public-use CPS has a set of variables that reports the share of responses that have been “allocated” by the Bureau. An allocated value is a type of imputation that is based on other information that the respondent or other members of their household have provided, or it is derived from other individuals and households in the same geographic area who have similar characteristics to the individual or household with the missing value or values.

If immigrants were increasingly unwilling to provide information about themselves due to a fear of enforcement, it seems likely that there would be a significant increase in allocated responses to immigration-related questions. Figure 6 reports the share of allocated responses for immigration-related questions based on the unweighted CPS from January to August of this year.⁷ Figure 7 shows the same information using weighted data. Both figures show that allocation rates for place of birth, parental place of birth, and citizenship are all low and stable in 2025.⁸ As discussed above, we would expect that the incoming rotation group would be especially sensitive to the recent increase in immigration enforcement as they are joining the survey for the first time and are asked all of the questions for the first time.⁹ Figure 8 and Figure 9, which use unweighted and weighted data, respectively, only report information from the first rotation group each month. Both figures show low, stable allocation rates from January to August.¹⁰ There is no indication in the public-use CPS that the foreign-born have become increasingly unwilling to answer immigration-related questions in 2025.¹¹

CPS Is Biased Toward Growth, but Shows Decline. The total population in the CPS is weighted to reflect the non-institutionalized population in the United States by a few key control variables such as race, Hispanic origin, age, and sex. Being foreign-born is itself not a control variable and is allowed to vary based on respondent answers, just like unemployment or educational attainment. The sample weights each month are effectively determined at the start of the year in January and are then carried forward based on estimates of deaths, births, and net international migration (NIM) each month. When constructing the weights for 2025, the Census Bureau assumed that NIM for 2025 would be large and positive based on past trends.¹²

Because of the positive and relatively high rate of NIM incorporated into the survey weights for 2025, the overall U.S. population in the survey in 2025 grows by about 150,000 each month. This population growth is primarily concentrated among Hispanics and Asians because such a large share of new migrants is from those groups.¹³ This means growth in certain groups in the population are locked in each month. As a result, the increase in the weighted size of Hispanics and Asians biases the survey toward showing an increase in the foreign-born population because these groups are disproportionately foreign-born.¹⁴ The key point here is that the assumption of significant positive NIM reflected in the weights this year, and the resulting increase in Hispanics and Asians each month in 2025, should, all other things being equal, result in a growing foreign-born population. The decline in the foreign-born in the CPS in the face of a weighting system that assumed significant positive net migration lends further support to the finding that the decline in the immigrant population is real.

Job Growth Shows Significant Downturn. As has been well covered by the media, revised data from the Current Employment Statistics (CES) survey of businesses shows job growth has slowed considerably in recent months. Figure 10 reports the number of jobs added each month based on the CES survey. In the last six months of 2024, employers added 171,000 jobs each month on average. In the first four months of this year, the average slowed to 123,000 a month; and for May to August the average was only 27,000, which likely will be revised downward, perhaps into negative territory, once July and August data are finalized.¹⁵ Figure 10 also shows that in two industries that can be examined separately using the CES survey where illegal immigrants make up a relatively larger share of workers — leisure and hospitality and construction — job growth also slowed.¹⁶ While the slowdown in job growth as measured by the CES survey is consistent with the fall-off in the foreign-born population in the CPS, it should be remembered that the two surveys measure different things. The CPS measures all workers where they live, while the CES asks employers about the number of employees they have.

Partly based on the CES data through July, Jeb Kolko, of the Peterson Institute for International Economics, a globalization-oriented think tank, is probably the most sophisticated analyst arguing that the decline in the foreign-born is not plausible. He states that if the decline is real, “nonfarm payroll employment would need to be revised downward by millions”. Further, in his view it would create an unrealistically low unemployment rate. In our view, this is not the case. First, the January to July 2025 CPS, the period Kolko is focused on, shows that the number of immigrant workers in the CPS fell 1.01 million — not “millions”.¹⁷ The decline is equal to only 0.6 percent of all workers. Equally important, we estimate that one-third of the decline in immigrant workers is made up of illegal immigrants paid under the table.¹⁸ Unlike the CPS, those paid off the books are not included in the CES survey.¹⁹ Finally, there were nearly 44 million working-age (16 to 64) U.S.-born Americans entirely out of the labor force at the start of the year — neither working nor unemployed.²⁰ Some share of the jobs vacated by immigrants were likely filled by working-age, U.S.-born workers whose labor force participation rate did improve some though July. As such, the unemployment rate need not have fallen as jobs can and often are filled each month by those not in the labor force in prior months, leaving the number unemployed unchanged.²¹ In our view, the CES jobs data, especially the lack of growth in recent months, is consistent with the decline in the foreign-born population.

The JOLTS Data. The Job Opening and Labor Force Turnover Survey (JOLTS) measures the number of people who voluntarily leave their jobs each month. These “quits”, as they are called, do not include those laid off or fired. Figure 11 shows quits in the accommodation and food services industry and Figure 12 shows them in construction. These two industries have a significant immigrant component that cannot be examined separately in the JOLTS data.²² Both figures show an increase in people leaving their jobs and tend to confirm the decline in the immigrant population, particularly illegal immigrants shown in the CPS. This is especially true if we ignore the preliminary data for July.²³ That said, the JOLTS, like the CES, does not include those paid under the table, which is the case for many immigrants. Further the JOLTS survey is relatively modest in size, surveying only about 21,000 businesses each month compared to the roughly 121,000 firms across over 600,000 work sites in the CES survey or the roughly 110,000 individuals in the CPS. Still, the JOLTS does seem to indicate that there has been an increase in people leaving two sectors that have relatively high concentrations of illegal immigrants.

Anecdotal Evidence. Since the start of this year, there have been too many stories to count in the media about immigrants, primarily illegal immigrants, leaving voluntarily or thinking about doing so to avoid being caught in stepped-up enforcement efforts. The New York Times has written more than one story about illegal immigrants fearful of immigration enforcement and thinking about leaving, as has the Los Angeles Times. Chicago Business reported on a Waukegan activist and pastor who said she has helped 200 illegal immigrants repatriate this year. The Associated Press wrote in May about families who have left or are planning to do so in the Denver area, where 3,323 students had withdrawn from school through mid-April. Fortune has also written on the impact on schools as illegal immigrants return to their home countries. The Guardian wrote in June about the “self-deportation” of a long-time communications director, an army veteran, and a chef and his wife. NPR wrote about a Kenyan man who had committed marriage fraud and has decided to leave. Most recently, the Wall Street Journal published a feature article on the huge decline in Venezuelans in a Miami suburb. Anecdotal evidence should always be interpreted with caution. However, the enormous number of such stories certainly buttresses the more systematic evidence discussed above.

Can the CPS Be Used to Measure Migration? Citing a Census Bureau working paper, Kolko argues that the CPS cannot really be used to measure immigration trends. It’s true that the working paper does caution against using the CPS to estimate the size of the foreign-born population “when other data is available”. But, of course, no other data is available to estimate migration patterns in the first half of 2025. The Census Bureau’s larger American Community Survey (ACS) is only released once a year, and the latest version of that survey is only available for 2024. Moreover, the working paper Kolko cites acknowledges that both the Congressional Budget Office (CBO), and the Social Security Administration (SSA), which relies on the CBO’s estimates, measure migration in part based on the CPS. The SSA even has a new paper out using the CPS to estimate illegal immigration in particular.

It is worth adding that the Census Bureau itself specifically lists the CPS as one of only two surveys providing information about the foreign-born. The Bureau also publishes detailed tables on the foreign-born based on the CPS, not just the ACS, every year and has done so for years.²⁴ The BLS also puts out employment information on the foreign-born every month based on the CPS. The Federal Reserve maintains a time series showing the 16-and-older foreign-born population back to 2007 from the CPS. Pew Research in August of this year published a report using the monthly CPS showing the large decline in the foreign-born population from January to June 2025. The bottom line is that the government and multiple outside research organizations have used the CPS to measure the size and characteristics of the nation’s immigrant population for years. To suggest otherwise is to ignore an enormous body of research.

Conclusion

Given the size of the 2.2 million decline in the foreign-born population since January, it is not unreasonable to question whether it is at least somewhat overstated. Further, the highly publicized increase in immigration enforcement in recent months could reasonably be expected to reduce the willingness of immigrants to take part in a government survey like the CPS or at least their willingness to identify as foreign-born if they do take part in it. If this is the case, then the number of immigrants in the country would only appear to have declined. For a number of reasons, we believe this is not what has happened. We think the evidence is strong that most or all of the decline in the foreign-born population shown in the CPS is real and primarily reflects a reduction in new arrivals and, in particular, a large increase in emigration.

We think the evidence is strong that most or all of the decline in the foreign-born population shown in the CPS is real and primarily reflects a reduction in new arrivals and, in particular, a large increase in emigration.

The CPS, while certainly not perfect, has been considered a reasonably reliable source of information on social and economic trends for many decades by numerous government agencies and outside researchers that measure migration. The falloff in the immigrant population since January of this year shows up in multiple months of data. The survey was able to capture the dramatic increase in the foreign-born, much of it illegal due to the border crisis, in the prior four years. So it stands to reason that the large falloff in the immigrant population over the first seven months of this year is an indication that the CPS is again capturing a real and sudden change in migration patterns. Also, for reasons we discuss at length in this report, the CPS is biased toward showing growth in the foreign-born population in 2025, yet it still shows a substantial decline since the start of the year.

Looking internally at the survey, we see no sudden decline in response rates. Equally important, based on publicly available allocation rates there has been no increase in the share of respondents refusing to answer immigration-related questions in the CPS. Also, surveys of employers, including both the Current Employment Statistics survey and the Job Opening and Labor Force Turnover Survey, indicate a slowdown in job growth and an increase in people leaving their jobs consistent with a decline in the foreign-born population. Finally, there have been an enormous number of stories in the mainstream media on individual immigrants or groups of immigrants who have left the country or who are considering doing so in response to stepped-up immigration enforcement since January. The large number of anecdotes of this kind further supports the findings in the CPS data.

We cannot entirely rule out the possibility that a large share of the decline in the foreign-born in the CPS in the first half of 2025 was due to a decrease in the ability of the survey to capture immigrants in an environment of increased immigration enforcement. Like any survey, the CPS has limitations. But we do believe that the weight of the evidence indicates that the decline is real. As more data becomes available, we will continue to evaluate the situation to either confirm or challenge this conclusion.

End Notes

¹ The term “immigrant” has a specific meaning in U.S. immigration law, which is all those inspected and admitted as lawful permanent residents (green card holders). In this analysis, we use the term “immigrant” in the non-technical sense to mean all those who were not U.S. citizens at birth. “Immigrant” and “foreign-born” are used synonymously in this report.

² Figure 1 shows a large jump in the immigrant population in January 2025 compared to December 2024. The January CPS was the first government survey to be weighted to more fully reflect the enormous increase in net international migration that the Census Bureau now observes due to the border surge during the Biden administration. The BLS reports that if the new weights are applied back to the December 2024 CPS, it adds 2.87 million people to the 16-plus population in the United States. The BLS is very clear that the increase in the population was due to the bureau’s new, dramatically higher estimate of migration reflecting the border surge from 2021 to 2024.

³ We discuss how the rotation groups in the CPS work in much more detail later in this report as part of the section devoted to that topic. But here it should be noted that individual households are part of one of eight “rotation groups”. Each group is part of the survey for four months, then the group leaves the survey and returns eight months later. The decline in the response rates since January 2025 shown in Figure 2 are primarily not in the first rotation groups — those joining the survey for the first time each month. Rather, the biggest decline is primarily in those being resurveyed, not those joining the survey for the first time each month.

⁴ If we compare the January 2022 to December 2024 slope (trend line) decline in response rates, then by July 2025 the actual response rate was 0.6 percent lower than the trend line would predict. This could be an indication that the drop in the foreign-born population from January to July of this year was caused in part by an acceleration in the long-term deterioration in the response rate due to more immigrants suddenly refusing to take part in the CPS. But 0.6 percent is not large in our view. Further, an increase in people leaving the country itself could lower response rates as the Census Bureau would be unable to find a larger fraction of those selected to be in the survey for the first time or when the bureau attempts to conduct a follow-up survey. In sum, a falloff in response rates in the context of significant outmigration would not be unexpected.

⁵ It should be noted that the number of immigrants in the survey declined from 12,708 in January to 11,600 in July. But in August it increased to 12,259. Though response rates specifically by nativity are not available, this increase could be an indication that the jump in response rates was caused by more immigrants being willing to take part in the CPS. However, the number of immigrants in the sample only returned to about the number it had been in May of this year and the weighted size of the foreign-born in August was still 1.2 million smaller than the May number. Further, the number of immigrants in the weighted data changed by a trivial 0.05 percent from July to August. It is unclear if the increase in immigrants in the August sample is meaningful. What we can say is that the increase in foreign-born respondents in the August sample did not result in an increase in the foreign-born once the data is weighted.

⁶ When looking at the rotation groups, it must be remembered that they do not represent the same group of people followed for eight months. Instead, each numbered rotation group reflects people moving through the survey so that the people in group 1 become group 2 in the following month and group 3 becomes group 4 and so on.

⁷ The CPS uses a different approach than the American Community Survey (ACS) and the decennial census long forms of the past to determine nativity and citizenship. The CPS combines responses to the place of birth and the parental place of birth questions to determine the foreign-born population. The CPS assumes all respondents whose place of birth is the United States or outlying areas of the United States, or whose parents were born in the United States, to be native-born American citizens. That is, they were U.S. citizens from the time they were born. These individuals are not asked their citizenship. The rest of the population is asked “Are you a citizen of the United States?” The two possible valid responses to the citizenship question are: citizen through naturalization and not a U.S. citizen. When results are processed by the bureau, responses to place of birth, parents’ place of birth, and citizenship are combined to produce a variable in public-use data that is very much like the citizenship variable in the 1990 and 2000 census long forms and the ACS, with five possible values — native-born, born in U.S; native-born, born in U.S. outlying area; native-born, born abroad of American parents; foreign-born, naturalized citizen; foreign-born, not a U.S. citizen.

⁸ The allocation variables in the public-use CPS specifically contain a response for those who refused to answer the survey. The share of those in the survey who specifically indicate that they refused to answer an immigration-related question is roughly 2 percent or less in both the weighted and unweighted data.

⁹ Looking at allocation rates for the first rotation group each month is especially important because respondents are asked their place of birth or parental place of birth only when they join the survey, typically when they are part of the first rotation group. New people who join a household in other rotation groups for the first time are also asked this question.

¹⁰ The allocation rates reported in the public-use data do not cover all forms of imputations. To the best of our knowledge, imputations other than allocations are not publicly available for immigration-related questions in the CPS.

¹¹ We have also looked at allocations for only those who are foreign-born and again find no increase in the share of immigrants whose responses to immigration-related questions were allocated for January to August of this year.

¹² NIM in 2024 was very high, and the Census Bureau population estimates are the basis of the population controls used for the CPS as opposed to other reasons for the missing data. The January 2025 CPS was the first to be weighted to fully reflect the border surge. BLS reported when the January 2025 data was released that if the new weights are applied back to the December 2024 CPS, it adds 2.871 million people to the 16-plus population, 2.121 million (74 percent) of whom were Hispanic and Asian. The BLS is very clear that the increase in the population was due to the new, dramatically higher estimate of migration for the past year.

¹³ From January to August 2025 the combined population of Hispanics and Asians in the monthly CPS grew by 1.01 million, from 90.6 million to 91.61 million, accounting for 94.5 percent of the 1.07 million increase in the total U.S. population in the survey. As already discussed, the 1.07 million increase in the total population in the CPS is a statistical artifact due to the sample weights, which assumed NIM would be large, positive, and primarily composed of Hispanics and Asians. If we accept that the huge decline in the foreign-born in the CPS over the first seven months of this year is real, there is no question that the total U.S. population must have declined from January to August. Natural population increase (births minus deaths) in the first seven months of 2025 should only have been roughly 291,000 and would be completely swamped by the outmigration of the foreign-born. But the Census Bureau will not release new estimates of NIM until December of this year. NIM is estimated by the bureau from July to July of each year. The sample weights in the CPS will then also be adjusted to reflect this situation starting in January 2026. All of this reflects the unpreceded nature of the massive increase in the foreign-born from January 2021 to January 2025, and then the subsequent decline under President Trump. The way the Census Bureau weights the CPS at the start of each year was never designed to account for such huge swings in migration.

¹⁴ In the August 2025 data, 72.5 percent of the foreign-born who said they arrived from 2020 to 2025 were Hispanic or Asian.

¹⁵ At the time of this publication, figures for July and August were preliminary. In recent months, revisions to this data have been downward, though there is no way to determine for certain if there will be significant revision for these two months or to know in which direction the revision will be.

¹⁶ While immigrants make up a large share of workers in the farm sector, agriculture makes up only a tiny portion of the overall labor force and only a small share of all immigrant workers, including illegal immigrants. Further, the CES does not capture farm jobs. The CPS shows some gain in farm jobs. This is not surprising because the winter months often represent the low point in agricultural employment, while employment in the sector often peaks in the summer and early fall.

¹⁷ More than half of the decline in immigrants is among children, older people, and even working-age people not working.

¹⁸ Our prior analysis through July found 72 percent of the decline in the foreign-born were illegal immigrants, which translates to a decline of about 720,000 among immigrant workers. If we further assume 60 percent of illegal-immigrant workers are paid on the books, either because they have work authorization (e.g. DACA, TPS, parole, and some asylum applicants) or provided their employer with false or stolen documentation, it still means that about 290,000 of the one million decline in immigrant workers in the CPS would be illegal immigrants paid off the books who would not show up in CES data.

¹⁹ The CES sample frame and survey weights are based on the BLS’s Quarterly Census of Employment and Wages (QCEW) program, which comes primarily from state unemployment insurance (UI) programs and only captures those covered by unemployment insurance. Those not covered by UI, such as illegal immigrants paid off the books, are not in the survey.

²⁰ To be considered unemployed, one has to report in the CPS that you have looked for a job in the prior four weeks.

²¹ For the reasons outlined above in the section entitled “CPS is Biased Toward Growth but Shows Decline”, the weighting scheme used in the CPS means that changes in the numerical size of the U.S.-born population cannot be taken entirely at face value given the decline in the foreign-born.

²² The CES and the JOLTS do not use the same industry classification systems, though there is some overlap.

²³ It should be noted that the JOLTS data for all industries does not really show an upward trend in the number of people leaving their jobs for the entire population in 2025.

²⁴ The Census Bureau estimates are from the March CPS, known as the Annual Social and Economic Supplement.