Mismeasurement of recent migrants probably isn't meaningfully distorting labor market statistics
The Current Population Survey seems to be doing a decent job reaching them
The United States has experienced an influx of migrants crossing the southern border since the beginning of 2021. If labor market surveys imperfectly capture these migrants, key labor market could be distorted. However, given the number of migrants who have arrived to date, the scope for such distortion is very limited. In practice, the Current Population Survey (CPS) probably does capture most of the influx of migrants.
Migrants have boosted recent population growth, potentially complicating labor market measures
Since roughly the beginning of 2021, an unusually-but-not-historically large number of migrants have entered the United States at the southern border without prior legal authorization, in many cases seeking asylum. The Congressional Budget Office (CBO) estimates that “other foreign national” migration, which includes people seeking asylum, contributed substantially to population growth in 2023 and projects that it will continue to do so in 2024.
Data from the Department of Homeland Security (DHS) suggest that, from the beginning of 2021 through June 2024, about 2.5 million people who are at least 16 years old have entered the United States at the southern border, been apprehended, and then been released into the country while legal processes related to their status progress.1
The arrival of these migrants could potentially complicate our understanding of the labor market because, ironically, the CPS is not designed to measure the size of the current population, only its characteristics. The survey, sometimes called the “household survey,” samples about 60,000 households/addresses each month, and responses are weighted to align with population projects that are established at the beginning of each year.
Given the rapid increase in migrant arrivals and the fact that they are not immigrating through formal channels, it is unclear how well they are captured by the CPS population projections. Since migrants may be able to receive work permits while waiting for their legal claims to be processed, it is also important to understand how they are engaging with the labor market. But if migrants are less likely to respond to the CPS (as one might think plausible for a group with tenuous legal status that is also the focus of intense domestic political sentiments) than native workers or other immigrants, we could be getting a misleading picture of things as fundamental as how many people there are in the country, how many have jobs, or how many want jobs. These are critical components of important labor market indicators like the unemployment rate and the employment-population ratio.
The most missing migrants could matter is not much
Reweighting existing CPS data to account for potential missing migrants can provide some sense of how much this issue could affect labor market statistics. For the sake of simplicity, consider the most extreme scenario, in which CPS population projections do not capture any of the recently arrived migrants, and all migrants who arrive remain in the country through the present. I rescale the CPS sample weights of foreign-born Hispanic non-citizens (FBHNCs) such that the full US population grows with the timing implied by the DHS data to about 2.5 million people larger than its current size (according to the population projections) in June 2024.2
These adjusted weights can also be used to estimate the number of employed and unemployed people.3 The changes in the trajectories of these measures, as well as the size of the labor force, are modest.
Unemployment in particular sees little change due to this adjustment. As of June 2024, the migrant adjustment increased the number of unemployed people by only about 100,000.
That increase in unemployment is much smaller than the increase in employment, which the adjustment increases by 1.8 million. As a result, the resulting changes in commonly used labor market indicators for the full population are quite small: labor force participation, which is the indicator that sees the largest shift due to the adjustment, increases by only about 0.09 percentage points.
One fact about the labor market that the migrant adjustment could help explain relates to the recent emergence of a gap between employment as measured by the establishment survey and the corresponding measure generated by the household survey. This “payroll concept” measure excludes (unincorporated) self-employment and counts jobs rather than people employed, so someone who has two jobs counts twice in this measure. Recently, growth in this measure has been stronger in the establishment survey than in the household survey, which some have suggested could be driven by recent migrants’ jobs being captured by the establishment survey but the migrants themselves being missed by the household survey.
The migrant adjustment closes this gap much more substantially than it shifts primary labor market indicators. The adjustment closes about 40 percent of the gap between the payroll-concept employment measures from the establishment and household surveys. This suggests there is some scope for recent migrants to be missing from the household survey.
Most migrants are probably not being missed
While it is impossible to directly assess what share of recent migrants are being captured by the CPS, we can try to investigate potential missingness indirectly. Consider the share of the population that are FBHNCs. Since the end of 2020, that share has increased by about 0.6 percentage points. If that increase represents only typical migration and growth within the population already in the United States, as the exercise above assumes, then the recent influx of migrants would increase this group’s population share by roughly another percentage point (i.e. +1.6 percentage points since the end of 2020).
We can also compare this group’s population share with its (unweighted) share of CPS respondents to try to get a sense of whether the composition of survey respondents is tracking changes in the population. If FBHNCs’ respondent share increases less than their population share, this could suggest that the CPS could be missing recent migrants even while the population projects incorporate some growth in the FBHNC population. If instead the respondent share increases more than the population share, the CPS could be capturing some share of recent migrants and then effectively down-weighting their responses because weights are based on slower population growth than has actually been realized.
After trending down over the years prior to the pandemic and then dipping sharply in 2020, FBHNCs’ population and respondent shares have been trending up clearly since the beginning of 2021, when the recent influx of migrants began. Importantly, the respondent share has increased by more than the population share. Between December 2020 and June 2024, the respondent share increased by 0.9 percentage points, while the population share increased by 0.6 percentage points.
The 0.9 percentage point increase in the respondent share represents a bit more than half of the 1.6 percentage point increase in the population share implied by the migrant adjustment above meant to reflect an extreme scenario in which the CPS has so far capture no recent migrants (and the adjustment has maximal effects on CPS-based estimates). The fact that we see an increase in the respondent share that is more than half of that maximal magnitude and begins exactly when migrants began arriving suggests we are probably not in that world, and that the CPS is capturing some recent migrants.
Assessing the share of migrants being captured by the CPS remains difficult, but it is likely greater than the 56% implied by comparing the actual and adjustment-implied changes in respondent shares (0.9/1.6). The migrant adjustment exercise assumed that all migrants remained in the United States from their arrival until the present, and that the population projections capture none of the recent increase in migrant arrivals. Both of those assumptions are likely false, meaning that the exercise overestimates the increase in the FBHNC population. The increase in the FBHNC share that we observe via the unweighted respondent share is therefore likely a larger share of the unobserved true increase in the FBHNC share than this exercise suggests at first blush.
Note that this number does not include people who enter the country without being apprehended. These people also contribute to population growth, but I do not know of a good way to incorporate them into the exercise conducted in the next section, and since they lack the ability to gain work authorization, they are less likely to have strong attachment to the formal labor market.
All migrants are foreign-born and non-citizens, and while not all are Hispanic, the large majority are. DHS data suggest that migrants are increasingly coming from countries other than Mexico, most of those countries are in Latin America, and the share coming from elsewhere remains small in absolute terms.
This exercise implicitly assumes that migrants have the same economic outcomes as FBHNCs already in the United States. This assumption is impossible to evaluate but seems reasonable enough for the bounding exercise at hand.