Center for Migration Studies Disputes PLOS ONE Estimates of US Undocumented Population

Robert Warren
Center for Migration Studies

Credit: Vector Illustration/Shutterstock

Center for Migration Studies Disputes PLOS ONE Estimates of US Undocumented Population

A report published in PLOS ONE[1] this month asserts that 22.1 million undocumented immigrants live in the United States. This estimate is twice the number reported by experts in the field. The Center for Migration Studies (CMS) has concluded that the new estimate is based on untenable assumptions and flawed data.

The essential problem with the new report’s methodology is that estimates of emigration are available only for those non-citizens that have established residence in the United States, and the report does not take into account the departure of short-term migrants. The lack of data on outmigration by short-term migrants caused the model to overestimate significantly the population in 2016. An estimate of 22.1 million would imply an undercount rate of undocumented in the 2016 American Community Survey (ACS) four times greater than any Census Bureau estimate of the undercount.[2] The CMS estimate of 10.1 million undocumented immigrants counted in the 2016 ACS is based on sound data and methods.[3] A population of 22.1 million would imply that 12.0 million (22.1 million minus 10.1 million) were omitted from the ACS. In that case, the undercount rate would be an astounding 54 percent.[4] For more information on this issue, see: http://cmsny.org/publications/warren-11million/.

The undercount rates used by researchers that yield estimates in the 11 to 12 million range are less than one-fourth the rate implied by the PLOS ONE estimate. Which is more plausible? Fortunately, there has been a real life test of this question. When the Immigration Reform and Control Act of 1986 (IRCA) was enacted, the former Immigration and Naturalization Service (INS) had to determine the number of likely applicants in each state in order to set up legalization offices. The INS Statistics Division adopted methods and assumptions developed by the Census Bureau (similar to the methods that currently produce estimates of 11 to 12 million), along with a range of assumptions about undercount, to project the number likely to apply at the national and state level. The number of applications for IRCA was within the projected range in nearly all of the states.

In summary, the estimate of 22.1 million undocumented residents in 2016 should be disregarded by policymakers and researchers. The estimates do not positively contribute to the ongoing discussion of undocumented immigration and, unfortunately, they may make it more difficult to resolve the thorny issue.


[1] Fazel-Zarandi, Mohammad; Feinstein, Jonathan S.; and Kaplan, Edward H. “The Number of Undocumented Immigrants in the United States: Estimates Based on Demographic Modeling with Data from 1990-2016,” PLOS ONE. September 2018.

[2] The highest undercount rate in the ACS reported by the Census Bureau was 15 percent (for foreign-born Hispanic males ages 30 to 49). The undercount rate for females in the same age group was 2 percent. The highest rate of undercount in any census reported by the Census Bureau was 14 percent for black males aged 30 to 34 in the 1990 census.

[3] The 2016 ACS counted 20.3 million noncitizens that arrived after 1981 (those that arrived earlier would be legal residents because of IRCA); 8.1 million of these were legal residents (that is, they were very likely to be here legally because they received Medicare, worked for various government agencies, were in the military, were foreign students or temporary workers, were refugees, etc. Of the remaining 12.2 million, CMS developed techniques to remove an additional 2.1 million legal residents, leaving 10.1 million undocumented counted in the 2016 ACS.

[4] If 12 million out of 22.1 million were omitted, as the estimate of 22.1 million suggests, the undercount rate of undocumented in the ACS would be 54 percent (12 million divided by 22.1 million), an implausibly high rate.