DHS’s 2014 Estimate of Undocumented Residents is About One Million Too High

In March 2018, the US Department of Homeland Security (DHS) released estimates of the undocumented population for 2014. As the result of flawed assumptions, the DHS estimates for 2014 are overestimated by about one million. This overestimate, along with the reported population increase of nearly 700,000 from 2012 to 2014, runs the risk of misleading policymakers and the public about the actual size and direction of change of the total undocumented population.

The flawed assumptions that led to the overestimate involved two important components of the DHS estimation procedure – undercount of the undocumented population and annual emigration of the legally resident population. The overestimate occurred mostly because DHS used constant rates (i.e., the same rates for different lengths of residence) each year to estimate emigration and undercount rather than using rates that decline with each year of residence. As shown below, the largest[1] part of the DHS overestimate was related to their adjustment of American Community Survey (ACS) data for undercount – about 500,000. The second largest problem with the DHS methodology – estimates of emigration of legal residents – accounted for nearly 300,000 of the overestimate.

For estimating both emigration and undercount, the appropriate procedure is to use rates that are highest for recent entrants and that decline with length of residence in the country. This reflects the fact that legal immigrants are less likely to leave the country, and undocumented immigrants are less likely to be undercounted by the Census Bureau the longer they live in the United States. Failure to recognize declines based on length of residence leads to overestimates of both emigration of legal residents and the level of undercount of undocumented residents. Further, the impact on the total estimated undocumented population are cumulative, and the effects accelerate when the number of arrivals is declining, as has been the case in recent years.  Ironically, DHS originally developed and used the concept of declining-rates-with-length-of-residence for both of these components in deriving its 2000 estimates, then changed to its present simpler, but flawed, method when it began making annual estimates in 2005.

The Pew Research Center – the only other organization that uses methodology similar to DHS – uses a set of graduated rates to make annual estimates of both emigration and undercount. Emigration is not a factor in the CMS estimates because estimating the legal population in not part of its methodology; however, CMS uses graduated undercount rates by length of residence (nearly identical to the rates DHS used in 2000) to adjust its estimates for undercount.

There is a more direct way to evaluate the 2014 DHS estimates that avoids assessing its assumptions, as well as the more than 30 years of administrative data, that DHS uses in the residual method.[2] The 2014 ACS counted a total of 19,974,000 noncitizens that arrived after 1981. That figure includes the entire undocumented population. Next, we can remove 7,913,000 that are very likely to be legal residents. [3] The remainder, 12,061,000, still contains the undocumented population, but it also includes a large number of legal residents that could not be removed in the previous step. Research by both CMS (see Warren 2014) and the Pew Research Center has established that roughly 85 percent or so of the remainder (i.e., total noncitizens minus likely legal residents) are undocumented residents. Multiplying 12,061,000 by .855[4] yields a total of 10,312,000 undocumented residents counted in the 2014 ACS. Adjusting the 10.3 million for undercount of 6.6[5] percent produces an estimate of 11,041,000 undocumented residents in 2014. There are no reasonable alternative assumptions that would raise the 11.0 million estimate computed here by even a few hundred thousand. The conclusion is that the DHS estimate of 12.1 million for 2014 is overestimated by about a million.

[1] Other aspects to the DHS methodology that caused the remainder of the overestimate of the total population could be improved fairly easily. For example, instead of beginning with 1982, the year after IRCA eligibility ended, DHS included 1980 and 1981 ACS data in the estimates. Not many undocumented residents that entered before 1982 were still living here and undocumented in 2014, and if a small number were, estimating them would require accurate administrative data as well as estimates of emigration and mortality for every year from 1980-81 to 2014; such data are not available.

[2] In the residual method, estimates of the legally resident foreign-born population are derived from administrative records and assumptions about emigration and mortality, for each year of entry, from arrival to the survey date. Then, the estimated legal population is subtracted from the total foreign-born population, usually from a survey or census, to estimate the number of undocumented residents counted. Finally, the difference, or residual, is adjusted to account for omissions from the survey or census.

[3] The figure of 7,913,000 includes legal non-immigrants (e.g., recently-arrived students and their families); selected immediate relatives of adult U.S. citizens; recipients of Social Security and Medicare; Medicaid (with limits and exceptions); from ‘refugee’ countries; government workers; in, or employed by, the Armed Forces; veterans; and those who work in occupations that usually require legal status (e.g., judges, air-traffic controllers).

[4] The percentage of .855 was used here, instead of the .85 figure mentioned in the text, because it is the percentage that CMS used to estimate number of undocumented residents counted in the 2014 ACS.

[5] The set of undercount rates by length of residence that DHS used to derive estimates in 2000 produced a total undercount rate of 10.0 percent. The total overstay rate – using the same set of undercount rates – dropped to 6.6 percent in 2014. This occurred because the number of new undocumented arrivals dropped steadily after 2000; therefore, relative fewer undocumented residents were new arrivals (subject to higher undercount rates) and relatively more have been here longer (and, thus, are subject to lower undercount rates). Failure to take the changing composition of undocumented immigration caused much of the overestimate in the DHS estimates for 2014.