What methods do Pew, Cato and CMS use to estimate the undocumented population and why do their results differ?
Executive summary
Three major independent centers—Pew Research Center and the Center for Migration Studies (CMS), along with multiple academic groups—rely on variants of the residual estimation method to produce nationwide counts of unauthorized immigrants, but they diverge because of different survey choices, adjustments for undercount, and post‑survey corrections for emigration, mortality and country controls; the available reporting does not document a Cato Institute methodology for estimating the undocumented population [1] [2] [3] [4].
1. What the “residual” method is and why both Pew and CMS use it
The fundamental approach—called the residual method—starts by estimating the total foreign‑born population from Census or large survey data, separately estimating the number of foreign‑born people lawfully resident (from administrative records and demographic accounting), and then treating the difference as the likely unauthorized population [1] [3]. Pew has long described this conceptual subtraction and the many refinements that go with it, and CMS likewise bases its published series on a residual framework applied to Census Bureau surveys [1] [5].
2. Key technical choices that create different headline totals
The differences between Pew and CMS largely trace to concrete methodological choices: which survey is used (CPS versus ACS), how the legal resident total is imputed, whether and how to adjust for survey undercount, and how to model emigration and mortality among immigrants—CMS explicitly builds independent population controls by country of origin and adjusts for emigration and mortality, while Pew documents an upward adjustment for expected undercount and its own imputation procedures for legal residents [5] [2] [6] [1].
3. Survey selection and timing: CPS versus ACS and provisional projections
CMS publishes ACS‑based estimates but has also produced CPS‑based provisional projections for more recent months by scaling CPS counts to its ACS baseline; those choices produce sensitivity to timing and sampling differences—CMS tested CPS‑based projections against ACS‑based CMS estimates and reported projection error within a narrow band for the years tested [5]. Pew historically uses the March CPS supplement for many of its series and explains how sample design and questionnaire timing affect sampling error and confidence intervals [7] [1].
4. Adjustments for undercount, emigration and logical edits
Pew and CMS both recognize that censuses and surveys miss people and therefore include upward adjustments, but they handle the corrections differently: CMS applies “logical edits” to survey records to identify likely legal residents and separately models emigration and mortality by country to control year‑to‑year fluctuations, while Pew documents its own upward adjustment to account for undercount in its final estimate and publishes confidence intervals tied to CPS sampling error [2] [6] [1] [8].
5. Magnitude of the differences and expert debate
Published point estimates for recent years are similar in scale—DHS, Pew and CMS produce estimates around 11 million for 2022 in some accounts—but earlier comparisons show differences on the order of several hundred thousand to more than a million in multi‑year changes, a gap commentators attribute to methodological choices on emigration and population controls rather than to wildly different raw data [3] [9]. Critics such as the Center for Immigration Studies have highlighted those discrepancies and questioned sensitivity to assumptions, and some researchers warn that over‑ or under‑estimating denominators can affect downstream analyses like crime‑rate comparisons [9] [10].
6. Institutional context, transparency and agendas to watch
Both Pew and CMS publish methodology appendices and data tools so users can inspect assumptions—Pew provides detailed methodological appendices and confidence bounds and notes its funding link to The Pew Charitable Trusts, while CMS hosts state and national data with notes about undercount adjustments and small‑sample cautions [6] [1] [8]. The public debate sometimes amplifies differences for political ends, and third parties (e.g., DHS, CBO) use similar residual approaches but with their own administrative inputs, producing broadly comparable but not identical series [3].
7. Bottom line: why numbers differ and what that means for users
Differences between Pew and CMS are not a mystery of completely different theories but of defensible choices about survey source, timing, imputation of legal status, undercount adjustments, and demographic controls—each choice shifts the estimate by hundreds of thousands depending on the year and subgroup [1] [2] [5]. The absent Cato methodology in the supplied reporting means no reliable claim can be made here about how Cato produces any unauthorized‑population estimate [4].