How do the Center for Migration Studies, Pew Research Center, and Migration Policy Institute calculate state-level undocumented population estimates and how do their methods differ?
Executive summary
Three leading estimators—the Migration Policy Institute (MPI), the Pew Research Center, and the Center for Migration Studies (CMS)—all start from the same problem (the U.S. Census and ACS do not record legal status) and largely rely on the residual estimation family of techniques, but they implement that backbone differently: MPI uses multi-source imputation and upweighting to produce state and county detail, Pew follows a classic residual approach tied to national administrative data, and CMS applies a modified procedure after 2010 with specific adjustments for undercount that remove the need for certain emigration assumptions [1] [2] [3].
1. The residual method: the common backbone
At base, the “residual method” compares the foreign‑born counted in a survey like the American Community Survey (ACS) or Current Population Survey (CPS) with independent estimates of the legally resident foreign‑born population; the difference—the residual—is attributed to unauthorized immigrants, and this framework underpins CMS, Pew, MPI and DHS estimates [1].
2. How MPI builds state‑level estimates: pooling, imputation and upweighting
MPI’s multi‑stage system assigns legal status to noncitizen respondents in the ACS by combining ACS microdata with SIPP and other administrative inputs, pools five years of ACS to boost sample size for states and counties, and then weights (upweights) the identified unauthorized counts to match country/region totals derived from administrative sources—an approach designed explicitly to produce reliable subnational estimates [2] [4] [5].
3. How Pew does it: a classic residual with national anchors
Pew also uses a residual framework—subtracting estimated legally present immigrants from total foreign‑born counts in survey data—and has produced widely cited national and state trend numbers; Pew has not publicly detailed every auxiliary assumption (notably emigration rates) in the same level of technical detail as MPI or some academic notes, and therefore its emigration treatment is not fully documented in the sources reviewed [1] [6] [3].
4. How CMS differs: a post‑2010 revision and an undercount adjustment
CMS began with the residual method for its 2010 benchmark but adopted a revised procedure for post‑2010 estimates that, among other things, adjusts explicitly for Census/ACS undercount and relies on CPS microdata in later cycles; that revision reduces reliance on external emigration rate assumptions and can produce state estimates that in a few cases approach or exceed the ACS noncitizen counts because of the undercount correction [1] [7].
5. The technical fault lines: emigration, pooling, and upweighting
Key methodological differences that drive divergent state estimates include how each group models emigration (MPI uses Social Security–based emigration rates from long historical records, DHS/CBO use different census‑based rates, Pew’s emigration treatment is less publicly specified, and CMS’s revised approach avoids needing emigration estimates) and whether analysts pool multiple ACS years or upweight to country totals to correct for coverage error—choices that matter most for state and county allocations [3] [5] [1].
6. Why state‑level tallies can diverge and why caution matters
Even with the same residual logic, state tallies differ because small‑area estimates amplify sampling error, choice of survey (CPS vs ACS), pooling windows, the method for assigning legal status to individual respondents, the upweighting procedure to correct undercount, and whether analysts impute or estimate emigration—all of which produce the modest but policy‑salient variation seen across CMS, Pew and MPI state tables [8] [5] [7].
7. Alternative viewpoints, transparency and potential agendas
All three organizations aim for methodological rigor, but tradeoffs reflect institutional priorities: MPI emphasizes subnational detail and transparent multi‑stage imputation; Pew produces concise national and state trend products but tabs some technical choices; CMS prioritizes an undercount‑corrected series and has publicly warned users about small counts and estimates that may exceed ACS noncitizen counts because of those corrections—readers should treat each set as an estimate informed by assumptions rather than a definitive headcount [2] [6] [7].
8. Bottom line
State‑level undocumented population estimates are not raw counts but modeled outcomes rooted in the residual method; differences among MPI, Pew and CMS come down to data inputs (ACS vs CPS, SIPP), emigration assumptions, pooling/upweighting practices, and explicit correction for undercount—understanding those technical choices is essential before drawing firm conclusions from any single state table [1] [3] [5].