How does Pew Research estimate unauthorized immigrants by state and how does that differ from MPI methods?
Executive summary
Both the Pew Research Center and the Migration Policy Institute (MPI) rely on variants of the well‑established “residual” approach—subtracting an estimated lawful foreign‑born population from total foreign‑born counts—but they implement that core idea differently in data sources, imputation steps, and adjustments for survey undercount and protected statuses, producing broadly similar national totals yet divergent state‑level profiles in some places [1] [2] [3]. These methodological choices—how lawful residents are reconstructed, whether and how legal status is probabilistically assigned to individual survey respondents, and how undercounts are corrected—explain most differences between Pew and MPI state estimates [4] [5].
1. The common foundation: the residual method everyone uses
The residual method underlies DHS, Pew, MPI and other independent estimates: demographers estimate the total foreign‑born population from Census surveys, estimate the number who are lawfully resident using administrative counts and demographic accounting, and treat the remainder as likely unauthorized—an approach described in government and academic reviews as the best current practice despite known limitations [1] [6].
2. Pew’s implementation: demographic accounting plus state subdivision
Pew constructs lawful resident counts by applying demographic methods to DHS counts of lawful admissions and projecting those cohorts forward by age, sex, origin and arrival period, with initial calculations done separately for six large states (California, Florida, Illinois, New Jersey, New York, Texas) and the balance of the country and then subdivided by origin groups to produce state estimates [2] [4]. After producing residuals, Pew explicitly adjusts for known survey omissions and applies probabilistic assignment where the group of “potentially unauthorized” exceeds the residual, using those adjustments to reconcile its individual‑level assignments with the residual totals [4] [7].
3. MPI’s implementation: survey imputation that assigns status to individuals
MPI’s three‑stage method imputes legal status directly onto noncitizen respondents in pooled American Community Survey (ACS) data by leveraging the Survey of Income and Program Participation (SIPP), which records green‑card status, and federal administrative counts to calibrate results; this lets MPI generate person‑level classifications and richer state and substate profiles by weighting pooled ACS years and adjusting for undercount [3] [5]. MPI explicitly documents assigning legal statuses in the ACS using classifiers trained on SIPP and administrative data and then slightly adjusting totals to account for ACS undercoverage and to match independent population benchmarks [3] [8].
4. Concrete methodological differences that change state estimates
Pew’s pathway emphasizes reconstructing the lawful population from admissions data and demographic flows and then reconciling survey residuals with probabilistic assignment in a way that preserves internally consistent state totals across time, while MPI’s process emphasizes individual‑level imputation in ACS combined with SIPP‑based model parameters and administrative calibration—differences that affect how Mexican and other origin groups are allocated across states, how short‑term protected populations are counted, and how undercounts are corrected [2] [3] [5]. MPI also documents inclusion of people in “liminal” statuses (DACA, TPS, parole, pending asylum) within its unauthorized‑definition and adjusts pooled ACS with SIPP to enrich characteristic estimates, which can shift state shares compared with Pew’s country‑and‑cohort based decomposition [5] [8].
5. What this means in practice: similar national totals, divergent granular portraits
Independent reviews and cross‑comparisons show that national totals from these approaches are broadly comparable, but state and subgroup differences emerge because of pooling choices, how omissions are adjusted, and whether status is assigned at the person level versus derived from cohort accounting—so policymakers should expect alignment on big trends but variation on county‑or‑state shares and demographic breakdowns [1] [6] [9].
6. Limits, transparency, and contested claims
Both approaches acknowledge core limits—survey omissions, undercount, emigration and mortality assumptions, and model dependence—and both publish methodological details and adjustments; critics who claim “hidden millions” outside these footprints have been challenged by MPI and others on empirical grounds because births, school enrollments and other administrative traces constrain plausible ranges [10] [1]. Where sources disagree, the divergence is methodological rather than conspiratorial: different, transparent choices about data sources, pooling, probabilistic assignment, and undercount adjustments produce different state‑level portraits from the same residual foundation [2] [3].