How do MPI and Pew estimate state unauthorized populations differently and which is more conservative?

1. How both start from the same core method but diverge in execution

Both MPI and Pew begin with the residual estimation logic: use Census Bureau survey data (ACS or CPS) to count all foreign‑born residents, estimate how many of those are lawfully present using administrative inputs or statistical imputation, and treat the remainder as likely unauthorized — a standard approach described in reviews and used across DHS, Social Security, and other researchers ^{[1] [2]}.

2. MPI’s state estimates: three‑stage imputation, SIPP calibration, and an undercount adjustment

MPI’s approach explicitly layers three data sources: the American Community Survey (ACS) for large state/local samples and characteristics, the Survey of Income and Program Participation (SIPP) to identify lawful permanent resident status, and federal administrative records (mostly DHS) to anchor counts by legal category; MPI then imputes legal status to ACS respondents, weights estimates to demographer Jennifer Van Hook’s population controls, and applies a small upward adjustment to account for the ACS undercount of unauthorized residents ^{[5] [6] [4]}.

3. Pew’s state estimates: detailed residuals by state, origin and arrival cohort, plus omission adjustments

Pew computes residuals for age‑gender groups in six large states and the remainder of the country, further subdividing by country/region of birth and period of arrival, and then adjusts those residuals for known survey omissions (separately treating Mexican versus other groups where sample size permits); after producing residual totals, Pew assigns individual survey respondents a likely status so the microdata align with the aggregate residuals — a process Pew says improves internal consistency and state comparability ^{[7] [8]}.

4. Why Pew’s national/state totals tend to be higher (and less conservative)

Pew’s recent methodology and data window extend through July 2023 and explicitly incorporate adjustments for survey omissions and fine‑grained cohort structure, producing a national total of about 14 million in 2023 — a larger figure than MPI’s mid‑2022 ACS‑based estimates (MPI reported about 11.3–11.4 million for 2021–2022 in its releases, though some MPI presentations that pool later ACS years give larger mid‑2023 snapshots) — meaning Pew’s state allocations will commonly be larger as well when applied to the same geography and time frame ^{[3] [9] [4] [10]}.

5. Methodological reasons that explain conservatism vs. expansiveness

Differences are traceable to timing and inputs: MPI’s reliance on SIPP and DHS administrative anchors, plus conservative adjustments for undercount and careful weighting to academic demographers’ controls, tends to yield lower, more conservative state-level totals when using the same ACS baseline; Pew’s more elaborate residual subdivisions, omission corrections, and rolling updates to mid‑2023 raise aggregate counts and thus state estimates ^{[5] [7] [8]}.

6. Caveats, competing perspectives, and what this means for interpreting state numbers

Neither product is the definitive count; both teams use accepted residual logic and both acknowledge limitations of survey timing, undercounts, and exits from the population—experts note residual‑method similarity across DHS, CMS, MPI, and Pew even as specific choices produce variance ^{[1] [2] [11]}. Importantly, part of the discrepancy reflects timing (MPI’s most recent ACS inputs may predate the record border encounters of 2023, which Pew’s July 2023 snapshot captures) and methodological tradeoffs between avoiding overcount and correcting for survey omissions ^{[9] [3]}.