How do Pew, CMS and MPI differ in methodology when estimating state undocumented populations?
Executive summary
Three leading sets of estimates—Pew Research Center, the Migration Policy Institute (MPI), and the Center for Migration Studies (CMS)—all start from the same basic “residual” idea (total foreign‑born in surveys minus estimated legal foreign‑born = likely unauthorized), but they diverge in the data combinations, treatment of emigration and undercounts, and how they allocate national totals to states, producing state-by-state variation despite broadly similar national totals [1] [2] [3].
1. The common starting line: residual method and survey bases
All three organizations rely on Census Bureau surveys as their raw counts—principally the American Community Survey (ACS) or the Current Population Survey (CPS)—and implement a residual approach that subtracts estimated lawful residents from the total foreign‑born counted in those surveys to infer the unauthorized share [1] [2] [3].
2. MPI’s three‑stage imputation: combining ACS with SIPP and demographic weights
MPI uses a bespoke three‑stage procedure that imputes legal status for noncitizen survey respondents by comparing ACS microdata with information from the Survey of Income and Program Participation (SIPP), then weights and pools multiple years of ACS and SIPP data to produce state and local estimates—MPI emphasizes detailed characteristic matching and pooled ACS samples to support subnational analysis [3] [4].
3. Pew’s residual with upward adjustment and less transparency on emigration
Pew follows a residual logic and makes explicit upward adjustments to account for survey undercounts, and it has described refinements to incorporate improved immigration flows; however, Pew has not published a transparent emigration estimation procedure in the same way other groups have, leaving a key component of demographic updating less documented in public materials [5] [6] [2].
4. CMS’s revision: a residual origin, a different update strategy, and limited emigration dependence
CMS historically used the residual method for its 2010 benchmark but revised its post‑2010 approach to update annual estimates in a way that, according to reviewers, does not require explicit emigration rate inputs in the same way other models do—a methodological departure that can produce different trend patterns and state allocations [1] [2].
5. Emigration assumptions: small differences with big effects
Estimates diverge in how they model emigration (people leaving the U.S.), and those choices matter for trends and state shares: DHS and CBO have relied on older emigration schedules, MPI uses Social Security‑based emigration rates, CMS’s revision reduces dependence on emigration estimates, and Pew’s public documentation lacks a clear emigration methodology—differences that explain part of the lift between competing totals [2] [7].
6. Undercount adjustments and upward corrections
Pew explicitly applies an upward correction for the known undercount of immigrants in surveys, noting that censuses and surveys tend to miss people and adjusting its totals accordingly; MPI and CMS also incorporate adjustments but use different calibration choices and data pooling that affect state‑level numbers [6] [3] [1].
7. Timing, pooling, and geographic allocation create state‑level spreads
Beyond methodological building blocks, estimates differ because of timing (which benchmark year and which pooled ACS years are used), sample sizes and pooling strategies for small geographies, and the rules used to allocate national residuals to states—factors that produce noticeable state‑by‑state variation even when national figures are broadly comparable [8] [9] [4].
8. Transparency, institutional context, and implicit agendas
Methodological transparency varies: MPI and CMS publish detailed procedures for assigning legal status to microdata, while Pew explains conceptual steps and applies upward adjustments but has not fully disclosed its emigration calculations; funders and institutional mandates matter too—Pew is supported by The Pew Charitable Trusts, which the organization discloses—readers should weigh methodological openness when comparing state estimates [7] [6].
9. Bottom line for state estimates: same toolkit, different knobs
The three groups use the same core residual toolkit anchored in ACS/CPS survey data, but they turn different “knobs”—choice of survey and years, how legal status is imputed or benchmarked with SIPP, emigration and undercount adjustments, and allocation rules to states—which produces the state‑level divergences observed in comparative tables and data tools [1] [3] [9].