How do DHS unauthorized immigrant estimates compare with Pew Research and MPI figures for 2016–2020?
Executive summary
From 2016–2020, the Department of Homeland Security’s (DHS) residual-method estimates of the unauthorized immigrant population largely fell in the same numerical band as independent estimates from the Pew Research Center and the Migration Policy Institute (MPI), though small year-to-year differences and divergent trend signals appear because of methodological choices, timing of data, and coverage adjustments [1] [2] [3].
1. Methods matter: all three rely on a “residual” backbone but diverge in implementation
DHS, Pew and MPI all start from the same basic residual approach—subtracting an estimated legally resident foreign‑born population from the total foreign‑born count in Census surveys—but each group applies different imputations, weights and data combinations that produce modest differences in totals and trends [4] [3] [5]. MPI’s published three‑stage method explicitly matches ACS counts to DHS administrative tallies, then uses SIPP and weighting adjustments to correct for coverage error, a multi‑step process that produces estimates compatible with but not identical to DHS and Pew outputs [3] [6]. Pew likewise uses variants of the residual method with reweighting and geographic stratification to harmonize survey series across years [5] [7].
2. The numbers: broadly similar ranges, small differences in trend signals 2016–2020
Comparative tabulations through the late 2010s show DHS, Pew and other demographers reporting unauthorized populations within a similar numeric range for overlapping years, and DHS itself acknowledges the broad similarity with Pew and MPI in resulting estimates [2] [4]. DHS analyses for 2015–2018 indicate that while the three organizations’ point estimates sat close together, DHS registered an upward trend from 2012 to 2016 that was either less pronounced or absent in Pew and CMS series—illustrating how a common range can still transmit different trend narratives [1] [8].
3. Why year‑to‑year differences appear—and why they matter
Small annual gaps arise from sampling variability in the American Community Survey (ACS), the timing of when population controls and census adjustments are applied, differing treatments of temporary parole programs and noncitizen subgroups, and how each group corrects for ACS coverage error and imputes legal status [2] [5] [3]. Analysts and policymakers should note that net change is a blunt measure—Pew and OIS/DHS generally do not fully disaggregate inflows and outflows (entries, visa overstays, removals, status adjustments, voluntary emigration), so similar totals can mask different underlying dynamics [9].
4. Contested claims and consensus: where the debate sits
Some academic exercises have claimed substantially larger unauthorized populations in the mid‑2010s, but MPI and Penn State researchers disputed those higher figures as relying on faulty assumptions and affirmed that rigorous residual‑method estimates place the population in a notably narrower range—roughly the 10.8–12.1 million band in earlier critiques—closer to DHS and Pew outputs [10]. DHS itself and independent reviews stress that methodological convergence across institutions strengthens confidence in the overall scale even as precise year‑to‑year movement remains sensitive to data revisions [4] [2].
5. Bottom line for 2016–2020: agreement on scale, modest divergence on trend detail and timing
For the 2016–2020 period the three leading producers—DHS, Pew and MPI—produced estimates that are substantively consistent in scale because they rely on the same residual foundation and overlapping source data, yet they can tell slightly different stories about short‑term increases or decreases because of choices about survey pooling, reweighting, administrative tallies and coverage error adjustments [1] [3] [5]. The available reporting does not provide a single year‑by‑year side‑by‑side table for every year 2016–2020 in these sources, so precise point‑by‑point comparisons are limited by what each institution has published and by later methodological revisions [1] [6].