How do DHS, Pew Research Center, and Center for Migration Studies estimates of undocumented immigrants differ?
Executive summary
The three leading estimates — the Department of Homeland Security (DHS), the Pew Research Center, and the Center for Migration Studies (CMS) — start from the same basic toolset (the Census Bureau’s surveys and the “residual” approach) but produce different headline totals because they use different micro‑adjustments, reference dates, and input data choices: DHS reported about 10.99 million as of January 2022 while Pew’s revised work shows 11.0 million for 2022 and a much larger 14 million for 2023 after reweighting, and CMS’s series produces slightly different mid‑range results (roughly 10.9–12.2 million depending on the year and method) — differences driven by undercount assumptions, which surveys are used (ACS vs. CPS), how temporary protection statuses are counted, and how recent migration flows are incorporated [1] [2] [3] [4].
1. Shared foundation: the residual method and survey data
All three organizations rely on variants of the residual estimation method that deducts the estimated lawful immigrant population from total foreign‑born counts in Census surveys, typically using the American Community Survey (ACS) or the Current Population Survey (CPS) as the base; this shared foundation explains why DHS, Pew and CMS often produce broadly similar magnitudes even when specifics diverge [1] [5].
2. Timing and reference dates create headline gaps
Some of the clearest differences come simply from the date each estimate refers to: DHS’s most‑cited figure is a January 2022 estimate (about 10.99 million), Pew produced a revised 2022 estimate of roughly 11.0 million but then reweighted ACS data to produce a 2023 estimate of 14 million, and CMS reports interim estimates that sit between those ranges depending on whether they use CPS‑based provisional updates or ACS‑based calculations [1] [2] [6] [4].
3. Undercount and adjustment choices matter
A central methodological divergence is how each group adjusts for the expected undercount of noncitizens and hard‑to‑reach respondents: DHS applies an undercount adjustment around 10 percent in some work, independent researchers and Brookings summarize undercount ranges of roughly 5–15 percent, and those judgment calls — plus different approaches to weighting and imputation — can add or subtract hundreds of thousands from national totals [7] [1].
4. Which populations are included or treated differently
Pew explicitly notes that its 2022 estimate includes millions of people with temporary protections who are counted as “unauthorized” in the residual framing; CMS and DHS may handle certain temporary statuses, asylum seekers, or very recent arrivals differently depending on whether their administrative data are available or whether the estimate is provisional, which produces variation in who shows up in the final tally [8] [3] [1].
5. Data sources and reweighting — ACS vs CPS and “newer migration data”
Pew’s large 2023 upward revision reflects reweighting the 2022 and 2023 ACS to align with newer international migration data and to make years comparable, a technical choice that can amplify recent flow signals; CMS has produced provisional CPS‑based estimates that can look different from ACS‑based residuals, and DHS blends administrative records with census inputs — so differences in which survey and what auxiliary admin data are used explain much of the numerical spread [3] [5] [6].
6. Magnitude of disagreement: more nuance than contradiction
While headlines sometimes present the estimates as sharply inconsistent, methodological reviews and comparisons show convergence for the same reference date: for January/July 2022 DHS, Pew and CMS estimates cluster around ~10.9–11.0 million, whereas Pew’s 2023 reweighted estimate (14 million) and some newer CMS provisional estimates (reported between ~11.7 and 12.2 million in different releases) diverge because they explicitly incorporate post‑2022 migration surges and different modeling choices [1] [2] [6] [4].
7. Reading the differences: what to take away
The practical lesson is that differences are methodological and temporal, not necessarily partisan or random: analysts point to undercount assumptions, inclusion rules for temporary protection, choice of ACS vs CPS and whether reweighting uses newer migration signals as the principal levers that move totals by hundreds of thousands or millions — so any single headline number should be read alongside its date, population definition, and methodological footnotes [1] [5] [2].