How did Pew Research estimate state unauthorized immigrant populations for 2023, and what are the methodological limitations?
Executive summary
Pew’s 2023 state-level estimates of the unauthorized immigrant population are built from U.S. Census Bureau survey data (primarily the American Community Survey), adjusted with administrative records and demographic corrections, and then reweighted to account for recent revisions in population and migration estimates [1] [2]. Significant methodological caveats — undercounts in surveys, changing definitions (including people with temporary protections), timing of reference dates, and sensitivity to reweighting choices — mean the 14 million national total and its state splits should be treated as the best-available, model-based estimates rather than direct headcounts [3] [4] [5].
1. What data Pew used and the time window
Pew based its 2023 estimates principally on the American Community Survey (ACS) microdata for 2021–2023 (and on earlier CPS-ASEC and decennial census data for prior years), using July 1, 2023 as the reference date for population counts drawn from the ACS [1] [6]. To produce internally consistent time series back to 1990, Pew harmonized these surveys with census benchmarks and incorporated administrative inputs from U.S. Citizenship and Immigration Services and immigration court records where relevant [2] [6].
2. The core estimation steps Pew applied
Pew’s method estimates the foreign-born population from ACS responses, then classifies people as lawful or unauthorized using a residual technique that subtracts known lawful populations (green card holders, visa holders, naturalized citizens) from the total foreign-born; the remainder is treated as unauthorized after further adjustments [2] [1]. Pew then applies upward “undercount” adjustments because surveys tend to miss immigrants — especially unauthorized immigrants — and augments the survey-based totals using administrative data and modeled flows of arrivals, departures, mortality and changes in legal status [3] [6].
3. How Pew produced state-level splits
State estimates come from the underlying ACS geography — Pew reweights and adjusts the ACS state samples so that state totals align with revised national population and net international migration estimates (notably the Census Bureau’s Vintage 2024 revisions), producing state-by-state totals that are comparable across years [1] [6]. For 2022–2023, Pew explicitly reweighted ACS data because the Census Bureau updated national population estimates without retroactively changing previously published survey weights, and Pew’s reweighting is intended to remove discontinuities caused by those revisions [5].
4. Key methodological limitations and uncertainties
Survey undercount is central: the ACS and CPS miss people, and unauthorized immigrants are disproportionately likely to be missed; Pew applies an upward adjustment but the magnitude of hidden populations is inherently uncertain [3]. Reweighting to new national estimates can introduce volatility — large Census revisions (Vintage 2024) required substantial adjustments that can change year-to-year trends and state allocations [1] [5]. Classification ambiguity is another limit: Pew’s “unauthorized” label includes people with partial protections (e.g., parole, pending asylum, temporary work authorization), and about 40% of those classified as unauthorized in 2023 had some form of protection, complicating interpretation [2] [7]. Data timing and family linkage matter too: reference dates (July 1 versus March 1) and the inability to link parents and children across households mean counts of children of unauthorized immigrants are understated relative to the true number who have unauthorized parents [6] [4].
5. Alternatives, competing estimates and potential agendas
Independent institutes produce somewhat different totals; for 2023 national estimates cluster roughly from about 12.2 million to 14.0 million, reflecting divergent assumptions about undercounts and which administrative sources to incorporate [8]. Media and political actors often seize on specific numbers for policy claims — either amplifying higher figures to argue for stricter enforcement or questioning them to push for tighter border narratives — so methodological choices (which surveys, how to reweight, how to define “unauthorized”) can have implicit agendas that shape public interpretation [7] [9].
6. Bottom line for reading Pew’s state estimates
Pew’s state estimates are a transparent, survey-plus-adjustment effort intended to reflect the best available evidence as of July 2023, relying on ACS geographic detail plus administrative augmentation and reweighting to align with revised population controls [2] [5]. They are not direct counts and carry notable uncertainty from undercounts, survey revisions, definitional complexity and model choices; the numbers provide a rigorous, reproducible baseline for comparison, but users should treat state splits as modeled estimates with known limitations rather than precise headcounts [3] [1] [4].