How do estimates of undocumented immigrant populations differ between research groups and how does that affect apportionment projections?
Executive summary
Estimates of the undocumented (unauthorized) immigrant population diverge because research groups use different data sources and methods—residual calculations from surveys like the ACS/CPS, administrative records, microdata modeling, and dynamic flow models—which produce point estimates ranging from roughly 11 million to more than 20 million in the literature [1] [2] [3]. Those methodological differences translate into materially different population baselines and trend projections that feed directly into apportionment models used by agencies such as CBO and state population estimates, meaning that which estimate policymakers and the Census Bureau treat as authoritative can change projected congressional seat allocations and federal funding formulas [4] [5].
1. Why the numbers differ: data inputs and the “residual” baseline
Most mainstream estimates start from the “residual method,” subtracting known lawful immigrant counts from survey counts of the foreign-born, but the inputs vary: Pew and Migration Policy Institute rely heavily on ACS-derived population weights and DHS/administrative tallies for consistent residuals, while Center for Migration Studies applies country-specific ratios and CPS microdata adjustments; researchers who adopt alternative administrative or modeling approaches (including some academic teams) produce higher or lower totals because they treat survey undercounts, overstay rates, and emigration assumptions differently [1] [6] [7] [2].
2. Modeling flows vs. stock counts: divergence in dynamics
Some analysts emphasize stocks—how many unauthorized residents are present at a point in time—while others model flows of entries, exits, removals and voluntary departures to produce monthly or annual net-migration estimates; the Dallas Fed and several modeling efforts use microdata on entries and court notices to produce monthly flow estimates and have recently found sharply negative net unauthorized flows in 2025, a result that contrasts with stock-based estimates showing a 2021–2023 surge [8] [9]. Those flow-focused projections matter because apportionment and budget projections depend on current-year population counts, not just long-run averages [8] [10].
3. High outliers and methodological controversy
A minority of high-end estimates, exemplified by the Yale modeling paper, argue for dramatically larger undocumented populations—twofold or more—based on alternative demographic and operational modeling; those results have been criticized in methodological reviews and do not reflect the consensus residual-based approach used by major centers like Pew, MPI and CMS [3] [2]. Advocacy or partisan groups also produce higher or lower tallies (for example, higher estimates cited by nonacademic sites), but those figures often rest on different assumptions about undercounting and are not adopted in official apportionment calculations [11] [2].
4. How differences cascade into apportionment projections
Agencies that project apportionment (for example CBO when it incorporates population control adjustments) use historical series and methods that absorb revised population estimates and different net-immigration assumptions; CBO explicitly attributes differences in its demographic projections to changes in historical data and to new methods for projecting net immigration, and notes that varying assumptions about removals, detentions and voluntary departures materially alter projected civilian population totals used for budget and apportionment analysis [4]. When the Census Bureau’s population controls changed in 2024–25—adding hundreds of thousands to the population and reclassifying many as foreign-born—research estimates of the unauthorized population shifted, producing downstream revisions in apportionment-sensitive projections [5] [1].
5. Practical stakes and the remaining uncertainty
The stakes are concrete: a few hundred thousand people can change which states gain or lose a congressional seat and redistribute billions in formula grants; yet data gaps (in removals, rapid border processing, undercounts in surveys and changing enforcement policy) leave considerable uncertainty, and alternative credible projections (CBO, Brookings, Dallas Fed, Pew/MPI/CMS) still disagree on 2024–2026 trajectories because they weight flows, administrative data and survey corrections differently [4] [10] [8] [9]. Reporting and policy debates should therefore be anchored to method transparency: cite which estimate and which assumptions (stock vs. flow, survey corrections, removals) are being used, because those choices determine the apportionment and fiscal consequences claimed [7] [12].