How do experts estimate the number of undocumented immigrants and what are the main uncertainties?
Executive summary
Experts chiefly use the “residual” approach — subtracting legally resident foreign‑born counts from total foreign‑born counts in surveys — and complementary methods like CPS-SSA nonmatch or administrative-data adjustments; Pew’s revised work put the unauthorized population at 14 million in 2023 and later revised some earlier estimates to 11.8 million for 2022 after policy changes [1] [2] [3]. Key uncertainties are survey undercount, assumptions about emigration and mortality, rapid changes in flows (including undetected border crossings), and differing methodological choices that produce wide-ranging estimates [1] [4] [5].
1. How researchers turn big surveys into an “unauthorized” headcount
Most academic and policy estimates begin with a Census Bureau survey count of all foreign‑born residents (ACS or CPS) then subtract estimated legally present immigrants (permanent residents, naturalized citizens, temporary visa holders, etc.). That subtraction — the residual method — yields the residual labeled “unauthorized,” and it remains the dominant approach described in a recent Social Security Bulletin review [1] [6]. Researchers refine that core step with administrative records and alternative statistical techniques, but the logic is the same: total foreign‑born minus legally counted foreign‑born equals an unauthorized residual [6] [1].
2. Other techniques and new twists on measurement
Analysts are developing and testing alternatives to improve timeliness and detail. One new method pairs CPS survey data with Social Security Administration nonmatches to flag likely unauthorized immigrants and to estimate characteristics; the SSA/NBER series compares that method with the residual approach and suggests it may enable continuous measurement [6]. Centers such as the Center for Migration Studies produce microdata‑based estimates using ACS microdata and additional statistical procedures, showing there are multiple technically plausible ways to get to a number [7] [8].
3. Where the biggest disagreements come from
Disagreement arises from assumptions needed to move from survey snapshots to population totals. Analysts must estimate how many foreign‑born were missed by surveys, how many unauthorized people subsequently left the country or died, and how to count recent arrivals and undetected entries. Critics have long highlighted that different assumptions about undercount rates, emigration, and visa‑status classification can push estimates dramatically apart; the residual method itself relies on contestable parameters such as assumed undercount rates [5] [1].
4. The role of rapid policy and flow changes
Estimates lag reality when migration flows shift quickly. Pew’s recent revisions — including an upward revision of its 2022 estimate after policy changes affecting large parole cohorts — illustrate how administrative policy and sudden inflows can force analysts to revise past figures and complicate trend interpretation [3] [2]. The San Francisco Fed and CBO work also show that new DHS and Border Patrol data about undetected entries and inflows can materially change net international migration and unauthorized‑flow estimates for 2024–25 [4].
5. Quantifying uncertainty: wide ranges and why they exist
Published estimates can range widely depending on method. Some think‑tank and advocacy groups publish far higher figures (e.g., FAIR’s 18.6 million figure), while academic and center estimates clustered lower; such disparities reflect methodological choice and reporting scope rather than a single “true” number [9] [10] [2]. The Social Security Bulletin and other reviews document how different models, treatment of emigration and undercount, and incorporation of administrative data produce divergent outcomes [1] [6].
6. Practical limits: what the data cannot (yet) tell us
Available sources do not mention a definitive way to observe every undocumented person in real time; instead, measurement depends on imperfect surveys, administrative feeds and modeled adjustments (not found in current reporting). Response rates, especially in smaller surveys like the CPS, are limited — CPS surveys about 60,000 households — and that sample size constrains precision for this subgroup; analysts note that declining response rates and possible nonresponse bias among immigrants add additional uncertainty [11] [2].
7. How to read new headline numbers
When a new number appears—whether from Pew, a federal agency, the CBO, or an advocacy group—read it as an estimate conditional on choices about undercount, emigration, and data sources. Pew’s updated headline — 14 million for 2023 — superseded earlier figures and reflects ACS‑based residual calculations updated for recent flows [2]. Analysts such as those cited by the San Francisco Fed show that revised DHS inputs and updated assumptions about undetected entries can change multi‑year net migration estimates by hundreds of thousands [4].
8. Bottom line for policy and reporting
Counting unauthorized immigrants is an inexact science shaped by transparent methodological choices; different, credible methods can and do produce meaningfully different totals [1] [6]. Reporters and policymakers should treat single point estimates as model outputs, scrutinize the assumptions behind them, and expect revisions when administrative data or migration patterns shift [2] [4].