Undocumented Immigrant Population Estimates

Limitations

These are four independent MODELED ESTIMATES of one quantity - the undocumented immigrant population living in Georgia - not counts of people. Each estimator uses its own method (residual or legal-status assignment over Census Bureau survey data), its own reference date and its own rounding, so ALWAYS filter to one estimator: never sum, average or chain values across estimators, and never read a change between two estimators' figures as a change in the population. The latest Georgia figures range from 340,000 (dhs_ohss, 2022) to 502,000 (cms, 2024): pew 450,000 (2023), mpi 479,000 (2023), cms 502,000 (2024), dhs_ohss 340,000 (2022). year is each estimator's OWN reference year - neither a federal fiscal year nor an event year - and there is no month. The reference dates differ: DHS counts the population on January 1, Pew and MPI describe mid-year figures, and CMS labels its release data year 2024. DHS 2022 is January 1, 2022, about 18 months before Pew's and MPI's mid-2023 figures. No cross-estimator grid is filled: a year an estimator did not publish has no row (Pew publishes no 2020 estimate, DHS none for January 2021, MPI and CMS one year each). A missing year is absent, not zero; never interpolate across it or borrow another estimator's value. pew = Pew Research Center: term 'unauthorized immigrant'; reference date the survey year of each estimate, a mid-year figure: American Community Survey 2005-2023, Current Population Survey 1995-2003 and the 1990 decennial census (Warren and Warren 2013); Pew publishes no 2020 estimate; rounded to the nearest 5,000 below 100,000, 10,000 from 100,000 to 250,000, 25,000 from 250,000 to 1 million and 50,000 above; the only estimator with a margin of error (plus or minus, 90% confidence), published for every year except 1990; years 1990, 1995, 1998, 2000, 2003, 2005-2019, 2021-2023. Source: Pew Research Center, report of 2025-08-21 on the U.S. unauthorized immigrant population reaching a record 14 million in 2023, detailed table 'Unauthorized immigrant population for states (and margins of error), 1990-2023' (https://www.pewresearch.org/race-and-ethnicity/2025/08/21/u-s-unauthorized-immigrant-population-reached-a-record-14-million-in-2023/). mpi = Migration Policy Institute (MPI): term 'unauthorized immigrant'; reference date mid-2023: MPI analysis of the 2019-23 American Community Survey and the 2023 (with 2008) Survey of Income and Program Participation, weighted to 2023 population estimates by Jennifer Van Hook (Penn State); rounded to the nearest 1,000; years 2023. Source: Migration Policy Institute, 'National and State Estimates' and 'National and County Estimates of the Unauthorized Immigrant Population, Mid-2023' (workbook MPI-2023_Unauthorized_Profiles-State-County-Topline_Estimates-FINAL.xlsx, uploaded 2025-10, retrieved 2026-09-23; https://www.migrationpolicy.org/data-tool/unauthorized-immigrants). cms = Center for Migration Studies of New York (CMS): term 'undocumented'; reference date data year 2024 as labeled by the CMS State and National Data Tool (the underlying American Community Survey vintage is not stated); rounded to the nearest 1,000; years 2024. Source: Center for Migration Studies of New York, State and National Data Tool, undocumented population by state, data year 2024 (https://data.cmsny.org/data/undoc_2024.csv); method: Robert Warren (2021), Journal on Migration and Human Security, https://journals.sagepub.com/doi/10.1177/2331502421993746. dhs_ohss = DHS Office of Homeland Security Statistics (OHSS): term 'unauthorized immigrant'; reference date January 1 of each year; no January 2021 estimate (DHS: limitations of the 2020 American Community Survey); the series changes basis at 2010, 2015 and 2018 (see limitations); rounded to the nearest 10,000 in both Table 3 and Appendix Table A2-1 (the appendix displays thousands; gold stores persons); the two agree for every Georgia year; years 2000, 2005-2020, 2022. Source: Bryan Baker and Robert Warren, 'Estimates of the Unauthorized Immigrant Population Residing in the United States: January 2018-January 2022', OHSS, April 2024 (Table 3, Appendix Table A2-1), superseding Bryan Baker, '... January 2015-January 2018', DHS, January 2021 (https://ohss.dhs.gov/topics/immigration/illegal/population-estimates). Terms: Pew, MPI and DHS say 'unauthorized immigrant'; CMS says 'undocumented', the term used here. Definitions differ at the margins: MPI counts people holding a liminal status (DACA, humanitarian parole, TPS) and pending asylum applicants as unauthorized (workbook note); DHS counts TPS and DACA beneficiaries, people awaiting removal proceedings and parolees not yet admitted as unauthorized (report definitions); CMS's undocumented estimate contains the people it flags with a liminal status (DACA, TPS, parole, SIJS, U/T visa, pending asylum). Margin of error: only Pew publishes one (plus or minus, 90% confidence), in estimated_population_moe; it is NULL for MPI, CMS, DHS and Pew 1990. The other estimates are uncertain too, by amounts their publishers do not quantify here. Pew revised its 2022 estimate (American Community Survey supplemented for immigrants added to Census Bureau population estimates), so it differs from Pew's earlier publications. Pew's series changes method twice: 1990 comes from the decennial census, 1995-2003 from the Current Population Survey, and 2005 onward from the American Community Survey. Compare Pew years within one method era, not across them. The CPS-era margins for Georgia are wide: plus or minus 80,000-95,000 on estimates of 55,000-275,000; the 90% interval includes zero for 1995 (55,000 plus or minus 85,000). DHS revisions and series breaks: gold carries the April 2024 edition's Appendix Table A2-1, which republishes 2000-2022 on three successive bases and prints each break year twice - the unrevised value continues the years before it, and the revised value (used here) starts the new basis: 2010* re-based to the 2010 Census, 2015** revised for DHS's updated methodology, and 2018** revised again for the April 2024 edition's update. Gold's DHS series therefore changes basis between 2009 and 2010, 2014 and 2015, and 2017 and 2018: a change across those years mixes population change with re-basing, so for a like-for-like change across a break compare against the unrevised value. Unrevised (superseded) values, recorded in the transform manifest: 2010 460,000 -> 430,000 (was in: April 2024 edition Table A2-1 column 2010, January 2021 edition Table A2-1 column 2010); 2015 390,000 -> 370,000 (was in: April 2024 edition Table A2-1 column 2015, January 2021 edition Table A2-1 column 2015); 2018 380,000 -> 390,000 (was in: April 2024 edition Table A2-1 column 2018, January 2021 edition Table A2-1 column 2018, January 2021 edition Table 3 column 2018). Table 3 of each edition is read as a cross-check and agrees with its edition's appendix in every year. County rows: only MPI publishes counties - 287 county or county-group estimates nationally (smallest 1,000), among them ten single Georgia counties: Gwinnett County 121,000, Cobb County 55,000, DeKalb County 47,000, Fulton County 35,000, Hall County 27,000, Clayton County 17,000, Cherokee County 16,000, Whitfield County 14,000, Forsyth County 12,000, Chatham County 8,000. They sum to 352,000, 73% of MPI's Georgia figure of 479,000. The other 149 Georgia counties are NOT PUBLISHED by MPI - not zero - and county rows exist only for estimator = mpi (mid-2023). MPI's state figure minus the ten counties is spread over those 149 counties and cannot be assigned to any one of them. Georgia only: the national figures (Pew U.S. total, MPI United States, CMS NATIONWIDE, DHS Total and 'All other states' / 'Other or Unknown') are used only as parse cross-checks, and other states are not included. Also not included: CMS's demographic breakdowns of its estimate and its separate liminal-status and naturalization-eligible populations; MPI's share-of-national columns; DHS's country-of-birth and age/sex tables. Every value is the estimator's own published aggregate; this platform applies no suppression or adjustment.