Unauthorized Immigration and Local Labor Markets: Post-Pandemic Evidence and the Compositional Effect
The measured retrenchment of the foreign-born population is now large enough to show up clearly in the monthly CPS — and the data-source picture (which series can even measure this, and when) has become a first-order methodological issue.
CPS: foreign-born down 2.9 million (Jan 2025 – Jul 2026); net migration plausibly negative
Steven Camarota and Karen Zeigler (Center for Immigration Studies, Sept 3, 2026) report (CIS report):
"Since January 2025 when President Trump took office the government's monthly household survey, officially called the Current Population Survey (CPS), shows an unprecedented 2.9 million decline in the foreign-born population — naturalized citizens, lawful permanent residents, long-term temporary visitors, and illegal immigrants."
This follows an 8.3 million increase in the prior four years (the post-2021 border surge). Their preliminary estimate puts the illegal immigrant population down 2.3 million, from 15.8M (Jan 2025) to 13.5M (Jul 2026); the entire decline is among non-citizen Latin Americans who arrived 1980 or later. Their labor-market-relevant decomposition:
"Only 44 percent of the decline in the foreign-born are workers; the rest are children, the elderly, and other non-workers. This fact, along with employment gains by the U.S.-born, mitigates the impact on the labor market and may help explain why the survey of businesses does not show a dramatic decline in workers."
So roughly ~1.3 million immigrant workers left the CPS counts — yet payroll employment has not shown a dramatic decline, consistent with the complementarity/absorption pattern in Immigration Enforcement: Native Employment Impacts and the Complementarity Trap and Low-Skilled Immigration: Complementarity, Technology Adoption, and Output Mix Adjustments.
The report's own methodological caveats (unusually explicit — weigh them)
- Margin of error: the monthly CPS foreign-born count carries ±1.038M (90% CI); at interval extremes the Jan-2025→Jul-2026 decline could be as small as ~763,000 — still large, but half the headline.
- Reweighting: the January 2026 population-control readjustment added +874,000 to the immigrant count vs. December 2025; within-year consistent weights show −1.7M (2025) and −2.0M (Jan–Jul 2026), which the authors argue shows the decline is not a weighting artifact.
- Nonresponse: the key alternative hypothesis — immigrants still present but avoiding the survey — is tested against response rates and finds no correlation; unweighted Hispanic respondent counts fell 7.4% (Jan 2025–Jul 2026), which the authors read as consistent with actual departure but which cannot fully rule out differential nonresponse.
- Mechanical artifact: "a decline in the foreign-born population creates an apparent increase in the U.S.-born population as a statistical artifact1" of the weighting — do not read rising U.S.-born CPS counts as native population growth.
- Source orientation: CIS is a restrictionist-leaning think tank; the "net migration negative for the first time since the 1930s" framing is theirs. But the CPS-based measurement approach is standard (BLS publishes monthly foreign-born employment series; CBO, SSA, Pew, and the Fed use the same survey), and the report is unusually candid about its own uncertainty. Directionally, the decline is corroborated by administrative data the report compiles: interior removals roughly tripled (+~207,000, UCLA Deportation Data Project) and voluntary departures rose from 11,977 to 89,494 (16-month comparison).
Data-source timing (relevant to the ACS watch)
The ACS cannot yet measure the retrenchment: "the most recent version of the ACS reflects only the population through July 2024... Even when the 2025 ACS is released later this year, it still will only reflect the population through mid-2025 and will not reflect the Trump administration's crackdown for the most part." CPS is the only timely government series — meaning ACS-based peer-reviewed replication of the surge-and-retrenchment cycle waits on the 2026 microdata2.
-
An instance of The measuring instrument, not the labor market, decides what immigration did. — The CPS's own weighting machinery manufactures population changes that never happened, the clearest demonstration that the instrument — not the market — decides what immigration did. ↩︎
-
An instance of The methods that count the population are now set by political order. — With the ACS knocked out by political direction, a single monthly survey carrying a ±1 million error bar becomes the sole timely measure of the largest foreign-born reversal in a century. ↩︎