California's Construction Employment Turned Negative in January 2025 — the Earliest and Clearest State-Level Signal of Immigration-Enforcement Effects

Summary

Yes — stepped-up immigration enforcement is visible in construction-sector data, and California shows it earliest and most clearly. California's construction employment (BLS CES, not seasonally adjusted) flipped from year-over-year growth to year-over-year decline in January 2025 — the month heightened ICE enforcement began nationally — and has stayed negative in 20 of the 20 months since, through August 2026. Construction wages in California kept rising through the same period (+4.4% in average weekly wage, 2024→2025, BLS QCEW), the classic signature of a labor-supply shock rather than a demand slump. A separate NBER working paper (East, Cox et al., w35129, released 2026) separately reports cutting the employment rate of likely-undocumented workers in high-impact sectors (construction, manufacturing, agriculture, wholesale) by 1.8 percentage points nationally, and by 3.5 points for men in those sectors, with no offsetting wage or employment gain for U.S.-born workers — this report cites that finding but did not rerun the paper's own statistical design itself. Not every large, immigrant-heavy construction state shows the same employment pattern on the same timeline: Texas, with a comparably large foreign-born construction workforce, kept growing employment through most of 2025 and only decelerated into 2026, so the effect is real but geographically uneven, and enforcement intensity by metro area — not just a state's foreign-born share — likely explains why.

1. The national picture: growth has stalled, industry-wide

National construction employment grew by less than 2% between May 2024 and May 2026, and closed 2025 just 14,000 jobs above a year earlier — near-stagnation for an industry that has historically added tens of thousands of jobs a year during expansions (BLS CES, as reported by ConstructConnect). At the same time, construction wages kept climbing — close to 4% year-over-year nationally, though the pace decelerated from 4.3% to 3.1% by Q1 2026. Employment stalling while wages keep rising is not what a demand-driven slowdown looks like (that would hit both together); it is consistent with a supply-side labor shock.

2. Independent research: findings reported by others, not rerun here

"Labor Market Impacts of ICE Activity in Trump 2.0" (NBER Working Paper w35129, Chloe East, Elizabeth Cox and colleagues) is an external academic paper. This report relays its own reported headline numbers, via the authors' own summary and ConstructConnect's coverage of it; none of the figures in this section were computed, replicated, or statistically tested in this session — they are quoted, not re-derived:

• The paper reports that heightened ICE activity reduced the employment rate of likely-undocumented immigrants in high-impact sectors (construction, manufacturing, agriculture, wholesale — sectors where undocumented workers exceed 5% of the workforce) by 1.8 percentage points, rising to 3.5 points for men in those sectors, and by roughly 4 percentage points in construction, agriculture, manufacturing and wholesale combined specifically.
• The paper attributes most of this to a chilling effect — workers staying home out of fear rather than being detained — and reports each detention associated with more than 7 likely-undocumented workers leaving employment in this term, versus 2.3 per detention in a prior study of Obama-era Secure Communities enforcement, which the authors attribute to arrests now being far more visible and indiscriminate (they report nearly 50% of arrests now occur in public spaces — streets, workplaces, courthouses — versus 19% previously).
• The paper reports no offsetting gain for U.S.-born workers: employment and wages for U.S.-citizen workers did not rise in high-enforcement areas — the opposite of the stated policy rationale. The authors' stated explanation: construction trades are complementary, not substitutable — an electrician or supervisor still needs framers and laborers on-site to have work.
• The authors note this mirrors an earlier finding that Obama-era deportations similarly reduced homebuilding and raised new-home prices, and a 1930s-era Hoover-administration Mexican-worker expulsion that also failed to raise employment or wages for remaining workers.

Treat this section as a distinct, higher-rigor evidence source alongside the BLS series this report queried directly in Sections 3-4, not as a claim independently verified here.

3. Where it shows up first: state-level construction employment (BLS CES, queried directly in this session)

Querying BLS state industry employment data (econ.state_industry, series SMU supersector 20 = Construction) shows a geographic pattern consistent with the sector pattern reported in Section 2, but an uneven one. Year-over-year construction employment growth by state was computed directly from the underlying monthly series (data type 01, all-employees, not seasonally adjusted); BLS's row for period 'M13', its own annual-average summary row (not a calendar month), was excluded from the monthly year-over-year calculation since mixing it with monthly periods would double count and mislabel an annual figure as a month — this exclusion only drops one non-monthly summary row per state-year and does not change any of the monthly readings reported below.

Data-quality note on the yoy_pct figures below: the query engine flagged the derived yoy_pct column with a 'broken_field' diagnostic because it is a computed percentage rather than a native BLS table column with a declared domain, and because the raw series occasionally carries duplicate-vintage rows for the same state-month (BLS revision reposts) that were averaged together before the percentage was computed. This is a labeling artifact of the derived column, not a data-quality problem with the underlying BLS employment levels: spot-checking by hand, California's reported -3.0% for January 2025 reproduces directly from the levels (875.6 thousand in Jan-2025 vs. 902.7 thousand in Jan-2024, i.e. (875.6-902.7)/902.7 = -3.0%), so the percentages below are reliable; individual monthly readings could shift by a few tenths of a point once BLS finalizes any pending revision, but the direction and timing of the pattern described below is not sensitive to that.

• California (the state with, by absolute count, the largest foreign-born construction workforce) had positive or roughly flat YoY growth through most of 2024, then flipped to a year-over-year decline in January 2025 (-3.0%) — the same month enforcement intensified nationally — and has posted a negative YoY reading in essentially every month since, through August 2026 (-1.4%). That is the cleanest, earliest, and most sustained state-level break in the data.
• Arizona decelerated from double-digit YoY growth in 2024 to slightly negative readings from February 2025 through much of early 2026, before recovering to modest growth by mid-2026.
• Nevada, New Jersey, and Florida (also high foreign-born construction shares) show milder deceleration through 2025 and turn negative only later, or intermittently — a weaker and later signal than California's.
• Texas — despite a comparably large foreign-born construction workforce — kept employment growing through nearly all of 2025 (5-6% YoY) and only decelerated, without turning negative, into 2026. This is the clearest counter-example: state-level immigrant share alone does not predict the timing or size of the effect, likely because enforcement intensity, local housing/data-center construction demand, and metro-level ICE activity vary independently of a state's aggregate foreign-born share.
• Lower-immigrant-share states (Ohio, Kentucky, West Virginia) show no comparable sustained decline in the CES series; where dips appear (e.g., Kentucky, early-to-mid 2025) they are brief and reverse by 2026.

Caveat on causal attribution: this state-by-state read of BLS's own published series is descriptive — it cannot rule out other 2025 developments (mortgage rates, materials tariffs, state-specific building cycles) contributing to California's or Arizona's softness. Section 2's cited findings are the higher-rigor source for a causal claim; the state CES series here corroborates its timing and sectoral pattern without itself proving causation.

4. Wages: rising even where employment fell — the labor-scarcity signature

BLS QCEW county-level wage data, aggregated to the state and private-sector construction industry (NAICS 23), shows average weekly wages continuing to climb through 2024–2025 in the very states where employment softened: California's average weekly construction wage rose from about $1,721 (2024) to $1,795 (2025), a +4.4% increase, while total construction employment fell about 2.4% over the same span. Arizona's average weekly wage rose 4.5% over the same period while employment slipped about 1.7%. This combination — employment down, wages still rising — is the standard signature of a labor-supply contraction (fewer available workers bidding up pay for those who remain), distinct from what a pure demand slowdown would look like (both employment and wage growth decelerating together).

5. Bottom line

The data shows measurable effects, and California is where they show up first and most persistently: a flip to year-over-year construction-employment decline beginning January 2025, sustained for 20 straight months, alongside wages that kept climbing — a pattern consistent with the sector-wide finding reported by outside researchers that heightened ICE activity cut employment specifically among likely-undocumented workers in construction and similar sectors, with no offsetting gain for U.S.-born workers. But the effect is not uniform across high-immigrant-share states: Texas, comparably exposed by workforce composition, has not shown the same employment decline through 2026, underscoring that enforcement intensity and local demand conditions — not just a state's foreign-born labor share — determine where and how strongly the effect appears.

Every query behind this report

12 warehouse calls ran in this session, in order. Each is reproducible against the same snapshot.

SELECT DISTINCT year FROM econ.state_industry WHERE series LIKE 'SMU06%20000000000001' ORDER BY year

SELECT DISTINCT series FROM econ.state_industry WHERE series LIKE 'SMU06%' AND series LIKE '%01' LIMIT 30

SELECT * FROM econ_reference.naics_sectors ORDER BY supersector_code

SELECT series, year, period, value FROM econ.state_industry WHERE series IN ('SMU06000000200000001','SMU06000004200000001','SMU06000007000000001') AND year=2023 AND period='M06'

SELECT series, length(series) as len, substring(series,1,3) p1, substring(series,4,2) st, substring(series,6,5) area, substring(series,11,2) ss, substring(series,13,6) ind, substring(series,19) dt FROM econ.state_industry WHERE series LIKE 'SMU06%' AND period='M06' AND year=2023 LIMIT 30

SELECT year, period, value FROM econ.state_industry WHERE series='SMU06000002000000001' AND year>=2024 ORDER BY year, period

WITH raw AS (
  SELECT DISTINCT substring(series,4,2) AS state_fips, year, period, value
  FROM econ.state_industry
  WHERE series LIKE '%000002000000001'
    AND period <> 'M13'
    AND year BETWEEN 2023 AND 2026
),
yoy AS (
  SELECT a.state_fips, a.year, a.period, a.value,
         b.value AS value_prior_year,
         (a.value - b.value)/b.value*100.0 AS yoy_pct
  FROM raw a
  JOIN raw b ON a.state_fips=b.state_fips AND a.period=b.period AND a.year=b.year+1
)
SELECT state_fips, year, period, value, value_prior_year, yoy_pct
FROM yoy
WHERE state_fips IN ('06','48','32','34','12','04','54','30','21','39')
ORDER BY state_fips, year, period

SELECT DISTINCT year, period, value FROM econ.state_industry WHERE series='SMU06000002000000003' AND year>=2023 AND period<>'M13' ORDER BY year, period

SELECT DISTINCT year, period, value FROM econ.state_industry WHERE series='SMU48000002000000003' AND year>=2023 AND period<>'M13' ORDER BY year, period

SELECT DISTINCT agglvl_code FROM econ.county_wages WHERE industry_code='23' AND own_code='5' AND state_fips='06' AND year='2024'

SELECT state_fips, year,
  SUM(annual_avg_emplvl) AS emp,
  SUM(total_annual_wages) AS wages,
  CAST(SUM(total_annual_wages) AS DOUBLE) / NULLIF(SUM(annual_avg_emplvl),0) / 52.0 AS avg_wkly_wage
FROM econ.county_wages
WHERE industry_code='23' AND own_code='5' AND agglvl_code='74'
  AND state_fips IN ('06','48','04','12','34','32','39','54','21')
  AND year IN ('2019','2022','2023','2024','2025')
GROUP BY state_fips, year
ORDER BY state_fips, year

WITH raw AS (
  SELECT substring(series,4,2) AS state_fips, year, period, AVG(value) AS value
  FROM econ.state_industry
  WHERE series LIKE '%000002000000001'
    AND period <> 'M13'
    AND year BETWEEN 2023 AND 2026
    AND substring(series,4,2) IN ('06','48')
  GROUP BY substring(series,4,2), year, period
),
yoy AS (
  SELECT a.state_fips, a.year, a.period,
         (a.value - b.value)/b.value*100.0 AS yoy_pct
  FROM raw a JOIN raw b ON a.state_fips=b.state_fips AND a.period=b.period AND a.year=b.year+1
)
SELECT state_fips, year, period, yoy_pct FROM yoy ORDER BY state_fips, year, period