What this research found
How often do US construction workers work through air that breaches the federal fine-particle limit? Joining EPA's 2023 daily county PM2.5 monitoring to Bureau of Labor Statistics construction payrolls produced an estimate of 18.95 million exposed worker-days above the 35 µg/m³ 24-hour standard, carrying roughly $5.68 billion in at-risk wages. Exposure proved highly concentrated, and 20 counties led by Cook (IL) and Hennepin (MN) were shortlisted as targets for a prospective worker-protection study.
- An estimated 18.95 million construction worker-days in 2023 fell on days when county-representative PM2.5 exceeded the 35 µg/m³ 24-hour standard, equivalent to about $5.68 billion in payroll on a 260-working-day basis.
- Exposure is very unevenly spread across the country, with a Gini coefficient of 0.846: the top 5% of counties account for 58.4% of exposed worker-days and the top 10% for 75.8%.
- Two-thirds of the estimate rests on direct measurement — 66.7% of exposed worker-days come from observed monitor readings, with the remaining 33.3% interpolated (29.1% spatial, 4.3% temporal).
- The ranking survives the main analytic choice. Switching from a median-of-monitors to a max-of-monitors county convention raises the national total from 18.95 M to 26.13 M worker-days, but county values under the two conventions correlate at r = 0.809.
- The 20 highest-priority counties by composite vulnerability index are led by Cook (IL), Hennepin (MN), Marion (IN), New York (NY), and Queens (NY) — a Midwest and Northeast pattern the authors tie to the 2023 Canadian wildfire smoke footprint.
How it was done
EPA's 2023 daily PM2.5 file (847,162 monitor rows) was deduplicated to 457,732 monitor-days and reduced to a county-day median across active monitors, giving 203,450 observed county-days for 636 monitored counties. Gaps were filled by linear interpolation over internal runs of five days or fewer, then by borrowing the nearest monitored county in the same state within 100 km, extending coverage to 2,674 counties; 554 counties had no qualifying neighbour and were left missing. The Bureau of Labor Statistics employment file (3,650,651 rows) was filtered to private-sector construction, and 310 counties whose figures were suppressed for confidentiality were imputed from state employees-per-establishment ratios. An inner join matched 2,656 counties, exceedance days were multiplied by county employment and daily wage bill, and an equal-weighted index over exposure volume, wage weight, and each county's 90th-percentile daily PM2.5 produced the priority ranking, written up as a 17-page paper with 5 figures and 31 references.
Data sources
- EPA Air Quality System — daily PM2.5 (parameter 88101), 2023; 847,162 monitor rows across 636 monitored counties
- BLS Quarterly Census of Employment and Wages — 2023 annual single file, 3,650,651 rows; 3,261 counties with private construction employment
- US Census Bureau 2023 county gazetteer — county centroids used for spatial interpolation and naming
- 31 DOI-verified references covering PM2.5 epidemiology, occupational and construction health, wildfire smoke toxicity, and exposure inequality
Limitations
One third of the national estimate rests on interpolated rather than directly monitored PM2.5, and county-average concentrations cannot capture variation within a county or whether workers were actually outdoors. Annual-average employment is applied to every exceedance day, and the wage figure is an upper bound on payroll at risk rather than a realized economic loss.
Figures from this analysis
How this research was produced
K-Dense Web planned and ran this public health investigation end to end — gathering the sources, carrying out the analysis, producing the figures, and drafting the report. The full session transcript, including every intermediate step, is available to view.
