Where the electricians went: Hyperscale data center construction and building trades membership in federal union filings, 2018 to 2025
Published employment figures for data-center construction come from the industry or from unions testifying on projects. This paper builds an independent series from federal LM-2 filings and IBEW jurisdiction maps: an eight-year panel of inside electrical construction membership for 89 metropolitan locals. Median treated local grew 33.7 percent between 2018 and 2025 against 13.2 percent elsewhere — a 25.8-percentage-point regression difference, randomization p below 0.001.
Working paper, version 0.1, September 2026, with its complete replication package.
Published employment figures for data center construction come from the industry, from agencies relaying an industry figure, or from unions testifying on projects. This paper builds an independent series from federal filings. Labor organisations above a receipts threshold report their membership annually to the U.S. Department of Labor under penalty of perjury on Form LM-2, and the International Brotherhood of Electrical Workers publishes maps of which locals hold inside construction jurisdiction. Joining the two gives an eight-year panel of inside electrical construction membership by metropolitan area for 89 locals. Treatment is a hyperscale campus published by one of four operators in the local's metropolitan area.
The median treated local grew 33.7 percent between its 2018 and 2025 filings against 13.2 percent elsewhere; the regression difference is 25.8 percentage points, 95 percent interval 7.8 to 43.8, clustered on 36 states, with a randomization p-value below 0.001. An event study with local and year fixed effects finds no differential trend in the two pre-period years the panel can see and a divergence significant from 2023. The effect is concentrated: removing the five fastest-growing locals leaves 9.0 points and no significance, and removing the four treated metros known to host a federally subsidised semiconductor or battery plant leaves 17.1. County employment of nonresidential electrical contractors from the Quarterly Census of Employment and Wages correlates with the membership series at 0.58 across 86 locals and shows the same treatment contrast independently, 19.6 points, p = 0.003. The paper is descriptive: no instrument, no runner-up control, and membership is not employment.