Counting a trade in one metropolitan area: union membership for inside electrical construction from federal filings, 2010 to 2025

Version 1.0, 20 September 2026. A measurement paper. Its companion research note, which reports the data-center application as its own line of work, is deposited under concept DOI 10.5281/zenodo.22866646. The data and code both rest on are deposited separately at 10.5281/zenodo.22862472. Union density is published for metropolitan areas and, within them, for private construction. It is not published annually for a single trade inside a metropolitan area, and the survey cannot supply that series. The Current Population Survey codes construction as one undifferentiated industry in the three classification vintages spanning these years, so the industry variable does not reach a trade at any sample size. The occupation variable does, but no table family the Union Membership and Coverage Database publishes crosses it with metropolitan area: the national sample behind the published electrician estimate runs from 695 to 940 observations a year, which caps the mean across the 253 to 260 published metropolitan areas below four in every year. The sector-by-metro cells one level up already run to a median of ten observations. Multi-year pooling is treated as the serious alternative it is, with the same arithmetic. This paper builds a membership count from administrative records instead. Every labor organisation covered by federal labor-reporting statutes files an annual report with the Department of Labor stating its membership, and the IBEW publishes maps of which locals hold jurisdiction over inside construction in each area. Joining the two yields a sixteen-year panel of 11,645 local-years covering 786 filing organisations, with membership, apprentice counts, headquarters geography and two vintages of the jurisdictional classification. The panel is released with the paper. It is set against two external series. The survey comparison fails, and the failure is the finding: the panel's count exceeds the survey's count for all of construction in 18.2 percent of metro-years, and within a metro the survey's figure moves eleven and a half times as much from year to year as the panel's. Against QCEW counts of electrical-contractor employment, which carry no sampling error, the two correlate in log levels in every year at 0.428 to 0.494 and 0.457 to 0.504 on two industry codes, and at 0.55 in growth over the full span against 0.23 year over year. They agree about where the trade is and how it moved over a decade, and disagree about what it did last year. Known defects are quantified, including a treatment-classification error found during preparation and reported in section 5.6 with its magnitude and a switch to reproduce the superseded specification. One application is reported as description rather than cause. JEL: C81, J51, J21, R23. Data and code. Not in this record. The panel, its codebook, the sixteen OLMS yearly extracts, both jurisdictional roster vintages, thirty analysis scripts, every deposited log and JSON, every retrieved source and the classification-error record are deposited as a separate citable dataset at 10.5281/zenodo.22862472, cited by this paper and by its companion alike. That record reproduces from its own inputs; its README.md gives the run order and names the one step that cannot be run from the package and why.