What MSHA's contractor employment data measures, and what it does not

Every quarter, mining contractors file their employee-hours and production with MSHA on the same forms operators use. We have now loaded that entire record: about 1.3 million quarterly filings from roughly 39,000 distinct contractors, covering 2000 through 2026. It is the exposure denominator that makes a contractor safety rate possible, and it is the piece the public incident data never shipped alongside the accidents.

With hours on one side and incidents on the other, a contractor's injury and fatality rate becomes computable. We use the standard MSHA and BLS convention: incidents times 200,000, divided by hours worked, which expresses harm per 100 full-time-equivalent work years. The join between the two datasets is clean. On a plain contractor-ID key, 96 percent of contractor-attributed accidents and 95 percent of contractor fatalities match a filing in the exposure record, with no key normalization needed.

The data has texture, and we handle it in the open. About 40 percent of the filings report zero hours for the quarter, a contractor that existed on paper but did no work that period; those rows are excluded from every rate denominator. A small number of contractors file the same quarter more than once with different hours, which our loader resolves by keeping the latest row. Across the whole file that trims roughly 2 percent of total hours, which does not move a normalized rate. And about 4 percent of contractor incidents come from contractors who never filed hours at all. Those are genuine non-filers, not a matching failure, so we surface them as an unmatched bucket rather than dropping them quietly.

What this record supports is a defensible per-hour safety rate for any contractor with enough recorded activity, and national aggregates across contractors and years. What it does not support is a stable fatality rate for a single contractor in a single year. There are 268 contractor fatalities in the entire 26-year record, so per-contractor-year fatal rates are too noisy to publish. We show rates only above an activity floor, and we aggregate when the counts are small.

This is the first in a series documenting each dataset we ingest: what it is, what it lets us measure, and where its limits are. The contractor exposure record sits behind every contractor page on the site: miningincidents.org