Died at Work

Every figure on this page is read from real BLS bulk files. Nothing here is estimated, modelled or curated. The counts come from the Census of Fatal Occupational Injuries - a census, not a sample: BLS substantiates each death against two or more independent source documents (death certificates, workers' compensation filings, OSHA and other agency reports) before it becomes a row. The rates and their margins of error come from the CFOI hours-based rates table. The wages come from the Occupational Employment and Wage Statistics survey for May 2024. The two are joined on the SOC occupation code. Vintage: 2024 reference year (CFOI released 2025-12; includes the BLS data correction of 2026-03-24). Wages: OEWS May 2024.

The rate is published, never derived

It is tempting to compute the fatality rate yourself: 5,070 deaths against 297,600 million hours worked, scaled to a 200-million-hour FTE year, gives 3.41 per 100,000. That number is wrong. BLS publishes 3.3, because the rate's numerator excludes volunteers, active-duty military personnel and workers under 16, while the headline count includes them. We ship the published rate everywhere and show the naive one only to name the mistake.

The wage join crosses two universes

This is the biggest caveat on the page, and it is the reason the central finding should be read as an order of magnitude rather than a decimal. CFOI's fatality rates cover everyone who works, including the self-employed, who are killed at 10.8 per 100,000 against 2.9 for wage and salary workers. OEWS covers employees only. The self-employed are in the numerator of the risk and absent from the pay.

The gap is not theoretical. OEWS publishes no wage at all for fishing and hunting workers - ranked #2 in the country for fatal risk - because it is not, for the most part, an employee job. We keep that row on the page with an empty wage rather than dropping it, because a silently missing row is the kind of thing that turns a caveat into a lie. The headline comparison therefore rests on the 9 of the 10 deadliest occupations that OEWS does cover: their median pay is $49,550 against $49,500 for all U.S. occupations.

The margins of error are large, and the ranking is soft

BLS publishes a margin of error on every rate, and for small occupations they are wide: logging is 110.4 +/- 34.88, fishing and hunting 88.8 +/- 42.31. Those intervals overlap. "The deadliest job in America" is a headline the data does not quite support, so we draw the whiskers and say so.

Occupations, not the groups that contain them

The rates table interleaves SOC major groups ("Construction and extraction occupations") with the detailed occupations inside them. Ranking them together would count the same workers twice, so we keep only the 47 detailed occupations that carry a published rate. If you see a count of 66 elsewhere, that figure includes the parent groups.

Three surveys, one census

CFOI is split across three BLS surveys by taxonomy era - FI (2003-2010), FW (2011-2022) and FA (2023 onward). They are separate because the SOC, NAICS and OIICS revisions broke comparability between them. The fourteen-year trend line uses the national total only, which does survive the break, and the break is drawn on the chart. Every detailed cut on this page - occupation, event, state - is FA-only. Do not concatenate the detailed categories across the eras.

The state rows double-count if you sum them

CFOI publishes New York three ways: the state excluding New York City, New York City, and the two combined. It also publishes Puerto Rico and Guam, which are not in the national total. Add up every area row and you get 5,309 against a true 5,070. We build New York from its two parts, drop the territories, and the build fails loudly if the 51 jurisdictions do not sum to the national total exactly.

What "died at work" includes

CFOI classifies a death as work-related on the circumstances of the event, not on its cause. So the 5,070 includes 263 suicides and 410 unintentional overdoses that occurred at work - together about 13% of the total. Whether a suicide in a workplace is a workplace death is a real question and we do not answer it for you. We do not remove those cases, and we do not bury the fact that they are in there.

What you are not seeing

Only fatalities. Non-fatal injuries and illnesses come from a different program (the SOII sample survey, with its own under-reporting and suppressed cells) and are out of scope here entirely. There is no state-level fatality rate on this page, because BLS does not publish one in this table and we will not manufacture one from a workforce denominator taken from another survey. 8 of the 5,070 deaths carry no event classification, so the event shares are shares of the 5,062 that do.

---