US fatal crash rates per mile by gender, 2022

• Window: calendar year 2022, the most recent year both the crash census and a national travel survey cover. Dates are UTC.
• Grain: one row per survey year, driver sex and age band, with an all ages and an all drivers roll-up in every combination.
• Cadence: the crash file is published once a year; the travel survey runs about every five years, so a new rate year arrives only when the survey does.

• survey_year — the year both sides describe, 2022.
• sex — male, female, unknown, or the all drivers roll-up.
• age_group — a band from under 16 to 80+, unknown, or the all ages roll-up.
• drivers_in_fatal_crashes — drivers of a motor vehicle in transport involved in a fatal crash that year. A count of every one of them, not an estimate.
• annual_vmt_miles — vehicle miles that group drove in the year, in miles, estimated by weighting each surveyed trip to a year of national travel.
• trips_sampled — how many surveyed trips the mileage estimate rests on.
• vmt_share — the group's share of all miles driven that year, between 0 and 1.
• fatal_involvements_per_100m_vmt — drivers in fatal crashes per 100 million vehicle miles that group drove.
• rate_note — empty when the rate is published, otherwise why it was withheld.

The rate is empty on three kinds of row, and rate_note says which: rows where a source did not record sex or age, because an unrecorded age in a crash file and a refused age in a survey are different populations; rows where the survey estimated no miles for the group; and rows resting on fewer than 30 sampled trips, which is too thin to divide by. Counts and miles are always present, zero included.

• Comparing men's and women's crash risk per mile rather than per head.
• Separating how much someone drives from how dangerously they drive.
• Age-by-sex risk curves for actuarial or road-safety work.
• A baseline for comparing other countries or later survey years.

FARS is NHTSA's census of every US crash involving a death, published as annual files; NHTS is the Federal Highway Administration's travel survey, published by Oak Ridge National Laboratory. Both are works of the United States federal government, in the public domain, and both were downloaded without a key. FARS counts fatal crashes only, so this is not a rate for all accidents. The survey measures about 1.85 trillion household vehicle miles in 2022 against a national total near 3.2 trillion that includes commercial traffic, so rates here run above figures built on that wider denominator; vmt_share is published so any reader can rescale to a total of their choosing.

The 2017 survey wave is not included: its trip file is too large for one decode, and its person file measures self-reported annual mileage rather than trip-diary miles, which would put two different measures in one column.