US road deaths per 100,000 residents, by county (2019–2023)
NHTSA FARS fatalities over Census county population, pooled across
5 years. 3,135 counties; 1,938 have at least 20 deaths over the
period and are treated as reliable.
Population is not exposure, and in small counties it is badly
wrong. Median rate by county size: 200 per 100k/yr in counties under
5,000 people, 8.3 in counties over 500,000 — a 24× spread.
Rural residents are not dying 24× more often per mile driven; they live
where other people’s driving happens. Loving County, Texas tops the raw
list at 9,767 per 100k from 21 deaths among 43 residents. Per-vehicle-mile is the
right denominator; county VMT is not published, so this uses what exists and says
so.
Hatched counties have fewer than 20 deaths over the whole
period — the NCHS threshold below which a rate is not considered reliable.
They are drawn rather than dropped, because a blank county reads as missing data
rather than as a small number. Connecticut is excluded (1,426 crash
records): FARS still codes it with the 8 legacy counties while the Census 2023
vintage uses 9 planning regions, and the two do not intersect at all.