Open data

Every death, one CSV

One row per person killed, 2010–2024, versioned in the open. No portal or login — a plain CSV you can cite, fork, or take to a city council meeting.

The dataset

Every pedalcyclist fatality in NHTSA's Fatality Analysis Reporting System, 2010–2024, cleaned and joined across the FARS crash, person and vehicle tables — 37 columns per death: date, day of week, location (99.5% geocoded), street and cross street, road class, weather, victim age and sex, where they died, light condition, striking vehicle class and speed against the posted limit, and the striking driver's age, license status, and prior crashes, DWI and speeding convictions from state records.

cyclist_fatalities.csv13,050 rows · 37 columns · 2.7 MB · CC BY 4.0
DATA_DICTIONARY.mdcolumn definitions & caveats
aggregates.jsonpre-computed chart data · JSON
cases.jsonnamed, cited recent cases · JSON

The CSV is anonymous because its source is: NHTSA strips all names and personal identifiers from FARS before release. Names appear only in the curated case files above, where each one is verified against published reporting under the rules in EDITORIAL.md. Driver names and court outcomes, traced through public records, are the next phase of the registry.

Methodology

  • Source: NHTSA FARS annual files (public domain), person type 6–7, fatal injury.
  • Validated: yearly totals match NHTSA published counts (2010: exact; 2023: 1,173 exact).
  • Hit-and-run, impairment and speeding are crash-level flags — "any vehicle in the crash," NHTSA's own convention.
  • 2024 is NHTSA's Annual Report File and will be revised; we re-release when FARS finalizes.
  • Pipeline code is open source — every transform is inspectable and re-runnable.

License

Data: CC BY 4.0 — use it for anything, credit "Last Ride Registry." Code: MIT. Upstream FARS data is US-government public domain.


The pattern

What fifteen years of data show

Fifteen years of federal crash data, one row per person. The numbers below reconcile with NHTSA's published totals to within a fraction of a percent, and every row behind them is downloadable above.