Data

FARS data explained

Name: US traffic deaths by state and year, from FARS
Creator: US Road Deaths
License: https://www.usa.gov/government-copyright

The Fatality Analysis Reporting System (FARS) is NHTSA's record of every crash on a US public road in which someone died within 30 days: about 35,000 to 45,000 crashes a year since 1975, with details of the crash, every vehicle and every person involved.

The data on this site

The latest year here is FARS 2024, published by NHTSA on 1 April 2026 and loaded on 30 September 2026. Totals go back to 1975, and every crash, vehicle and person record is loaded for 2020–2024.

Data last updated 30 September 2026. Which years each measure covers

What is in FARS

Each year is a set of linked files. The three main ones are:

Accident: one row per crash, with the date, time, location, road, weather, light and number of deaths.
Vehicle: one row per vehicle, with its make, model, year, body type and factors such as speeding.
Person: one row per person, with their role (driver, passenger, pedestrian, cyclist), age and injury.

Other files add detail, including NHTSA's statistical estimates of drivers' blood alcohol where no test result was recorded.

Fields this site relies on
Field	File	What it holds
ST_CASE	All	The case number that links a crash to its vehicles and people. Unique within a year.
STATE, COUNTY, CITY	Accident	Where the crash happened, as FIPS-style geographic codes.
FATALS	Accident	How many people died.
LATITUDE, LONGITUD	Accident	The crash location, from the police report. Sometimes missing or approximate.
PER_TYP	Person	The person's role: driver, passenger, pedestrian, cyclist and so on.
INJ_SEV	Person	Injury severity; 4 means the person died.
BODY_TYP	Vehicle	The kind of vehicle, from passenger car to motorcycle and heavy truck.
SPEEDREL	Vehicle	Whether police recorded the driver as speeding (from 2009).
DRDISTRACT	Distract	What police recorded as distracting each driver, if anything (comparable from 2012).
DR_DRINK	Vehicle	Whether police reported the driver had been drinking.
A1 to A10	MIDRVACC	Ten estimates of the highest driver blood alcohol in the crash (from 1982).

How NHTSA releases it

Each year's file comes out more than a year after the year ends, once states have finished their records. About a year later NHTSA publishes a final version of that year alongside the next one. The final file can change the counts slightly as late records come in, so the most recent year is always the most likely to be revised.

Getting the raw data

NHTSA's FARS page: the official overview, with links to the files and documentation.
The FARS download folder: every year's files as CSV or SAS, from 1975. Best if you want to work with the records yourself.
FIRST, NHTSA's query tool: build tables in the browser without downloading anything.
NHTSA crash statistics: published reports and Traffic Safety Facts, including the analytical user's manual that documents every field.

Download tables from this site

These CSV files are built from FARS and updated when new data is published. Each ends with a source line.

Every year, nationally and for each state
Deaths, crashes, population, rate and deaths by type, one row per place and year.
US traffic deaths by year
Deaths and fatal crashes for every year.
US deaths per 100,000 people by year
From 2000.
Traffic deaths by state, latest year
Deaths, rate, five-year average rate and change for all 51.
Motorcycle deaths by year
From 1975.
Motorcycle deaths by state
Latest year, with rates and shares.
Drunk driving deaths by year
From 1982.
Drunk driving deaths by state
Latest year, with rates and shares.
Pedestrian deaths by year
From 1975.
Pedestrian deaths by state
Latest year, with rates and shares.
Bicycle deaths by year
From 1975.
Bicycle deaths by state
Latest year, with rates and shares.
Speeding deaths by year
From 2009.
Speeding deaths by state
Latest year, with rates and shares.
Distracted driving deaths by year
From 2012.
Distracted driving deaths by state
Latest year, with rates and shares.

Every state, county and city page also has a CSV of its crashes under the crash list, and every chart on the site has its own CSV.

Using FARS correctly

Deaths are not crashes. One crash can kill several people, so there are always more deaths than fatal crashes.
The 30-day rule. Someone who dies 31 days or more after a crash is not counted.
Codes changed over time. Person types were recoded in 1982 and vehicle body types in 1982 and 1991, so compare old and new years with care.
Blood alcohol is often estimated. Many drivers are never tested. NHTSA's alcohol-impaired figures use statistical estimates; police-reported drinking is a narrower, different measure.
Factors are as recorded by police. Speeding or drinking does not establish fault or cause.
Rates per person are not rates per mile. Places where people drive further have higher per-person rates.
Small numbers swing. A county going from 2 deaths to 4 is a 100% rise that may mean nothing.

How this site counts and defines everything

Reuse and citation

FARS is a US government work and free to reuse. If you use tables from this site, a citation like this helps readers find the source:

US Road Deaths (https://usroaddeaths.com), tables built from NHTSA's Fatality Analysis Reporting System, 1975–2024. Accessed 1 October 2026.