The Crash Report

NHTSA FARS national data

The Fatality Analysis Reporting System (FARS) is a census of all fatal motor vehicle crashes in the United States, maintained by NHTSA. FARS covers all crashes nationally and can be normalized by vehicle miles traveled (VMT) — but only at the broad vehicle-class level (passenger cars, light trucks, motorcycles), not per make/model.

VMT data comes from the FHWA Highway Statistics Table VM-1, which estimates total miles driven annually by vehicle type. Dividing FARS fatalities by VMT yields the "fatality rate per 100 million VMT" — the standard metric used in NHTSA Traffic Safety Facts publications.

• 2024 data is an early estimate based on NHTSA's preliminary projections and is subject to revision.
• Fatalities by road user type (2014–2023) are final FARS counts.
• Per-class VMT rates use FHWA VM-1 data matched to FARS occupant fatality counts for the corresponding vehicle type.

FARS per-model estimated rates

The FARS per-model section aggregates all occupant fatalities across 2014–2023 from NHTSA FARS bulk CSV downloads, grouped by make/model. This data includes:

• All occupant fatalities (drivers + passengers), not just driver deaths
• All model years on the road, not a single MY cohort
• All vehicles involved in fatal crashes, regardless of registration volume

Since per-model VMT data does not exist publicly, estimated fatality rates use a proxy method:

• Fleet estimate: publicly reported average annual US sales × fleet multiplier (12.5 yr average vehicle age × 0.70 survival discount ≈ 8.75 effective fleet years)
• Annual VMT estimate: estimated fleet × NHTS class-average annual miles (sedans: 11,500 mi; SUVs: 12,500 mi; pickups: 13,500 mi; vans: 11,800 mi; sports cars: 8,000 mi)
• Rate: 10-year total deaths ÷ (estimated annual VMT × 10 years ÷ 100,000,000)

Key caveats:

• Sales figures ≠ registrations — fleet size estimates are approximate
• All vehicles within a class are assumed to drive the same annual miles
• Does not account for driver demographics, geographic variation, or vehicle age distribution
• Includes models with 50+ deaths or significant annual sales (>1k) for rate comparison

Impaired driving analysis

Impairment data comes from the FARS PERSON.csv file, filtered to drivers only (PER_TYP = 1). Each driver record is joined to its vehicle record via ST_CASE and VEH_NO.

• Alcohol positive: DRINKING = 1 (police-reported) OR ALC_RES (BAC test result) between 1–94 (BAC > 0.00 g/dL)
• Drug positive: DRUGRES1/2/3 values 100–295 (specific drug detected in toxicology) OR DRUGS = 1 in older files
• Any impairment: alcohol positive OR drug positive

Key caveats:

• Toxicology testing rates vary significantly by state and jurisdiction — some states test nearly all fatally-involved drivers, others test far fewer
• Untested drivers are coded as unknown, not negative — actual impairment rates are likely higher than reported
• Only models with 100+ drivers in fatal crashes are shown to ensure statistical significance

Model year analysis

The MOD_YEAR field from FARS VEHICLE.csv identifies the model year of each vehicle involved in a fatal crash. Deaths are aggregated by (make, model, model year) across the 2014–2023 observation period.

• Older model years have more cumulative years of exposure on the road during the observation period, creating a natural age-related skew
• This chart reflects fleet-age composition and crash involvement, not inherent safety differences between model years
• Model years with fewer than 5 deaths are excluded
• Invalid model year values (0, 9998, 9999, pre-1980, or future years) are excluded

Scope: fatalities only

This dashboard covers fatal crashes only from the FARS census. NHTSA also maintains the Crash Report Sampling System (CRSS), which covers all police-reported crashes (including injuries and property damage) — but CRSS is a probability-based sample, not a census, and is not incorporated here.

Limitations

• NHTSA/FARS data: VMT normalization is only available at the vehicle-class level, not per make/model. Class-level rates mask variation within each category.
• FHWA VMT estimates are modeled from traffic counts and may not perfectly reflect actual travel.
• FARS per-model: Estimated rates depend on sales-as-fleet-proxy assumption. Vehicles with much higher or lower than average usage will have distorted rates.

Sources

NHTSA FARS database →  |  NHTSA Traffic Safety Facts →
FHWA Table VM-1 →
FARS bulk CSV downloads →  |  NHTS (National Household Travel Survey) →
FARS/CRSS Coding and Validation Manual →