NHTSA Says Roads Are the Safest in a Decade. Your Car Might Disagree by 71x.

That 1.10 is a weighted average of 337 vehicle models, and using it to judge your personal risk is like checking the average temperature across all 50 states to decide whether you need a coat in Fairbanks.

I decomposed the national rate using FARS fatal crash data from 2014 through 2023, cross-tabulating per-model fatality rates against the fleet-wide benchmark. The Mazda CX-5 sits at 0.12 deaths per 100 million VMT while the Hyundai Veloster sits at 8.54. Both vehicles, both rates, both collapsed into the same reassuring 1.10 headline. The spread between them is 71 to 1.^[2]

Seventy-one to one.

The math NHTSA did not show you

The national fatality rate is calculated as total deaths divided by total VMT: 36,640 fatalities over roughly 3.33 trillion miles in 2025.^[1] That formula works beautifully for policy dashboards and terribly for anyone trying to understand whether their Tuesday commute is dangerous. Because the national rate is a sum: each vehicle model contributes its own death rate multiplied by its share of total miles driven, and models with enormous fleets and low rates (the RAV4 at 0.19, the CR-V at 0.21) pull the average down so hard that models with catastrophic rates essentially vanish into statistical noise.

The five safest vehicles in FARS data are the Mazda CX-5 (0.12), Honda HR-V (0.13), Ram 1500 (0.13), Ford Transit (0.14), and VW Tiguan (0.14). The five deadliest are the Hyundai Veloster (8.54), Chevrolet Tracker (7.83), Toyota Land Cruiser (6.27), Ford Mustang (6.02), and Chevrolet Impala (5.00).^[2] Those top-five safest vehicles exist in every price bracket from $27,000 to $55,000, which demolishes the assumption that safety requires spending luxury money.

Why the improvement might not be improvement

The national rate dropped from 1.19 in 2024 to 1.10 in 2025, a decline NHTSA framed as the fifth-largest percentage decrease in FARS history.^[3] But fleet composition shifted during the same period: crossover SUV registrations climbed while older sedan registrations fell as vehicles aged out of the fleet. SUVs post an average lethality ratio of 0.524 in FARS versus 0.645 for sedans. If enough Impalas and Corollas from the danger decade of 2000 through 2008 left the road and were replaced by Tiguans and CR-Vs, the national rate would drop without a single driver changing behavior or a single road getting safer.^[2]

NHTSA cannot disentangle this because the 2025 preliminary data lacks model-level breakdowns, and FARS model-level data currently ends at 2023. The agency celebrated a number it cannot explain, and reporters repeated it without asking what was inside the average.

The strongest case for the national rate

Epidemiologists defend population-level metrics precisely because individual variation is noisy, and they have a point that deserves full voice. NHTSA sets nationwide standards for crashworthiness, AEB performance, and seatbelt use. Those standards require a population-level benchmark, and the trend line from 1.19 to 1.10 accurately reflects that the average American mile was marginally less lethal in 2025 than in 2024. The national rate was never designed as a consumer tool. Conflating its policy purpose with personal risk assessment is a category error, and the blame falls partly on journalists and advocates who use the number in consumer-facing contexts without qualification.

What you should actually do

Stop treating the national rate as your risk. It was never meant for that, and NHTSA should say so explicitly but does not.

Look up your vehicle in FARS data and check where it falls. If your model's estimated rate is below 0.5 per 100 million VMT, you are driving something meaningfully safer than the fleet average. If it is above 2.0, you face roughly double the national rate, and no amount of defensive driving fully compensates for the physics of a lighter, less structurally rigid vehicle in a collision with a 5,000-pound SUV.^[4]

The single fastest way to reduce your personal fatality risk is not a driving course or a dashcam or a bumper sticker about sharing the road but rather driving a different vehicle entirely. Vehicles with rates below 0.5 per 100 million VMT exist at every price point from the $27,000 Mazda CX-5 to the $55,000 Ram 1500, and the choice between a 0.12-rate vehicle and a 5.00-rate vehicle is a 42-fold difference in fatal crash likelihood that the national average papers over completely.

Limitations

FARS captures only fatal crashes. The 36,640 deaths in 2025 are a fraction of approximately 6.7 million total crashes per year, and a vehicle with a low fatality rate might have high injury rates that this analysis does not detect. Per-model estimated rates use nationally averaged VMT assumptions from the National Household Travel Survey rather than actual odometer readings, introducing roughly plus or minus 15% uncertainty for low-volume models.^[5] High-rate vehicles like the Veloster and Mustang may also attract more aggressive driving behavior, confounding the vehicle-versus-driver attribution. The 2025 national rate is preliminary; NHTSA's final figure could differ. And critically, I cannot compute the exact fraction of the 1.19-to-1.10 decline attributable to fleet composition shifts versus genuine safety improvements because model-level FARS data for 2024 and 2025 does not yet exist.