A Mazda CX-5 Kills 15 Owners Per Million Per Year. A Chevy S-10 Kills 652.

NHTSA publishes fatality rates per 100 million vehicle miles traveled. IIHS publishes driver death rates by vehicle class. Neither gives you the number that actually matters: if a million people own your car, how many of them die in it each year?

We computed it. For every model in the FARS database with at least 100,000 registered vehicles and 100 or more fatalities over 2014–2023, we divided annual deaths by estimated fleet size and multiplied by one million. One number. Annual deaths per million owners.

Mazda CX-5 scores 15. Fifteen people per million CX-5 owners, per year, die in their vehicle. Chevrolet S-10 scores 652. Same country. Same roads. Same decade of data. Forty-three times the risk.

The Safest Ten

Every vehicle in the bottom ten is a modern crossover SUV, a heavy-duty truck, or a commercial van. Mazda CX-5 (15), Honda HR-V (16), Volkswagen Tiguan (17), Ford Transit (17), Ram 1500 (17), Ram 2500 (17), Ford F-350 (21), Chrysler Pacifica (23), Toyota RAV4 (24), Chevrolet Traverse (25). These aren't performance machines. They're the grocery-run and job-site fleet. And they're the safest vehicles on American roads by the metric that should matter most to anyone making a monthly payment.

The Deadliest Ten

Chevrolet S-10 (652), Nissan Maxima (588), Chevrolet Cobalt (587), Chevrolet Impala (575), Toyota Solara (489), Ford Mustang (482), Ford Ranger (392), Dodge Dakota (353), Chevrolet Trailblazer (353), Honda Accord (353). Cobalt carries the ignition switch scandal's statistical scar.^[4] Impala hauled rental fleets into guardrails for two decades. But Honda Accord at number ten will sting. America's most trusted nameplate. 353 deaths per million owners. Fourteen times deadlier than a CX-5.^[1]

Blame demographics, not engineering. Millions of 1998–2008 Accords still circulate, driven hard, maintained variably, owned by people who bought the cheapest reliable sedan on Craigslist. A vehicle that gets you through college is not a vehicle that protects you in a 55-mph offset.

Why This Metric Disagrees With Per-Mile Rates

Consider the Dodge Challenger. Per-VMT death rate: 1.00 per 100 million miles. Middling by that measure. But its owner risk is only 80 per million. Why? Challenger owners barely drive them. Weekend cars, car-show queens, garage trophies. Low mileage suppresses total exposure. Mediocre engineering, saved by habits.^[2]

Pickups show the opposite pattern. Ford Ranger posts a per-VMT rate of 2.91, but an owner risk of 392. Ranger owners rack up miles. Work trucks get driven. Per-mile rate looks reasonable; per-owner rate reveals the full cost of that driving pattern.

Per-VMT answers a narrow question: how dangerous is one mile in this vehicle? Owner risk answers the question buyers actually face: how likely am I to die in this thing over the next year?

The Heavy-Truck Surprise

Ram 2500: 17 deaths per million owners. Ford F-350: 21. These are 7,000-pound machines that obliterate anything they hit. By the aggressor metric, they're terrifying. By the owner-risk metric, they're the safest vehicles on the road. Mass protects occupants. Full stop. Occupants of a Ram 2500 are extraordinarily unlikely to die in it. Occupants of the Civic it T-bones are a different calculation.^[3]

Methodology

Owner risk = (total FARS deaths 2014–2023 / 10) / estimated registered fleet × 1,000,000. Fleet estimates derive from cumulative US sales data. We filtered to vehicles with 100,000+ registrations and 100+ FARS deaths to avoid small-sample distortion. VMT estimates use NHTS average annual mileage by vehicle class.

Limitations

Fleet size is estimated from sales, not actual DMV registration rolls. A model with high attrition (junked, exported, parked) would have an inflated fleet denominator, making it look safer than it is. Conversely, a durable model that stays on roads longer might have more registrations than sales data suggests. The metric also conflates vehicle design quality with driver behavior. The S-10 scores badly partly because its remaining drivers skew toward demographics with higher crash rates. Disentangling vehicle from driver requires micro-data FARS doesn't make public.^[1]

The Strongest Case Against This Metric

A vehicle could rank as "safe per owner" simply because its owners don't drive. The Corvette's per-VMT rate is 1.52, worse than the fleet average, but its owner risk is only 122 because most Corvettes log 3,000 miles a year. You could argue the per-VMT rate better reflects the vehicle's actual crashworthiness. Fair enough. But the person writing a check doesn't care about one hypothetical mile. They care about the year ahead.

What You Can Do

If you're shopping used, cross-reference the model against FARS data before you buy. A $6,000 Chevy Impala and a $6,000 Kia Soul are in the same price bracket but not the same risk universe: 575 vs. 64 deaths per million owners. Check your VIN for open recalls at nhtsa.gov/recalls. If you own a vehicle in the top ten deadliest, prioritize defensive driving courses and verify your insurance deductibles match the actuarial reality of your car. If you're buying new, the crossover SUVs that dominate the safest list start under $28,000. Safety isn't a luxury anymore. It's a segment.

Source: NHTSA FARS 2014–2023. Owner risk calculated as annual deaths / estimated registered fleet × 1,000,000. Fleet estimates from cumulative US sales data; actual registrations may differ. See methodology for caveats.