The 10 Soberest Vehicles in Fatal Crashes Are All Family Cars. 4,079 People Died in Them.

Run the FARS toxicology database through a ranking algorithm, sort every vehicle by the percentage of fatally crashed drivers who tested positive for any alcohol or drug, and the leaderboard’s last-place finisher is the Subaru Ascent. Just 8.2% of Ascent drivers in fatal crashes returned a positive result on any substance, against a dataset average of roughly 20% across all 307 models with 200 or more tested drivers.^[1] Below the Ascent on the sobriety standings: nobody. Dead last in impairment and first in something worse.

Four hundred and twenty-six of those deaths belong to the Ascent alone—the soberest vehicle in a dataset full of sober vehicles that keep killing people.

Here is the full bottom ten, ranked by any-positive impairment rate among drivers with 200 or more toxicology screens: Subaru Ascent (8.2%), Toyota Land Cruiser (8.9%), Chevrolet Tracker (12.7%), Ford Transit (13.6%), Dodge Nitro (14.0%), Nissan Quest (14.5%), Dodge Grand Caravan (15.3%), Honda Odyssey (15.4%), Lexus GX (15.4%), and Lincoln MKX (15.6%).^[1] Read that roster and listen to what it sounds like: three-row crossovers, minivans, work vans, and premium SUVs built for car-pool lanes and college move-in weekends, the kind of vehicles that come pre-loaded with Cheerio crumbs and orthodontic copay anxiety. The only bar associated with these vehicles is the one you ratchet across a car seat. Combined occupant deaths: 4,079, with an average impairment rate of 13.4%.

A note on what “impairment” means here, because the definitions matter and most reporting conflates them: FARS toxicology flags any positive result for alcohol (BAC > 0.00) or any drug, from cannabis to prescribed benzodiazepines. That is a lower bar than NHTSA’s headline statistic, which counts only drivers at or above the legal limit of BAC ≥ 0.08.^[2] Under that stricter definition, about 30% of annual traffic fatalities involve a legally impaired driver. Under the broader FARS screen, the dataset average is about 20% any-positive. The Ascent’s 8.2% clears both thresholds by a margin wide enough to park a three-row SUV in.

Flip the standings and the top of the impairment leaderboard is exactly where you’d expect if you’ve spent any time in this dataset: Buick Park Avenue at 31.7% any-positive, Chevy C/K pickup at 28%, Audi A3 at 27.1%.^[1] The relegation zone is old sedans with sun-bleached dashboards and bumper stickers from casinos that closed in 2014, pickup trucks with aftermarket exhausts and empty tallboy cans rolling around the bed, and aspirational German imports bought at auction by buyers who could afford the monthly payment but not the maintenance schedule. Impairment tracks who buys the car, not how the car is built, which nobody needed a federal crash database and a decade of toxicology screens to tell them. But the inverse is the finding nobody talks about: if sobriety correlated with survival, every Ascent driver would walk away. Every single one. They don’t. That 8.2% impairment rate coexists with 426 deaths in 759 fatal crashes—a lethality ratio of 0.561, meaning 56% of the time an Ascent appears in a fatal crash, one of its occupants dies.^[1]

Who gets toxicology-tested introduces its own bias: protocols vary by jurisdiction, at-fault drivers are screened more often than victims, and some states test nearly everyone while others barely bother. There is also a methodological break: FARS expanded its drug-reporting protocol in 2018, meaning the Ascent’s entire five-year window falls under the enhanced screen while older models straddle pre- and post-expansion datasets.^[1] All of which means the 8.2% is a snapshot of tested drivers in fatal crashes, not a census. The ranking could shift if every jurisdiction tested uniformly and every model year used the same protocol, but it probably wouldn’t shift much. When your sample is 305 drivers and the next-nearest competitor is nearly a full percentage point higher, the Ascent isn’t winning on noise. It’s winning on car seats.

When the driver isn’t drunk and isn’t high and still dies, the variables left standing are crash structure, vehicle mass, impact speed, and road geometry. Every one of those is an engineering problem with an engineering budget, and every one of them has had engineering solutions on the shelf since at least the first federal crash test in 1979. NHTSA has spent forty years and uncountable taxpayer dollars perfecting the “Drive Sober or Get Pulled Over” slogan and its seasonal PSA rotations, all of which does exactly nothing for the 80% of fatally crashed drivers who were already sober and already driving a vehicle that cleared every federal crash standard on the books. Bumper stickers are cheaper than interchange redesigns, and nobody in Washington has proposed otherwise. The Ascent’s 8.2% is not a drinking problem: it is a dying-sober problem, and nobody has a campaign for that.