One in Four Drivers Who Died in These Cars Was Impaired. They Are Some of the Safest Cars on the Road.

I cross-tabulated two FARS tables that nobody cross-tabulates: driver toxicology against fatality rate, all 307 models. Four vehicles landed in the quadrant that should not exist. About one in four drivers killed in them was alcohol- or drug-impaired, yet their death rates sit at or below the fleet median.^[1] I ran the numbers, then I ran them again, and the quadrant stayed put.

Meet the drunk tank index: the Ford Five Hundred (26.4% of fatal-crash drivers impaired, 0.57 deaths per 100M miles), the Toyota FJ Cruiser (25.3%, 0.43), the Jeep Commander (24.9%, 0.28), and the Buick Verano (24.7%, 0.46).^[1] Fleet medians are 20.2% impaired and 0.54 dead, so every one of these clears the impairment bar by a wide margin while matching or beating the lethality bar. For calibration, the fleet-wide drug-positive rate is 8.7%; the Verano's drivers test drug-positive at 13.1%, the Five Hundred's at 13.0%.^[1]

The obvious reading says these cars attract impaired drivers, which is wrong, or at least incomplete. FARS impairment percentages are conditional on the crash being fatal, and that conditioning does quiet, enormous work. In a well-built car, the sober driver walks away from the crash and never enters the dataset at all, which means the remaining fatal-crash pool gets enriched for impaired drivers automatically. Under this lens, sky-high impairment paired with rock-bottom lethality is not a drunk-magnet signature but a protection signature: these are cars where it takes an impaired driver to die in them.

The crash-test receipts back that reading, starting with the Five Hundred, which earned IIHS Top Safety Pick for 2006 with Good ratings in the frontal, side, and rear tests, plus five NHTSA stars front and side.^[2] Verano took the 2012 Top Safety Pick with Good across the board and a five-star NHTSA overall score.^[3] FJ Cruiser scored Good in IIHS frontal-offset and side tests with five NHTSA stars for the driver up front and in side impacts.^[4] The Commander is the outlier of the four: never IIHS-tested, but NHTSA gave it five stars frontal for driver and passenger, and every Commander shipped with three rows of side curtains on rollover sensors plus stability control with roll mitigation.^[5] Four different engineering departments, one shared outcome: structures that forgive almost everyone.

Now the strongest case against my own index, stated at full strength. Toxicology testing is not random: police order screens more often when impairment is suspected, so the percentages skew high in ways that vary by state, and the states where Commanders and FJ Cruisers die may simply test more. Sample sizes are thin where it matters most. Forty-six Commander deaths means a handful of misclassified cases swings the percentage by points. I cannot fully exclude differential testing, and neither can anyone else working from FARS. What I can say: the protection story is the only explanation that also fits four independent sets of crash-test ratings, while the testing-bias story has to get lucky four times.

What this means if you are shopping used, which is the only actionable question here. All four nameplates are deep into depreciation, all four aced the crash testing of their era, and the FARS record says their structures forgive even impaired drivers, a brutal real-world test no lab can run. None of that excuses impaired driving, which remains the single most reliable way to defeat the engineering. Run any VIN through nhtsa.gov/recalls before money changes hands.^[6] Buy the structure. Skip the impairment.

What I did not prove, because the data will not let me: FARS records only fatal crashes, roughly 40,000 a year against some 6.7 million total, so low fatality rates can mask high injury rates. My rates divide by estimated vehicle miles traveled, not odometer readings, which injects meaningful uncertainty into low-volume models like the Commander. Drug panels differ by state and year, and a positive screen measures presence, not causation. Treat the index as a signal about structures, not a verdict on drivers.