The Most Dangerous Drivers on the Road Aren’t Drunk. They’re Sober.

We built an entire moral framework around impaired driving. MADD. Billboards. Ignition interlocks. Drunk. High. Reckless. But when you cross-tabulate FARS toxicology with kill ratios across 182 vehicle models over a decade, a different pattern surfaces, one that has nothing to do with impairment and everything to do with mass.

Two FARS fields make the metric: one measures how often a vehicle’s drivers in fatal crashes test positive for alcohol or drugs, and the other measures how often the dead person is the vehicle’s own occupant versus someone else. Invert both, multiply, and you get the Sober Killer Index — the joint probability that a given fatal crash involving this model features a sober driver and a dead person outside it.^[1]

The Ram 2500 scores 0.639. Of its drivers in fatal crashes, 80.3% are completely sober; of the dead, 79.5% are not inside the Ram. A 7,000-pound truck doing exactly what physics demands when it meets a 3,200-pound sedan at speed. Ford Transit follows at 0.597, then GMC Acadia (0.538), Ram 1500 (0.525), BMW X3 (0.517, ranking on exceptionally high sobriety rates more than mass alone). Class averages: vans 0.397, SUVs 0.388, pickups 0.382, sedans 0.279. At the bottom sits the Corvette at 0.086 — 88.4% of the time, the only fatality is its driver.^[1]

Fifty years of deterrence architecture, aimed at the drunk driver. Meanwhile, the statistically modal fatal outcome on American roads is a sober adult in a heavy vehicle killing someone in a lighter one, facing no charge because nothing illegal happened.^[2]

Counterarguments deserve weight. Mass differential is physics, not negligence: FARS encodes no fault, so a law-abiding Ram driver struck by a red-light-running Civic scores identically to one who caused the crash, and many heavy-vehicle drivers need the truck for work, not image, logging rural highway miles where any collision turns fatal. Part of the SKI is also a mathematical artifact: heavy vehicles simultaneously lower own-occupant death risk and raise the other party’s, so the product trends high regardless of driver behavior. “Sober” itself means only “no substances detected” — FARS toxicology varies by state and omits fatigue entirely.

All of that is true. The dead are still disproportionately outside the truck.

Limitations: SKI measures outcome distribution conditional on a fatal crash occurring, not per-mile crash probability, which differs sharply by vehicle class and driving environment; single-vehicle crashes are included and depress SKI for classes prone to them; fleet estimates carry ±15% uncertainty for lower-volume models; and the 2013–2022 window spans major redesigns.

What to do with this: If you drive a top-20 SKI vehicle, treat sobriety as a baseline, not a credit. At 45 mph, your kinetic energy already exceeds what most compacts can structurally absorb. Leave more following distance than feels necessary, because your margin for error is not what you learned in a Corolla.