The Cheaper the Audi, the Drunker the Driver. BMW Works the Other Way Around.

According to the toxicology reports, and there are a lot of them, Audi sells a sobriety ladder. The A3 sits at the bottom: of 203 A3 drivers in fatal crashes, 27.1 percent tested positive for alcohol or drugs, the highest rate of any Audi in a decade of federal data. Climb the lineup and the drivers get cleaner at every rung: Q5, A4, A6, Q7, five models, five steps down, zero exceptions.

That staircase is the prettiest thing this desk has published all year. Full disclosure, because the lab demands it: rung to rung, the gaps are only marginally significant. The A3-to-Q7 spread carries a p-value around 0.06, which in statistics translates to "interesting, come back with more bodies." Still. Five consecutive steps in the same direction is a pattern, not a coincidence. Somebody at Audi is selling cheaper cars to drunker people, or drunker people keep buying cheaper Audis, and the toxicology screens cannot tell those apart.

Tesla runs the same direction: Model S drivers test positive at 24.0 percent, then the X at 21.3, the 3 at 20.1, the Y at 17.6, a clean descent from flagship to entry. Honda does it quieter: Civic 20.4, Accord 20.0, Pilot 19.4, CR-V 17.6, which makes three brands with the same moral. Last week this desk argued the price tag sorts the drivers. Inside Audi, Tesla, and Honda, the price tag sorts the drivers.

Then BMW walks in and flips the table. Its drunkest drivers buy the flagship: the 7 Series posts 26.1 percent impairment, a full seven points above the 5 Series at 19.0, and that gap clears significance at p around 0.02, which means this one is real. Meanwhile the M5 contributes its own wrinkle, with a drug-positive rate of 15.6 percent nearly matching its 17.4 percent alcohol rate, one of the most drug-leaning sedans in the set. Mercedes refuses to play entirely: the mid-price E-Class leads the brand at 23.5 percent while the S-Class loafs in the middle at 21.9. Toyota inverts too: the Avalon, the most expensive Toyota sedan, has the drunkest Toyota sedan drivers at 21.7 percent, while the Land Cruiser, the most expensive Toyota of any kind, posts just 8.9 percent, among the lowest in the database.

So much for a universal law, and the revised theory, after arguing with myself for an afternoon, is that price never sorted anyone. Cohorts sort themselves, and a badge means different things to different tribes. An A3 is the cheapest ticket into the Audi club, which skews young, and NHTSA's 2024 data puts drivers aged 21 to 24 at the top of the impaired-driver standings at 28 percent.^[2] A 7 Series is the most expensive way to announce you have arrived, which skews toward men with expense accounts and late dinners; nationally, male drunk drivers outnumber female ones four to one.^[3] Different tribes, different bars, same lab results.

Now the strongest case against everything above, stated at full strength: I may have measured youth, nightlife geography, and inconsistent state testing regimes, then labeled the whole stew "price." Some rungs rest on small samples, 109 M5s and 150 Model Xs, which is worth remembering before crowning anyone. Drug-positive is not the same as impaired, since THC lingers in blood for days after the high fades, and NHTSA warns explicitly against conflating the two.^[3] FARS captures only fatal crashes, the worst 40,000 of roughly 6.7 million annual wrecks. Nothing here proves a car causes anything; it proves cohorts differ, which is all this desk ever claimed.

The practical question is what you actually do with it. If you are shopping the cheapest badge in a luxury lineup as a first car for a teenager, know which cohort you are joining: entry-luxury sport sedans are cheap enough for young risk-takers and fast enough to kill them. IIHS guidance for teen drivers says skip high-horsepower vehicles and treat electronic stability control as non-negotiable.^[4] Whatever you buy, run the VIN at nhtsa.gov/recalls before money changes hands.^[1] You cannot buy sobriety. You can avoid buying into its fan club.