Cannabis Legalization Raised Fatal Crashes 4.1%. One in Four Impaired Drivers Is Invisible to the Breathalyzer.

According to the toxicology reports, and there are a lot of them, the great legalization experiment just received its bill. Five states that legalized recreational cannabis watched injury crash rates climb 5.8% and fatal crash rates climb 4.1% against six states that kept prohibition, after adjusting for seat belt use and unemployment.^[1] Colorado's injury rate jumped 17.8% while Nevada's fell 6.7%. Same plant, opposite outcomes.

The part nobody expected is the geometry of the damage. Fatalities rose less than injuries, because cannabis does something to drivers that alcohol never does: it makes them careful. Researchers call it the compensation effect. Reaction time slows, so drivers slow down, leave bigger gaps, avoid maneuvers. Stoned drivers still crash more, but they crash at speeds that maim instead of kill. Colorado and Oregon paid the fatality tax anyway, up 1.4% and 3.8%, while Washington, California, and Nevada did not. The early states stumbled while the later ones watched and learned, which makes this a policy result rather than a plant result.^[1]

Meanwhile the enforcement apparatus is fighting the last war. I ran FARS toxicology across the fleet and found 98,348 impaired drivers in fatal crashes, of whom 24,169 had no alcohol in their system at all. A breathalyzer scores zero on nearly a quarter of impaired fatal crashes. And per-se THC limits are theater: THC lingers in blood for weeks after use, so "cannabis-involved" is as much a test artifact as a risk factor. Charles Farmer, the IIHS researcher behind the new study, says it outright: there is no objective measure of how impaired a cannabis user has become.^[1] America built a drunk-driving enforcement machine, then aimed it at a drug it cannot see.

The demographic surprise: the druggest cars are not the Corvettes. The Buick Park Avenue, with 16.6% of its fatally-involved drivers drug-positive, plus the Verano at 13.1% and the Ford Five Hundred at 13.0%, proves that grandma sedans lead the league in chemistry. Now the strongest case against everything I just said: the measured cannabis surge may be a testing artifact, because legalize and cops test more. THC stays detectable for weeks, so rising "involvement" measures rising use plus testing enthusiasm, not necessarily rising impairment. A separate 19-year FARS study found cannabis-involved deaths doubled from 9.0% to 21.5% between 2000 and 2018, and found those crashes increasingly paired with alcohol, even below the legal limit.^[2] The DRUID crash-risk data put THC-positive drivers at 1 to 3 times the crash risk of sober drivers, roughly a glass of wine, while a 0.15 BAC is twelve times.^[3] In that reading the 4.1% tax is real, but the killer is still mostly alcohol wearing a leaf costume.

Caveats, honestly stated: FARS records drug presence, not intoxication, and the per-model data cannot split cannabis from stimulants or opioids, so the 24.6% figure covers all drug classes. Testing rates vary by state and year. Farmer's study is correlational by his own admission. What to do with all this: if you drive, the risk that compounds is polysubstance use, nearly a third of impaired fatal-crash drivers in the fleet data carry two or more substances. If you legislate, spend on roadside impairment assessment, drug recognition experts and oral-fluid testing, rather than per-se THC thresholds that measure last month's choices. The tax is real, and so is the blindness.