America Stopped Arresting Drunk Drivers and Deaths Dropped Anyway

According to the toxicology reports, the United States ran an accidental experiment starting in 2020: DUI arrests collapsed, traffic stops evaporated, and sobriety checkpoints vanished in a matter of months. NORC researcher James Fell documented the enforcement cratering in a 2025 study, noting that arrests went down "dramatically" during COVID, and they never came back.^[1]

Predictable result: impaired fatalities surged to 13,617 in 2021, the worst in a generation.^[1] Unpredictable result: they started falling, and kept falling. Overall traffic deaths dropped 6.7% in 2025 to 36,640, while speeding deaths plunged 11% back to pre-pandemic levels.^[2][3] Colorado's impaired-crash share dropped from 31% to 22% year-over-year without a corresponding rise in enforcement activity.^[4] Q1 2026 posted a fatality rate of 0.99 per 100 million VMT, while Americans drove 11 billion more miles than the year before.^[3] More driving, fewer deaths, same cops doing less.

Not policing, and certainly not a sudden outbreak of sobriety. Electronic stability control, mandated since 2012 and now virtually fleet-universal, cuts fatal single-vehicle crashes 56% and rollovers 78%.^[5] A drunk driver who overcorrects at 1:30 AM in a 2024 sedan is dramatically less likely to spin out than in a 2010 model. Automatic emergency braking went from luxury option to near-standard equipment across every major automaker. Ride-hailing reached saturation in every metro above 100,000 people, reducing alcohol-involved fatalities roughly 6% in urban markets.^[6] None of these required a single officer to leave the station.

Fell's NORC research catalogs interventions that do involve officers, and they demonstrably work: sobriety checkpoints reduce impaired fatalities about 20%.^[1] Enforcement is real, and the United States still runs impaired fatality rates "significantly higher" than peer nations where enforcement never collapsed.^[1]

Fair counterargument: this is just the pandemic hangover wearing off. Benchmark against 2019 instead of 2021 and progress looks modest, which is a legitimate critique that concedes the larger point: if deaths reverted to baseline without enforcement recovering, the mechanism everyone assumed was essential proved dispensable. FARS toxicology across 490,736 drivers shows impairment rates steady by class: sports cars 22.5%, sedans 20.4%, pickups 20.1%.^[7] People are still driving impaired at the same rates, but cars are catching more of their mistakes before those mistakes become body bags.

If you drive a pre-2012 vehicle without ESC, you are sitting in the era before the safety floor existed, and you should check whether your car has it at nhtsa.gov/equipment/electronic-stability-control. If you are shopping used, ESC and AEB availability should be your first filter, not horsepower. Fell's NORC study documents interventions like sobriety checkpoints that still produce measurable reductions on top of the engineering baseline, so both levers work, but only one kept working when nobody was paying attention.