Summary
Waymo says its analysis found that sober drivers’ fatal-crash risk at night was 3.1 to 3.9 times the daytime risk. A separate analysis of US fatal crashes found that safety risks often overlap rather than arise from human behaviour alone.
Waymo says an analysis of sober driving across the 50 most populous US urban areas found fatal-crash risk at night was 3.1 to 3.9 times higher than during the day. The company presented that finding alongside a second study of fatal crashes, arguing that road safety depends on more than whether a driver is impaired.
The first study has been accepted for publication in the Journal of Safety Research, according to Waymo. Its researchers estimated miles driven by people with zero blood alcohol concentration (BAC) and by impaired drivers, then used those estimates to compare fatal-crash involvement rates. The second analysis examined 37,654 fatal cases in the 2023 Fatality Analysis Reporting System (FARS) database.
A sober-driver baseline still shows elevated nighttime risk
National traffic databases do not record vehicle miles traveled by a driver’s BAC, Waymo says. To estimate that exposure, its researchers cross-referenced toxicological data from FARS with real-world telematics. The resulting estimates allowed them to calculate local baselines for sober and impaired human drivers.
Across the urban areas studied, removing alcohol-involved drivers reduced the average fatal-crash involvement rate from 1.42 to 1.10 incidents per 100 million miles—a 23% decrease, according to Waymo. The size of the decrease varied by city: it was 31% in Houston and 13% in Minneapolis. Those differences illustrate why local comparisons can reveal patterns obscured by a single national average.
The nighttime finding addresses a related question: whether alcohol impairment alone explains higher overnight crash risk. Waymo reports that even among sober drivers, estimated fatal-crash risk at night remained 3.1 to 3.9 times the daytime level. The company points to reduced visibility, inadequate lighting, fatigue, road layouts that encourage speeding and interactions with unpredictable road users as risks that sobriety cannot remove. Sober drivers may also still have to respond to impaired people sharing the road.
Fatal crashes involve overlapping risks
Waymo’s second analysis uses the Safe System Approach, which treats road safety as a shared responsibility involving safer people, vehicles, speeds and roads. In the 37,654 fatal cases it evaluated, more than 77% involved at least one of three factors: impairment, speeding, or seat-belt or helmet misuse or non-use. In 9.2% of cases, all three were present. Waymo says only 8.2% involved human-behaviour factors alone.
These figures describe how often the factors appeared in the fatal cases examined; they are not rates of risk per mile. The analysis highlights how human behaviour can combine with road and vehicle conditions in fatal crashes. For example, Waymo describes an unlit road, a mismatch in vehicle size and unexpected speed as factors that may converge in one event.
The company says its autonomous-driving analysis estimates that such systems could address up to 83% of system failures contributing to passenger-vehicle-involved fatal crashes in the United States. That figure is Waymo’s estimate of potential, rather than a measured reduction in crashes. The blog does not detail how the estimate was calculated. The studies’ broader contribution is to set out different ways of examining risk: a sober-human baseline for comparison, and a system-level view of how factors can compound.