A study by the Insurance Institute for Highway Safety (IIHS) indicates that Waymo autonomous vehicles were involved in 91% fewer rear-end collisions and 81% fewer injury-causing crashes than human drivers.
The researchers documented 1.28 police-reportable crashes per million driverless miles for Waymo. In comparison, human drivers in the same cities and timeframes averaged 4.06 crashes per million miles, establishing a 68% lower overall crash rate for the autonomous fleet. Waymo vehicles also demonstrated an 85% reduction in single-vehicle crashes, a 77% reduction in sideswipes, a 61% reduction in T-bone and angle impacts, and a 40% lower rate of being rear-ended by other vehicles.
The study analyzed 1,692 federal crash records involving Level 4 automated vehicles from operators including Waymo, Cruise, and Zoox between July 2021 and December 2024. After removing duplicate records, the dataset was reduced to 977 distinct incidents. The researchers further excluded private property collisions, manual driving incidents, and non-crash events, narrowing the total to 736 incidents.
Three researchers evaluated each report to determine if the collision met the criteria for a police report based on impact severity, vehicle damage, injuries, towing requirements, and airbag deployment. A consensus of at least two researchers was required for classification. Through this process, 159 incidents (22%) were classified as police-reportable or potentially police-reportable. This total comprised 89 Waymo crashes, 50 Cruise crashes, 10 Zoox crashes, and 10 from other operators.
Because only Waymo provided sufficient mileage data for a direct statistical comparison, the final analysis focused on its operations. Researchers assessed 64 police-reportable crashes across nearly 50 million fully driverless miles and compared them against 900,509 police-reported crashes spanning approximately 222 billion human-driven miles.
Geographically, Waymo’s crash rate was 76% lower than human drivers in Phoenix, 71% lower in Los Angeles, and 35% lower in San Francisco. Conversely, the rate was 4% higher in Austin, though researchers noted the sample size there was limited to approximately 555,000 Waymo miles.
The scope of the study excluded interstates and freeways, as Waymo did not operate on these roadways during the tracking period. Consequently, the findings are strictly applicable to driverless Waymo vehicles operating within specific zones of the four analyzed U.S. cities.
