IIHS Finds Lower Waymo Crash Rates
The Insurance Institute for Highway Safety reported on July 23 that Waymo's driverless Level 4 vehicles had 68% fewer police-reportable crash involvements per mile than human drivers in four U.S. cities. The comparison covered San Francisco, Phoenix, Los Angeles and Austin, while IIHS warned that federal reporting and mileage data are inadequate for continuous monitoring as autonomous deployments grow.
The Insurance Institute for Highway Safety reported on July 23 that Waymo's driverless Level 4 vehicles had 68% fewer police-reportable crash involvements per mile than human drivers in San Francisco, Phoenix, Los Angeles and Austin.
The result applies only to Waymo's driverless operations and the cities and years studied. It should not be generalized to every automated-driving system or to consumer driver-assistance products.
What the study measured
IIHS researchers compared federal automated-driving crash reports with state records for human drivers in the same regions and years. They manually reviewed incident narratives to decide whether a reasonable person would have reported each crash to police, then calculated rates using vehicle miles traveled.
Waymo's driverless vehicles covered about 50 million miles during the study period, while the matched human-driver exposure was about 222 billion miles. The overall Waymo rate was 68% lower. The difference varied by city: 76% lower in Phoenix, 35% lower in San Francisco, 71% lower in Los Angeles and 4% higher in Austin, where the Waymo sample was relatively small.
The study also found 85% fewer single-vehicle crashes and 81% fewer injury crashes per mile for Waymo. Its rate of rear-ending another vehicle was 91% lower, while its rate of being rear-ended was 40% lower. Electrek reported the underlying overall rates as 1.28 police-reportable crashes per million miles for Waymo and 4.06 for human drivers.
Why the comparison is difficult
Automated vehicles and human drivers enter different reporting systems. Companies must report many incidents involving automated systems, while human drivers often do not report minor crashes. Until a 2025 federal change, automated-vehicle reporting also captured some very small incidents that would not ordinarily appear in police data.
IIHS began with 736 public-road crashes in which automation was engaged. After removing duplicate, off-road and non-crash records and manually coding severity, researchers judged 22% police-reportable or possibly police-reportable. Mileage data created another constraint: Waymo publishes driverless miles voluntarily, but most other operators do not, preventing equivalent rate calculations.
The paper says this manual narrative-coding method is not sustainable as deployments expand. IIHS called for consistent national crash definitions and mileage reporting so regulators and researchers can monitor changes over time rather than reconstructing comparable datasets after the fact.
What the result does and does not show
Waymo's vehicles are SAE Level 4: the system performs the driving task without human supervision inside defined operating conditions. That differs from Level 2 products such as GM Super Cruise and Ford BlueCruise, which require an attentive human driver. The study therefore does not establish a safety rate for supervised consumer automation.
For autonomy and ML teams, the methodological lesson is as important as the headline number. A credible fleet comparison needs a shared event definition, a mileage denominator, matched operating conditions and auditable incident records. Without those pieces, a lower crash count may reflect exposure and reporting differences rather than system performance.
Key Points
- 1IIHS calculated a 68% lower police-reportable crash rate for Waymo's driverless Level 4 fleet across four operating cities.
- 2Manual incident coding reconciled incompatible reporting thresholds, but IIHS concluded that approach cannot scale with wider autonomous-vehicle deployment.
- 3Comparable autonomy safety evaluation generally requires public mileage denominators, matched operating domains, and consistent crash definitions across fleets.
Scoring Rationale
The IIHS analysis provides an independent, mileage-normalized safety comparison for a major U.S. driverless robotaxi deployment with public exposure data. Its methodological caveats are highly relevant to teams building, evaluating, and governing autonomous systems, although the results remain limited to Waymo's Level 4 operating domains.
Sources
Primary source and supporting public references used for this report.
Practice with real Ride-Hailing data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Ride-Hailing problems
