IIHS study: Waymo robotaxis involved in 68% fewer crashes than human drivers across four cities, 2021–2024

Headline Finding

A July 2026 Insurance Institute for Highway Safety report found that Waymo driverless vehicles were involved in 68% fewer crashes than human drivers operating in the same areas and years — the most comprehensive comparative safety dataset yet published for a commercial robotaxi fleet. Across 2021–2024 in four cities, Waymo logged 1.28 reportable crashes per million vehicle miles traveled (VMT) in driverless mode, against 4.06 crashes per VMT for human drivers. Specific crash types showed even starker gaps: Waymo’s rate of rear-ending another vehicle was 91% lower, and its rate of being rear-ended was 40% lower. The findings directly bear on the legitimacy lever — they provide the evidentiary foundation that regulators, cities, and the public need to calibrate permitting and public-trust instruments — while simultaneously exposing a critical governance gap: no national VMT reporting requirement exists, making comparable analysis impossible for any other AV operator.

Detail

The IIHS report, published in July 2026, drew on Waymo’s operational data from 2021 to 2024 across four cities where the company runs driverless commercial service. In driverless mode, Waymo vehicles recorded 1.28 reportable crashes per million VMT, compared with 4.06 crashes per million VMT for human drivers in the same areas and years — a 68% reduction. Disaggregated by crash type, the rear-end-striking rate was 91% lower than human drivers, and the rate of being struck from behind was 40% lower.

The IIHS report concluded that Waymo “is maintaining a level of safety as its technology develops” within its current operational design domain. The report explicitly cautioned, however, that findings may not generalise to future geographies where Waymo might expand, nor to other robotaxi operators. The caveat is structurally significant: other AV companies “do not release VMT figures or release only very limited ones,” and there is currently no national requirement to report VMT, making cross-operator or cross-city safety comparisons impossible.

Governors Highway Safety Association CEO Jonathan Adkins framed the findings against the baseline of human driving, noting that human drivers kill more than 36,000 people on U.S. roads every year, and argued that “progress, not perfection, must be the standard for AVs.” He also called for federal, state, and local safety officials to have “the tools and information they need to monitor performance, identify emerging safety issues and respond effectively.” The IIHS report acknowledged that Waymo vehicles have been observed driving into flooded streets, passing stopped school buses, and impeding emergency vehicles — incidents that may disproportionately shape public perception relative to the aggregate safety record.

Governance Implication

1. The legitimacy lever is moved in Waymo’s favour: a peer-reviewed, multi-year, multi-city dataset showing a 68% crash reduction gives regulators and city governments an empirical anchor for performance-based permitting decisions, shifting the burden of proof toward those who would restrict deployment rather than those who seek to expand it.
2. The VMT reporting gap is the sharpest governance finding: because no national VMT disclosure requirement exists and other operators release little or no data, the IIHS analysis is structurally limited to Waymo — meaning cities cannot make evidence-based comparative assessments of competing AV operators, undermining the data-sharing-mandate instrument that Chapter 6 identifies as foundational to accountable deployment.
3. The two-clocks problem is visible in the data itself: Waymo accumulated enough VMT across four cities over four years to produce statistically robust safety findings, yet the governance infrastructure — mandatory VMT reporting, standardised crash taxonomy, multi-operator benchmarking — does not yet exist at the federal or state level, meaning the technology clock has outrun the regulatory clock on the very metric (safety) that legitimises deployment.
4. The IIHS caveat on operational design domain generalisation is a direct signal for spatial-allocation governance: cities considering expanding Waymo’s service area, or permitting new entrants, cannot assume that safety performance demonstrated in current geographies will transfer — making ODD-scoped, performance-based permitting (rather than blanket city-wide licences) the appropriate instrument.
5. The gap between aggregate safety data and high-salience individual incidents (flooded streets, school buses, emergency vehicles) illustrates the public-trust dynamic at the heart of Chapter 4: legitimacy is not automatically conferred by statistics, and cities need communication and accountability architectures that contextualise both the aggregate record and the edge-case failures.

Sources

https://www.smartcitiesdive.com/news/waymo-robotaxis-safer-than-human-drivers-iihs-report/826320/ https://www.iihs.org/topics/advanced-driver-assistance/automated-vehicles https://www.ghsa.org/resources/news-releases/2026/IIHS-Waymo