The Debate Over Robotaxi Safety

Public opinion regarding Waymo's autonomous taxis remains starkly divided, ranging from viewing them as a major public health advancement to seeing them as significant safety hazards. The Insurance Institute for Highway Safety (IIHS) recently weighed in, suggesting that driverless vehicles are involved in fewer accidents than human-operated cars. However, the institute acknowledges that reaching this conclusion required significant statistical adjustments, as direct comparisons between human and machine drivers remain difficult due to fragmented data.


The Challenge of Data Disparity

While crash data for human drivers is widely available and standardized, equivalent metrics for autonomous vehicles (AVs) are harder to pin down. Although authorities like California and the NHTSA mandate incident reporting, the specific criteria for these reports differ from those applied to human drivers. A critical factor in calculating safety rates is the total vehicle miles traveled (VMT). While government bodies track human driving mileage extensively, similar reporting is not compulsory for AV firms.

Waymo is the only company that proactively shares its mileage data, which is why the IIHS study focused exclusively on their performance rather than that of competitors like Zoox or Tesla.


Statistical Filtering and Methodology

A primary obstacle for researchers was the quality of existing AV data from 2021 to 2024. The IIHS had to undertake a rigorous process to cleanse the dataset, removing duplicates and incidents where automation was not active or no actual crash occurred. They applied strict criteria—such as identifying incidents equivalent to those reportable to police—to ensure a fair comparison.

The results of this filtering process were notable:

  • Less than 25% of the initial autonomous crash data was deemed suitable for analysis.
  • Researchers identified only 22% of 736 reported incidents as likely being police-reportable.
  • Of those, 89 incidents involved Waymo vehicles.

The IIHS Conclusion

Following their analysis, the IIHS reported that Waymo's crash involvement rate was 68% lower than that of human drivers. Specifically, the institute noted significant reductions in injury-related crashes and single-vehicle accidents. However, these figures varied significantly by region, showing higher performance in Phoenix compared to more modest improvements in other cities.


Are the Findings Statistically Significant?

Despite the findings, questions regarding the reliability of the comparison remain. The disparity in scale is massive: the study contrasted millions of human-related crashes against a handful of AV incidents. Furthermore, the IIHS notes that the vast difference in miles traveled—roughly 50 million for Waymo compared to 222 billion for humans in the same regions—creates a difficult gap to bridge statistically.

As the IIHS itself concluded in its study, current reporting methods may not be sustainable as autonomous fleets expand:

«The method of coding narratives is unsustainable as these deployments and the number of incidents continue to grow. Practical application: This study identifies ways in which national crash and VMT data collection for L4 vehicles can be improved for more timely and accurate safety evaluations.»