Data from insurance claims demonstrate the unmatched safety of autonomous vehicles over human drivers.
A new study analyzing 25.3 million miles of fully autonomous driving data from Waymo has found that their driverless cars are significantly safer than human-driven vehicles. The study, conducted in partnership with Swiss Re, one of the world’s largest reinsurers, examined Waymo’s auto liability insurance claims and compared them to human drivers. The results were striking, demonstrating an 88 percent reduction in property damage claims and a 92 percent reduction in bodily injury claims compared to the overall driving population.
The study builds upon a growing body of research investigating the safety implications of driverless vehicles. The methodology employed in this study sets it apart from others by leveraging a unique data source: insurance claims. This approach provides a more comprehensive and robust way to assess safety performance, addressing key limitations found in previous studies.
Examining liability claims
Insurance claims data offer a valuable window into the real-world performance of autonomous vehicles. Unlike crash reports, which can vary in completeness and reporting thresholds, insurance claims provide a consistent and standardized measure of responsibility. When a claim is filed, it indicates that an incident has occurred where one party is alleged to be at fault, providing crucial information about the circumstances surrounding a collision.
In this study, researchers focused explicitly on third-party liability claims, which are claims filed against Waymo for damages caused to another party. This focus allowed them to assess Waymo’s performance in situations where their vehicles allegedly had some responsibility for the incident. By comparing the frequency of these claims to those filed against human drivers, the researchers could directly understand how Waymo’s autonomous driving system performed relative to human drivers in similar situations.
Comparing Waymo to human drivers
To establish a fair comparison, researchers developed two benchmarks representing human driving performance:
- Overall Driving Population: This benchmark included claims data from all vehicles registered in Waymo’s operational areas (San Francisco, Phoenix, Los Angeles, and Austin) between 2017 and 2022.
- Latest-Generation HDVs: Recognizing the advancements in vehicle safety technology, researchers created a second benchmark focusing specifically on claims data from vehicles with model years between 2018 and 2022. This benchmark aimed to capture the performance of cars with newer safety features, such as automatic emergency braking (AEB) and lane-keeping assist.
The study found that Waymo’s autonomous driving system significantly outperformed human driver benchmarks. Compared to the overall driving population, Waymo experienced a nine-fold reduction in property damage claims and a twelve-fold decrease in bodily injury claims.
Waymo’s performance remained statistically significant even when compared to the latest-generation HDV benchmark, which exhibited lower claim rates than the overall driving population.
Waymo’s performance remains consistent as operations scale
This study demonstrates the superior safety performance of driverless cars and highlights the consistency of this performance as operations expand. The researchers compared Waymo’s claims data from their previous study, which analyzed 3.8 million miles of driving, to the current 25.3 million-mile dataset. While, apparently, there is no statistical difference between the two periods due to the smaller sample size of the earlier study, the researchers observed that Waymo’s claims rates remained very low, with only nine property damage claims and two bodily injury claims reported over 25.3 million miles. This finding underscores the robustness and scalability of Waymo’s safety performance as they continue to expand their driverless services.
Future applications of insurance data
The use of insurance claims data opens up promising avenues for future research in assessing the safety impact of autonomous vehicles. The study’s authors highlight several potential areas for further investigation, including:
- Liability in ADS-involved collisions: Exploring the frequency and types of collisions that lead to liability claims for autonomous and human-driven vehicles.
- Claims adjudication with safety models: Using advanced safety envelope and human behavioral reference models to establish normative references for driving behavior and potentially informing claims adjudication processes.
- Defining additional human driving subsets: Leveraging claims data to create a distribution of human driving performance, enabling more nuanced comparisons between ADS performance and specific cohorts of drivers, such as those with excellent safety records.
The findings of this groundbreaking study, with its innovative use of insurance claims data, provide compelling evidence for Waymo’s fully autonomous driving system’s superior safety performance.
By demonstrating substantial reductions in property damage and bodily injury claims compared to human drivers, this research offers a promising glimpse into driverless technology’s potential to significantly improve road safety.
As companies such as Waymo continue to innovate and expand their autonomous operations, insurance data will be crucial for monitoring and evaluating the safety impacts of these transformative technologies.
From the Study: Do Autonomous Vehicles Outperform Latest-Generation Human-Driven Vehicles? A Comparison to Waymo’s Auto Liability Insurance Claims at 25.3M Miles. Di Lillo, L., Gode, T., Zhou, X., Scanlon, J. M., Chen, R., & Victor, T. (2024).
