Research Question/Objective:​ Subaru’s EyeSight is a camera-based driver assistance system that includes the ability to brake automatically when it detects a crash-imminent situation involving a pedestrian. The objective of this study was to evaluate whether Eyesight was preventing vehicles from striking pedestrians. Insurance claims under bodily injury (BI) liability coverage without associated first-party (collision) or third-party (property damage liability [PDL]) claims for vehicle damage were used as a surrogate measure of pedestrian crashes.

Methods and Data Sources:​ EyeSight is an option on various 2013–17 Subaru series. The presence or absence of EyeSight is discernible from information encoded in the vehicle identification numbers (VINs). The number of BI-only claims per insured vehicle year was compared for Subaru vehicles with and without EyeSight, using regression analysis to account for other factors also known to affect insurance claim frequency. A Poisson distribution was used to model claim frequency. Covariates included calendar year, garaging state, vehicle density, age group, gender, marital status, risk classification, and vehicle model year and series. Twenty percent of the insured-vehicle-year exposure came from vehicles equipped with EyeSight. Prior investigation has shown that injury-only BI claims are consistent with pedestrian or other nonoccupant injuries. Even so, these may include some nonpedestrian crashes, and some pedestrian crashes that were severe enough to also cause vehicle damage may be excluded. This study was based on vehicles with overlapping BI liability, collision, and PDL insurance in U.S. states with traditional tort liability insurance laws.

Results:​ When all series were combined, Subaru vehicles with EyeSight showed a statistically significant 35% reduction in BI-only claim frequency. When the Subaru Forester, Legacy, and Outback were separated by generation, results also showed statistically significant reductions of 33% for the first generation and 41% for the second generation. When the vehicle series were modeled individually, claim frequency reductions ranged from 18 to 57%, although only the Legacy (57%) and Outback (34%) results were statistically significant.

Discussion and Limitations:​ Claim frequency reductions for the EyeSight generations are similar, but it is promising that the second generation is showing a larger reduction. There are limitations to the data used in this analysis. Covariates describing driver characteristics are generally those of the primary driver of each vehicle and not necessarily the driver involved in the crash claim. Likewise, geographic covariates describe where the owner of the insured vehicle lives and not necessarily where crashes occurred. Nevertheless, these variables are consistently predictive in explainable ways.

Conclusions and Relevance to Session Submitted:​ Subaru’s Eyesight system is associated with a lower BI-only claim frequency than the same Subaru vehicles without EyeSight. In 2016, there were nearly 6,000 pedestrian fatalities in the United States, up 9% from 2015, and an 11% increase in bicyclist deaths. Pedestrian detection systems like Subaru EyeSight have the potential to effectively reduce these numbers, and efforts to promote similar systems will help protect vulnerable road users.