Addressing Potentially Missing Relevant Information on Attitudes and Other Behavioral Elements as Unobserved Heterogeneity in Highway Safety Studies

Abstract
Understanding highway safety requires an assessment of physical and behavioral factors that influence the occurrence and resulting injury severity of crashes. This assessment is exceedingly difficult because it involves an understanding of wide variety of factors including human responses to external stimuli and complex interactions between the vehicle, roadway features, roadway conditions, traffic-related factors, and environmental conditions. In addition, attitudes and other factors that potentially affect drivers’ safety-risk profiles play a key role in the occurrence and resulting injury severity of crashes. Recognizing that it would be virtually impossible to collect all of the data that could potentially influence the occurrence and severity of crashes, the safety field has moved forward by addressing omitted data as unobserved heterogeneity – using mixing distributions to arrive at appropriate statistical estimations. This study explores mixing distribution approaches to gain insight on the potential role that attitudes and other behavioral elements may play. This is done by analyzing pedestrian injuries over time (using data from Kansas) and assessing the frequency and severity of crashes using data from roadway segments in multiple states. The findings indicate that mixing distributions provide an approach that can capture important attitudinal and other behavioral elements when such data cannot be realistically collected, and that mixing distributions should be used in highway-safety practice as quickly as possible through their use in the American Association of State Highway and Transportation Official’s Highway Safety Manual.

Research Products and Implementation

Scope of Work

​​Final Report