Crash causal factors: Crash frequency, crash severity and crash collision models

This study examined the variables related to roadway geometry, environmental, driver and traffic factors to identify crash causal factors. It relied on three years of crash data from the Arkansas Highway Transportation Department (AHTD) and analyzed nonjunctions of rural and urban US highway systems. In the first part of this study, negative binomial modeling technique was used to model the frequency of crash occurrence. To further analyze the crash factors this study also analyzed crash severity and collision types. The second part identified the factors responsible for severe crashes and fatalities including using the binary logistic regression modeling technique. The third part used the multinomial logistic regression modeling technique to identify the factors associated with specific collision types (single vehicle, head-on, rear-end, sideswipe-same, and sideswipe-opposite direction). The crash data were analyzed statistically, and the factors significant for crash frequency proved to be surface width, roughness, left and right shoulder widths, road segment length, and Annual Average Daily Traffic. Driver related factors such as age, gender, restraint type, and alcohol consumption were significant in severe crashes. Variables such as horizontal and vertical road curvature, wet road surface, and darkness differentiated single-vehicle collisions from multi-vehicle collisions. This study clearly indicated the importance of using different analysis techniques to identify the main factors responsible for crashes --Abstract, page iii