Explanatory models for crashes at high-risk locations

Abstract

The purpose of this paper is to identify the most important underlying factors that determine the number of injury crashes in a sample of dangerous intersections. To this end, negative binomial models are fit for a dataset of 601 intersections. Data were collected about 36 potentially important variables that describe the traffic volumes, geometric characteristics, road environment and transportation planning context. The models show that traffic volume is generally not a very strongly determining variable for the crash count at these intersections, as opposed to most literature about crash prediction models. Furthermore, intersections with major road category “secondary” systematically show a higher crash count than intersections at primary/main and local roads. Also the number of lanes on the minor road has an important influence on the number of crashes. Intersections with a one-lane minor road have the lowest crash count, while intersections with two-lane or three-lane minor roads show the highest crash counts. The presence of a median on the major road corresponds in most models with a lower number of injury crashes. Finally, a number of models indicate a higher crash count at intersections with crossing facilities for vulnerable road users, especially at unsignalized intersections. However, the latter variables may act as a proxy for the unknown exposure of these road user types.