On factors related to car accidents on German Autobahn connectors

We make an attempt to identify factors that explain accidents on German Autobahn connectors. To find these factors we perform an empirical study making use of count data models with fixed and random coefficients. The findings are based on a set of 197 ramps, which we classify into three distinct types of ramps. For these ramps, accident data is available for a period of 3 years (January 2003 until December 2005). The negative binomial model with some random coefficients proved to be an appropriate model in our cross-sectional setting for detecting factors that are related to accidents. The most significant variable is a measure of the average daily traffic. For geometric variables, not only continuous effects were found to be significant, but also threshold effects indicating the exceedance of certain values.

► Overdispersion and unobserved heterogeneity make the negative binomial regression with random coefficient the appropriate model for the number of accidents on German highway connectors. ► Average daily traffic is the most important accident factor. ► Various geometric characteristics of the ramps and the ratio of trucks to total traffic help explain accidents. ► The identified factors do not allow for simple short term improvements of ramps, but should be useful when constructing new ones.