Effects of changes in road network characteristics on road casualties: An application of full Bayes models using panel data

•A panel data set for road network is used.•Exposure variable at aggregate level is estimated using origin–destination data.•Better connected and more complex road network experiences more crashes.

In order to ensure a high level of road safety, road network planning needs to be based on the best knowledge available of the effects of road design on road safety. In this study, we look into how changes in road network characteristics affect road casualties. An approach based on traffic assignment is proposed in order to estimate the traffic exposure at ward level. We apply a widely used approach for before–after evaluation studies, the Bayesian method. We also use a panel semi-parametric model to estimate the dose–response function for continuous treatment variables. The result suggests that there are more casualties in areas with better connectivity and accessibility, where more attention should be paid to safety countermeasures.