Explaining variation in safety performance of roundabouts.

The conversion of an intersection into a roundabout has been proven to reduce generally the number of crashes with injuries or fatalities. However, evaluation studies frequently showed considerable individual differences in safety performance of roundabouts or particular groups of roundabouts. The main purpose in the present study was to explain the variance in safety performance of roundabouts through the use of state-of-the-art cross-sectional risk models based on crash data, traffic data and geometric data of a sample of 90 roundabouts in Flanders-Belgium. Poisson and gamma modelling techniques were used, the latter one since underdispersion in the crash data was observed. The results show that the variation in crash rates is relatively small and mainly driven by the traffic exposure. Vulnerable road users are more frequently than expected involved in crashes at roundabouts and roundabouts with cycle lanes are clearly performing worse than roundabouts with cycle paths. Confirmation is found for the existence of a safety in numbers-effect for bicyclists, moped riders and - with less certainty - for pedestrians at roundabouts.