Spatial characteristics of bicycle vehicle accidents in Christchurch, New Zealand: A case-control approach

Abstract

This paper aims to predict the probability of bicycle motor vehicle (BMV) accidents occurring on a network wide level in Christchurch, New Zealand based upon the spatial characteristics present in the road environment. To achieve this, logistic regression was undertaken with a binary dependant variable (accident/non-accident) using a case-control strategy, with case sites being locations of reported BMV accidents whilst control sites were sampled from the road network in proportion to where people cycle. Due to the uncertainty in cycling flows in Christchurch, five logistic regression models were undertaken based upon different route selection preferences. The results identified that BMV accident probability increased with the same magnitude across all five models due to the presence of drive ways and intersections, identifying that these characteristics are associated with an increase in BMV accident probability. In contrast to this, there is significant variation across all five models relating to how BMV accident probability is influenced by the presence of painted on road cycle lanes, road classification and urban panning zones.