Applying the Palm distribution for bicycle crash risk assessment and informed policy making

Abstract

This study analyses factors associated to bicycle accidents by a notion of Palm theory for traffic conditions. The method allows for the comparison of the distribution of conditions as seen by an arbitrary cyclist (the Palm distribution) with those seen by a cyclist subject to an accident (the accident distribution). This allows for a straight forward assessment of the relative risk change given different conditions as well as their statistical significance. The study is based on accident data from police reports of bicycle accidents over 4 years, in Copenhagen and Frederiksberg (N = 1136). Relative risk change compared to the overall risk (Palm frequency) was evaluated given time, weather and seasonality. Relative risk increase was found to be significant at night (0-4) as well as in morning and afternoon peak hours. Upon further evaluation, the night effect is only significant in the weekend, while morning and afternoon peak are only associated to significant increase during the week. This demonstrates how using over-aggregated explanatory variables might cause misleading conclusions with regard to interventions. Furthermore results indicate that precipitation is associated to a general increase in the riskiness. Overall the Palm distribution for traffic conditions offers a novel way to capture factors associated to bicycle accidents.