This study evaluates the influence of traffic volume on accident frequency and investigates other possible factors that contribute to accident risk, using vehicle kilometers (VKM) as a proxy of exposure. Information on traffic volume and road design factors was obtained from the Hong Kong Annual Traffic Census (ATC), and accident data were extracted from the Hong Kong Accident Database System (TRADS) respectively. These data were incorporated into a road network map of Hong Kong using the Geographical Information System (GIS). Count data models were employed to the analysis of accident frequency. As the data are subject to over-dispersion, a negative binomial regression method was deployed to measure the association between accident frequency and traffic volume and to control for the effects of factorssuch as temporal variation and road environment. The results indicate that greater traffic volume leads to a less than proportionate increase (a coefficient estimate of 0.62) in accident frequency and thus a decrease of accident risk. The interaction effects by traffic volume and other possiblefactors on accident occurrence are also revealed.
Abstract