The screening of dangerous road locations is essential in further road safety improvement interventions. The common definition of a dangerous road location by traffic administration authorities around the world uses raw crash count or deaths, but these two measures are always believed to contain randomness. In this paper, we try to detect dangerous intersections at city level using three approaches, and test the consistency of these approaches in terms of site, method, and relative ranking. The results reveal the underlying spatial and urban/rural distribution of intersection hot spots in Hong Kong. Moreover, the Empirical Bayes method seems to be the most reliable and consistent method in terms of site and method. Finally, some problems of the current approaches and possible future directions are also discussed.

CICTP