The identification of unsafe road locations (blackspots) and theevaluation of treatment effectiveness should be based on the numberof accidents expected in the long run (accident potential) rather than on the short-term count. A method for estimating the underlying accident potential of ontario road sections, using accident and other data, is presented. The method first uses regression models to produce an initial estimate of a section's accident potential on the basis of its traffic and geometric characteristics. This estimate is then refined by being combined with the section's accident count, using an empirical bayesian procedure. The results indicate that the empirical bayesian estimates are superior to those based on the accident count or the regression prediction by themselves, particularly for sections that might be of interest in a program to identify and treat unsafe road locations. This paper appears in transportation research record no. 1327, Visibility, rail-highway grade crossings, and highway improvement evaluation 1991
Abstract