The magnitude of the regression to the mean effect in traffic crashes

Abstract

Regression to the mean has been recognized as a phenomenon that influences road safety evaluations and should be accounted for. However, some doubts have risen about the necessity to implement rather sophisticated techniques such as the empirical Bayes method to correct for regression to the mean whereas the use of a sufficient long before-period could reach the same objective. Present study examines the existence and the magnitude of the regression to the mean effect in crash data from 169 intersections for the injury crashes and 143 for the more severe crashes in Flandrs-Belgium for whom regression to the mean was likely to occur as they were selected based on their crash history. The presence of a RTM-effect was investigated by comparing the crash numbers of this period with the crash numbers in the next three years, during which no traffic safety measure was applied. Two comparison groups were used. The results demonstrate the existence of a substantial regression to the mean effect in the investigated sample of intersections. The magnitude of the regression to the mean effect is estimated to be almost 9% for injury crashes and 37% for the most severe crashes. From this can be concluded the correction for regression to the mean in evaluation studies is highly recommended in cases when locations are selected based on their crash history. This can be applied through the Empirical Bayes method.