The accident history of two urban routes, each of approximately 40 km in length, has been examined. The routes included roads of all categories, from motorway to back streets and some 2,700 injury-accidents occurred on the routes during the 62 month period analysed. Data were available, which described the physical characteristics of the routes in great detail. Some traffic data were available through the local authorities concerned and those were supplemented by further traffic and pedestrian movement surveys. These data were used to determine statistical relations between various measures of accident rates and road or traffic variables. The factors found to be significantly related to accidents were used as criteria by which to disaggregate the accident data into categories for which multiple regression equations could be formulated as a basis for predicting accident rates. This was done for junction and non-junction accidents and for both routes separately and combined. The accident data sets for each route were subdivided up to five times on the basis of these criteria in order to achieve reasonable predictions of accident rates. By further regression on both sets of data combined it is hoped that a set of equations has been found which can be used for more general prediction of urban accident rates. As the equations include more than just traffic flow variables and relate to specific types of accident, they may prove to give more accurate predictions than the very general accident rates commonly used.
Abstract