A statistical model of individual accident risk prediction using driver record, territory and other biographical factors

Abstract

The objectives of the study were: (1) to determine the relative importance of territory, prior driving record and other variables in predicting future accident involvement: and (2) to determine whether a driver's area of residence is a fair and actuarially sound rating factor. Starting with two separate random samples totaling more than 90,000 drivers, various prediction models were developed using multiple regression techniques to predict subsequent three year accident involvement frequency, Although both territory and prior driving record proved to have some validity in predicting a driver's accident risk, the accuracy of the prediction was low, with multiple correlations ranging from 0.08 to 0.25. Prior driving record, particularly a driver's previous number of traffic convictions, was a much better predictor than was territory. Although absence of accident cost (insurance loss) information precludes precise validation of insurance rate setting practices, both territory and prior driving record appear justifiable as rate factors. However, the relatively small unique predictive contribution of territory suggests that territory may be less important than previously believed.