The Distribution of Automobile Accidents--Are Relativities Stable Over Time?

Data on the distribution of automobile accidents typically reject the hypothesis that accident rates are the same for all members of a group. Given these findings, policy analysis is usually based on models that assume that accident proneness differs among individuals of a group and that the differences are stable over time. The analysis presented in this paper is aimed at assessing the validity of these assumptions. A simple model that allows for variability in both proneness and exposure level is used to estimate the potential contribution of variability in exposure levels to total variability in accident rates. Data indicate that variability of exposures may have a substantial bearing on the variability of accident rates. Data from several groups of drivers from California and North Carolina are used for direct tests of the stability of relative accident rates. When the California data are used, the tests do not lead to a rejection of the hypothesis that relative accident rates are stable. When the North Carolina data, based on a larger number of observations, are used, the tests clearly reject the hypothesis. The implications of these findings for economic and policy analysis are discussed. Keywords: Classifications, LOB-Auto Liability