One of the primary missions of the California Department of Motor Vehicles is to protect the public from drivers who represent unacceptably high accident risks. Optimum fulfillment of this objective requires the development and implementation of strategies for identifying high-risk drivers. One such system in California is the department's negligent operator point system. This system assigns points to moving violations and accidents and authorizes the department to take driver control actions against drivers who meet the prima facie definition of a negligent operator. The present study explored the viability of predicting accidents from equations constructed to predict convictions for the general driving population. Equations or models that better identify drivers at increased risk of future accident involvement would increase the number of accidents prevented through post license control actions. Although the results did not support prior findings that equations keyed to citations do as well as or better than equations keyed to accidents in predicting subsequent accident involvement, a canonical correlation approach considering subsequent accident and citation rates simultaneously produced a 14.9% improvement in the classification accuracy or "hit rate" for identifying accident-involved drivers.
Since 1964 the California Department of Motor Vehicles has issued several monographs on driver characteristics and accident risk factors as part of a series of analyses known as the California driver record study. A number of regression analyses were conducted of driving record variables measured over a 6-year time period (1986 to 1991). The techniques presented consist of ordinary least squares, weighted least squares, Poisson, negative binomial, linear probability, and logistic regression models. The objective of the analyses was to compare the results obtained from several different regression techniques under consideration for use in the in-progress California driver record study. The results are informative in determining whether the various regression methods produce similar results for different sample sizes and in exploring whether reliance on ordinary least squares techniques in past California driver record study analyses has produced biased significance levels and parameter estimates. The results indicate that, for these data, the use of the different regression techniques do not lead to any greater increase in individual accident prediction beyond that obtained through application of ordinary least squares regression. The methods produce almost identical results in terms of the relative importance and statistical significance of the independent variables. It therefore appears safe to employ ordinary least squares multiple regression techniques on driver accident count distributions of the type represented by California driver records, at least when the sample sizes are large.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.