For predicting accident frequencies, a succession of log-linear models for Poisson data, some of which include nested random effects, is introduced. By applying maximum likelihood and empirical Bayes estimation techniques to these models, one can incorporate the actuarial notions of risk classification, model-based smoothing, credibility theory, and experience rating under a unified statistical approach to loss prediction. The performance of these methods is evaluated by using accident data from California.

Actuarial statistics, Automobile insurance, Credibility theory, Empirical Bayes, Log-linear model, Random-effects model
Journal of the American Statistical Association
Sprott School of Business

Tomberlin, T.J. (1988). Predicting accident frequencies for drivers classified by two factors. Journal of the American Statistical Association, 83(402), 309–321. doi:10.1080/01621459.1988.10478600