MODELING THE FREQUENCY OF CLAIMS IN AUTO INSURANCE WITH APPLICATION TO A FRENCH CASE

The aim of this paper is to present the different models for count data used in the actuarial literature. In addition to the Poisson regression, Negative Binomial and Zero-Inflated models are applied to an auto insurance portfolio of a French insurance company. Statistical tests to evaluate the performance of the models are explained taking into consideration the difference between the nested and the non-nested models. The comparison between the nested models is performed using specification tests and the Vuong test is used to compare the fitting of non-nested models.