Use and extension of count data models in the determination of relevant factors for claims in the automobile insurance sector

Abstract

Using real, Spanish data, different specifications of zero-inflated models are provided in this paper to estimate the number of accidents declared by policyholders. These count data models seem to be the most appropriate solutions to study this question. Our work is completed with the estimations of the number of clients that do not declare their actual accidents and the number of these accidents. The analysis of all these factors could become useful for insurers to improve their efficiency. We conclude with a final theoretical discussion on the possible advantages given by other alternative models, like the so-called thinned models.