The Ammeter Risk Model

Abstract

A Poisson distributed number of claims is not dispersed enough to fit real data. One therefore often uses a negative binomial distribution for models in a single period. This distribution can be constructed by mixing the Poisson parameter with a Gamma distribution. We therefore choose annually a new mixing parameter for the Cramér-Lundberg model. The asymptotic results obtained for the classical risk model can then be generalised to this more general risk model, both for small and large claims.