Modelling The Risk Probability with Compound Generalized Poisson Distribution

Abstract

Insurace is a form of risk management that is used for protection against loss. In insurance companies, the pricing of insurance premium should be based on fair principles, i.e. in accordance with the value of class of risk. The higher the risk, the higher amount of insurance premium will be. The class of risk can be estimated by insurance claim. Modelling the claim frequency is one of the most important areas in risk theory. Generally, insurance claims can be modelled with Poisson or negative binomial distribution. Gossiaux and Lemaire (1981) and Willmot (1987) have considered generalized Poisson distribution as an alternative to Poisson or negative binomial disribution. Generalized Poisson distribution will be applied if there is overdispersion in data. Generalized Poisson distribution can be viewed as a compound Poisson (𝜃) and sum of Borel (𝜆) distribution. If a random variable N (the number of claims) has a generalized Poisson distribution, then the total number of claims has a compound generalized Poisson distribution. A recursive algorithm is needed to model the risk probability of this distribution. First step of this algorithm is to compute the coefficients of the moment generating function, which sum to the probability function of the compound Borel (𝜆) distribution. The next step of this algorithm is to compute the probabilities by using Panjer’s recursion formula for the Poisson (𝜃) distribution.