CAPTURING EXCESS ZEROS IN MODELINGAUTO-INSURANCE CLAIMS IN AN INDIGENOUSINSURANCE FIRM USING ZERO INFLATED MODELS ANDHURDLE MODELS

Abstract

Count data occur naturally in a number of disciplines ranging from economics and social sciences to finance as well as medical sciences. Count data could be plagued with over-dispersion and excess zeros making it difficult to employ the use of classic linear models. Different models have been proposed to capture this peculiarity in count data, A number of classical regression models such as the generalized Poisson and negative binomial have been used to model dispersed count data. Hurdle and zero-inflated models are also said to be able to capture over-dispersion and excess zeros in count data. In this paper, we compare the performance of Poisson and Negative Binomial hurdle models, zero-inflated Poisson and Negative Binomial models, classical Poisson and Negative Binomial regression models as well as the zero-inflated compound Poisson generalized linear models to modeling frequency of auto insurance claims in a typical emerging market. The model parameters are estimated using the method of maximum likelihood. The models performances are compared based on some model selection criteria, including: Akaike and Bayesian information Criteria (AIC and BIC), and the model lift which was obtained by the Gini index score. The zero-inflated compound Poisson generalized linear models performed better than the other models considered.