Modelling of Claim Frequency for a Motor Hull Insurance

Abstract

In recent decades, the insurance companies usually set an annual premium to match the number of engines and the size of the policy holder’s area of residence, although some insurance companies also consider the age of the customer. As a result, annual premiums are increasingly determined based on the risks assumed to generate insurance claims or the risk behavior (risk characteristics) of policyholders. Due to the current insurance trends, the Generalized Linear Models (GLM) have become a popular statistical tool that can analyze and model claims frequency and severity. The aim of this thesis is to propose the count model appropriate to the modelling of claim frequency for a motor-hull insurance, that is, the Poisson regression model and Negative binomial regression model, then compare the Poisson regression model and Negative binomial regression model to find out which one is better. This thesis is divided into five chapters. The first chapter is the introduction which is the outline of whole thesis, it will set the framework for the whole thesis.The second chapter of this thesis is the theoretical introduction of risk and insurance, involving the related concepts of risk, the related concepts of insurance as a risk management method and the introduction of auto-hull insurance. The third chapter of this thesis is the theoretical introduction of Poisson regression model and Negative binomial regression model used in the modeling research of this thesis. The fourth chapter, we will apply our data to the model in Stata according to the theoretical part in Chapter three. And the last chapter gives the final conclusion of this thesis.