Generalized linear model for deductible pricing in non-life insurance

Abstract

One common feature of a non-life insurance policy is deductible. When deductible is introduced, it changes both the distribution of frequency and severity due to the censoring and truncation of the random variables. Therefore, different premiums are required for different deductibles. This leads to the need of determining the relativities over possible deductible levels. There are a few methods for calculating relativities. In practice, some of them present difficulty in statistical estimation for the unobserved random variables. An alternative method, the regression approach, is more practical in application. This paper focuses on the regression approach in relativity calculation. For some parametric severity distributions, improvements can be made by incorporating correct covariates to the model.