Claim size estimation with deductibles as covariates

Specialeforsvar ved Maria Mikkelsen

Titel: Claim size estimation with deductibles as covariates

Abstract: We study the performance of generalized linear and generalized additive models where deductibles are used as a covariate in the model to facilitate computations. We apply these models to shipping data. The aim of the investigation is to compare The claim size models using deductible as a covariate with the "correct model" which account for the deductibles in a mathematically correct way. Furthermore we investigate the differences the models imply from a premium perspective. A simulation study is carried out to further investigate the difference between the "correct model" and the generalized linear models using deductibles as a covariate.
It is found that the generalized linear model might lead to premiums very different from the "correct" premiums. It is also found that the use of the generalized linear model might lead to the very undesirable property of increasing premium with increasing deductible. The generalized additive models are found to improve the generalized linear models but only significantly in some cases when adapting it to a non-life insurance set-up. The pros and cons of the models using deductibles as a covariate are discussed.

Vejleder: Jostein Paulsen
Censor:   Mette Havning