Determination of Key parameters in Generalized Linear Models for claim reserving - General Insurance

Abstract

Based on Generalized Linear Models, we are able to establish the key parameters to be used for determining the future claim frequency through the adoption of appropriate logical variable selection method given various variable interactions in the model. With ample calibration, we prove the robustness of the model by analyzing its predictability power through evaluation of the dispersion parameters. This facilitates the efficiency in claim reserving given the scarcity of resources and time to run the saturated models. We adopt the statistical based variable selection method for the model as opposed to experience selection. The general insurance industry could embrace the study to economically maintain and improve the certainty in their claim reserving. Relying on the sample analysis, general insurers and particularly motor vehicle insurers could estimate their future claim frequency at about 85% confidence interval basing on the model of the car and the age bracket of the policyholder.