Pricing Insurance Product With Generalized Linear Models And Price Re-Balancing Using Credibility Theory

Abstract

The insurance industry is one of an essential section of economic which contributes significantly to the improvement and development of modern society, and the development of the insurance industry in Vietnam is active and active with the participation of many Insurance companies from abroad and in the country. In response to this trend, the thesis wishes to provide a process for simulating the insurance pricing system with Generalized Linear Model (GLMs) and price re-balancing system by using Credibility Theory which belongs into the primary purpose to provide the reader a technical vision about the insurance industry. After few chapters providing some the introduction of the insurance industry and the literature review, the conceptual background of both system is accessed directly by illustrating from basic concepts to the advanced estimating method to produce the technical price. GLMs builds the relationship between the outcome variable and the predicted parameters on a set of training data explicitly, and then according to this result model, we can estimate and forecast the target variable on predictive data sets. The Credibility theory gives the idea of rebalancing the structure of the estimated price of specific risk classes, and from that we build up the new premium structure. The thesis report focuses mainly on two approaches to Credibility theory namely Classical Credibility and B¨uhlmann-Straub Credibility. About the data used for illustrating, this paper uses the possible data sources of the vehicle insurance product with 30750 data points of 11 parameter fields to demonstrate the application process. The specific example and realistic application of this theory are included in this document. The calculation was executed entirely by an automatic prediction, which was self-programmed by the author in R programming language. Keywords: Insurance Pricing System, Price Re-balancing System, Generalized Linear Model (GLMs), Credibility Theory, Classical Credibility, B¨uhlmann-Straub credibility model, R programming language, practical data sources.