Automobile Ratemaking Based On Hierarchical Generalized Linear Models

Since insurance industry recovered, automobile insurance has taken a large market share in non-life insurance sector. In these ten years, it maintained a market share of more than 60% and reached 73.1% in 2014. But the operating condition of automobile insurance business is generally not good. In recent years, the average loss ratio was seldom less than 50%. The high loss ratio reflects that insurance companies have a cognitive problems of risk in one side. On the other hand, it means that insurance premium rate is not reasonable. The former mainly stems from the risk classification is not detailed and the classification principle of different people, region and environment is not clear. The latter is mainly derived from the backward actuarial technology, unreasonable models of rating, related risk parameters are not accurate and so on.In recent years, some insurance companied tried to use Generalized linear model and its extended models for determining rate. Its loose assumption condition has made great progress in the model structure and model diagnosis, and overcomes some defects of traditional rating method. Chapter 2 introduces generalized linear models, generalized linear mixed models and hierarchical generalized linear models including the basic structure of models and methods for estimating parameters.Based on chapter 2, in chapter 3, I use these three models to analysis a real insurance data. For other parts, chapter 1 introduces the background and significance of my research, and also describes research methods, ideas and innovation points of this paper. Chapter 4 summarizes the full paper, analyzes its characteristics and shortcomings. In the end, it puts forward suggestions for future research.