Study And Application Of Quantile Regression In Non-life Insurance

The importance of insurance products'rate-making cannot be over emphasized, especially with the further development of reform and opening process. China has promoted market making-rate system in the whole country since 2003, which increases the competition among insurance companies. Facing with increasingly stronger price competition, accurate and reasonable determination of rates becomes even more crucial to the insurance companies'survival and development . For non-life insurance companies, the pricing process is actually building rate models based on historical loss data of the insured subjects. Unlike life insurance companies, non-life insurance company cannot depend on reliable rating basis such as life table, therefore in most cases, the pricing of non-life insurance products requires actuarial professionals to combine historical data of claims with their own experiences. There are arguments among actuaries about different rating results due to different experiences and methods. In recent years, both foreign and domestic researchers have been actively exploring effective solutions in this area, and have made some progress. This paper studied ideologies of previous researchers, and estimated the risk premiums of non-life insurance companies directly by introducing risk metrics VaR into quantile regression model in statistics and combining pure premiums with security attachment.This paper attempted to make innovation based on methods from previous researches; in addition to the theoretical method mentioned in addition to discussion and comparison about current theoretical rating methods, this paper tried to find more reasonable and practical approaches to cope with real world problems. This article tried to break down the complex variable total claim payment into two variables, claim payment per exposure and number of claims, and selected proper regression method according to their respective characteristics. In the final part, a real auto-insurance case of a non-life insurance company was provided to explain the quantile regression method specifically. For the purpose of demonstrating clear comparison and relation between the quantile regression method and traditional methods, two traditional models were used first in practice to determine the premium rate, then the advantages and disadvantages of the quantile regression method were analyzed and illustrated by contrasting the results.This thesis was divided into four main chapters, and the first chapter stated the background and significance of the study and summarized related foreign and domestic academic works in this field. The second chapter briefly reviewed and evaluated three primary categories of current non-life insurance risk premium rating methods, and developed the classification rate which is the third category as the subject of this paper. Then, this paper introduced VaR risk metrics quantile regression method, and expounded its ideas, estimates and applications in detail. In the third chapter, this paper introduced the four kinds of classification rate methods, compared the quantile regression with the traditional methods, and pointed out its advantages. In addition, at the end of this chapter, the paper gave a necessary explanation of its application in real world practice, and made innovative modifications to the realization of the method. The fourth chapter computed empirical results of a set of real data gained from a certain auto-insurance company by quantile regression method and two other traditional methods seperately, comparing and analyzing the strengths and weaknesses of each method, thereby illustrated the advantages and values of Quantile Regression method. In the end, this paper gave a summary of the empirical results of the quantile regression model, and made some suggestions on non-life insurance products premium rating in the light of the current situation of China's insurance industry.