Research On Pure Premium Pricing Of Auto Insurance Based On Physical Collision Test

As the largest property insurance in China,the lack of accuracy rating determination has been a pain point for the industry.At present,China mainly adopts the car insurance rating determination pattern based on the car,and the purchase price of the new car determines which rate level the car is in.In this rating determination pattern,different payout differences leaded by the car’s own crashworthiness and repairability economy can not match the motor vehicle premium.In the US,Japan,UK and other developed foreign insurance markets,the collision safety level and its repairability have been the important rate setting factors for commercial auto insurance,and there are relevant institutions specializing in auto collision tests.In order to fill the gap in this area in China,CIIR and China Automotive Engineering Research Institute(CAERI)cooperated with the China Insurance Association(CIA)conduct low-speed crash tests,simulate the common front-end collision and rear-end collision to evaluate the crashworthiness and economic repairability of cars,and construct the China Insurance Automotive Safety Index(C-IASI).This paper firstly provides a detailed introduction of collision tests in different countries and models used,on the basis of which an empirical analysis is established.The paper then introduces the basic information of the data used and the meaning and grouping of each risk factor,and visualizes its impact on claim frequency,claim intensity and pure premium through single factor analysis.The empirical analysis is developed through two aspects of the rating determination factors and rating determination models.And the impact of adding the collision tests based C-IASI on rating determination is investigated by using the GLM model for Poisson regression of claim frequency,gamma regression of claim intensity,and GAMLSS model for ZIP regression of claim frequency and GB2 regression of claim intensity,as well as directly modeling the pure premium using the Tweedie regression model.The empirical results show that the using of physical collision results in auto insurance rating determination can significantly improve the accuracy of pure premium prediction,especially adding scoring rules of C-IASI.Finally,according to the empirical conclusions and the actual accident situation,this article puts forward some relevant improvement suggestions for C-IASI.All in all,the ZIP-GB2 model(GAMLSS)with C-IASI scoring rules has the best prediction effect on the pure premium pricing of auto insurance.