Applications Of GLMM For Vehicle Insurance Rate Making In A Property Insurance Company

In March 2015,the Work Plan for Deepening the Pilot System Reform of the Administration of the Commercial Vehicle Insurance Clauses and Premium Rates was formulated by CIRC,which has been fully implemented across the country by June 2016.The new rate structure,considering the risk factor of vehicle,is different from the old one in the process of calculating pure risk premium.In particularly,the new rate structure has constructed the first ?life table? for automobile physical damage insurance,contributing to the steady development of the vehicle insurance market.Moreover,with the development of large data technology and the accumulation of vehicle insurance industry data,using data mining technology to identify the risk factor which could impact premium rates is a new direction for vehicle insurance premium rate making.At present,GLM has been used for vehicle insurance premium rate making,however,its limitations promote the exploration of the new model.Based on the comparative analysis of automobile physical damage insurance key business indicators,such as frequency,severity,loss ratio,discount rate etc.,before and after the rate reforming,this paper finds that the rate reform has positive effects on controlling the claiming cost and improving profitability.Under the background of rate reforming,changes of policy data structure,transform of NCD and risk classifications of rating factors have been dissected.Then,this paper uses a series of vehicle damage insurance loss data of a property insurance company,traced from 2011 to 2015,as the object of empirical study,and applying GLM,which is the representative of the traditional rate making model,on pure risk premium modeling of the vehicle damage insurance.After that this paper explores the advantage and disadvantage of GLM and put forward the necessity of the application of generalized linear mixed model(?GLMM?).Finally,by constructing Poisson GLMM for frequency,Gamma GLMM for severity and Tweedie GLMM for pure risk premium,and by comparing with corresponding GLMs,this paper concludes that GLMM has advantages in the fitting of the multi factor car families and it is better to improve the risk identification of the rate making.