Auto Insurance Rate Determination Models Based On Longitudinal Data

Commercial vehicle insurance is a necessary insurance to ensure the driver's daily travel,and it's also one of the most popular types of insurance.In car insurance business,premium rate making is one of the most important part.Whether the pricing of auto insurance is fair and reasonable is not only related to the consumer experience,but also the stable operation of the property insurance company,and also the healthy development of the whole industry.Since the first pilot project of ‘commercial fare reform' on June 1st,2015,the right to pricing and business scope of insurance companies have gradually expanded.This puts forward higher requirements for abilities of risk identification and risk control,which further increases the status of auto insurance pricing.Insurance companies began to explore more scientific and effective rate setting patterns,which is the significance of this paper.Currently,the method of premium rate making model of car insurance is mainly based on the generalized linear models.This paper puts forward that the application of the estimation method of longitudinal data in car insurance rate setting model,and the related empirical studies were made.Firstly,the current research background of vehicle insurance rate setting model is analyzed.This paper introduces the relevant theories of commercial vehicle insurance rate,and expatiates on the generalized linear model,the generalized estimation equation of longitudinal data,and the basic theory of generalized linear mixed effect model.In the empirical research part,these models are used to fit the auto insurance claim frequency.Firstly,the poisson distribution model and the negative binomial distribution model of the generalized linear model are selected to fit the vehicle insurance claim frequency.Secondly,by using the generalized estimation equation method,the working correlation matrix with the lowest QIC value is selected to model the longitudinal data of vehicle insurance.Finally,in the generalized linear mixed effects models,by using three methods to estimate the likelihood function,namely the Laplace approximation method,the adaptive Gauss-Hermite quadrature method and Bayesian inference method.In the conclusion part,the estimation effect of the three models used in the paper is compared,and the model applicable to commercial vehicle insurance rate is obtained.