The Study Based On State Space Model Of Claims Reserve

Outstanding claims reserve is a major liability for the insurance companies. Extracting sufficient or not directly affects the company's financial position, and Estimating for outstanding claims which is too high or too low would adversely affect the company.So it is essential to choose a suitable method to estimate the outstanding claim reserve.For estimates of the outstanding loss reserve, in currently, we mainly make use of deterministic methods in domestic insurance companies in real work. Such as the chain ladder method, the payments per claim incurred(PPCI) method and the B-F method. A few insurance companies use the lognormal model and generalized linear models. However, these methods are either too simple, or can not meet the more complex situation.In this paper, we use Bayesian theory and methodology for the outstanding claims reserve and introduce the state space model which takes full account of internal relations about factors in the progress year and factors in the development year.AS the core content, the state space model is stochastic model, what's more, It is made up of the chain ladder method and kalman filter .In order to predict the outstanding claims reserve, based on a province's auto insurance company data about China Continent Property&Casualty Insurance Company LTD, we use WinBUGS software and take MCMC methods and Gibbs sampling into the simulation, updating the iteration of the samples continuously ,completing the estimates of the outstanding claims reserve which is about a crime incident occurred in the auto insurance close to article 7 delayed years and the claims which is occurred in prior years of insurance accident and delayed to next year.Then, in the empirical analysis phase, we compare the state space model with the chain ladder method, the B-F method. This model can not only be an estimate of outstanding claims, but also estimate its error and distribution density function. In addition, compared with a log-normal model, the state space model takes full account of internal relations about factors in the progress year and factors in the development year, the predicted results closer to reality.Finally, we comprehensively consider the influence of the compensation amount and compensation rate on outstanding claims reserve. Based on state space model, we join the claims count in it. This is a improved model. Theoretically, the improved model has more advantages.