Researches On Outstanding Claims Reserving Based On Stochastic Models

Non-life liability reserving has always been an important issue for insurance compa-nies. Firstly, whether an insurance company draws adequate reserves has a direct effect on its solvency. Secondly, an accurate valuation of the reserves provides an important basis for the management in its operational decision-making. There is a considerable variety of difficulties in the valuation of outstanding claims reserve both in theory and practice while that of unearned premium reserve is relatively easy. In practice, an insurance company usually takes deterministic approaches to value the outstanding claims reserve. While these approaches can give an estimate for the outstanding claims reserve, they cannot tell the accuracy of this estimate. Many researchers abroad have realized the importance of stochastic models in relation to valuating the accuracy of the estimate. Various stochastic models and methods have been studied in relevant research papers published. But there is little discussion on the comparison among these models and further research is really needed. And there is still room for further improvement in the current methods.This article consists of six chapters. Chapter 1 gives the overall framework as well as the research background. Chapter 2 and Chapter 3 focus mainly on three parts:firstly, the introduction of the traditional bootstrap methods:secondly, the application of smooth bootstrap methods in stochastic models; thirdly, the implementation of the smooth boot-strap methods in time series model by Monte Carlo simulation, and the comparison in the simulation results of both bootstrap methods and smooth bootstrap methods. Chapter 4 and Chapter 5 introduce the Bayesian models and verify the feasibility of MCMC method applied in Bayesian models with a particular example. Chapter 6 contains the summary of this article, mainly the results of the comparison of the two stochastic models, and also gives some basic rules in model-choosing. Besides, the two resample methods are also discussed and compared in this final chapter.The major contribution of this article is to apply the smooth bootstrap methods in stochastic models, and to compare the smooth bootstrap methods with the traditional bootstrap methods. Another contribution is to set the standards in model-choosing and to compare the characteristics of the mentioned models according to the simulation results.