未决赔款准备金不同评估模型的研究

发布时间：2018-05-21 16:48

本文选题：未决赔款准备金 + GJM　；参考：《山东大学》2015年硕士论文

【摘要】：未决赔款准备金是保险准备金的一种,是指保险公司在会计年度决算时,对上一个会计年度已经发生的保险事故,应付而实际没有支付的赔偿或者给付保险金所作的资金准备,即是指保险公司从当年收入的保费中,提取出相应的部分转入下一个会计年度,做为已发生保险事故而尚未赔付部分的资金来源。提取未决赔款准备金的目的在于保证保险公司承担将来的赔偿责任或给付责任,切实保护被保险人及其受益人的权益。未决赔款准备金不是保险公司的营业收入而是保险公司的负债。自然和人为引起的巨灾损失频繁发生以及市场竞争日益激烈,这就要求保险公司提高自身竞争力,一方面体现在产品的设计和销售,另一方面体现在准备金的计提上,投资的策略。提取准备金过多就会降低了公司的盈利能力,而过少的提取准备金将使保险公司不能给予足额索赔,可能导致保险公司陷入财务危机或破产。所以合理计提准备金异常重要。通过合理的方法计算出来的非寿险责任准备金大大地增加了非寿险责任准备金评估的充足性和准确性,从而增强了财产保险公司履行保险赔偿的能力。所以,对非寿险责任准备金的评估不仅仅是保监会的重点监管目标之一,而且也是理论研究的重要课题。经典的非寿险未决赔款准备金评估方法也就是确定性方法,例如链梯法、分离法、案均赔款法、准备金进展法、B-F法由于简单易于操作的优点等被广泛地被保险公司接受,但是确定性方法只能得到索赔准备金的均值估计,无法度量索赔准备金的波动性,而随机变量法可以度量估计的不确定性,在确定索赔准备金负债的估计的波动性、索赔准备金最佳估计与区间估计等方面都提供了重要的理论依据。本文首先对于未决赔款准备金的几种常用的方法予以介绍,并给出了三种随机模型。在接下来的研究中,用一些实例论证各种随机方法确定未决赔款准备金的合理性,将各种未决赔款准备金模型进行实证分析,用SAS、Matlab、VBA软件做数据处理,比较各种最新方法的优越性。如用Bootstrap方法求解对数正态模型、MCMC法求解贝叶斯正态模型、利用SAS求解广义线性模型求解未决赔款准备金的比较研究。本文的创新点在于改进了贝叶斯对数正态模型MCMC求解的模型进行改进,并且得到的结果优于之前的贝叶斯对数正态模型,与之前文献[1]相比,误差更小,精确度更高,参数估计值更加接近链梯法所求的结果。
[Abstract]:A reserve for outstanding claims is a kind of insurance reserve, which refers to the insurance company's financial preparation for an insurance accident that has occurred in the previous fiscal year but has actually failed to pay, or for the payment of insurance gold, when the insurance company accounts for the accounting year. That is to say, the insurance company draws the corresponding part from the insurance premium of the current year and carries it into the next accounting year, as the source of funds that has occurred the insurance accident but has not been paid. The purpose of drawing the reserve for pending claims is to ensure the insurance company to assume future liability or liability for compensation and to protect the rights and interests of the insured and its beneficiaries. The reserve for outstanding claims is not the operating income of the insurance company but the liability of the insurance company. Natural and man-made catastrophe losses occur frequently and the market competition is increasingly fierce, which requires insurance companies to improve their competitiveness, on the one hand, in the design and sale of products, on the other hand, in the provision of reserves. An investment strategy. Too much reserve will reduce the profitability of the company, and too little reserve will make the insurance company unable to make full claim, which may lead to the financial crisis or bankruptcy of the insurance company. Therefore, reasonable provision of reserves is extremely important. The non-life liability reserve calculated by the reasonable method greatly increases the adequacy and accuracy of the evaluation of the non-life insurance liability reserve and thus enhances the ability of the property insurance company to fulfill the insurance compensation. Therefore, the evaluation of non-life insurance liability reserve is not only one of the key regulatory objectives of CIRC, but also an important subject of theoretical research. The classical non-life insurance pending claim reserve evaluation method is also known as deterministic method, such as chain ladder method, separation method, case average compensation method, reserve progression method and B-F method, which are widely accepted by insurance companies because of their simple and easy-to-operate advantages. But the deterministic method can only obtain the average value estimate of the claim reserve, and can not measure the volatility of the claim reserve, while the stochastic variable method can measure the uncertainty of the estimate, and determine the volatility of the estimated liability of the claim reserve. The optimal estimate and interval estimation of claim reserve provide important theoretical basis. In this paper, several common methods of pending claim reserve are introduced, and three stochastic models are given. In the following research, some examples are used to demonstrate the rationality of various stochastic methods to determine the reserve for outstanding claims, and the models of the reserve for outstanding claims are empirically analyzed, and the data processing is done by using the software of SASMA Matlab / VBA. Compare the advantages of the latest methods. For example, the Bootstrap method is used to solve the logarithmic normal model and the SAS method to solve the Bayesian normal model. The innovation of this paper is to improve the MCMC model of Bayesian logarithmic normal model, and the result is better than the previous Bayesian logarithmic normal model. Compared with previous reference [1], the error is smaller and the accuracy is higher. The parameter estimates are closer to the results obtained by the chain ladder method.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：F840.4

【参考文献】