基于Bayesian Bootstrap方法的准备金风险度量研究

发布时间：2018-01-12 12:07

本文关键词：基于Bayesian Bootstrap方法的准备金风险度量研究　出处：《天津财经大学》2014年硕士论文　论文类型：学位论文

【摘要】：保险公司的风险主要来自承保风险、操作风险、违约风险等,其中承保风险包括准备金风险和保费风险。由于未来赔付的不确定性,准备金的评估值在实际值附近波动,提取的准备金与未来赔付可能并不相等,保险公司为了保证有充足的资本应对非预期风险,在技术准备金的基础上增加一部分作为偿付能力资本要求中的组成成分,这一部分就是准备金风险。因此,准备金风险是用来抵御准备金不充足的风险,是准备金与未来赔付之间差距的量化。常用的准备金风险为最终风险,衡量的是保单在未来全部进展年的不确定性,在随机准备金评估方法中,可以用均方根误差来进行度量。欧盟偿付能力Ⅱ提出要基于一年期对准备金风险进行度量,用来抵御保单在未来进展12月准备金不充足的风险,时间范围设定为1年。一年期准备金风险比最终准备金风险小,且更符合实际需要,更有利于风险监管。一年期准备金风险的度量方法可分为解析法和随机模拟法。Wiithrich(2008)基于Mack模型,推导了一年期准备金风险的解析方法,解析方法理论性强,不足之处是推导复杂,且未能考虑尾部进展因子。Ohlsson Lauzeningks(2009)提出了一种更加实用的一年期准备金风险度量方法——随机模拟Re-reserving方法,该方法与解析法相比,简便易行,且可以考虑尾部进展因子的影响,随机模拟Re-reserving方法主要应用MCMC原理和Bootstrap方法。文中首先介绍了欧盟偿付能力Ⅱ背景,进而明确了一年期准备金风险的概念；并介绍了解析式方法和随机模拟方法；然后在随机模拟的框架下,将Bootstrap方法扩展为Bayesian Bootstrap方法,并将尾部因子考虑进来,并可以获得赔付进展结果预测分布置信水平为99.5%的VaR值以量化一年期准备金风险；最后利用数据进行了实证分析,并将Bayesian Bootstrap方法与Wiithrich Merz解析方法(MW方法)、Bootstrap方法(2008)进行比较。结果表明：Bayesian Bootstrap方法与MW方法结果相近,且比Bootstrap方法标准差更低,过程误差和估计误差也更低。
[Abstract]:The risk of insurance companies from underwriting risk, operational risk, risk of default, which includes risk premium and underwriting risk reserve risk. Due to the uncertainty of future payment and reserve evaluation value in the actual value fluctuated around the reserve and future loss may not be equal to the insurance company, in order to ensure adequate capital to deal with unexpected the risk increase, as part of a solvency capital requirement in the composition on the basis of the technical reserves, which is part of the risk reserve. Therefore, the risk reserve is used to resist the risk reserve is not sufficient, is to quantify the gap between the reserve and future loss. The common reserve risk as the ultimate risk, is a measure of policy in the future all the progress of years of uncertainty in stochastic reserving method, can be used to measure the RMS error compensation of EU. The ability to pay one year to II proposed to reserve risk measurement based on the policy to resist the risk of future developments in December reserves are not sufficient, the time range is set for 1 years. One year reserve risk than the final reserve risk is small, and is consistent with the actual needs, more conducive to the risk supervision. Measure the one-year reserve the risk can be divided into analytical method and stochastic simulation method.Wiithrich (2008) based on the Mack model, the analytical method is the one-year risk reserve, the analytical method of strong theory, the deficiency is is complex, and failed to consider the tail factor in.Ohlsson Lauzeningks (2009) proposed a one-year risk reserve is more practical the measure method of stochastic simulation Re-reserving method, compared with analytical method, this method is simple and feasible, and can consider the influence factor in the tail, the stochastic simulation method Re-reserving The main application of MCMC principle and Bootstrap method. This paper firstly introduces the background of the EU Solvency II, and defines the concept of the one-year risk reserve; and introduces the analytical method and stochastic simulation method; and then in the framework of stochastic simulation, the Bootstrap method is extended to Bayesian Bootstrap method, and the tail factor into account, and can obtain compensation in distribution prediction results of 99.5% confidence level VaR values to quantify the one-year reserve risk; finally, using empirical analysis, and the Bayesian Bootstrap method and Wiithrich Merz analysis method (MW method), Bootstrap (2008) were compared. The results showed that the Bayesian Bootstrap method and MW method the results are similar, and the standard deviation is lower than the Bootstrap method, the process of error and error estimate is also lower.

【学位授予单位】：天津财经大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F840.31;F224

【参考文献】