基于RJMCMC的准备金风险随机模拟算法
发布时间：2018-08-30 19:32
【摘要】：准备金是一般保险公司最大的负债项目,提留充足的准备金对保险公司的经营非常重要。随着各种因素不确定性的增加,仅仅单一的准备金计提结果并不能满足监管机构的监管要求和公司风险管理需要。需要一种方法来刻画准备金风险。传统的随机准备金模型一方面可以给出准备金估计结果,另一方面还可以给出准备金估计不确定性的度量。绝大多数的随机准备金模型,可以依据数理统计推导方法,也可以依据bootstrap方法,还可以依据贝叶斯分析来评估准备金的不确定性。 可以从多种角度审视准备金风险。在传统的随机准备金模型中,是通过计算所有进展年未决赔款的不确定性来刻画准备金风险的。为了与其他风险度量方法相一致,偿付能力Ⅱ下要求将时间范围限制在一年以内。如何对一年以内的准备金风险进行度量,随机模拟技术是最佳选择。往往是在基础的随机准备金模型基础上,对未来一年的赔付进行模拟,然后计算准备金。整个的风险内容包括未来一年期风险和未来估计参数风险,比较接近于偿付能力Ⅱ对准备金风险的定义。 本文以超散布泊松分布模型为基础,构建随机准备金评估模型。采用贝叶斯估计方法,使用的后验分布模拟工具是RJMCMC(Reversible jump MCMC)。使用该方法可以寻找最优的尾部进展年水平拟合曲线。在此模型基础上,采用重复准备金(Re-reserving)技术用赔付进展结果和重复准备金两个随机变量刻画准备金风险。这种度量模式可以捕获一年期准备金风险,为保险公司资产负债管理提供决策依据。
[Abstract]:Reserves are the largest liability of an insurance company, and it is very important to keep sufficient reserves for the operation of an insurance company. With the increasing uncertainty of various factors, a single reserve can not meet the regulatory requirements and corporate risk management needs. A method is needed to characterize reserve risk. On the one hand, the traditional stochastic reserve model can give the result of reserve estimation, on the other hand, it can also give the measure of uncertainty of reserve estimation. The vast majority of stochastic reserve models can be derived from mathematical statistics, bootstrap methods and Bayesian analysis to evaluate the uncertainty of reserves. Reserve risk can be viewed from a variety of perspectives. In the traditional stochastic reserve model, the reserve risk is described by calculating the uncertainty of all outstanding claims in the progress year. In order to be consistent with other risk measures, solvency II requires a time limit of less than one year. How to measure the reserve risk within one year, stochastic simulation technique is the best choice. On the basis of the stochastic reserve model, the compensation in the next year is simulated, and then the reserve is calculated. The whole risk includes future one-year risk and future estimated parametric risk, which is close to the definition of reserve risk in solvency II. In this paper, a stochastic reserve evaluation model is constructed based on the superscattered Poisson distribution model. Using Bayesian estimation method, the posteriori distribution simulation tool used is RJMCMC (Reversible jump MCMC). This method can be used to find the optimal horizontal fitting curve of tail progress year. On the basis of this model, duplicate reserve (Re-reserving) technique is used to describe the reserve risk with two random variables, the compensation progress result and the duplicate reserve. This measurement model can capture one-year reserve risk and provide decision basis for asset liability management of insurance companies.
【学位授予单位】：天津财经大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F840.3;F224
本文编号：2214072
[Abstract]:Reserves are the largest liability of an insurance company, and it is very important to keep sufficient reserves for the operation of an insurance company. With the increasing uncertainty of various factors, a single reserve can not meet the regulatory requirements and corporate risk management needs. A method is needed to characterize reserve risk. On the one hand, the traditional stochastic reserve model can give the result of reserve estimation, on the other hand, it can also give the measure of uncertainty of reserve estimation. The vast majority of stochastic reserve models can be derived from mathematical statistics, bootstrap methods and Bayesian analysis to evaluate the uncertainty of reserves. Reserve risk can be viewed from a variety of perspectives. In the traditional stochastic reserve model, the reserve risk is described by calculating the uncertainty of all outstanding claims in the progress year. In order to be consistent with other risk measures, solvency II requires a time limit of less than one year. How to measure the reserve risk within one year, stochastic simulation technique is the best choice. On the basis of the stochastic reserve model, the compensation in the next year is simulated, and then the reserve is calculated. The whole risk includes future one-year risk and future estimated parametric risk, which is close to the definition of reserve risk in solvency II. In this paper, a stochastic reserve evaluation model is constructed based on the superscattered Poisson distribution model. Using Bayesian estimation method, the posteriori distribution simulation tool used is RJMCMC (Reversible jump MCMC). This method can be used to find the optimal horizontal fitting curve of tail progress year. On the basis of this model, duplicate reserve (Re-reserving) technique is used to describe the reserve risk with two random variables, the compensation progress result and the duplicate reserve. This measurement model can capture one-year reserve risk and provide decision basis for asset liability management of insurance companies.
【学位授予单位】：天津财经大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F840.3;F224
【参考文献】
相关期刊论文 前4条
1 张连增;段白鸽;;基于GLM的未决赔款准备金评估的随机性链梯法[J];财经理论与实践;2012年01期
2 陈希孺;广义线性模型(十)[J];数理统计与管理;2004年02期
3 卢志义;刘乐平;;广义线性模型在非寿险精算中的应用及其研究进展[J];统计与信息论坛;2007年04期
4 孟生旺;;未决赔款准备金评估模型的比较研究[J];统计与信息论坛;2007年05期
相关博士学位论文 前1条
1 卢志义;基于广义线性模型的损失准备金估计方法研究[D];天津财经大学;2008年
，本文编号：2214072
本文链接：https://www.wllwen.com/jingjilunwen/bxjjlw/2214072.html
