残差相关条件下非寿险准备金风险分析  

Non-life Reserving Risk Analysis with Residuals Dependable Structure

作　　者：刘乐平[1] 高磊[1] 丁东洋[2] 

机构地区：[1]天津财经大学统计系 [2]南昌大学公共管理学院

出　　处：《统计研究》2015年第10期82-91,共10页Statistical Research

基　　金：国家自然科学基金"Solvency Ⅱ框架下非寿险准备金风险度量与控制研究"(71171139);"多重风险相依情形下的最优保险问题研究"(71371138);"Basel Ⅲ框架下商业银行监管资本套利识别研究"(71303169);"逆周期资本监管框架下考虑跳跃行为的信用风险度量研究"(71401069);研究生科研资助计划(2014TCB03)的资助

摘　　要：本文研究Solvency II框架下非寿险准备金风险的度量与控制问题。针对精算实务中赔付数据残差三角形不满足独立性假设情形,首先基于多重假设检验理论与错误发现率(FDR)控制过程,对残差三角形是否具有相关性进行识别和检验;然后提出两阶段分区域Bootstrap方法,模拟获得未决赔款的预测分布,分析残差相关性对准备金风险波动的影响;最后,保险公司真实索赔数据的实证结果表明:消除残差三角形相关性的影响,可以增强准备金估计的稳健性,从而有效提高准备金风险度量的准确性。Bootstrap method has become a popular tool for assessing the MSEP of claims reserving and simulating the prediction distribution of outstanding liabilities. Based on Over-dispersed Poisson model,this paper makes a research on the dependable residual problem in the application of Bootstrap method,and designs a two stage block Bootstrap method to make a improvement on the traditional Bootstrap method. The paper finds that independent assumption of the real claims payment data maybe violated and the residuals may have a dependable structure. We can identify the dependable structure in the residuals using runs test and FDR control process which is the most used multiple hypothesis testing approach. This research has important theoretical significance and practical value for Bootstrap method to be introduced and developed into non-life insurance companies.

关 键 词：非寿险准备金风险 残差相关性 FDR 两阶段分区域Bootstrap方法 

分 类 号：O212[理学—概率论与数理统计]