摘要
文章就中国商业银行操作风险内部欺诈损失样本数据较小的情况,采用贝叶斯推断增加推断效果。损失频率以泊松分布和负二项分布为例,对损失强度以对数正态分布和极值分布为例,利用先验信息分析损失的联合先验分布,其中,极值分布没有共轭分布,针对参数分别推导后验估计。并进行了实证分析,得出四种分布的联合分布。结果表明,由于采用了先验知识,能够比较准确地进行推断,极值分布对损失强度尾部估计较好,中国商业银行内部欺诈损失巨大,采用贝叶斯方法有助于解决操作风险损失数据,尤其是年度极值数据不足问题。
Internal fraud is the most serious operational risk in Chinese commercial banks. As for the loss frequency, taking the Poisson distribution and Negative Binomial distribution as the examples, and the loss severity, taking the Log - Normal distribution and Extreme distribution as the examples, this paper analyzes the joint prior distribution of the loss with the prior information, deduces the posterior estimate and then makes an empirical analysis of the four joint distributions. The results prove that using the priori knowledge can precisely infer that the loss is enormous due to the internal fraud in Chinese commercial banks and applying the Bayesian inference helps solve the problem of the lack of data of the operational risk, especially the annual data.
出处
《山东财政学院学报》
2013年第1期12-18,共7页
Journal of Shandong Finance Institute
基金
国家自然科学基金面上项目"银行操作风险度量模型与风险资本测定研究"(70701033)
山东省自然科学基金高校科研单位联合专项项目"基于损失属性的中国商业银行操作风险管理研究"(ZR2010GL011)
关键词
金融风险
内部欺诈
贝叶斯推断
损失分布法
极值分布
finance risk
internal fraud
Bayesian inference
loss distribution approach
extreme distribution
