首页 | 新闻公告 | 作者指南 | 编委会 | 关于杂志 | 订阅 | 相关链接 | 下载区 | 联系我们

基于非参数方法的银行操作风险度量
Operational risk measures for banks based on nonparametric methods
摘要点击 15  全文点击 6    
查看全文  查看/发表评论  下载PDF阅读器
中文关键词  操作风险; 非参数估计; 区间估计; VaR
英文关键词  operational risk; nonparametric; interval estimate; VaR
基金项目  国家自然科学基金资助项目(71171083) ; 教育部人文社会科学研究基金资助项目(09YJC630075) ; 上海市教育委员会科研创新资助项目(14ZS058) .
学科分类代码  
作者单位
汪冬华 华东理工大学商学院,上海200237 
徐驰 华东理工大学商学院,上海200237 
中文摘要
      金融业全球化竞争和金融管制放松,导致商业银行面临的操作风险不断增加,操作风险已成为金融监管的焦点.因此,对商业银行和其它金融机构来说,可靠的操作风险度量正变得越来越重要.文章采用基于厚尾分布的非参数方法度量我国商业银行操作风险,给出了VaR的点估计方法和3 种区间估计方法( 正态近似法NA,经验似然法EL,数据倾斜法DT) .此方法的优点在于不用假设操作风险损失分布,这样可以消除参数化模型设定差异而带来的估计偏差.同时,根据厚尾分布的特征,提出了新的厚尾分布样本均值求法,调整后均值更注重对尾部的描述和刻画.实证结果表明: 调整后的厚尾分布样本均值大于简单算术平均值,更符合右偏厚尾的分布特征; 非参数方法得到的VaR 点估计和区间估计考虑了厚尾的因素,解决了传统VaR 低估风险的问题,更接近真实情况; VaR 3 种区间估计的方法能够提升对风险衡量的准确性,其中DT 方法所得到的区间估计最为准确.
英文摘要
      With global competition of the financial sector and financial deregulation,commercial banks are facing increasing operational risk which has become the focus of attention. Therefore,reliable operational risk measurement is becoming increasingly important for commercial banks and other financial institutions. In this paper,nonparametric methods based on heavy-tailed distributions are applied to operational risk measurement.The main advantage of these nonparametric methods is that there are no assumptions made about the shape of loss distributions. It avoids estimate deviation caused by unwittingly mis-specified models. Meanwhile,according to the characteristics of heavy-tailed distributions,a new method to estimate the mean of loss distributions is put forward,and the adjusted mean focuses more on the tail part of loss distributions. The empirical results demonstrate that the adjusted mean exceeds the sample mean,which is in more conformity with the right heavy-tailed distributions’characteristics. This paper employs non-parametric approaches and constructs a consistent and unbiased point and interval estimates for VaR. It has overcome the weakness of underestimating of traditional VaR. We discuss three methods of estimating confidence intervals to improve the accuracy of risk measurement. As a consequence,DT ( Data Tilting) interval estimates turned out to be the best.
关闭

版权所有 © 2007 《管理科学学报》
通讯地址:天津市南开区卫津路92号天津大学第25教学楼A座908室 邮编:300072
联系电话/传真:022-27403197 电子信箱: jmstju@263.net