企业违约概率模型表现的监管标准研究  

The Regulatory Criterions for Performances of Corporate Default Probability Models

作　　者：章彰[1] 

机构地区：[1]中国银行风险管理总部新资本协议办公室

出　　处：《金融监管研究》2013年第12期16-30,共15页Financial Regulation Research

摘　　要：在内部评级法实施过程中,企业违约概率模型的表现是国内外监管机构持续关注的重点。企业违约概率模型表现的监管指标体系应以AR值(Accuracy Ratio,精确率)为区分力指标;以在一定的容忍度下,主标尺下各个级别的实际违约率围绕该级别违约概率中间值落入的设定区域为准确性指标和审慎性指标;以子样本AR值和不同时点上AR值的变化为稳定性指标。监管机构除了评估上述四个维度指标值的合理性以外,还需要结合银行主标尺设置、评级企业数量、违约企业数量、模型方法论等因素,整体判断企业违约概率模型的表现。The performance of corporate default probability models （ DPM ） in internal ratings is of main con-cern among regulatol3, authorities. The regulatory criterions for the performance of corporate DPMs should use Ac-curacy Ratio as an indicator for discriminatory power, apply the frequency of realized default probability falls in a pre-set range of the middle value of forecasted default probability at a certain confidence level as indictors for accu-racy and discretion, and empoly the AR values of a sub-sample and the changes of AR values at different time as a stability indicator. Besides assessing the qualifications of these four indicators, regulatory authorities should also consider the main scale of AR set by banks, corporate numbers, default corporate numbers, and methodology of models e. t.c. to judge the overall performances of corporate DPMs.

关 键 词：违约概率 模型表现 信用风险 

分 类 号：F830.5[经济管理—金融学]