【作者】 秦昌贵；

【导师】 田新时；

【作者基本信息】 华中科技大学 ， 金融学， 2005， 硕士

【摘要】 对相关违约风险进行建模,在当前是风卷信用市场的一种新现象。由于在我国没有类似美国Moody和Standard & Poor那样的评级服务公司提供较全面完整的违约概率统计资料,因而,本文采用新华远东资信评估有限公司截止到2003年6月对113家A股上市公司的公开资信评级(Pi),以及这113家上市公司的历史股价,分别计算出它们违约概率的时间序列,并且基于常数失效率假定,将违约概率转换成失效率。本文利用计算得到的6个信用等级的113个上市公司的总失效率数据,并且基于这些总失效率所对应的超额水平,画出关于超额相关性的描点图,主要分析了我国A股上市公司的违约相关的三个特征:违约风险的水平、违约相关性的非对称性,以及联合违约的尾部关联性。关于超额相关性的描点图反映了高质量的证券比低质量的证券有更大的违约相关性。但是,低级别的债务似乎比高级别的债务有更多的尾部相关。这两个结论也是很直观的:高级别的证券更会由大的企业发行,而这些企业经常遭受更大的系统风险,而低信用质量的企业会承受更多的特有风险,低质量的证券更容易遭受传染,导致更大的尾部相关。其次,关于相关的非对称性有明显的证据,即当失效率的变化为正时,其相关性水平要高很多。因此,对相关违约的建模不能单纯建立在相关的对称性假设上。本文还提出了如何健全我国信用评级的一些建议。更多还原

【Abstract】 Modeling correlated default risk is a new phenomenon currently sweeping through the credit markets. Laking of credit rating companies like Moody or Standard & Poor providing comprehensive risk management services, without data set of default probabilities (PDs) about the firms in the stock market of China, we use the stock prices of 113 firms, and the public rating information about these firms coming from Shanghai Far East Credit Rating Co.,LTD, to calculate their PDs of time series. We convert PDs into hazard rates using an assumption of constant hazard rates. Using the total hazard rate (THR) across issuers at each point in time as calculated, we graph the exceedance correlations for the different rating classes, and analyse three features of default correlation: correlation levels, correlation asymmetry, and tail-dependence. The exceedance plot shows that stocks within high quality ratings have greater default correlation than stocks within low quality ratings. However, tail dependence is higher for lower rated stocks. Both results are intuitive – high grade debt tends to be issued by large firms that experience greater amounts of systematic risk, and low grade firms evidence more idiosyncratic risk; however, when economy-wide default risk escalates, low grade bonds are more likely to experience contagion, leading to greater tail dependence. There is clear evidence of correlation asymmetry, so we conclude that we can’t model default correlation under the assumption of symmetry. At last in this paper, we also give some advises how to complete the credit rating job in China.更多还原