基于结构化模型的违约概率期限结构研究

Study on term structure of default probability based on structural models

研究国内信贷市场上违约概率的期限结构问题.首先,基于结构化模型基本原理,提出了累积违约概率和边际违约概率模型;然后,利用国内某银行的数据,测算了1～20年期边际违约概率曲线以及年度累积违约概率和年度边际违约概率;最后,通过情景分析的方法研究了波动率σ和违约阈值 B 对违约概率期限结构的影响.研究发现:1)在1年期违约概率的预测方面,结构化模型优于 Z′评分模型;2)违约概率的变化幅度随期限的增加而逐渐降低.但对于风险较高的贷款,5年期以上的年度边际违约概率的变化幅度不应被忽略,需要对目前的信用风险度量和定价方式加以改进;3)风险越高,近期违约概率越大,中远期违约概率先增后减;时点距离当前越远,边际违约概率对这些风险因素的敏感性越低.

This paper studies the term structure problem of default probability （PD） in China＇ s local debt market. Firstly, models of cumulative PD and marginal PD were put forward by following the basic rules of structural models ; Then, using local banks＇ data, the 1 - 20 year marginal PD curves and 1 -20 cumulative PD and the yearly marginal PD were calculated; Finally, the effects of volatility or and default threshold B on PD were studied through scene analysis. It is found that： l ） the structural models＇ prediction power on one-year＇ s PD is better than Z＇ score models; 2） the longer the term is, the less the change scope of PD is. However, the change scope of high-risk debts whose term is more than the 5-years should not be ignored. So, the current method of debt pricing for more than 5-year debt should be improved; 3） the higher the risk is, and the bigger the short- term PD is, both the medium-term and the long-term PD increase firstly and then decrease; the longer the term is, the less the sensitivity of PD to those risk factors is.