信用违约风险模型中违约概率的统计推断  

Statistical inference of default probability in credit risk models

作　　者：周勇[1] 谢尚宇[1] 袁媛[2] 

机构地区：[1]中国科学院数学与系统科学研究院,北京100190 [2]上海财经大学统计学系,上海200433

出　　处：《系统工程理论与实践》2008年第8期206-214,共9页Systems Engineering-Theory & Practice

基　　金：国家973项目子项目(2007CB814902);国家自然科学基金委统计重点项目(10731010);国家自然科学基金委杰出青年基金B项目(10628104);国家自然科学基金(10721101)

摘　　要：简约化模型是目前研究信用风险的一种重要模型,应用信用违约风险模型最重要的步骤是计算违约概率.在简约模型中,可以假定违约是外生的,由强度函数(风险率函数)可以方便地描述违约机制,从而为考虑多因素复杂违约模型提供了基础.本文通过利用一些在违约风险上有用的生存分析方法,并对违约概率进行估计.所提出的风险模型既能自然地体现各种风险因素,又能很好地解释各种风险因素对风险的影响,同时所提出的违约概率模型也能考虑因素的动态性和交互作用.In studies of credit risk, reduced model is very important and useful. Default probability is the most important quantity in order to apply the reduced model. In reduced model, we suggest that default is exogenous, so that we can easily use many statistical methods to compute defatult probability. In this paper we propose some hazard rate models to analyze default risk by some methods of statistics. These modeLs can take into account various risk factors and excellently explain the effect of those factors on the default probability. Meanwhile, these models can deal with dynamic effect and interaction.

关 键 词：信用风险 违约风险 删失数据 风险率(强度)函数 COX模型 Logist模型 Cox变系数模型 

分 类 号：O212.7[理学—概率论与数理统计] F224[理学—数学]