非线性多重约束的内部评级违约概率模型的设计研究  

A Study of the Design of Nonlinear Multi-constraint Internal Rating Models of Default Probability

作　　者：李海涛 

机构地区：[1]恒丰银行总行风险计量中心,新资本协议项目实施办公室

出　　处：《金融监管研究》2014年第11期78-91,共14页Financial Regulation Research

摘　　要：2014年4月,经银监会核准,国内6家银行实施资本计量高级方法,其显著特色是过渡期后利用内部评级初级法下违约概率结果计量风险加权资产和监管资本。目前关于违约概率估计的研究大都存在未能严格遵循监管要求建立识别违约样本的标准,未能建立包括校准环节等在内的完整内部评级模型开发框架,以及未能将专家的业务经验纳入模型设计等方面的不足。鉴此,本文建立了一个满足监管合规要求的模型开发框架,设计了一个非线性多重约束的内部评级违约概率模型,并引入分类器原理评估模型的可靠性。基于某银行业务数据的实证研究表明,该模型效果良好。Six domestic banks have been approved to implement advanced methods of capital measurement in April 2014, whose significant characteristic is to measure risk-weighted assets and regulatory capital based on the probability of default after the transition period. However, present research findings have some shortcomings, such as failing to identify default samples according to regulatory requirements, failing to establish a development framework （including calibration etc.） of the internal rating model and being incapable of fully bringing banking business experience into the model design. Therefore, the paper establishes a model development framework meeting regulatory compliance requirements, innovates to design a nonlinear multi-constraint internal rating model of default probability, and implements an empirical study with the business data of a bank, and introduces the classification principles to evaluate the model＇s reliability. Empirical study results show that this model has a good performance.

关 键 词：新资本协议 内部评级 违约概率模型 资本计量高级法 

分 类 号：F830.5[经济管理—金融学]