商业银行信用风险评估的生存分析模型及实证研究  

A Survival Analysis Model for Credit Risk Appraisal of Commercial Banks and an Empirical Study

作　　者：宋雪枫[1] 杨朝军[1] 徐任重 

机构地区：[1]上海交通大学管理学院 [2]申能股份有限公司财务部

出　　处：《金融论坛》2006年第11期42-47,共6页Finance Forum

基　　金：国家自然科学基金(70373053)

摘　　要：企业发生财务危机,不能归还到期贷款是商业银行信贷资产的主要风险来源,商业银行如何构建恰当的信用风险评估模型来预测企业的财务危机,从而避免这类信用风险的出现就显得尤为重要。本文以我国上市公司为研究对象,结合杜邦分析法建立了基于生存分析的信用风险评估模型,模型对于随机选取的预测样本,其提前1年、2年和3年的预测准确率分别达到86%、72%和68%。通过与Altman模型、Ohlson模型预测结果的比较和鲁棒性检验的结果发现,该模型同时具有可以使用时间序列、无需样本配对、中远期预测能力强和高鲁棒性的特点,这些特点特别对于商业银行中长期信贷风险管理具有较高的应用价值。That an enterprise in financial crisis is unable to repay the due loans constitutes the major source of risks to credit assets of commercial banks. Thus it is of paramount importance for commercial banks to predict a company＇ s financial crisis with an appropriate credit risk appraisal model so as to keep such credit risk at bay. This paper studies listed companies in our country and establishes a credit risk appraisal model based on survival analysis using Du Pont approach. The model provides 86% ,72% and 68% prediction correctness respectively for 1,2 and 3 year randomly selected samples. By comparing with Altman model and Ohlson model and through robustness test, it is found that this model which can use time sequence and demands no sample match demon- strates high medium-and-long term forecastability and robustness. These characteristics put commercial banks in good stead in terms of medium-and-long term credit risk management.

关 键 词：商业银行 信用风险 企业财务危机 生存分析 COX模型 多期预警 

分 类 号：F832.4[经济管理—金融学] F224