摘要
在评估商业银行整体信用风险时,债务人的信息一般不会传递到风险管理部门,导致在缺少违约数据时传统方法的分析十分复杂甚至难以进行。基于贝叶斯方法的潜在因素模型可以有效解决无法获得特定债务人信用质量的问题,并能够在宏观经济环境变动时准确评估违约风险强度变化,从而避免低估风险。利用MCMC模拟方法对商业银行数据的实证分析表明,潜在因素模型不仅推断方法及模拟途径简洁清晰,估计结果更加精确,而且在贝叶斯框架下具有较强的灵活性,适合在不同的数据约束条件下应用,便于国内风险分析人员采用。
In assessing the overall credit risk of commercial banks, the information of debtor generally will not be passed to the risk management department, which complicate and impede the inference of traditional methods in the absence of default data. Latent factor model, based on Bayesian Methods, can effectively resolve the lack of access to the credit quality of a particular debtor, and be able to assess the default risk intensity accurately when the macroeconomic environment change, thereby avoid underestimating the risk. The example of commercial banks data using MCMC simulation shows that the relative inference and simulation methods are not only very vivid, get more accurate estimates, but also has strong adaptability and flexibility in the Bayesian framework, fit the data in different condition, thereby is very conducive to internal risk analyzer.
出处
《统计与信息论坛》
CSSCI
2009年第8期24-29,共6页
Journal of Statistics and Information
基金
国家自然科学基金项目<非寿险经验费率模型>(70771108)
天津财经大学科研发展基金项目<非寿险经验费率的客观贝叶斯模型>(Y0804)