基于Cox比例风险模型的商业银行信用风险度量研究

发布时间：2018-05-06 19:59

本文选题：生存分析 + Cox比例风险模型　；参考：《山东大学》2014年硕士论文

【摘要】：信用风险作为银行业的主要风险之一,表现形式越来越多样化,影响也逐渐扩大,不仅会给金融行业带来巨大的冲击还会影响整个社会的稳定,因此,其危害性也得到了越来越多的重视。为了有效的管理信用风险,商业银行也加快了信用风险研究的步伐。发达国家的信用风险管理起步早,发展较成熟,逐渐实现定性向定量的发展。而我国的商业银行信用风险管理还处于定性分析的阶段,缺乏成熟的定量分析模型和技术。然而,随着金融行业的不断发展,传统的信用风险的评级方法和传统的信用风险模型已经很难适应发展需要,因此准确合理的度量商业银行的信用风险对商业银行的信用风险管理尤其重要。目前研究信用风险的模型较多,应用较早有Logit模型、研究单一信用风险的KMV模型及测度组合信用风险的Credit Risk+等模型,这些模型在一定条件和假设下取得了较好的结果。然而Credit Risk+等模型需要企业信用评级的数据,而我国缺乏企业信用评级的历史数据,所以这些模型不太适用我国的金融市场。随着生存分析理论的发展和应用的不断推广,基于生存分析的Cox比例风险模型在信用风险度量研究中显示出一定的优越性,Cox比例风险模型不需要信用评级数据,而是研究对生存时间有影响的财务数据,因此可以用来分析国内企业的财务状况。本文尝试进行Cox比例风险模型的实证研究。本文基于现有的信用风险研究结果,首先,介绍了生存分析的基本理论并分析生存分析在信用风险度量中的作用。其次,详细介绍了基于生存分析的Cox比例风险模型,并给出了模型参数和非参部分的估计方法。最后,基于实际数据建立Cox比例风险模型,进而得到上市公司的违约概率,并对模型结果进行分析,根据模型结果我们可以分析增加上市公司财务风险的因素(即危险因素)和减少上市公司风险的因素(即保护因素)。通过分析模型的时点预测能力及准确性,判断模型可行,商业银行可以根据模型结果判断自身面临的信用风险情况。本文通过CAP曲线、ROC曲线及KS检验验证了模型的有效性和稳定性,为模型在实际中的应用提供了基础。
[Abstract]:As one of the main risks of the banking industry, credit risk is becoming more and more diversified and its influence is gradually expanding. It will not only bring huge impact to the financial industry but also affect the stability of the whole society. Therefore, its harmfulness has also been paid more and more attention. In order to manage credit risk, commercial banks also accelerate their credit The pace of risk research.
The credit risk management in developed countries is early, mature and gradually realized to the qualitative and quantitative development. The credit risk management of commercial banks in China is still in the stage of qualitative analysis, lack of mature quantitative analysis model and technology. However, with the continuous development of the financial industry, the traditional credit risk rating method and tradition The credit risk model has been difficult to adapt to the development needs, so the accurate and reasonable measure of the credit risk of commercial banks is particularly important for the credit risk management of commercial banks.
At present, there are many models to study credit risk, the Logit model is used earlier, the KMV model of single credit risk and the Credit Risk+ model of measuring combination credit risk are studied. These models have obtained good results under certain conditions and assumptions. However, Credit Risk+ and other models need the data of enterprise credit rating, but China is short of enterprise. The historical data of the industry credit rating, so these models are not very suitable for China's financial market. With the development and application of survival analysis theory, the Cox proportional risk model based on survival analysis shows some advantages in the research of credit risk measurement. Cox does not need credit rating data, but it does not need the credit rating data. The financial data that affect the survival time can be used to analyze the financial situation of domestic enterprises. This paper attempts to conduct an empirical study on the Cox proportional hazards model.
Based on the existing research results of credit risk, this paper first introduces the basic theory of survival analysis and analyzes the role of survival analysis in credit risk measurement. Secondly, the Cox proportional risk model based on survival analysis is introduced in detail, and the model parameters and non parametric estimation methods are given. Finally, based on the actual data, the Cox is established. According to the results of the model, we can analyze the factors of increasing the financial risk of the listed companies (i.e., the risk factors) and the factors to reduce the risk of the listed companies (that is the protection factor). It is feasible that commercial banks can judge their credit risk according to the results of the model. In this paper, the validity and stability of the model are verified by CAP curve, ROC curve and KS test, which provides a basis for the application of the model in practice.

【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F832.33;O213

【参考文献】