【作者】 徐杰；

【导师】 聂富强；

【作者基本信息】 西南财经大学 ， 统计学， 2008， 硕士

【摘要】 伴随着世界经济一体化进程的加快,特别是在我国加入WTO之后,我国银行业将更多地融入金融全球化的进程。因此,我国银行正面临着前所未有的发展机遇;同时也面临着参与国际竞争的严峻挑战。商业银行在经营活动过程中,面临的风险包括信用风险、市场风险、操作风险和流动性风险。由于我国尚未实行利率和汇率的市场化,因利率和汇率变动所产生的市场风险和流动性风险对我国银行业影响甚微,所以面对的主要风险为信用风险和操作风险,而其中信用风险又是当前我国商业银行所面临的最重要的风险。一、研究目的和意义信用风险是银行经营活动中的主要风险,我国商业银行积累的信用风险是中国经济健康运行的“悬剑”,也是困扰金融体制改革的难题之一。研究分析信用风险对信用风险的防范与化解有重要的意义。而我国商业银行内部的信用风险管理和控制方法仍然比较落后,缺乏科学有效的手段来控制风险,特别在信用风险度量方面与国际先进技术差距最大。长期以来以定性分析为主,缺乏量化分析,与国外银行大量运用数理统计模型、金融工程等先进方法相比差距很大。在本文的研究的主要部分中,一方面,VaR信用风险模型作为一种符合未来风险度量管理发展方向的综合性风险度量方法,得到了世界范围内的主要商业银行的认可和支持。本文结合我国实情,尝试将现代信用风险度量模型运用到我国商业银行的风险管理分析中;另一方面,作为一种新兴的数学理论,Copula理论正得到越来越多的关注。通过Copula理论,我们可以用一种较为简单的方法来处理相对复杂的金融问题。随机变量的联合分布完全决定了其各个分量的分布以及相依结构;反过来,联合分布是由边缘分布以及分量的相依结构两个组成部分共同决定。在维数比较高的时候,对于分析多2维随机变量的联合分布,将边缘分布与相依关系区分开来分别进行研究是必要的。一个多维分布的连接函数(Copula)就是描述了各个分量是种怎样相互依赖的函数关系,它不考虑边缘分布的影响,而只聚焦于相依结构,这种吸引人的性质使得我们可以用连接函数来考虑各种相互影响的风险因子是如何变动最终决定风险值大小的。本文根据此思路在VaR的基础之上对信贷风险度量改进做了一定的尝试研究:对信贷资产组合中各资产的边缘分布(margins)与资产间的相依结构(dependent structure)做了一定的假设和不同的选择,并对最后的结果做了比较分析。在论文的研究过程中,作者深深的感受到,了解、使用并发展新方法、新技术是我们认知事物发展规律的重要源泉。只有了解自然才能征服自然,同样只有了解金融市场的发展变化规律才能规避金融风险。因此,只有不断的发现、了解和使用新方法、新技术,才能使我们走近金融世界的自由王国。二、本文的基本框架和内容本文立足于如何提高我国商业银行信用风险量化准确性的角度,由于我国现行的信用风险管理手段仍然主要采用定性分析、静态管理的方式,因此研究并开展量化管理势在必行。而现有的条件下,必须发挥后发优势,借鉴国外相对成熟的VaR度量模型,控制研发成本。本文第一章为相关研究综述,首先介绍了研究的背景和意义,接着概述了国内外的研究现状和发展趋势。这些经典的理论综述为本文的研究提供了思路和方向。在此基础上,简要阐述了本篇文章所要研究的问题和创新点。本文第二章介绍了风险、信用风险与信贷风险的定义,以及信贷风险的特征。规定本文主要以信贷风险作为讨论的重点,凡无特别说明,信用风险均指信贷风险。这些概念都是分析信贷风险的基础,也是文章结论的支撑。本文的前两章为全文的基础部分,为后面的比较分析与实证研究做了铺垫和引导。第三章中,首先阐述了VaR的基本理论,在模型借鉴方面,介绍了国际上流行的四个模型,分别是:KMV穆迪信贷组合模型、CreditMetrics模型、麦肯锡CPV信贷组合模型和CreditRisk+信贷风险附加法模型。接着,在介绍了基于VaR的四个信贷风险度量模型的基础上,根据中国的实际情况对模型的优缺点进行了比较分析,得出CreditRisk+模型在我国的适用性优于其他三个模型的结论。在第四章中,接着上面的结论,首先详细分析了CreditRisk+模型的数理结构和模型内容,为下一节的模型实证提供数学基础。紧接着便是实证部分:利用CreditRisk+模型计算信贷组合的损失分布,应用CreditRisk+模型对我国某家商业银行的一组信贷组合数据进行实证研究,其中的一些参数根据我国商业银行已有的数据条件做了一定代替。实证研究结果表明:用贷款风险度代替CreditRisk+模型中单个借款人的违约概率具有可行性,经过改进后的CreditRisk+模型可以在我国商业银行现有数据基础上进行贷款损失的计量。第五章是对第四章实证研究的假设改进和拓展研究。上面介绍的现代信贷风险度量模型大多基于多元联合正态分布,运用方差—协方差法来求解信贷组合的VaR。虽然传统方法已公式化,具有运算简单的特点,但在实际应用中,资产收益分布并非完全正态,而是呈现“尖峰厚尾”的特征,同时,组合中几项资产之间的非线性关系也不是传统的相关系数矩阵所能表达。因此,需要拓展新的方法来更好地计算信贷组合的VaR。在本章中,首先对“连接函数”的理论原理和应用发展进行了较为全面的概述。Sklar提出连接函数将相关性与边缘分布分开研究,其主要特点是只需分别考虑连接函数和边际分布,就可以计算出信贷组合的VaR,而不需要确定联合分布的具体形式。这不仅扩大了方法的适用范围,而且提高了计算的效率和准确性。紧接着便是两个关于利用连接函数对信贷风险度量改进研究的实证。在实证1中,主要研究的是如何选择合适的连接函数进行信贷风险分析;实证2主要研究的是Copula理论改进后计算信贷组合的VaR。这两部分验证了借助连接函数理论,能够取得比传统的多元联合正态分布及线性相关假设模型更符合实际的VaR数据。使用连接函数方法对联合分布的相依结构进行改进,提出用t—连接函数代替正态连接函数。可以看到经过改进后使得信贷组合的损失分布发生较大变化,分布的尾部更厚,集中趋势也降低了。本文最后结束语包括三个方面:一是研究结论;二是政策建议;三是不足与展望,为今后进一步研究提供思路。三、本文的主要贡献1、我国信贷风险分析仍以定性分析为主,缺乏定量分析。国外经典现代信贷风险度量模型在国内的应用还没有普遍展开,原因一方面为缺乏一定的比较分析,各模型在中国的适用性分析不够。针对这一点,在本文第三章中结合中国的实际情况对模型的优缺点和实用性做了详细的比较分析,找出了相对而言更适合我国商业银行的信贷风险度量模型;原因其二是缺乏充足的历史数据,针对数量分析模型对数据要求高的特点,作者尝试结合我国商业银行现有的数据做了一定的替代:借鉴我国已有的关于贷款风险度的定义,代替计量模型中关于单个贷款人违约概率的数据要求。这两方面虽然是改进的一小步,但对于我国银行的管理重点逐渐从传统的资产负债管理过渡为以风险计量和风险优化的全面风险管理有一定的实证探索意义。2、本文最后所研究的Copula函数在商业银行信贷风险度量中的应用,旨在寻求一种能够更好度量相关性,把握尾部特征的工具。由于Copula应用于金融市场在国内还处于新兴起步阶段,用Copula研究我国商业银行信贷风险大多还都停留在定性分析,而不像Copula方法应用于股票资本市场那么成熟。但这种滞后性并不是代表着此技术的落后,反而体现着一种研究价值和现实应用要求。本文选题正是出发于对我国商业银行面临的急待解决的信贷风险及其管理理论与实践的思考,并对我国银行业的健康发展提供决策依据,因此,作者认为选题是具有比较重要的理论意义和现实意义的,也具有一定的前瞻性。由于数据的可得性受到一定的限制,而且不同于股市资本市场,Copula方法应用于商业银行信贷风险分析的经济研究和定量分析文献较少,加之笔者的能力有限,文章难免有疏漏和不足,恳请各位老师和同学指正。更多还原

【Abstract】 With the development of international economic integration,especially after China joining WTO,the banking of China is being in the process of finance globalization. So, the banking will be facing the unprecedented opportunities. At the same time, domestic banking will be facing the severe challenge of international competition.The commercial bank mainly faces includes Credit Risk, Market Risk, Capital Risk, Liquidity Risk in business activities, Foreign Exchange Risk and Operational Risk. Because our major banks are state owned banks and China is adopting strict Foreign Exchange control, we have been seldom facing Liquidity Risk and Foreign Exchange Risk. Among the remaining risk, the Credit Risk is the most important risk we are dealing with.1、Aim and MeaningCredit Risk is the principle risk in the bank operating activity. The credit risk accummulated in the past years by the state-owned banks is danger sword of the sound economics system and is one of the difficult problems halting the finaancial system reform. It is very important to analyze the credit risk for preventing and dealing with the credit risk. But our commercial bank’s inner credit risk management and controlling technology is still laggard, has not found efficient and scientific method to control risk. Our commercial bank has been using qualitative analysis for long, and in short of quantitative analysis method, so cannot be scientific in risk identification,measurement and monitoring. We have dropped behind the international advanced technology.In this paper, on one hand, The models of VaR credit risk, as a kind of comprehensive risk tolerance method that accords with future risk management development direction,have got the approval and support of the worldwide main commercial bank. According to the condition of our state, this paper has tried to apply modern credit of measurement model to the commercial bank of our country in risk management. On the other hand, as a fresh statistic tool, Copula Theory gets more and more attention, and more and more accomplishments of Financial Econometrics are based on the Copula Theory. With Copula Theory, we could use an easier method to deal with more complicated finance problems. The joint distribution of random variables fully incorporates the information of margins and dependent structure. But when analyzing high dimensional joint distribution, it is necessary to separate margins and dependent structure for the purpose of simplicity. A copula function of a joint distribution is a function that describes how and to what extent do random variables dependent on each other without considering their marginal distributions.This attractive property makes it possible for us to employ copula to do research on the joint occurrence of risk events. According to this thought, this article did some trier research on the estimate of credit risk. It made some assumptions on the relationship between the margins of the credit portfolio and the dependent structure of the assets, and then did some comparative analysis on the final results.During the research process, the author deeply felt, realizing, using, and developing new method and technology, is the primary resource that we realize the development rules of things. Only knowing the nature, we could overmaster it. As the same, only knowing the development and changing rules of the financial market, we could avoid the financial risk effectively. So we should keep on discovering, knowing, and handling new method, new technology, that we could get into the freedom field of the finance.2、Basic Framework and Content of the PaperThis article stands on how to improve domestic commercial bank credit risks quantitative management. Internal management had weakness such as static management and qualitative analysis, so it is necessary to refer to international quantitative management models.The first chapter is the summary about the relevant researches. First, it introduced the background and meaning of the research, then the research status and trend in domestic and international. These classic theory summaries offered the thought and direction for the research of this article, based on which, it also generally described the problem and innovation that this article will work over.The second chapter introduced the definitions of the risk, credit risk, and credit risk(loan), and also the characters of credit risk(loan). These definitions are the foundation for analyzing the credit risk, and also the support of the article. The first two chapters are the basic part, which served for later comparative analysis and demonstration analysis.In the third chapter, first of all, This paper explains the basic theories of VaR and intoduces four prevalent models, they are: Moody’s KMV EDFS, CreditMetrics, Credit Portfolio View and CSFP CreditRisk+. Then, based on the introduction of these four models which are based on VaR, it did the comparative analysis on the advantages and disadvantages of these models, then got the conclusion that the CreditRisk+ Model’s applicability in our country is better than other three models.In the next chapter, the forth chapter uses CreditRisk+ model to carry on positive research commercial bank of our country. The result shows:It is a feasible plan to refer to loan risk degree to determine CreditRisk+ model of individual probability in breach. And improving CreditRisk+ model can carry on the measurement with lost loan on the basis of existing data of commercial bank of our country. The fifth chapter is the most important part. The current credit risk measuring models introduced above are mostly based on the multi-joint normal distribution, and using the variance and covariance to work out the VaR of the credit portfolio. The traditional method had been formulized, and easily to compute, but in the practice, the asset prices are not perfectly normal distributed, but usually with the character of“sharp peak and thick tail”. Meanwhile, the non-linear relationship between the several asset prices of the portfolio can not be expressed by the traditional correlation matrix either. So we need to develop new method to computer the VaR of credit asset better. In this chapter, it firstly introduced the Copula Theory and then did two demonstration based on this. In the demonstration1, the primary research is how to choose proper connectivity function to analyze the credit risks. And the demonstration is primarily about computing the VaR of the credit portfolio after adjusting by the Copula Theory . We verify that copula method can obtain more pratical VaR value than traditional model of multivariate normal distribution.This article proposes to use t-copula to revise these models.The conclusion is that using new copulas instead of normal copula,the tails of loss distributions will be heavier and scattered.In the end, the paper concludes with research achievements and brings forward unsetted problems, paving the way for future research. 3、Major Contributions of the PaperOur commercial banks have been using qualitative analysis for long, and in short of quantitative analysis method, and have dropped behind the international advanced technology. There are many reasons. In this paper, based on the introduction of these four models which are based on VaR, it did the comparative analysis on the advantages and disadvantages of these models, then got the conclusion that the CreditRisk+ Model’s applicability in our country is better than other three models. Then use improved CreditRisk+ model to carry on positive research of certain commercial bank in China.In the last part of this article, we did some research on the application of the Copula Function on the estimate of the credit risks in the commercial banking, which is in order to find a tool that can measure the correlation and the characters of the tail better.The biggest and most emergency problem our commercial banking in our country facing now is how to manage the credit risk. That’s why this topic was chosen. It is a trying to offer some evidence to help the commercial banking industry when they make decisions. So the author think there are some important theory and practice meaning of this topic, and also some forecast meaning.Whereas limited literature in the quantitative analysis and economic research , access in data and the author’s capability, suggestions and recommendations are welcome for the paper from teachers and students.更多还原