基于CreditMetrics模型对我国信贷组合信用风险度量研究

发布时间：2018-04-24 08:07

本文选题：商业银行 + 信用风险　；参考：《西南财经大学》2013年硕士论文

【摘要】：上世纪七十年代由于金融自由化发展和金融管制放松,使各金融机构的业务迅速发展,趋利性的金融衍生产品迅速出现。金融业的相关风险不断暴露,不断引发金融危机,如亚洲金融危机、欧洲货币危机、美国次贷危机。尤其是美国次贷危机,被前美联储主席格林斯潘认为是“百年一遇”的全球性金融风暴。有多方面原因导致该金融危机的出现,但究其根源,主要是信用风险管理方面的失控。这些由于金融创新而爆发的金融危机使世界经济不断遭到巨大的影响。为此,国际金融界各监管部门对风险监管提高了重视。国际银行监管机构于2010年9月出台的《巴塞尔新资本协议》,对银行风险的控制更加关注,提出的资本和流动性监管标准比以前更加严格。从国内目前的研究现状来看,我国银行业信用评级体系只可达到基本的内部评级法的要求,但是与高级内部评级法的规定还相差较远,远达不到规定的要求。我国商业银行成立较晚,银行业监管方面不完善,信用风险管理水平落后,技术不发达,体系不健全,没有更好的方法量化信用风险的大小。因此,借鉴并研究学习国际银行业信用风险管理的相关先进技术对我国银行业抗风险能力的提高和信用风险管理的加强具有重要意义,在将来的国际竞争中处于有利地位。结合我国国有商业银行的具体情况以及面对当前上市公司非交易性金融信贷资产,应用CreditMetrics模型进行研究和计算,对于我国银行业的信用风险水平度量具有重大意义。本文以CreditMetrics模型的实证分析与应用为导向,先后分析我国商业银行信用风险方面的现状、存在的问题和对我国银行业在信用风险管理方面实践的简要回顾基础上,围绕商业银行如何利用信用风险的模型方法去度量其面对的非交易性金融资产信贷组合,如针对银行贷款和非公开性私募债券的价值和风险进行度量和分析。最后提出在我国如何更好的加强社会各层面的信用风险管理,以至于减少商业银行及其他金融结构等信用风险,并对此提出一些政策和建议。文章主要采用实证分析和数理模型分析的研究方法。主要基于金融学、信用风险管理、数理金融学等方面的相关理论知识和模型工具。通过对我国商业银行信用风险管理现状的分析,利用CreditMetrics模型及其他相关数理模型辅助工具,如蒙特卡洛模拟方法、Cholesky分解方法、Nelson-Siegel方法以及VaR方法和ES方法。在假设的非交易性金融信贷资产组合头寸的信息下,来实证分析如果信贷质量发生变化而导致2013年信用评级变化的信用风险水平,并结合目前我国当前各层面存在的信用风险管理问题提出相应的建议。在信用风险度量的实证部分,基于我国十二家上市公司相关数据组成的信贷资产组合,应用CreditMetrics模型进行该信贷资产组合的信用风险水平度量。但是,由于贷款不能公开交易信息,无法准确获得贷款市值及其一年内贷款价值的波动大小。对此,可以通过利用该债务公司的一些其他公开信息来估计其贷款市值。这些需要搜集的信息包括：选取的12家债务公司2012年的历史股价,债务公司主体长期信用等级及其资产回收率,信用风险转移矩阵和债券市场上的各信用评级的公司债券信息。 CreditMetrics模型最先由J.P.摩根同其他合伙人于1997年提出,用于计量贷款组合信用风险的新型内控模型,考虑资产价值随经济的时间变动和信用等级的变化引起的资产总价值变动。该模型基本思想是通过收集该债务公司的其他公开信息,并基于债务公司在一定期限内(通常是1年)的某个市场风险因子的变动情况,研究下一年因受违约事件或债务公司信用质量导致的信用等级转移、降级、升级,从而影响资产组合的价值,以此计量该资产组合在第2年期末的市场价值。进一步根据期末损失分布,求出一定置信水平下信贷资产组合可能发生的最大价值损失。 CreditMetrics模型对于信贷组合信用风险的度量应用需要一些理论假设条件,这些假设包括：信贷资产的股票收益率呈标准正态分布,并服从标准的几何布朗运动。不良贷款回收率与违约率不相关。非交易性金融资产组合中各笔贷款头寸在研究期间保持不变。论文具体研究了以下几个问题：第一,我国商业银行信用风险的现状和存在的问题,并简要回顾了我国已经对此做出的管理实践。由于我国银行业还没有建立起先进的信用风险资料的历史数据库,数据库存在问题,导致信用风险的计量难度较大。商业银行在长期经营中也暴露出大量信用风险,仅从财务报表中就可以体现出商业银行的不良贷款数额巨大而且在急剧上升,贷款呆帐准备金比率较低,资产负债比例偏高,贷款量略大。在商业银行的业务中,经营监管各环节仍能反应出信用风险管理的缺失。鉴于国际银行业频繁的发生动荡并出现危机,国际监管机构不断发布并出台一些法律法规,以此约束并规范银行业的信用风险以及内部控制。按照《巴塞尔新资本规定》的要求,我国一些商业银行开始进行内部控制制度建设。在我国,中小企业作为特殊群体,近些年迅速发展。在促进我国经济繁荣以及多元化发展的过程中,对中小企业发展的风险管理也采取了一系列的保护方法。第二,基于CreditMetrics模型的信用风险度量实证分析与研究过程。在信用风险度量的实证部分,在我国上市公司和资本市场中选取数据,建立了信贷风险组合的度量框架。通过计算债务公司的历史(2012年)股价收益率,利用蒙特卡洛方法模拟出与上年相关水平一致的收益率。利用穆迪公司发布的信用风险转移矩阵,计算出由于信用评级变化对应的不同收益率临界值。通过Nelson-Siegel方法计算资本市场不同信用级别的债券收益率,从而得到2013年底信贷资产评级变化后的资产价值。再经过搜集债务公司主体长期信用等级和对应的资产回收率,计算VaR值和ES值来度量信用风险水平。实证结果认为CreditMetrics模型对于研究我国信贷资产组合的信用风险水平的度量可行,但是仍然需要进一步改进。第三,结合当前我国各层面存在的信用风险管理问题提出相应的建议。评级机构应尽快建立信用风险数据库,积极研究发明适合我国的内部评级模型和体系,建立全面可靠的信用风险数据库,并统计出适合国内使用的风险转移矩阵、资本回收率和债券远期贴现率等。本文采用国际权威并流行的度量信用风险水平的CreditMetrics模型,由于历史数据的搜集需要较强的数据库支持,因此这种模型目前在我国只被少数大型银行使用,还未在国内广泛推广。笔者为此度量方法做尝试,并验证易于计算并可行。本文可能存在的创新内容有：第一,文章对于信贷资产组合之间的相关系数用Cholesky方法分解,将每对债务公司之间的相关特征用Cholesky分解矩阵来表示。利用该分解矩阵将第一年债务公司之间的相关特征转移到模拟的第二年(2013年)债务公司收益率的时间序列中,从而对模拟的2013年信贷资产的收益率进行调整。相关性水平利用该上市公司的股价收益率相关性代替资产总价值的相关性,与经济变动密切相联系,时刻反应经济的市场变动,具有较强的时效性。第二,在计算不同信用评级的信贷资产远期利率水平时,没有采用传统方法的零息票国库券利率去贴现信贷资产的现金流,而是采用Nelson-Siegel模型,这样可以针对不同信用评级的信贷资产分别计算其价值。为了使本文研究国内信贷资产更有针对性,所以样本数据选取的国内各信用评级的公司债券,以此来计算国内公司一年后变换到其他评级的价值。第三,在计算信用风险水平时,本文不仅采用了VaR值的计算,还计算了ES值。考虑了所有信贷资产损失超过VaR值的小概率事件,对超过VaR值的所有信贷资产的损失值同样重视。在运用CreditMetrics实证研究过程中发现,该模型度量风险的精确性高度依赖于信用评级转移矩阵和资本市场的债券远期利率的准确性,所以评级机构应尽快建立相关数据库并对此进行统计计算。计算结果有可能低估了《巴塞尔新资本协议》中经济资本的8%要求,因为度量过程中的置信水平选取较低,而且债券的到期日都在三年以上,而本文只研究了债券信贷资产发行一年后的信用风险水平。而且本文认为CreditMetrics模型的损失分布函数有可能存在一个厚尾分布,今后可以基于此进一步考虑CreditMetrics模型的实证计算。本文的CreditMetrics模型使用股价市值,计算结果具有客观性和前瞻预期性,贷款信息紧跟市场变动而变动。而且不仅考虑了贷款违约的风险,也考虑了信贷资产质量变化的风险。不仅可以用来度量信贷组合的信用风险,也可度量单一贷款的信用风险。在度量信用风险时,不仅利用VaR值来表示,而且还用ES值对VaR值进行补充。指出银行业的信用风险是我国商业银行面临的主要风险之一,当前我国国内银行业要依据《巴塞尔新资本协议》的要求,完善我国银行业的体制改革,制定前瞻性的银行发展策略,从而引导银行的改革与建设。
[Abstract]:Since the development of financial liberalization and the relaxation of financial regulation in the 1970s and the 1970s , the rapid development of the business of financial institutions and the rapid emergence of financial derivatives . The risks associated with the financial industry have been continuously exposed to the global financial crisis , such as the Asian financial crisis , the European monetary crisis and the US subprime crisis . Especially in the US subprime crisis , the former Federal Reserve Chairman , Greenspan , is regarded as a global financial storm of " one hundred years . " However , it is the root cause of the financial crisis , which is mainly the control of credit risk management .

The financial crisis triggered by financial innovation has caused the world ' s economy to continue to suffer . For this reason , regulators in the international financial sector have given greater attention to risk regulation . The Basel 2 Basel Capital Accord , issued in September 2010 , is more concerned about the control of banks ' risks , and the proposed standards of capital and liquidity supervision are more stringent than before .

In view of the present research situation in China , the credit rating system of China ' s banking industry can only meet the requirement of the basic internal rating method , but it is far from the requirement of the advanced internal rating method , but it can ' t meet the requirement . The bank ' s banking supervision is not perfect , the credit risk management level is backward , the technology is not developed , the system is not perfect , and the credit risk is not quantified .

Therefore , it is of great significance to learn from and study the relevant advanced technology of credit risk management in China ' s banking industry . It is of great significance to strengthen the anti - risk ability and credit risk management in China ' s banking industry in the future .

Based on the empirical analysis and application of the Credit Metrics model , this paper analyzes the current situation and existing problems of the credit risk of commercial banks in China , and measures and analyzes the value and risk of the bank loans and non - public private equity bonds . Finally , it puts forward some policies and suggestions on how to strengthen the credit risk management at all levels in our country so as to reduce the credit risk of commercial banks and other financial structures .

Based on the analysis of the current situation of credit risk management in China ' s commercial banks , this paper makes an empirical analysis on the credit risk level of credit rating change in 2013 based on the analysis of the current situation of credit risk management in China ' s commercial banks , such as Monte Carlo simulation method , cholesky decomposition method , Nelson - Siegel method , VaR method and ES method .

In the empirical part of the credit risk measure , based on the credit assets combination of twelve listed companies in our country , the credit risk level measure of the credit asset portfolio is made based on the Credit Metrics model . However , due to the fact that the loan cannot disclose the transaction information , it is impossible to accurately obtain the loan market value and the fluctuation size of the loan value within one year . The information that needs to be collected includes the historical stock price of the 12 debt companies selected in 2012 , the long - term credit rating of the debt company principal and its asset recovery rate , the credit risk transfer matrix and the corporate bond information of each credit rating on the bond market .

Credit Metrics Model is first proposed by J . P . Morgan and other partners in 1997 . It is used to measure the credit risk of loan portfolio . The basic idea of this model is to study the value of the asset portfolio at the end of the second year by collecting other public information of the debt company and based on the change of the credit rating caused by the credit quality of the debt company .

The Credit Metrics model requires some theoretical assumptions for the measurement of credit portfolio credit risk . These assumptions include : the stock yield of credit assets is standard normal distribution and follows the standard geometric Brownian motion . The non - performing loan recovery rate is not related to the default rate .

In this paper , the following problems are studied :

First , the present situation and existing problems of credit risk in commercial banks in China are briefly reviewed , and the management practice has been briefly reviewed .

Because China ' s banking has not established the advanced historical database of credit risk information , there is a problem in the database , which leads to the great difficulty of credit risk measurement . In the long run , commercial banks also exposed a large amount of credit risk . Only from the financial statements , the amount of non - performing loans of commercial banks is huge and the loan amount is slightly larger . In the business of commercial banks , the management and supervision links can still reflect the lack of credit risk management .

In view of the frequent turbulence and crisis in the international banking industry , the international regulatory agencies have issued and issued some laws and regulations to restrict and regulate the credit risk and internal control of the banking industry . In accordance with the requirements of the Basel 2 new capital requirement , some commercial banks in China have started the internal control system construction . In our country , small and medium - sized enterprises are developing rapidly in recent years . In the process of promoting the economic prosperity and diversification of our country , the risk management of the development of small and medium - sized enterprises has also taken a series of protection methods .

Secondly , based on the Credit Metrics model , the empirical analysis and research process of credit risk measurement .

In the empirical part of credit risk measurement , the data is selected in China ' s listed company and capital market , and the measure frame of credit risk combination is established . By calculating the yield of stock price in the history of the debt company ( 2012 ) , the yield of different yield corresponding to the previous year is simulated by Monte Carlo method . By means of the credit risk transfer matrix issued by Moody ' s Company , the asset value after the credit rating change has been calculated . Through the collection of the long - term credit rating and the corresponding asset recovery rate of the debt company , the VaR and ES value are calculated to measure the credit risk level .

The empirical results show that the Credit Metrics model is feasible to study the credit risk level of our country ' s credit portfolio , but we still need to improve further .

Thirdly , the paper puts forward some suggestions on the management of credit risk in all aspects of our country .

The rating agencies should establish the credit risk database as soon as possible , actively study the internal rating models and systems suitable for our country , establish a fully reliable credit risk database , and calculate the risk transfer matrix , capital recovery rate and long - term discount rate suitable for domestic use .

In this paper , based on the Credit Metrics model of the international authoritative and popular measure credit risk level , since the collection of historical data requires stronger database support , this model is only used by a small number of large banks in our country , and has not been widely promoted in China . The author attempts to do this and verifies that it is easy to calculate and feasible . The possible innovations in this paper are as follows :

First , the correlation coefficient between the credit asset portfolio is decomposed by the cholesky decomposition matrix , and the correlation between each pair of debt companies is represented by the cholesky decomposition matrix . By using the decomposition matrix , the correlation characteristics between the first year debt companies are transferred to the simulated second year ( 2013 ) debt company yield time series , so as to adjust the yield of the simulated 2013 credit assets .

Secondly , when calculating the forward interest rate level of credit assets with different credit rating , we do not adopt the traditional method ' s zero coupon treasury bond interest rate to discount the cash flow of the credit assets , but adopt the Nelson - Siegel model so that the value of the credit assets with different credit rating can be calculated respectively . In order to make the research study domestic credit assets more targeted , the domestic credit rating of the sample data is selected as the corporate bond of each credit rating , so as to calculate the value of the domestic company after one year after conversion to other ratings .

Thirdly , when calculating credit risk level , this paper not only adopts VaR calculation , but also calculates ES value . Considering the small probability event that all credit asset losses exceed VaR value , the loss value of all credit assets that exceed VaR value is also paid .

It is found that the accuracy of the model measure risk depends on the accuracy of the long - term interest rate of the credit rating transfer matrix and the capital market , so the rating agency should establish the relevant database as soon as possible and calculate the risk .

It is possible to underestimate the 8 % requirement of economic capital in Basel 2 Capital Accord , because the confidence level in the measurement process is lower , and the maturity of the bond is more than three years , and the credit risk level after one year of the issuance of the bond credit asset is studied .

The credit risk is one of the main risks faced by commercial banks in China . It is pointed out that the credit risk of banking is one of the main risks faced by commercial banks in China .

【学位授予单位】：西南财经大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F832.4;F224

【参考文献】