金融学外文翻译西方商业银行信用风险的度量.doc

外文原文 Tolerance of the Credit Risks ofCommercial BankCommercial bank main model and method including tradition credit risks, credit of management measure method and based on VAR modern credit risks measure model. Among them, traditional credit risks measure method include expert system camphor tree law , credit point system and neural network model, and based on VAR modern credit risks measure model include KMV model , surtax model , model of Credit-Metrics and credit risks of CSFP. Expert system model law before one year, financial institution is it analyze or determine the nature analytical method not to weigh credit risks, enterprise of loan subjectively to rely on mainly, this kind of method is called the expert system models. Such as classical 6C law - By the morality about experts foundation debtor (character), ability (capacity), capital (capital), pledge (collateral ) , management environment (condition ) and continuity (Continuity ) of undertaking six factor evaluate creditworthiness and comprehensive refund ability their, determine whether to grant the loan finally or not. Whether classical credit is it stores some defects in to analyze, mainly showing in the following aspects. Risk that the human factor brings. Following the serious bureaucratic style of work.Credit point system to reflect debtor economic situation or influence debtor several indexes (such as the financial rate of the borrowing enterprise), credit of state entrust to certain weight, receive and reflect credit the dividing value or the value of probability in breach synthetically of the credit state through some specific methods, and is it is it pay loan grant and loan fixed price to determine to come compared with basic value it. Whether Z value model propose by Altman, adopt five financial indexes (5 financial index these on 1968. Xl =Net working capital / total assets, X2 =Retain the incomes / total assets, X3 =EBIT / total assets, X4 =Benefit market value / the book value of the debt, X5 =The income from sales / total assets) calculate the weighting, implement credit to mark to the borrowing enterprise, compare total points with critical value, distinguish the bankrupt company and not go bankrupt in the company, will not grant the loan to the bankrupt enterprise. 1977, Altman, Hardeman, Mahayana is it set up categorized accuracy higher ZETA model to expand to go on to primitive Z value model. 1995, as to private company, Altman revise to Z value model, counting by Z value model. At present, the main defect of this kind of model is lacking the essential historical record material.The neural network model roughly imitates the thought process of human brain and artificial intelligence system of the learning method. Neural algorithm of network whether one group input, carry on mathematics is it produce through transfer function one export to change and then. Foreign Altman and Vrettos, Coats and Fans, etc all try to use this law, receive certain result. Someone uses such methods as the neural network, etc. to carry on the appraisal of risk of credit to the commercial bank too in our country. But neural model heavy shortcoming that their randomness of working is relatively strong most, and need to debug artificially. Consume a large amount of manpower and time.The nerve network model imitates the persons artificial intelligence system of the brain thinking process and the study method mostly. The calculate way of the nerve network is an importation, and then pass to convert the function to carry on mathematics conversion to produce an exportation of. Altman and Vrettos, Coates and Fans etc.excesses of the abroad all tried to make use of this method, being subjected to the certain result. Someone applies also the nerve network etc. the method carries on the credit risk evaluation to the commercial bank. However, the biggest weakness of the nerve model is the random that it works stronger, and need the artificial to adjust to try. Waste a great deal of manpower and times. The Model of the KMV is the estimate to the borrow funds the business enterprise default all rate of method. First, it makes use of the Black- Stoles three option list price formula, the according to motion of the market value, the property value of the business enterprise property, expire time, the calm insurance to borrow interest rate and owe debt faces to be worth to estimate a market value of business enterprise ownership of a share and it undulates sex, compute a default implement( the Default Exercise Point, for business enterprise a year following the value of short- term obligation plus dont long- term obligation of pure face the value of half) of company according to the liabilities of the company again, then the calculation borrow funds the persons default distance, the end is apart from to break contract the rate( EDF) its with expectation according to default of the business enterprise of beg an expectation default rate of business enterprise towards should relate to.The model of Credit-Metdcs was credit risk that J P root develops in 1997 to calculate model. It is an establishment on the usefulness foundation of reputation rating system of. The usefulness of reputation rating system means that investment in enterprise failure, profits descend, margin outlet dried up etc. Reputation affairs appear towards fulfill contract influence of ability can in time and fittingly through variety body of its reputation grade. Because of the occurrence of reputation affairs, will produce influence to reputation grade of business enterprise, the market value of its reputation tool also takes place homologous variety by all means. That model is an angle that combined from the property, not angle of the one property to treat the credit risk. It passes to contrast the combination to contribute (the average limit risk contribute= augmentative risk of combination because of increase one some reputation tool, the market of reputations tool value) in the limit risk of each reputation tool, the reputation grade of a reputation tool, expose degree with related coefficient and its risks of other properties etc. end turn basis for the quantity that the letter loan of investor makes policy to provide science.The CSFP credit risk affixture calculates the Credit-Metrics dissimilarity that the model and conduct and actions stare at the city model( MTM), it is a break contract model( DM), it is not the rise and fall of the reputation rating and change with this related reputation excess fare to see as a part of VAR( credit risk) for lend money, but see do only is the market risk, it at any period consider to break contract and dont break contract these two kinds of affairs appearances only, loss that calculate to expect and did not expect, but be unlike in the Credit- Metrics value that generous character expect and the value variety that did not expect. In the CSFP credit risk affixture calculate model, break contract all the rate is no longer long-lost, but was certainly change in to have by the model all the rate distribute of change quantity continuously. The Each loan was see to do small all the rate breaks contract the affairs, and the default of each loan all rates are all independent the in other loans, thus, the lending money the combination the default all the rate distributes to near the loosen to distribute.CSFP credit risk affixture calculates the model consideration break contract all the different from lose indetermination of the size, calculate all rate and lose the size and can get ally of indetermination of the rate the segment distributes lose, adding to all losses of segment of total for lend money the combination of loss distribute.Above-mentioned three differentiations of the models can induce for six aspects of the following: First, definition aspect in the risk, the model of Credit-Metrics belongs to the model of MTM; The CSFP credit risk affixture calculates the model to belong to the model of DM; And model of KMV since can be consider as the model of MTM, also can be consider as the model of DM.Second, drive the factor aspect in the risk, in model of KMV and Credit Metrics, the risk drives the factor is business enterprise property value and it undulates sex; But in the CSFP credit risk affixture calculate model, the risk of key drives factor is a variable default rate to all be worth in the economy. Third, the motion aspect in the reputation affairs, in the Credit Metrics, break contract all the rate was change into by the model according to the history data of fixed, or long-lost value; But different from CSFP credit risk affixture calculate in the model, default all rate is variable, but obedience in the model of KMV all the rate distribute. Fourth, relativity aspect in reputation affairs, each model has different relativity structure, the model of KMV and Credit Metrics are to change quantity more positive; But the CSFP credit risk affixture calculate the model to suppose or break contract relativity of the rate with expectation independently. Fifth, at recovery rate aspect, in the simple form of the model of KMV, recovery rate is a constant; In the CSFP credit the risk affixture calculate model, losing serious degree was gather together integral combines demarcation as different segment of recovery rate is the constant in the segment of; In the latest version of the model of KMV, recovery rate is random; In Credit Metrics and be willing to the tin model, the recovery rate also is random. Sixth, aspects of calculating the method, Credit Metrics to individual loan or loan the combination adoption analyzes the method to carry on calculate, combining to then adopt Monte Carlo to the large-scale loan emulation technique carries on calculate; the model of KMV and CSFP credit risk affixtures calculate the model adoption analysis method to carry on calculate.Ever since that time in 1996, the Basel agreement ruled to use for assurance that an internal model of capital of the risk misted be to take Vary as the basal model, the Vary became most popular risk to manage the model currently. Not only can carry on quantitys turn to the credit risk of debtor of the one, more important can carry on measure to concentrated risk of debtor of whole line and reputation of its related communities, but also can press profession, term, category.etc. to carry on quantity to turn measurement decomposition risk source according to the Vary credit risk generous character model. The risk degree that can contribute to manager to control trade and invest in time and accurately thus. In addition, the traditional credit risk of bank analysis mainly pays attention to loss that the obligation faces to make under default condition, but neglected loss that the debtor reputation natural intelligence the dynamic state variety cause, but thought and operations of the model have to certainly draw lessons from the meaning towards improving the commercial bank credit risk manage.中文译文西方商业银行信用风险的度量商业银行信用管理的主要模型和方法包括传统的信用风险度量方法和基于VAR的现代信用风险度量模型。其中，传统的信用风险度量方法又包括专家系统模型法、信用评分法和神经网络模型，而基于VAR的现代信用风险度量模型又包括KMV模型、Credit Metrics模型和CSFP信用风险附加模型。专家系统模型法是在20年前，金融机构主要依赖主观分析或定性分析方法衡量企业贷款的信用风险，这种方法称为专家系统模型。如经典的“6C法”由有关专家根据借款人的品德(character)、能力(capacity)、资本(capital)、抵押品(collateral)、经营环境(condition)和事业的连续性(Continuity)等六个因素评定其信用程度和综合还款能力，决定是否最终发放贷款。经典的信用分析存在着一些缺陷，主要表现在:人的因素所带来的风险。伴随着严重的官僚主义作风。信用评分法是将反映借款人经济状况或影响借款人信用状况的若干指标(如借款企业的财务比率)赋予一定权重，通过某些特定方法得到反映信用状况的信用综合分值或违约概率值，并将其与基准值相比来决定是否给予贷款发放以及贷款定价。Z值模型由Altman于1968年提出，采用五个财务指标(这5个财务指标是:Xl=营运资本/总资产，X2=留存收益/总资产，X3=EBIT/总资产，X4=权益市场值/债务的账面值，X5=销售收入/总资产)进行加权计算，对借款企业实施信用评分，并将总分与临界值比较，区分破产公司和非破产公司，对于破产企业将不发放贷款。1977年，Altman、Hardeman和Nahayana对原始的Z值模型进行了扩展建立了分类准确度较高的ZETA模型。1995年，对于非上市公司，Altman对Z值模型进行了修改，得到了Z计值模型。目前，这类模型的主要缺陷是缺乏必要的历史纪录材料。神经网络模型是大致模拟人脑思维过程及学习方法的人工智能系统。神经网络的算法是一组输入，再通过转换函数进行数学转换产生出一个输出。Altman和Vrettos、Coats和Fans等都尝试过运用此法，受到一定效果。有人也应用神经网络等方法对商业银行进行信用风险评价。然而，神经模型的最大缺点就是其工作的随机性较强，而且需要人工调试.耗费大量人力与时间。KMV模型是估计借款企业违约概率的方法。首先，它利用Black-Stoles三期权定价公式，根据企业资产的市场价值、资产价值的波动性、到期时间、无风险借贷利率及负债的账面价值估计出企业股权的市场价值及其波动性，再根据公司的负债计算出公司的违约实施(Default Exercise Point，为企业一年以下短期债务的价值加上未清偿长期债务账面价值的一半)，然后计算借款人的违约距离，最后根据企业的违约距离与预期违约率(EDF)之间的对应关系，求出企业的预期违约率。Credit-Metrics模型是JP摩根1997年开发的信用风险计量模型。其是建立在信用评级体系的有效性基础上的。信用评级体系的有效性是指企业投资失败、利润下降、融资渠道枯竭等信用事件对履约能力的影响都能及时恰当地通过其信用等级的变化体现出来。由于信用事件的发生，对企业的信用等级会产生影响，其信用工具的市场价值也必然发生相应的变化。该模型是从资产组合的角度，而不是单一资产的角度来看待信用风险的。它通过对比组合中各信用工具的边际风险贡献(平均的边际风险贡献=组合因增加某一信用工具而增加的风险，该信用工具的市场价值)，进而分析每种信用工具的信用等级、与其他资产的相关系数以及其风险暴露程度等各方面因素，可以看出各信用工具在整个组合的信用风险中的作用，最终为投资者的信贷决策提供科学的量化依据。CSFP信用风险附加计量模型与作为盯市模型(MTM)的Credit-Metrics不同，它是一个违约模型(DM)，它不把信用评级的升降和与此相关的信用价差变化视为一笔贷款的VAR(信用风险)的一部分，而只看作是市场风险，它在任何时期只考虑违约和不违约这两种事件状态，计量预期到和未预期到的损失，而不像在Credit-Metrics中度量预期到的价值和未预期到的价值变化。在CSFP信用风险附加计量模型中，违约概率不再是离散的，而被模型化为具有一定概率分布的连续变量。每一笔贷款被视做小概率违约事件，并且每笔贷款的违约概率都独立于其他贷款，这样，贷款组合违约概率的分布接近泊松分布。CSFP信用风险附加计量模型考虑违约概率的不确定性和损失大小的不确定性，并将损失的严重性和贷款的风险暴露数量划分频段，计量违约概率和损失大小可以得出不同频段损失的分布，对所有频段的损失加总即为贷款组合的损失分布。上述三模型的区别可归纳为以下六个方面:第一，在风险的界定方面，Credit-Metrics模型属于MTM模型；CSFP信用风险附加计量模型属于DM模型；而KMV模型既可被当作MTM模型，也可被当作DM模型。第二，在风险驱动因素方面，在KMV模型和Credit-Metrics中，风险驱动因素是企业资产价值及其波动性；而在CSFP信用风险附加计量模型中，关键的风险驱动因素是经济中可变的违约率均值。第三，在信用事件的波动性方面，在Credit-Metrics中，违约概率被模型化为基于历史数据的固定的、或离散的值；而在KMV模型和CSFP信用风险附加计量模型中，违约概率是可变的，但服从于不同的概率分布。第四，在信用事件的相关性方面，各模型具有不同的相关性结构，KMV模型和Credit-Metrics是多变量正态；而CSFP信用风险附加计量模型是独立假定或与预期违约率的相关性。第五，在回收率方面，在KMV模型的简单形式中，回收率是不变的常数；在CSFP信用风险附加计量模型中，损失的严重程度被凑成整数并划分为不同的频段，在频段内回收率是不变的；在KMV模型的最新版中，回收率是随机的；在Credit-Metrics麦肯锡模型中，回收率也是随机的。第六，在计量方法方面，Credit-Metrics对个别贷款或贷款组合采用分析方法进行计量，对大规模贷款组合则采用蒙地卡罗模拟技术进行计量;KMV模型和CSFP信用风险附加计量模型采用分析方法进行计量。自从1996年，巴塞尔协议规定用于确定风险的资本充足率内部模型必须是以VaR为基础的模型，VaR成为目前最为流行的风险管理模型。基于VaR的信用风险度量模型不但可以对单一的债务人的信用风险进行量化，更重要的是可以对全行的债务人及其相关群体的信用集中风险进行测度，而且可以按行业、期限、种类等进行量化测定分解风险来源。这样可以有助于管理人员及时、准确地掌握交易和投资的风险程度。此外，银行传统的信用风险分析主要注重违约条件下债务账面制的损失，而忽视了债务人信用资质动态变化所引起的损失，而模型的思想及操作对改进商业银行信用风险管理有一定借鉴意义。