1、 1 外文翻译 原文 Modelling non-performing loans probability in the commercial banking system: efficiency and effectiveness related to credit risk in Italy Material Source: Springer-Verlag 2010 Author: Bernardo Maggi Marco Guida Abstract In this article, we model the effect of the non-performing loans on t
2、he cost structure of the commercial banking system. With this aim, we comment on an increase in the non-performing loans by studying the consequences of such a change on the cost function and compute the probability of failure of maintaining a performing loan as such. In doing so, we are convinced t
3、hat geography does matter and evaluate the risk propensity of the bank towards the non-performing loans accordingly. We nally stress that traditional efciency indicators of cost elasticity do not t properly with such a problem and propose a measure based on the costs for managing and monitor- ing th
4、e loans which, according to the related density function, will reveal effectively as non performing. Introduction This article deals with the problem of loans that do not perform (NPL from now on). Though such a problem is dated back in the literature on banking systems, little space has been given
5、to it for the difculty both to obtain the necessary data and to embed it appropriately in the methodology till now developed on the bank efciency. As for the former, acknowledged data bases reveal the absence of information on the most important variable for this study, i.e. the bad loans. As for th
6、e latter, the question is how to account for the bad loans in a specic function to obtain reasonable results from the efciency point of view. The puzzle is that, as argued in Berger and De Young (1997), there exists a simultaneity between NPL and inefciency but the link represented by bad loans is t
7、o be excluded from the explicative variables of the function used to avoid diminishing the inefciency measured in the regression residuals, in the case of bad loans not due to internal factors. In this respect, the attempts to address the problem,such as that of Hughes 2 and Mester (1993), though pi
8、oneering, fail in that NPL are included as a whole in cost function without distinguishing between bad loans due to bad management and those due to external factors which, not depending on the bank, should not be counted in the inefciency. As a consequence, only the former effect should be included
9、in the cost function specication. From their side, Berger and De Young (1997) focused on this problem but without modeling the NPL in the cost function, rather they use the cost function only to compute the inefciency and then leaving to a causality analysis the job of explaining their relationship.
10、 However,the conclusions from this method, though correct, are not exhaustive in that: rst state only gross statistical association and secondly are general, i.e. are not referable to a specic bank. Only Pastor (1999, 2002) and Pastor and Serrano (2005) face frontally the problem and distinguish bet
11、ween NPL deriving from internal and external causes,calculating efciency accordingly. However, Pastor (1999, 2002) uses a non parametric approach with the caveats implied for the deterministic techniques, that do not allow inference on the functions for the bank management. He uses a DEA analysis at
12、 several stages to evaluate the composition of problem loans in conjunction with the two mentioned causes pictured by apposite variables. Pastor and Serrano (2005)for rst used a parametric approach. They skip the problem of simultaneity recurring to ad hoc instrumental variables to represent the int
13、ernal and external causes of NPL and then inefciency. As a consequence, the evaluation of the cost of the non-performing loans, as well as the structure of the cost function, is strongly affected by the discretion both in the choice of such instruments and in the use of their possible combinations.
14、Then, the two authors do not enter in the cost structure questions and jump straightly to the use of the cost and prot function as a frontier for the measurement of the efciency. Other recent researches started studying this problem but with particular reference to both developing countries and emer
15、gent economies (Hauner and Peiris(2005); Matthewes et al. (2007), for the implications due to the insolvency of the banking sector on the economic growth and the performance of the banks, respectively. Such studies consist either in netting the loans of the non performing part or in considering NPL
16、as an exogenous variable useful to pick downward shifts in the profits. We go more into depth and propose a theoretical and empirical analysis to place the problemof the non-performing loans in the cost efciency framework.We develop an analysis based on the probability on NPL to occur, governed by s
17、ome parameters representing the structural and local factors (both exogenous) that concern our problemwhile the internal factors depend 3 on the aimof the bank of capturing loans. In such a way our methodology is easily interpretable and more quickly applicable given that the effects of the exogenou
18、s factors are represented by the just mentioned parameters.Our intention is to gauge how much of the efforts in managing and monitoring loans succeed in preserving the bank from the non-performing loans. In order to answer this question, we formalize a general methodology, independently of the conte
19、xtual economic situation of the country and capable to capture the linkage between NPL and cost structure of the commercial banking system. We refer to the widest European bank and, specifically, to the Italian case where problem loans did matter. We take into consideration the marginal effect of th
20、e non-performing loans on the operating costs by using and estimating a transformation function, that connects loans with NPL. By exploiting this indirect relationship we also show a possible way out of the dilemma of the simultaneous estimation between the cost function and NPL which,though logic i
21、n principle, would alter the analysis with an overestimated efciency for the additional variable (NPL) in the cost function (Berger and De Young (1997). We represent the link between loans and non-performing loans by stressing the role of the geographical position of the bank.1 Such a choice is also
22、 motivated by the importance that the economic specicities, associated to the geographical localization of the loans, experimented especially by the US, played in the banking sector recent international nancial crises of 2007 and 2008. In doing so, we evaluate and use the density function of a loan
23、to become non performing and another measure of probability that connects loans with the more general concept of uncertain loans.We then study where, on average, a bank is positioned with respect to the maximum sustainable level of NPL thus evaluating the propensity of the bank to protect itself fro
24、mthe credit risk. Finally, we conclude by proposing an index that summarizes the issues stemming from the question we moved from of efciency and effectiveness,related, respectively, to the management of the loans and to the capacity to prevent non-performing loans, according to the mentioned probabi
25、lity. Alternative ways to measure the effect of NPL may be found in the efciency lit-erature of the undesired outputs. In such a context, the problem becomes to nd the correspondent NPL shadow price following Pittman (1983) and Fare et al. (1993). The two authors, propose a deterministic approach so
26、lving a maximization problem for, respectively, a prot and a revenue function dened in terms of the distance function. In the rst case, shadow prices derive from apposite constraints 4 representing the regulation, and in the second case shadow prices derive directly from the maximi- zation problem e
27、xploiting Shephards dual lemma and assuming that, for marketable outputs, market and shadow prices are the same. Our choice fell on the indirect cost function approach for preference of the parametric approach as argued above and for the additional information from the NPL probability. This article
28、is organized as follows. In the second section, we present the basic theoretical relationships to which we refer for the estimation. In the third section, we reason on the variables to include in the cost function in connection with the scope of our analysis and describe our data set. In the fourth
29、section,we present the econometric analysis of such a function according to the literature of the exible functional forms. In the fth section, where the focus is centered on the problemof the NPL, we estimate their trade-off with loans and evaluate the attitude to the credit risk of the bank. The si
30、xth section reports the nal calculation of the effect of NPL on the marginal cost and some other efciency and effectiveness indicators. The seventh section concludes. Conclusions and further remarks In this study, we move from the consideration of the lack, in the literature of efciency in the banki
31、ng sector, of an accurate dealing with the NPL. We formalize a model for the placement of the NPL in the cost function thus trying to ll somehow the hiatus in this respect and giving further potential to the tools so far available.We answer the question of how much of the efforts in terms of costs f
32、or managing and monitoring loans succeed in preserving the bank from the non-performing loans i.e. we measure the change in the cost function due to new incoming L and select, proportionally, the expenses associated with the generated NPL. Our framework is based on the consideration that NPL is not
33、directly an item of the cost function but can be viewed as a transformation of the loans that stands for the banks trade-off between these two. From this representation we derive the density function of a loan to become non performing which gives the possibility to study how, on average, the bank is
34、 positioned with respect to the maximum sustainable level of NPL and to evaluate the propensity of the bank to protect himself from the credit risk. This is done by assigning a particular role to the geographical aspects that are found to be relevant both in the transformation and cost function esti
35、mation. This allows for the possibility to make considerations on the efciency and effectiveness of the costs management referred to both loans and non-performing loans and, consequently, to control costs with reference to the 5 regions considered.We nd that traditional output-elasticity indicators
36、do not t properly with the problem of NPL and propose more representative measures. We nd that the effect of a change in the probability of an uncertain loan to become non performing is extremely costly for the banking system thus further encouraging the research in this eld with an analysis that, f
37、rom the one hand, disentangles the specic categories of loans and their degree of risk and, from the other hand, could test the effectiveness of such credit policies. 译文 商业银行系统的不良贷款率模型：意大利效率性和效益性有关的信贷风险 资料来源：施普林格出版社 2010年 作者：贝尔纳多马贵大 马吉 摘要 在本文中，我们模拟不良贷款对商业银行系统成本结构的影响。本着这一目标，我们通过研究成本函数的变化结果评论不良贷款的增加 ，
38、并 计算出维持这样的一个不良贷款的失效概率。 如此一来 ，我们认为区域因素对银行不良贷款的风险倾向有一定的影响。最后，我们强调的是，传统的成本弹性效益指标不是很适合这个问题，基于管理和监测的贷款的成本提出措施， 并 根据相关的密度函数，有效地 揭示不良贷款。 引言 这篇文章主要处理不良贷款问题。 虽然这个问题在银行系统文献中有记载，但直到银行业 高效 发展的今天，由于我们在获取必要的数据并 采取 适 当的方法方面还存在一定的困难，因此还是一片空白。 至于前者，确认的数据库显示了本研究最重要的变量信息缺失，即不良贷款。而对于后者，问题是如何解释在一个特定函数下的不良贷款，从效率的角度以获取合理的
39、结果。 让人不解的是，在伯杰和代扬（ 1997）认为：在不良贷款和低效率之间存在同时性，但以不良贷款为代表的联系同时性将被排除在函数解释性变量之外，在不良贷款 形成原因 不是内部因素的情况 下 ，该函数用来避免减少回归误差测量的低效。 在这一领域 ， 许多学者都做过研究， Hughes 和 Mester(1993)也曾 作 过研究 ， 但他们没有 将 由管理不善 引起的和 由 外部因素 引起的不良贷款 做出区别， 而是将他们作6 为一个整体在成本函数中计算， 这些外部因素，不 是 银行 自身问题引起的 ，不应计算在内。因此，只有前者 的影响 应包括在成本函数中。 依他们看来，伯杰和代扬（ 19
40、97） 也 关注过此问题，但没有在成本函数 模型中考虑 不良贷款， 他们 运用成本函数只是为了计算低效和 利用 因果分析解释他们之间的关系。 然而，这种 方法的结论，虽然正确， 但 并不详尽：第一 ， 只 有 总的数据联系 ； 第二 ，只是 一般概括性 的，没有参考 具体 的 某家 银行。只有 Pastor (1999, 2002)和 Pastor and Serrano (2005)正视了此问题， 区分了不良贷款产生的 内 因和外因 ， 并据此计算相应的效率。 其中 ， Pastor (1999, 2002)使用 了一种具有确定性的非参数方法， 其不允许参考银行管理的函数。他在几个阶段使用数
41、据包络分析来评价问题贷款的构成 ， 以及由两个适当变量描述两个提及的因素。 Pastor and Serrano (2005)第一次使用参数方法。他们跳过同时性问题，循环的特设工具变量来表示不良贷款的内外因以及低效。因此，不良贷款成本的评价以及成本函数的结构，其深受工具的选择以及他们可能的 组合使用的判断力的影响。那么，两个作者没有涉及成本函数问题而直接使用成本和利润函数来对效率的测量。其他最近的研究开始研究此问题但偏向于参考发展中国家以及新兴经济体 (Hauner 和 Peiris(2005); Matthewes et al. (2007)作为参考 -由于银行的破产 -分别在经济增长和银行
42、的性能上。此研究包括不良贷款部分的静赚资产或考虑不良资产作为外部因素，该因素在利润上对重整下滑趋势有利。我们深入剖析以及提出理论研究与实证分析来提出在成本框架上不良贷款的问题。基于要发生的不良贷款可能性的分析， 我们提出分析，该分析由一些参数支配，这些参数代表结构和当地的因素（都是外部的）和我们的问题相关同时内部因素决定银行截获贷款的目的。这样，我们的方法容易解释且更加便捷适用，只要外部因素的影响有以上提及的参数来表示。我们的目的是测量管理和监控贷款付出的努力，成功保护贷款免受不良贷款。为了回答这个问题，我们正式提出一种普遍的方法，独立于国家的情景经济环境并且能够捕获不良贷款和商业融资系统成本
43、函数的链接。我们参考西方欧洲的银行，特别是意大利的例子，那里的问题贷款的确是个问题。我们通过使用和评价变形函数。考虑不良贷款 边际效应对操作成本的影响。该函数链接贷款和不良贷款。通过使用这个非直接的关系我们也可以看到在成本函数和不良贷款之间走出同时估算的困境的可能性。成本函数和不良贷款，虽然原则上有逻辑，对于成本函数 (Berger and De Young (1997)中附加的变量（不良贷款），将因为对效率的过高评价而改变分析。我们。通过强调银行区域位置，在贷款和不良贷款之间使用链接。此举也是由经济规范，与区域位置相关的贷款，在 2007年和 2008年美国特别的经历过扮演银行业的国际金融危
44、机的重要性所推动的。在这种情况下，我们评 价和使用贷款的密度函数成为不良贷款以及其他的可能性的测量，将贷款和不确定贷款的宏观概念连接。7 我们然后，平均地，有关不良贷款的最大可持续水平定位为银行，评价银行保护使其免受信用危机倾向。最后，我们提出指数 -该参数总结问题，来源于我们提及的 -从有关效率和效果到贷款的管理以及到阻止不良贷款能力，根据提及的可能性。 我们可能会在没有公开的文献中 发现 一些 方法来测量 模型因素对 不良资产的影响。在这种情况下，按照 Pittman (1983) 和 Fare et al. (1993)，问题是发现客户的不良资产的影子价格。两个 作者，各自提出解决最大化
45、问题的确定性方法，利润和收益函数规定了函数功能。在第一种情况下，影子价格来源于代表条例的适当的限制。且在第二种情况下，影子价格直接来源于利用 Shephard双重引理的最大化问题并且认为，对于市场化的输出，市场和影子价格是一样的。我们的选择落到用于参考以上提出的参数方法以及来源于不良贷款可能性的附加信息的非直接成本函数方法上。 本文章 结构主要 如下。在第二部分，我们呈现基本理论关系，该关系我们喜欢评估。在第三阶段，我们对变量作出推理 ， 包括成本函数 中的 分析范围 和描述的数据集合。在第四阶段 ，根据灵活函数的形式我们提出像此函数的计量经济分析。在第五阶段，当焦点落在不良贷款的问题的地方，
46、我们评价贷款的比较和评价银行信用风险的态度。第 6部分报告不良贷款的最后计算对边际成本的影响以及一些其他的效率和效果的指数。第七部分得出结论。 结论 在 这项研究 中，银行 部门文献中 缺少 一种正确处理不良贷款的方法 。我们建立了一个关于 不良 贷款的成本 函数 模型 ， 并试图去 填补这 方面 的空白 ，为以后的研究提供依据 。 我们回答了要成功维护银行不良贷款在管理和监控上的成本所需付出的努力，即根据吸纳新贷款所需的成本费用和产生不良贷款的费用之间按比例对成本函数的影响进行测算 。 我们的 模型 考虑 到 不良贷款不是成本函数的 直接影响因素， 但 它 可以视为代表 上述 两者 之间 的 权衡 。 通过这种表示，我们 从 贷款的密度函数衍生 出 不良贷款，它 给出了 有关不良贷款最大可持续 和评 估 银行免受信用危机倾向的可能性。通过 在 特定 区域操作 ，发现 地理位置的不同和 成本函数 的 评估 有一定的关联 。 这使有关贷款和不良贷款的成本管理的效率和效果成为可能， 从而 控制有关区域的成本。 我们发现传统的产出 弹性 指标不适宜不良贷款问题 ， 而提出更加有代表性的措施。 我们发现不确定贷款成为不良贷款对于银行系统来说成本很 昂贵。因此，进一步鼓励在这一领域的研究，一方面 ， 解决贷款的详细分类以及他们风险的程度。从另一方面，可以检验此信用政策的效果。
