信用评分系统的建模及其验证研究

《巴塞尔新资本协议》鼓励银行采用内部评级法评估信用风险,但要求所采用的评分系统必须提供建模过程说明与验证结果。本文按照这一要求,在采用ROC曲线、变量序别化转换等不同于以往国内研究的统计方法对信用评分系统进行建模的同时,还提供了相应的验证结果。实证研究结果表明,依据方法所建立的信用评分卡具有较好的预测力和稳定性,可实现对借款者信用等级和违约概率的快速有效评估,具有实际可操作性。     

基于Logistic模型的单个信用资产违约概率预测

在信用风险管理领域,由于中国商业银行信贷数据只满足Logistic模型的要求,因而预测单个信用资产违约率只能以Logistic模型为主线建模。以中国某商业银行1999~2005年的信贷数据为样本,实证分析得出,企业本身、宏观经济、地区及行业四方面因素对企业违约概率存在显著相关性。通过以上述四因素为变量,所构建的预测电力、公路、城镇建设三个行业信用资产违约概率的Logistic模型分析与预测单个信用资产违约的结果来看,四要素模型对中国商业银行的信用风险管理具有参照价值。     

引入非财务因素的中小企业信用风险预警模型实证研究

相当多的研究已证明财务因素能有效预测借款人违约风险,而非财务变量的预警能力仍不明确。为检验非财务变量预警能力,本文建立了三个假设,利用调查数据,建立了Logistic回归模型检验假设,结果表明：1．财务变量提前预警效果明显;2．加入非财务变量后,模型预警能力得到显著提高;3．预警模型分类效果显著优于随机分类。     

信用等级变动下企业破产概率的研究

本文建立了企业信用等级变动的模型。该模型包含了Yang(2003)考虑到的情形,也包含了违约情形。本文在假设信用等级转换遵循具有吸收状态的齐次马氏链前提下,推导了公司破产后赤字的分布的递推公式,并通过给出数值算例使得结果更加清晰。     

信用等级变动下企业破产概率的研究

本文建立了企业信用等级变动的模型。该模型包含了Yang(2003)考虑到的情形,也包含了违约情形。本文在假设信用等级转换遵循具有吸收状态的齐次马氏链前提下,推导了公司破产后赤字的分布的递推公式,并通过给出数值算例使得结果更加清晰。     

违约风险模型对违约定义的敏感性研究

本文根据贷款信用风险四级分类和五级分类定义三种违约,运用商业银行的贷款企业会计数据分析了Logistic违约风险模型对违约定义的敏感性。实证研究表明,在三种违约定义下,违约模型的结构相似,但模型选择的变量和变量的显著水平存在差异,违约模型对违约定义具有敏感性。     

如何利用CDS市场数据拟合不同存活率模型

信贷衍生工具评估的关键要素是违约时间模型。此类事件隐含的不确定性通过违约概率分布（也称为违约期限结构）模型获取,此模型对发行人在未来给定时间间隔内的违约概率进行建模。违约概率估算具有不同的类型：历史违约概率采用公司的信用等级转换或评分模型进行估算;结构模型综合市场以及资产负债表信息来计算违约概率;风险中性概率采用简化模型通过单纯的市场数据进行推导。     

内蒙古商业银行房地产贷款风险的实证研究

运用CPV模型探究宏观经济变量和房地产信贷违约率的关系,证明CPV模型可以预测商业银行房地产贷款的违约率.为减少商业银行风险损失,建议构建商业银行内部的风险管理衡量模型,完善房地产业的监管制度、改善行业风险评估体制,加强银行内部信贷管理制度、严控信贷发放流程,加强复合型风险管理人才的培养.     

德国公司违约概率预测及其对我国信用风险管理的启示

内部信用评级是新巴塞尔资本协议的核心,而违约概率的预测又是内部评级的基础。本文利用具有出色分类功能的非线性支持向量分类(SVC)方法来预测德国公司的违约概率,识别其信用风险。结果显示,SVC模型的预测能力优于基准的logit模型;而且非线性SVC模型能够捕捉线性logit模型所不能识别的影响信用风险的重要变量。本文虽然分析的是德国公司数据,但是同样对我国商业银行和公司构建全面风险管理体系有着直接的指导意义。     

基于Logistics回归算法的证券客户流失预测模型及应用

本文以Logistics回归算法为基础,结合证券行业经纪业务的特点,选择了总资产比例、交易次数比例、总资产连续下降等12个指标建立证券客户流失预测模型;并通过历史数据计算,确定了模型的常变量及回归变量。相关理论方法评估及真实数据验证表明,该模型的预测准确度较高,可以基于客户历史行为反应其流失倾向。一、客户流失预测的基础模型简介     

Exploring the components of credit risk in credit default swaps

In this paper, we test the influence of various fundamental variables on the pricing of credit default swaps. The theoretical determinants that are important for pricing credit default swaps include the risk-free rate, industry sector, credit rating, and liquidity factors. We suggest a linear regression model containing these different variables, especially focusing on liquidity factors. Unlike bond spreads which have been shown to be inversely related to liquidity (i.e., the greater the liquidity, the lower the spread), there is no a priori reason that the credit default swap spread should exhibit the same relationship. This is due to the economic characteristics of a credit default swap compared to a bond. Our empirical result shows that all the fundamental variables investigated have a significant effect on the credit default swap spread. Moreover, our findings suggest that credit default swaps that trade with greater liquidity have a wider credit default swap spread.     

A PSYCHOLOGICAL APPROACH TO MICROFINANCE CREDIT SCORING VIA A CLASSIFICATION AND REGRESSION TREE

Microfinance institutions' (MFIs') peculiar lending methodology is characterized by an unchallenged decision‐making predominance from the part of loan officers. Indeed, the latter are in charge of providing a great deal of diagnostic information regarding the entrepreneur's psychological traits likely to help them run a business. This paper constitutes an initial attempt towards exploring the role of borrowers' psychological traits in predicting future default occurrences. It builds on a nonparametric credit scoring model, based on a decision tree, including borrowers' quantitative behavioural traits as input for the final scoring model. On applying data collected from a Tunisian microfinance bank, the major depicted result lies in the fact that borrowers' psychological traits constitute a major information source in predicting their creditworthiness. Actually, the variables deployed have helped reduce the proportion of bad loans classified as good loans by 3.125%, which leads to a decrease in MFIs' losses by 4.8%. In addition, the results indicate that the scoring model based on a classification and regression tree (CART) outperforms the classic techniques. Actually, implementing this CART model might well help MFIs reduce misclassification costs by 6.8% and 13.5% in comparison with the discriminant analysis and logistic regression models respectively. Our conceived model, we consider, would be of great practical implication for microfinance and may provide a means for securing competitive advantage over other MFIs that fail to implement such a methodology. Copyright © 2014 John Wiley & Sons, Ltd.     

A re-examination of the impact of credit ratings and economic factors on state bond yields

In this article we re-examine the impact of credit ratings and economic factors on state bond yields using a two-step model. In the first step, we adopt an ordered probit technique to obtain consistent estimates of state bond default risk. In the second step, we estimate state bond risk premiums using a regression analysis with a categorized risk variable obtained from the first step. Similar to Terza (1987) and Hsiao (1983), the model involves a categorized ordinal explanatory (rating) variable. However, our two-step model deals with a case where category thresholds are unknown and dependent on economic factors. The model provides consistent estimates for the effects of ratings and economic factors on state bond yields. Contrary to previous findings, we find that state bond yields are mainly affected by fundamental economic variables.     

A two-stage credit scoring model based on random forest: Evidence from Chinese small firms

Small firms are major contributors to most economies, often supported by government policies. However, the credit scoring of small firms is complicated and costly, making it a challenging field of research. Using loan data from 3045 small firms in China, we design a two-stage expert system for default prediction that quantifies the variables and thresholds that have a key impact. Firstly, we use SMOTE to deal with the imbalanced data and secondly, we employ random forest to build predictive credit features. Dominance analysis shows that, when making default assessments on Chinese small firms, it is important to consider not only financial factors, but also non-financial and macroeconomic factors. In particular, the net cash profit, the firm's legal disputes and the per capita disposable income of urban residents are key factors in credit scoring. Robustness tests show that our proposed methodology performs better than other machine learning models, and this result is robust with observations from other countries.     

A two-stage dynamic credit scoring model,based on customers’ profile and time horizon

As credit card usage has expanded rapidly worldwide, credit scoring has become a very important task for banks, which can benefit from reducing possible risks of default. Credit scoring models help decision makers to decide whether to issue a credit card to a new applicant on the basis of both financial and nonfinancial criteria. The scope of the current study is to develop a dynamic scoring model that (a) estimates the credit performance of an applicant using generalised linear models and (b) accommodates the changes of a borrower's characteristics after the issuance of the credit card and forecasts the time of default using survival analysis.     

A practical approach to validating a PD model

The capital adequacy framework Basel II aims to promote the adoption of stronger risk management practices by the banking industry. The implementation makes validation of credit risk models more important. Lenders therefore need a validation methodology to convince their supervisors that their credit scoring models are performing well. In this paper we take up the challenge to propose and implement a simple validation methodology that can be used by banks to validate their credit risk modelling exercise. We will contextualise the proposed methodology by applying it to a default model of mortgage loans of a commercial bank in the Netherlands.     

A simulation study of Basel II expected loss distributions for a portfolio of credit cards

Credit scoring models have been used traditionally as the basis of decisions to reject or accept credit applications. They are also used to categorize applicants or existing accounts into risk groups. Based on estimates of probability of default (PD), the risk groups may seem well separated. However, by considering distributions on risk elements such as model estimation uncertainty, exposure at default and loss given default, a simulation approach is used to compute Basel II expected loss distributions for a portfolio of credit cards. These show that discrimination between risk groups is not as clear as is immediately suggested simply by PD estimates. Based on these distributions, we also show that measuring extreme credit risk with Value at Risk can lead to considerable underestimation if distributions on these risk elements are not entered into the computation.     

Credit cycles and macro fundamentals

We use an intensity-based framework to study the relation between macroeconomic fundamentals and cycles in defaults and rating activity. Using Standard and Poor's U.S. corporate rating transition and default data over the period 1980–2005, we directly estimate the default and rating cycle from micro data. We relate this cycle to the business cycle, bank lending conditions, and financial market variables. In line with earlier studies, the macro variables appear to explain part of the default cycle. However, we strongly reject the correct dynamic specification of these models. The problem is solved by adding an unobserved dynamic component to the model, which can be interpreted as an omitted systematic credit risk factor. By accounting for this latent factor, many of the observed macro variables loose their significance. There are a few exceptions, but the economic impact of the observed macro variables for credit risk remains low. We also show that systematic credit risk factors differ over transition types, with risk factors for downgrades being noticeably different from those for upgrades. We conclude that portfolio credit risk models based only on observable systematic risk factors omit one of the strongest determinants of credit risk at the portfolio level. This has obvious consequences for current modeling and risk management practices.     

Is there any dependence between consumer credit line utilization and default probability on a term loan? Evidence from bank-customer data

We study the relationship between two financial instruments through the simultaneous analysis of personal credit line utilization and default probability on a personal term loan. We model both dependent variables in a system of simultaneous equations and find strong evidence of dependence between the two financial instruments. Individuals in the default state draw their credit line by 9 percentage points more and, depending on the specification, a 10 percentage point increase in credit line utilization decreases the default probability by 0.09 to 0.41 percentage points, on a base default rate of 1.08%. This provides evidence that borrowers may use the liquidity of the credit line to pay down the term loan in periods of financial distress and suggests that banks should manage both financial instruments simultaneously.     

Modelling Credit Risk for SMEs: Evidence from the U.S. Market

Considering the fundamental role played by small and medium sized enterprises (SMEs) in the economy of many countries and the considerable attention placed on SMEs in the new Basel Capital Accord, we develop a distress prediction model specifically for the SME sector and to analyse its effectiveness compared to a generic corporate model. The behaviour of financial measures for SMEs is analysed and the most significant variables in predicting the entities' credit worthiness are selected in order to construct a default prediction model. Using a logit regression technique on panel data of over 2,000 U.S. firms (with sales less than $65 million) over the period 1994–2002, we develop a one-year default prediction model. This model has an out-of-sample prediction power which is almost 30 per cent higher than a generic corporate model. An associated objective is to observe our model's ability to lower bank capital requirements considering the new Basel Capital Accord's rules for SMEs.