Evaluating borrower’s default risk in peer-to-peer lending: evidence from a lending platform in China

Recent years have witnessed the popularity of online peer-to-peer lending, which allows individuals to borrow from and lend to each other on an Internet-based platform. Using data from a large P2P platform in China, this article explores the factors that determine the default risk based on the demographic characteristics of borrowers. Moreover, we propose a credit risk evaluation model, which can quantify the default risk of each P2P loan. Empirical results reveal that gender, age, marital status, educational level, working years, company size, monthly payment, loan amount, debt to income ratio and delinquency history play a significant role in loan defaults. Finally, we analyse the relationship between default risk and these contributory variables, and the possible causes are also discussed in this study.     

基于K-S检验与距离相关分析的网络借贷信用评价指标体系构建

网络借贷作为一种新型互联网金融模式,提升了金融资源使用效率,缓解了小企业融资难的困局。构建合理的网络借贷信用评价指标体系,从而对网络借贷的潜在风险及时甄别与预防,对互联网金融健康持续发展意义重大。本文根据K-S检验与距离相关分析相结合,筛选对借款客户违约状态甄别能力强的指标,建立了网络借贷信用评价指标体系,通过P2P网络借贷(peer to peer lending,个人对个人借贷)平台LendingClub交易数据进行实证研究,结果表明:不仅借款金额、借款利率等借款标的特征对借贷者违约具有显著相关性,借款者年龄等个人特征、借款者年收入等财务特征以及借款者违约次数等信用特征均对借贷者违约风险产生显著影响。投资者在出借资金时,往往青睐于已婚、年龄适中、具有一定工作经历、历史违约次数较少的借款人。因此,风险监管部门应构建网络借贷违约风险评估模型,对P2P平台进行风险监测,同时建立关键信息共享机制,融合多源数据,明确审查范围,实现P2P网络借贷行业健康有序发展。     

LES DÉTERMINANTS DE LA PERFORMANCE DE REMBOURSEMENT DES MICROCRÉDITS EN TUNISIE

Our work is part of the quest for sustainability of MFIs. To ensure sustainability, an MFI must be effective. It must, among others, face many risks and in particular the default risk. A question arises: what are the determinants of portfolio quality of MFIs? In other words, what are the determinants that influence delays in payment of MFIs? The interest of this research is twofold. The first is practical: knowing the importance of performance measures of microcredit as a precondition for the efficiency and financial performance of MFIs, this study identified determinants of the reimbursement rate in an MFI of Tunisia. It may even give an evidence for banks to intervene on the market of micro credit. The second is methodological: to find the determinants of reimbursement behaviour in MFIs, a binary logistic regression is used, while differing in explanatory variables between key variables and moderators’ ones, which was not met previously in the literature. The results showed that among the variables related to the characteristics of the loan, the loan amount and the borrower's experience with the MFI affect the rate of reimbursement. The credit increases the likelihood of default. Unlike, when borrower's experience with the MFI increases, the risk of default decreases. A relationship between the industry and the repayment rate has not been verified. For variables related to the characteristics of the borrower, age, educational level and marital status of the borrower appear to affect the relationship between the risk of default and the amount of credit. The effect of the credit amount on the default risk decreases as age and level of education increases. In addition, this effect is more important for married compared to the unmarried. For gender, we found that the effect of maturity on the risk of default is not the same for men or women borrowers. It seems that when the maturity increases, men provide reimbursement rates lower than women.     

Heterogeneity of farm loan packaging term decisions: a finite mixture approach

This article revisits the minority borrowers’ discrimination issue in farm lending by departing from traditional loan approval-rejection or default rate-based analytical models to focus on loan packaging decisions. This study analyses such decisions using a Finite Mixture Model that optimally separates the borrowers into two sub-classes allowing for a priori unspecified heterogeneity in borrowers’ data, which has not been accounted for in previous loan discrimination analyses. Results show that non-white farm borrowers tend to receive larger loans among those in the lower loan latent class, but receive relatively lower loans in the larger loans borrower category. These farmers are also charged higher interest rates vis-à-vis their peers in both the low and high interest rate latent classes. This study’s results also indicate that male borrowers are accommodated with larger loans and longer maturities in all loan amount and maturity latent classes. This study validates the interplay among significant trends in loan packaging terms for racial and gender minority borrowers that seems logical from the lenders’ credit risk management perspective.     

Lending policies of informal, formal and semiformal lenders

This paper compares lending policies of formal, informal and semiformal lenders with respect to household lending in Vietnam. The analysis suggests that the probability of using formal or semiformal credit increases if borrowers provide collateral, a guarantor and/or borrow for business‐related activities. The probability of using informal credit increases for female borrowers. It also appears that the probability of using formal credit increases in household welfare up to a certain threshold, but at a decreasing rate. In addition, the paper discerns the determinants of probability of default across lender types. Default risk of formal credit appears to be strongly affected by formal loan contract terms, e.g., loan interest rate and form of loan repayment, whereas default risk on informal loans is significantly related to the presence of propinquity and other internal characteristics of the borrowing household. Overall, the study raises several important implications for the screening, monitoring and enforcement instruments that may be employed by different types of lenders.     

The Definition of the Grading Scales in Banks' Internal Rating Systems

An internal risk rating system can be defined as the process used to classify bank borrowers into categories of different credit riskiness. Most of the related literature has investigated various aspects of this process, but the problem of defining the categories and the distribution of borrowers into the different classes or grades has received rather less attention, other than noting that the number of grades and their dispersion should achieve a meaningful differentiation of risk. An appropriate definition of the grading scale is of primary importance because the probability of default associated to each grade is the key inputs of capital allocation systems at many best-practice banks and is the core of the January 2001's new proposal of the Basel Committee for the calculation of capital requirements. Statistical techniques such as cluster analysis can help in identifying distinct subgroups of borrowers possessing the same creditworthiness. We use a logit model to estimate individual default probabilities for four categories of borrowers and apply cluster analysis to assign borrowers to each grade. However, since cluster analysis is not a purely mechanical process, but requires examination of the nature of observations and of the objective of clustering, the ultimate choice of the most appropriate grading scale for a given portfolio relies on empirical grounds. A sufficient granularity and an appropriate quantification of risk must be balanced.
(J.E.L.: G21, G22, G33)     

Measuring systemic risk using vine-copula

We present an intuitive model of systemic risk to analyse the complex interdependencies between different borrowers. We characterise systemic risk by the way that financial institutions are interconnected. Using their probability of default, we classify different international financial institutions into five rating groups. Then we use the state-of-the-art canonical (C-) and D-vine copulae to investigate the partial correlation structure between the rating groups. Amongst many interesting findings, we discover that the second tier financial institutions pay a larger contribution to the systemic risk than the top tier borrowers. Further, we discuss an application of our methodology for pricing credit derivative swaps.     

Evaluating credit risk and loan performance in online Peer-to-Peer (P2P) lending

Online Peer-to-Peer (P2P) lending has emerged recently. This micro loan market could offer certain benefits to both borrowers and lenders. Using data from the Lending Club, which is one of the popular online P2P lending houses, this article explores the P2P loan characteristics, evaluates their credit risk and measures loan performances. We find that credit grade, debt-to-income ratio, FICO score and revolving line utilization play an important role in loan defaults. Loans with lower credit grade and longer duration are associated with high mortality rate. The result is consistent with the Cox Proportional Hazard test which suggests that the hazard rate or the likelihood of the loan default increases with the credit risk of the borrowers. Finally, we find that higher interest rates charged on the high-risk borrowers are not enough to compensate for higher probability of the loan default. The Lending Club must find ways to attract high FICO score and high-income borrowers in order to sustain their businesses.     

Credit card cash‐out as informal financing: Evidence from China

Using a proprietary account‐level database from a commercial bank in China, we document that credit card holders can ease their credit constraints through the practice of cash‐out based on bogus transactions using credit cards. We find that such behaviour might be beneficial to both cardholders and banks. First, we find that a 1% increase in the cumulative number of credit card cash‐out transactions lowers the probability of default by 9.59%. Second, for private businesses, a 1% increase in the number of abnormal cash‐out transactions lowers overdue risk by 13.45%. Third, by lowering the overdue risk, the card‐issuing bank earns a larger profit. Our results are consistent with the notion that unconventional credit card cash‐out can mitigate the extent of capital misallocation in emerging markets.     

Adverse selection as an explanation of credit rationing and different lender types

An adverse selection model is utilized to demonstrate that informational asymmetry may make it wealth optimal for the financial intermediary (FI) to credit ration and to rationalize the existence of different lenders in the credit market. The crucial assumption is that borrowers differ in their tolerance for a lender-imposed default penalty, the severity of which also varies with the lender. The credit rationing portion proves that the FI will: 1) be forced by a binding regulatory constraint to overinvest in capital; 2) ration its worst risk class borrowers; 3) establish its optimal loan interest rate on the basis of the average quality of its loans and the interest rate elasticity of the borrower demand in its best risk category; and 4) decrease the total loan volume and increase the loan interest rate due to an increase in the capital requirement, but the effect on the default risk quality of its loan portfolio is ambiguous. The existence result is that if a lender has a high default penalty, he can charge a lower rate and attract only “good” borrowers, i.e., heterogeneous lender types encourage the screening of borrowers and vice versa.     

Parametric versus nonparametric methods in risk scoring: an application to microcredit

The importance of credit access to improve economic opportunities in developing markets is well established in the literature. However, there exists a strong need to mitigate adverse selection problems in microlending. A risk scoring model that more accurately predicts the likelihood of repayment of potential borrowers can help address this market imperfection and to benefit both lenders and borrowers. This paper compares the performance of nonparametric versus semiparametric and traditional parametric risk scoring models based on default probabilities. We show the advantages of relying on less structured, data-driven methods for risk scoring using both simulated data and data from credit loans granted to small and microenterprises in rural Peru. The estimation results indicate that nonparametric methods lead to a better evaluation of credit worthiness and can help prevent including potential “bad” borrowers and excluding “good” borrowers from sensitive microcredit markets.     

中国商业银行体系信用风险评估——基于宏观压力测试的研究

宏观压力测试,作为压力测试方法在宏观经济分析中的具体运用,可以提供极端事件对金融体系影响的前瞻性信息.随着各国金融监管当局对系统性风险的日趋重视,宏观压力测试方法逐渐成为检验一国银行体系的脆弱性、维护金融稳定的首选工具.本文主要研究宏观压力测试在银行信用风险评估中的应用,并在已有的模型成果的对比分析基础上,建立适用于我国的宏观压力测试模型并以此进行实证分析.本文以贷款违约率作为评估银行系统信用风险的指标,选取对银行信贷违约风险构成冲击的宏观经济变量,通过多元线性回归模型将其整合成为一个综合性指标.研究结果发现:名义国内生产总值(NGDP)和通货膨胀率指标(CPI)对银行体系的贷款表现冲击力较强.在此基础上构建了两种宏观经济极端情境,在关于NGDP大幅下降和CPI骤升的压力情境设定下,银行体系的贷款违约率都出现了不同程度的大幅度提高.尤其在关于通货膨胀率的情境设定下,贷款违约率的增幅高于其在NGDP下降情境下的增幅.     

Information technology and the rise of household bankruptcy

Several studies have attributed the rise of household bankruptcy in the past two decades to the decline of social stigma associated with default. Stigma explanations, however, cannot account for the large increase in the use of unsecured credit during this period. I explain the simultaneous increase in bankruptcy rates and unsecured credit as the result of improvements in credit-rating technologies. Using an environment where borrowers face heterogeneous default costs (unobservable by creditors), I show that such improvements will lead to agents with high default costs, i.e., “safe” borrowers, being able to borrow more. A quantitative example illustrates that this increased access to credit can be large enough to raise both equilibrium borrowing and default rates.     

Convergence Clubs: Some Additional Evidence

The paper shows that the joint presence of moral hazard and repudiation risk generates an important interaction effect. In order to provide the proper incentives to borrowers, the optimal financial contract under moral hazard calls for all available resources to be paid to the lender in the event of a poor realization for output. Repudiation risk limits the size of this transfer, as the debtor has the option to default. This upper bound on the resource transfer exacerbates the moral hazard problem, reducing lending and the equilibrium level of investment and output.     

North–South Lending with Moral Hazard and Repudiation Risk

The paper shows that the joint presence of moral hazard and repudiation risk generates an important interaction effect. In order to provide the proper incentives to borrowers, the optimal financial contract under moral hazard calls for all available resources to be paid to the lender in the event of a poor realization for output. Repudiation risk limits the size of this transfer, as the debtor has the option to default. This upper bound on the resource transfer exacerbates the moral hazard problem, reducing lending and the equilibrium level of investment and output.     

Two stages credit evaluation in bank loan appraisal

Traditionally, banks conduct standard credit evaluation such as credit scoring following the receipt of loan request and make the accept/reject decision accordingly. This research explores the possibility of two stages credit evaluation in lending process. When the evaluation cost drops below the trigger cost, it pays to conduct the second-stage loan appraisal. We derive two trigger cost thresholds for borrowers who are rated as credible and default in the first stage, respectively. Contingent on the share of good borrowers relative to the bad ones, the optimal strategy of the bank can be differentiated to implement second-stage evaluation on either (1) both types, or (2) only one type, or (3) neither type of the borrowers. We find that during severe economic contractions or in geographic areas/industries which are in deep troubles, whilst the borrowers who repay the loan are out-numbered by the borrowers who fail to pay, the trigger cost for good borrower is higher than that of default borrower. In this scenario, the banks are more inclined to undertake the second-stage credit evaluation on good borrowers. On the other hand, if the percentage of credible borrowers is higher than that of default borrowers, the trigger cost for good borrower lies below the trigger cost of default borrower. As a result, the banks are less inclined to undertake the second-stage evaluation on good borrowers.     

网络借贷中羊群效应的存在性、驱动机制与投资者投资效率*

本文探究了网络借贷中羊群效应的存在性、背后的驱动机制以及对投资者投资效率的影响。实证结果表明，在控制了标的流标风险和时间固定效应后，我国网络借贷投资者群体中存在显著的羊群效应。进一步，标的羊群效应程度与借款人信息以及投资者类型紧密相关，借款人的还款能力越低，参与投资的投资者风险厌恶程度越低，则该标的的羊群效应越显著，说明羊群效应是“风险厌恶程度较低的投资者期望通过模仿他人的投资选择来消除由于信息不对称带来的违约风险”造成的结果。最后，网络借贷中的羊群效应有利于提升投资者的投资效率，具体表现在羊群效应提高了投资者成功投资的概率，并且有助于投资者在低质量标的中将资金投资到潜在违约风险更低的标的之中。     

A costly state verification model with diversity of opinions

Summary. In simple models of borrowing and lending with ex-post asymmetric information, Gale and Hellwig (1985) and Williamson (1986) have shown that optimal debt contracts are simple debt contracts where borrowers repay a fixed interest rate whenever possible and lenders seize all the profit when borrowers default. In this note, we depart from their works by assuming that borrowers and lenders have heterogeneous beliefs, and show that simple debt contracts do not necessarily survive as optimal contracts.JEL Classification Numbers: G3, D8.     

A DSGE model of banks and financial intermediation with default risk

This paper takes the view that a major contributing factor to the financial crisis of 2008 was a failure to correctly assess and price the risk of default. In order to analyse default risk in the macroeconomy, a simple general equilibrium model with banks and financial intermediation is constructed in which default-risk can be priced. It is shown how the credit spread can be attributed largely to the risk of default and how excess loan creation may emerge due to different attitudes to risk by borrowers and lenders. The model can also be used to analyse systemic risk due to macroeconomic shocks which may be reduced by holding collateral.     

LGB-BAG模型在P2P网贷借款者信用风险评估中的应用

立足于P2P平台,利用P2P平台个人借款人的信息建立了一套系统的信用风险评估指标体系来甄别可能违约的借款人。本文基于LightGBM（一种基于决策树的Boosting模型）和Bagging而提出一种新的LGB-BAG模型,有效结合了Boosting和Bagging的优势。结果表明,在N增大到一定程度的时候,LGB-BAG的F1均值（预测效果）要高于LightGBM和随机森林;并且LGB-BAG的F1方差也要小于其余两种模型。LGB-BAG的F1均值最高可达到0.71175,且LGB-BAG模型能够显著提高信用风险预测效率。