High-risk Borrowers Research Articles

Feature selection in peer-to-peer lending based on hybrid modified grey wolf optimization with optimized decision tree for credit risk assessment

ABSTRACT Lending platforms operating on a peer-to-peer (P2P) basis encounter the intricate challenge of assessing borrower creditworthiness to minimize the risk of defaults. This study addresses this challenge by proposing an advanced approach to feature selection that leverages the Grey Wolf Optimizer (GWO) in conjunction with a finely tuned Decision Tree (DT) model. The main objective is to enhance the precision and efficiency of feature selection processes within P2P lending datasets. The study begins by fine-tuning DT hyperparameters using Genetic Algorithms (GA), yielding an optimal configuration: ‘max_depth’ = 40, ‘min_samples_leaf’ = 20, and ‘criterion’ = ‘entropy’. Subsequent phases involve the application of GWO and modified GWO (nGWO, cGWO, and lGWO) to conduct feature selection under distinct Search Agent (SA) setups (SA = 5, SA = 20, SA = 50). Particularly impressive is the performance of the lGWO model with the SA = 50 setup, achieving a remarkable 91% accuracy while selecting 80.55% of the total 36 features. This study significantly improves how lenders manage risks in P2P lending by identifying high-risk borrowers more effectively, helping lenders reduce financial risks and benefiting all parties involved.

Group lending as a mechanism for self-insuring default risk

AbstractWe show that banks can provide loans at low costs to high-risk borrowers in the form of a group lending contract in which all members are jointly liable for their loans. By providing such contracts borrowers self-insure against some of the default risk the bank faces. We determine the optimal group size in a competitive banking system and find that it is reasonably small and borrowers internalize an increasing fraction of the risk the higher their risks are.

Fraud-Agents Detection in Online Microfinance: A Large-Scale Empirical Study

Online Microlending, a new financial service, focuses on small loans without any sort of collateral. It provides more flexible and quicker funding for borrowers, as well as higher interest rates of return. For platforms that provide such services, an essential task is to adequately evaluate each loan’s risk so as to minimize the possible financial loss. However, there exists a special group of borrowers, namely <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">fraud-agents</i> , who gain illegal profits from inciting other borrowers to cheat, i.e., they help the high-risk borrowers evade the risk evaluation by crafting fake personal information. The existence of fraud-agents poses a severe threat to the risk management systems and results in a huge financial loss for lending platforms. In this article, we present the first machine learning-based solution to detect fraud-agents in online microlending. The key challenge of this decade-long problem is that it is unclear how to construct effective features from multiple behavior logs such as phone call history, address book, loan history and activity logs of borrowers. To address this problem, we first conduct an empirical study on over 600K borrowers to gain some insights on the adversarial behaviors of fraud-agents comparing to normal borrowers and benign-agents. Based on the study, we are able to design a total of 26 features, falling into four groups, for fraud agent detection. Then, we propose a two-stage detection model to address the challenge of limited number of labeled fraud agent examples. The evaluation results show that our method can achieve a precision of 94.30%. We deploy our method on a real large online microlending platform with 11,953,273 borrowers, and we identify 29,727 fraud-agents from them. The domain experts from the platform confirm that 95.59% of them are real fraud-agents, and have added them to the platform’s internal blacklist. We further conduct a measurement study on those fraud-agents to share deeper insights on their adversarial behaviors.

GSIB status and corporate lending

Global Systemically Important Banks (GSIBs) face additional capital requirements and closer supervision. We study how closer supervision affects corporate credit supply and investigate consequences for firms. GSIB designations reduce lending on average by 5.9% but to risky firms by 7.2%. The consequences are lower asset, sales, and investment growth, especially among high-risk borrowers, and reduced R&D expenditures among all GSIB-dependent firms. Closer supervision therefore reduces banks' risk-taking but has potentially unintended implications for firms' ability to finance innovation, which seems to crucially depend on bank credit. The supervision-induced effects are larger than those attributed to GSIB-specific capital surcharges.

Effect Of Loan-To-Asset Ratio On Non-Performing Loans In Deposit Money Banks In Nigeria.

This study examined the effect of loan-to-asset ratio (LAR) on Non-Performing loans in deposit money banks in Nigeria. The study used an ex post facto research design. The population of the study comprised all the 14 quoted deposit money banks in Nigeria with a sample size of eight (8) deposit money banks licensed with international authorities consisting of Access Bank Plc., Fidelity Bank Plc., First City Monument Bank Plc., First Bank Nigeria Ltd., Guarantee Trust Bank Plc., Union Bank of Nigeria Plc., United Bank for Africa Plc and Zenith Bank Plc. Data was generated from the annual financial reports and accounts of the sampled deposit money banks, the Central Bank of Nigeria, and the Nigerian Exchange Gro up respectively. Descriptive statistics, correlation tests and Panel regression were used for analysis. Panel regression was used to analyze the data and the findings showed that loan-toasset ratio has a negative and significant effect on non-performing loans of deposit money banks in Nigeria. Furthermore, bank size has a positive and insignificant effect on non-performing loans of deposit money banks in Nigeria. The study recommends that the CBN should constantly update its supervisory skills and strategies while enforcing its lending policy guidelines to achieve stability in the banking sector. Similarly, management of Deposit Money Banks should avoid charging high-interest rates on loans and advances as well as lending to high-risk borrowers to bring cases of loan defaults to the barest minimum. This measure will no doubt curtail the excessive accumulation of bad loans and financial crises in the banking sector.

Monopolistic Screening Under Mental Accounting: Applications to Loan Markets with Collateral

AbstractThis paper extends the standard model of monopolistic screening to allow for some consumers who engage in narrow framing, a prominent behavioral bias of mental accounting. Narrow framing generates a bias toward high quality-price ratios, which induces even high-type consumers to choose a menu that targets low-type consumers. To strategically account for narrow framing, when the monopolist induces the high-type consumers to stay with the more expensive menu, there arises a downward quality distortion even at the top and a smaller downward distortion at the bottom. Then, we apply this model to optimal loan contracts to screen heterogeneous borrowers with different default risks. We find that the lender optimally reduces the collateral requirement for low-risk borrowers when some high-risk borrowers are subject to narrow framing. This result empirically implies that narrow framing may lead to smaller credit rationing and lower monitoring intensity in the lending markets.

MICROFINANCE AND GENDER DISCRIMINATION IN CREDIT ALLOCATION: AN ECONOMIC ANALYSIS OF INCENTIVE CONTRACTS

In this paper, we develop a model of discrimination by effort between women and men in credit markets. The specificity of our model compared to the current literature consists in the level of effort effected by the agents who benefited of a microcredit is endogenous. We assume that borrowers have financial constraints and that they want to obtain funds to carry out their projects whose success depends on costly and unobservable efforts. Assuming that women exert a higher level of effort than men, we show that when information is perfect, all projects are financed and the collateral is completely eliminated. Women benefit of a lower interest rate than men and at the same time receives a loan of the same amount as men. In a situation of imperfect information, when the level of the project manager’s effort is private information, we show that in a competitive credit market where risks are not mixed, the riskier borrower obtains the same contract as in a perfect situation. When the state can guarantee highrisk borrowers (low efforts), their welfare improves. On the other hand, when government can provide guarantees for lowrisk borrowers (high efforts), it reduces collective welfare.

Increasing microfinance risk tolerance through revenue sharing: an experiment

ABSTRACT Microfinance has been found to be less effective for high risk/return borrowing groups. We report a group liability microfinance lab experiment that tests a mechanism to raise repayment rates among such borrowers. The mechanism offers partial revenue sharing among groups of borrowers, agreed to before individual business outcomes are realized and loan repayment is due. Such revenue sharing makes loan repayment optimal under more outcome states, increasing the expected benefit to each borrower of repayment to qualify for future loans. We further test the effect of allowing borrowers to renege on revenue sharing agreements after learning their business outcomes, prior to loan repayment decisions. Our results illustrate the problem that exogenously higher risk/return borrowing groups achieve lower loan repayment rates than lower risk/return borrowing groups. We find evidence that optional revenue sharing significantly increases high risk borrowers’ repayment rates, but that most of this gain is lost if they can renege on revenue sharing agreements.

Internal Credit Risk Models and Digital Transformation: What to Prepare for? An Application to Poland

Purpose: The digitization of credit risk through machine learning technology is becoming more attractive, especially nowadays. The article aims to analyze the performance of models estimated with Machine Learning (ML) algorithms in predicting the risk of default compared with standard statistical models such as logistic regression (benchmark model). Design/Methodology/Approach: The indicated models were estimated using an original dataset, including financial information and the credit history of non-financial Polish enterprises. The dataset is also enlarged 20-fold to obtain a set of the so-called Big Data that could also be accepted. The out-of-sample performance comparing one-year-ahead PD estimates and observed default data for the 2015-2020 period was verified about the models under consideration. The period above also includes that associated with the COVID-19 pandemic. Findings: Based on the results obtained, practical information was supplied to credit-risk researchers. Where only a limited data set is available, and where this is confined to financial indicators only, models based on ML are seen to offer a significant increase in discriminant power and precision as compared with statistical models, this being especially the case with an artificially generated set of so-called Big Data. Practical Implications: Models estimated with ML algorithms can benchmark the probability of default obtained using more apparent statistical models. In practice, this is useful when estimates under the two types of model prove notably different. Application is handy with, for example, more significant or higher-risk borrowers. Originality/Value: The article seeks to ascertain how the market expansion of a bank's product and digital divisions might be supported without the speed and quality of credit-risk assessment is limited. The inclusion here of the COVID-19 (exogenous economic shock) period ensures the particular usefulness of recommendations for credit-risk analysts.

Lending Standards and Borrowing Premia in Unsecured Credit Markets

&lt;ns2:p&gt;Using administrative data from Y-14M and Equifax, we find evidence for large spreads in excess of those implied by default risk in the U.S. unsecured credit market. These borrowing premia vary widely by borrower risk and imply a nearly flat relationship between loan prices and repayment probabilities, at odds with existing theories. To close this gap, we incorporate supply frictions – a tractably specified form of lending standards – into a model of unsecured credit with aggregate shocks. Our model matches the empirical incidence of both risk and borrowing premia. Both the level and incidence of borrowing premia shape individual and aggregate outcomes. Our baseline model with empirically consistent borrowing premia features 45% less total credit balances and 30% more default than a model with no such premia. In terms of dynamics, we estimate that lending standards were unchanged for low risk borrowers but tightened for high risk borrowers at the outset of Covid-19. Borrowing premia imply a smaller increase in credit usage in response to a negative shock, which this tightening reduced further. Since spreads on loans of all risk levels are countercyclical, all consumers use less unsecured credit for insurance over the cycle, leading to 60% higher relative consumption volatility than in a model with no borrowing premia.&lt;/ns2:p&gt;

Financial Globalization: Effects on Banks’ Information Acquisition and Credit Risk

English Abstract: Financial liberalization accelerates global banks’ entry into new markets where host countries hope to spur investment and economic growth. However, banks sometimes retreat from their global ambitions and exit these new markets. This study demonstrates how difficulties of foreign banks in new markets may emerge due to disadvantages in the ability to assess credit quality compared to that of established domestic banks. We present a duopoly model where two banks conduct investigations in the search of qualified loan borrowers. The model assumes that the domestic bank has a cost advantage in evaluating a borrower’s credit quality compared to the competing foreign bank. Despite the cost heterogeneity, an equilibrium exists in which two such banks coexist in the market. Specifically, the information cost advantaged bank orchestrates a cream skimming strategy which entails lower-price commitments of loan products and higher investigation levels to screen and find low-risk borrowers. In contrast, the information cost disadvantaged bank chooses a bottom fishing strategy which consists of higher-priced loan offers with lower investigation levels. This results in high acceptance rates of high-risk borrowers in the foreign bank’s loan profile and correspondingly, higher default rates. We analyze the results to derive implications for development policy.

The IRB Approach and Bank Lending to Firms

This paper examines the impact of the regulatory approach adopted to calculate capital requirements on banks’ lending policies. Since the capital absorption of loans to high-risk borrowers is greater under the internal ratings-based (IRB) method than under the standardized approach (SA), IRB banks may raise interest rates and reduce credit to riskier borrowers following their regulatory regime shift. The analysis examines banks’ lending policies around each of the shifts that occurred between 2007 and 2017. First, in a context of declining rates and credit growth, banks adopting the IRB approach decreased interest rates (credit) less (more) for riskier than for safer borrowers when compared with SA intermediaries. Second, an existing credit relationship with a high-risk borrower is more likely to end after the shift. Third, the results at the firm level suggest that high-risk borrowers partly compensated the reduction in bank credit by obtaining funds from SA institutions, but that they were not able to offset the rise in their average cost of credit because of the significant costs involved in switching lenders.

High Cost Lenders and the Geographic Concentration of Foreclosures

We define high cost lenders as lenders that issue a disproportionate number of high cost loans. We develop a shift-share measure to capture the market representation of these high cost lenders in housing submarkets. After conditioning on housing submarket fixed effects, origination year fixed effects and trends over origination years based on housing submarket attributes, the magnitude of the estimated relationship is very stable as detailed controls for borrower attributes, credit score and loan terms are added. The relationship between the representation of high cost lenders and foreclosure is broad based across borrowers and types of loans, but is strongest for loans originated by high cost lenders whether or not the loans themselves are high cost. We investigate three potential mechanisms: reverse causality where high cost lenders respond to an increase in demand from higher risk borrowers, the types of mortgages issued when high cost lenders increase their market presence, and the behavior of loan servicers when a cohort of loans contains a large number of loans issued by high cost lenders. While we do not have direct information on loan servicers, our evidence points towards foreclosure decisions during the crisis as the primary mechanism behind our findings. Institutional subscribers to the NBER working paper series, and residents of developing countries may download this paper without additional charge at www.nber.org.

THE ROLE OF SAVINGS AND CREDIT COOPERATIVES (SACCOS) ON FINANCIAL INCLUSION IN ZIMBABWE

This paper investigated the role of Savings and Credit Cooperatives (SACCOs) in Zimbabwe using a national baseline survey. Several economic challenges over the past two decades left ordinary citizens unbanked and financially excluded. SACCOs are viewed as a viable option towards financial inclusion. Classical and modern theoretical literature as well as experiences from other African countries were discussed. Data was collected using a triangulation method of questionnaires, interviews and focus group discussions (FGDs) across 147 SACCOs. It was observed that there are around 6,028 cooperatives with only 4.8% (289) being SACCOs and majority of members cannot distinguish between cooperatives and SACCOs. More than 90% cited that economic and political conditions eroded their savings value thereby reducing confidence in thrift models. 77 SACCOs (52.8%) do not own assets, and only 6.2% have over 1,000 members. Discrepancies emanate from inconsistent subscriptions, unfavorable and outdated government by-laws. SACCOs are classified as high-risk borrowers due to inconsistent subscriptions and poorly audited financial statements resulting in high interest rates and adverse selection. The study recommends additional training to increase administrative capacity and the scope of SACCOs, increased coordination of operations to reduce systematic risk through consistent by-laws and economic policies that stimulate stability and restore confidence in the financial sector.

Fintech Borrowers: Lax Screening or Cream-Skimming?

Abstract We study the personal credit market using unique individual-level data covering fintech and traditional lenders. We show that fintech lenders acquire market share by lending first to higher-risk borrowers and then to safer borrowers, and rely mainly on hard information to make credit decisions. Fintech borrowers are significantly more likely to default than neighbor individuals with the same characteristics borrowing from traditional financial institutions. Furthermore, they tend to experience a short-lived reduction in the cost of credit, because their indebtedness increases more than non-fintech borrowers after loan origination. However, fintech lenders’ pricing strategies are likely to take this into account.

CF-NN: a novel decision support model for borrower identification on the peer-to-peer lending platform

With the prevalence of online individual micro-loans, an increasing number of peer-to-peer lending platforms have been established during the last ten years. One main problem for these platforms is to accurately identify the ‘bad’ applicants with high default risk. In this paper, we propose a CF-NN model that combines neural network and collaborative filtering for identifying high-risk borrowers. It is demonstrated in the experimental analysis that the CF-NN model significantly outperforms other widely used data mining models on the identification of bad borrowers. Moreover, the experimental results show that, to achieve the best performance in borrower identification, the CF-NN model should be equipped with parameters of intermediate values.

Fintech Borrowers: Lax-Screening or Cream-Skimming?

We study the personal credit market using unique individual-level data covering fintech and traditional lenders. We show that fintech lenders acquire market share by first lending to higher-risk borrowers and then to safer borrowers, and mainly rely on hard information to make credit decisions. Fintech borrowers are significantly more likely to default than neighbor individuals with the same characteristics borrowing from traditional financial institutions. Furthermore, they tend to experience only a short- lived reduction in the cost of credit, because their indebtedness increases more than non-fintech borrowers a few months after loan origination. However, fintech lenders' pricing strategies are likely to take this into account. Institutional subscribers to the NBER working paper series, and residents of developing countries may download this paper without additional charge at www.nber.org.

Unequal and Unstable: Income Inequality and Bank Risk

We provide evidence that regions in the U.S. with higher income inequality tend to have a riskier banking sector. However, not all banks are more risky, as reflected in a higher dispersion of bank risk. We show how a model based on risk-shifting incentives where banks channel insured deposits into subprime loans can account for both findings. In equilibrium, a competition to risk-shift emerges, leading to a subprime lending boom in which loans to high-risk borrowers carry negative NPVs. Some banks engage in risk-shifting by lending to high-risk subprime borrowers, while the rest specialize in lending to low-risk prime borrowers.

Lending Next to the Courthouse: The Salience of Adverse Events and Mortgage Lending Decisions

Adverse housing market events can affect credit conditions not only by hurting the financial fundamentals of banks, but may also by changing the lending behaviors of individual loan officers. In this paper, we test the latter mechanism by analyzing loan-level mortgage applications from 2008 to 2016 and utilizing a distinctive setup in the home foreclosure process: foreclosure auctions are typically held live at the county courthouse, thus making loan officers working nearby more exposed to news about the county-wise foreclosures. By comparing lending outcomes across loan applications within the same county and across branches of the same bank, we find that mortgage lending standards are more stringent when the decisions are made in bank branches next to county courthouses, resulting in higher rejection rates and smaller approved loan size. Accordingly, aggregating at the branch level, credit supply is significantly tighter at bank branches next to the courthouses relative to other branches of the same bank within the same county. We show that these effects on lending decisions and credit supply increase in the county-wise foreclosure intensity and are especially pronounced among high-risk borrowers and for branches belonging to small-sized banks. Our results are not driven by fundamental differences in borrower and branch characteristics of neighborhoods next to the courthouses and remain robust in a matched sample of branches and loan applications. Overall, our findings provide micro-level evidence that adverse news can have significant impacts on credit market outcomes by changing the risk preferences and beliefs of individual financial decision makers.

Borrowing costs and the role of multilateral development banks: Evidence from cross-border syndicated bank lending

Cross-border bank lending is a growing source of external finance in emerging and developing countries and could play a key role in infrastructure and development financing. This paper looks at the role of multilateral development banks (MDBs) on the terms of syndicated loans, focusing on loan pricing. The results show that MDBs’ participation is associated with higher borrowing costs and longer maturities—signaling a greater willingness by MDBs to finance risky projects which may not be financed by the private sector. We also show that MDBs are more likely to lend to borrowers located in countries with high credit and financial risk. Overall, our findings suggest that MDBs could play a role in easing access to credit for high risk borrowers.