Consumer Credit Scoring Research Articles

Impact of Credit Record Revalidation on Consumer Credit Scores: Implications for Lending Decisions

Impact of Credit Record Revalidation on Consumer Credit Scores: Implications for Lending Decisions

Impact of Credit Record Revalidation on Consumer Credit Scores: Implications for Lending Decisions

Impact of Credit Record Revalidation on Consumer Credit Scores: Implications for Lending Decisions

An explainable machine learning model for consumer credit scoring in Mexico

An explainable machine learning model for consumer credit scoring in Mexico

Credit Card Fraud Detection Using LSTM Algorithm

With the rapid growth of consumer credit and the huge amount of financial data developing effective credit scoring models is very crucial. Researchers have developed complex credit scoring models using statistical and artificial intelligence (AI) techniques to help banks and financial institutions to support their financial decisions. Neural networks are considered as a mostly wide used technique in finance and business applications. Thus, the main aim of this search is to help bank management in scoring credit card clients using machine learning by modelling and predicting the consumer behavior with respect to two aspects: the probability of single and consecutive missed payments for credit card customers. The proposed model is based on the bidirectional Long-Short Term Memory (LSTM) model to give the probability of a missed payment during the next month for each customer. The model was trained on a real credit card dataset and the customer behavioral scores are analyzed using classical measures such as accuracy, Area Under the Curve, Brier score, Kolmogorov–Smirnov test, and H-measure. Calibration analysis of the LSTM model scores showed that they can be considered as probabilities of missed payments. The LSTM model was compared to four traditional machine learning algorithms: support vector machine, random forest, multi-layer perceptron neural network, and logistic regression. Experimental results show that, compared with traditional methods, the consumer credit scoring method based on the LSTM neural network has significantly improved consumer credit scoring.

Credit Building or Credit Crumbling? A Credit Builder Loan’s Effects on Consumer Behavior and Market Efficiency in the United States

AbstractA randomized encouragement design yields null average effects of a credit builder loan (CBL) on consumer credit scores. But machine learning algorithms indicate the nulls are due to stark, offsetting treatment effects depending on baseline installment credit activity. Delinquency on preexisting loan obligations drives the negative effects, suggesting that adding a CBL overextends some consumers and generates negative externalities on other lenders. More favorably for the market, CBL take-up generates positive selection on score improvements. Simple changes to CBL practice, particularly to provider screening and credit bureau reporting, could ameliorate the negative effects for consumers and the market.Authors have furnished an Internet Appendix, which is available on the Oxford University Press Web site next to the link to the final published paper online.

Credible risk: private credit bureaus and the work of loan officers in West Africa

AbstractIn 2015, the Central Bank of West African States selected a private credit bureau to build a digital platform to collect and analyse financial data across the eight-country West African Economic and Monetary Union, including the activities of credit cooperatives and microfinance agencies that serve many West African borrowers. The credit-reporting platform promises to produce new and valuable knowledge about borrowers’ creditworthiness and risk. But valuable in what ways and for whom? Private credit bureaus rely on vast arrays of personal and financial data in order to produce credit reports and consumer credit scores that claim to represent the creditworthiness of individual borrowers. Cooperative loan officers, instead, understand creditworthiness to be highly variable and rely on relational metrics in order to determine risk as well as carefully overseeing existing loans to ensure their timely repayment. Based on ethnographic fieldwork with loan officers at a small-scale credit cooperative in Lomé, Togo, this article examines how the impending arrival of a private credit bureau and new credit-reporting technologies highlights the distinction between the ways in which credit bureaus and loan officers understand the nature of financial risk, as well as the centrality of loan officers and their management of borrower debt in sustaining the field of small-scale lending in Togo.

A CWGAN-GP-based multi-task learning model for consumer credit scoring

In consumer credit scoring practice, there is often an imbalanced distribution in accepted borrowers, which means there are far fewer defaulters than borrowers who pay on time. This makes it difficult for traditional models to function. Aside from traditional sampling methods for imbalanced data, the idea of using rejected information to one’s benefit is new. Without historical repayment performance, rejected data are often discarded or simply disposed of during credit scoring modeling. However, these data play an important role because they capture the distribution of the borrower population as well as the accepted data. Besides, due to the increasing complexity in loan businesses, the current methods have difficulties in addressing high-dimensional multi-source data. Thus, a more effective credit scoring approach towards imbalanced data should be studied. Inspired by the state-of-the-art neural network methods, in this paper, we propose a conditional Wasserstein generative adversarial network with a gradient penalty (CWGAN-GP)-based multi-task learning (MTL) model (CWGAN-GP-MTL) for consumer credit scoring. First, the CWGAN-GP model is employed to learn about the distribution of the borrower population given both accepted and rejected data. Then, the data distribution between good and bad borrowers is adjusted through augmenting synthetic bad data generated by CWGAN-GP. Next, we design an MTL framework for both accepted and rejected and good and bad data, which improves risk prediction ability through parameter sharing. The proposed model was evaluated on real-world consumer loan datasets from a Chinese financial technology company. The empirical results indicate that the proposed model performed better than baseline models across different evaluation metrics, demonstrating its promising application potential.

Predictive Value of Credit Score on Surgery Resident and Fellow Academic and Professional Performance.

IntroductionSurgical specialties consistently remain among the most competitive residency and fellowship programs with some of the highest rates of unmatched applicants. Attrition in surgical specialties is as high as 30% and particularly problematic given the extended duration of training and limited number of positions. Applicants are traditionally evaluated using a streamlined set of objective metrics, such as board scores, class rank, leadership, letters of recommendation, research productivity, and volunteerism. Consumer credit scores have been shown to be predictors of personality and work performance, however, the literature has yet to explore consumer credit histories in the context of surgical resident and fellow performance. This study aims to determine whether consumer credit scores of surgery residents and fellows are predictive of academic and professional performance.MethodsThis is a multi-institutional observational survey study across all American Council of Graduate Medical Education and Royal College of Physicians and Surgeons accredited surgical residency and fellowship programs in the United States and Canada. Ninety-nine surgical residents and fellows with educational status of post-graduate year two or higher participated in this study. Dichotomous (yes or no) survey items were formulated to assess performance indicators in the domains of notable achievements and awards, research output, written examination performance, professionalism, and surgical/technical skills. Three-digit Fair Isaac Corporation (FICO) credit scores, a widely accepted consumer reporting score, were collected to avoid calculation variability between algorithms. ResultsSurgical residents and fellows reported credit scores between 611( fair) and 853 (exceptional) with a median (interquartile range) of 774 (715-833). The majority of participants 51.5%(51) reported very good credit scores. Those with higher credit scores (very good/exceptional) were 377% more likely to have one or more positive performance indicators OR (95% CI) = 3.77 (1.43-9.97). Similarly, residents with lower credit scores (fair/good) were only 40% more likely to have one or more negative performance indicators. The credit score has a moderate ability to distinguish between the presence and absence of positive performance indicators (area under the curve {AUC} = 0.70, p = 0.001). The use of 753 as a credit score cutoff is 78.9% sensitive and 52.4% specific for discerning surgery residents and fellows with one or more positive performance indicators. The credit score did not significantly discern those with negative performance indicators.ConclusionsWhile credit score was significantly functional in discerning those with and without positive performance indicators, sensitivity and specificity rates leave much to be desired. Although our data may have indicated that higher credit scores may be associated with increased residency academic performance on examinations and research productivity we are not recommending any implementation of using credit scores as a metric for selecting individual surgical candidates for a position in residency or fellowship due to extensive socioeconomic variables and historical context of credit scores, which must be taken into consideration. Additional studies are needed to assess this utility on a larger scale.

Modelling customers credit card behaviour using bidirectional LSTM neural networks

With the rapid growth of consumer credit and the huge amount of financial data developing effective credit scoring models is very crucial. Researchers have developed complex credit scoring models using statistical and artificial intelligence (AI) techniques to help banks and financial institutions to support their financial decisions. Neural networks are considered as a mostly wide used technique in finance and business applications. Thus, the main aim of this paper is to help bank management in scoring credit card clients using machine learning by modelling and predicting the consumer behaviour with respect to two aspects: the probability of single and consecutive missed payments for credit card customers. The proposed model is based on the bidirectional Long-Short Term Memory (LSTM) model to give the probability of a missed payment during the next month for each customer. The model was trained on a real credit card dataset and the customer behavioural scores are analysed using classical measures such as accuracy, Area Under the Curve, Brier score, Kolmogorov–Smirnov test, and H-measure. Calibration analysis of the LSTM model scores showed that they can be considered as probabilities of missed payments. The LSTM model was compared to four traditional machine learning algorithms: support vector machine, random forest, multi-layer perceptron neural network, and logistic regression. Experimental results show that, compared with traditional methods, the consumer credit scoring method based on the LSTM neural network has significantly improved consumer credit scoring.

A graph-based semi-supervised reject inference framework considering imbalanced data distribution for consumer credit scoring

Credit scoring has been attracting increasing attention in the Chinese consumer financial industry. Traditional approaches are easily influenced by sample selection bias because they use accepted applicant samples only, while the applicant population also includes rejected applicants. Reject inference is a technique to infer good/bad labels for rejected applicants, which can overcome biases in credit scoring. However, previously proposed reject inference methods usually ignore the imbalanced distribution in accepted data, which means that good applicants are much more than bad ones in most practical consumer loan applications. Both the neglect of rejected data and the imbalanced distribution in accepted data weaken the performance of current credit scoring models. In this paper, we propose a novel reject inference framework that takes into account the imbalanced data distribution for consumer credit scoring. First, we use an advanced graph-based semi-supervised learning algorithm to solve the reject inference problem, which is called label spreading. Second, faced with an imbalanced distribution of good and bad samples in accepted applicants, we conduct imbalanced learning using a modified Synthetic Minority Over-sampling Technique before reject inference. Then, six binary classifiers are studied in our proposed framework for credit scoring modeling. Finally, we present the results of four exact experiments as well as online A/B tests for performance evaluation using data provided by a leading Chinese fintech company. Empirical results indicate that the proposed framework performs better than traditional scoring models across different evaluation metrics, representing a progressive method that promotes credit scoring research as well as improving fintech practices.

Piercing through Opacity: Relationships and Credit Card Lending to Consumers and Small Businesses during Normal Times and the COVID-19 Crisis

We investigate bank relationships in a rarely-considered context – consumer and small business credit cards. Using over one million accounts, we find during normal times, consumer relationship customers enjoy relatively favorable credit terms, consistent with the bright side of relationships, while the dark side dominates for small businesses. During COVID-19, both groups benefit, reflecting intertemporal smoothing, with more benefits flowing to safer relationship customers. Conventional banking relationships benefit consumers more than credit card relationships, with mixed findings for small businesses. Important identification issues are addressed. CARES Act consumer-delinquency reporting impediments reduce the informational value of consumer credit scores, penalizing safer borrowers.

Modeling and forecasting of COVID-19 using a hybrid dynamic model based on SEIRD with ARIMA corrections.

The outbreak of novel coronavirus (COVID-19) attracted worldwide attention. It has posed a significant challenge for the global economies, especially the healthcare sector. Even with a robust healthcare system, countries were not prepared for the ramifications of COVID-19. Several statistical, dynamic, and mathematical models of the COVID-19 outbreak including the SEIR model have been developed to analyze the infection its transmission dynamics. The objective of this research is to use public data to study the properties associated with the COVID-19 pandemic to develop a dynamic hybrid model based on SEIRD and ascertainment rate with automatically selected parameters. The proposed model consists of two parts: the modified SEIRD dynamic model and ARIMA models. We fit SEIRD model parameters against historical values of infected, recovered and deceased population divided by ascertainment rate, which, in turn, is also a parameter of the model. Residuals of the first model for infected, recovered, and deceased populations are then corrected using ARIMA models. The model can analyze the input data in real-time and provide long- and short-term forecasts with confidence intervals. The model was tested and validated on the US COVID statistics dataset from the COVID Tracking Project. For validation, we use unseen recent statistical data. We use five common measures to estimate model prediction ability: MAE, MSE, MLSE, Normalized MAE, and Normalized MSE. We proved a great model ability to make accurate predictions of infected, recovered, and deceased patients. The output of the model can be used by the government, private sectors, and policymakers to reduce health and economic risks significantly improved consumer credit scoring.

The American Way — Until Machine Learning Algorithm Beats the Law? Algorithmic Consumer Credit Scoring in the EU and US

Algorithmic consumer credit scoring has caused anxiety among scholars and policy makers. After a significant legislative effort by the European Union, the General Data Protection Regulation (GDPR) that has provisions tailored to automated decision-making (ADM) was implemented. When the EU Commission and the US Department of Commerce negotiated for US organizations to whom data from EU data controller is transferred to comply with the key principles of EU Data Protection Law under the EU-US Privacy Shield (PS) Framework, the Department of Commerce refused to incorporate the GDPR principles governing ADM in the PS Framework. The EU Commission accepted this refusal reasoning that where US companies make automated decisions with respect to EU data subjects, such as in consumer credit risk scoring, there are laws in the US that protect the consumer from adverse decisions. This view contradicts recommendations for implementing GDPR-Inspired law in the US to tackle the challenges of automated consumer credit scoring. This article argues that despite the differences in the approach to the regulation of automated consumer credit scoring in the EU and the US, consumers are similarly protected in both jurisdictions. Furthermore, US consumer credit laws have the necessary flexibility to ensure that adverse automated decisions are tackled effectively. This article, through analyzing statutes, cases, and empirical evidence, demonstrates that the seemingly comprehensive legal rules governing ADM in the GDPR do not make the EU consumers better off. In addition, the challenges presented by the increasing sophistication of Artificial Intelligence (AI), especially machine learning, place both the EU and the US legal regimes in similar position as neither jurisdiction is equipped to respond to autonomous, unpredictable, and unexplainable algorithms making decisions. While the EU’s risk-based approach to AI regulation adopted by the Draft AI Regulation which also contains provisions on regulatory sandboxing is a significant improvement, it does not significantly change the rules regarding algorithmic consumer credit scoring. Nevertheless, this is the approach that regulation should primarily adopt for the future.

From the Black Swan, to the Snowball. Risks of COVID-19 Pandemic for Consumer Credit Scores in the Lack of a Harmonized Regulatory Intervention

In light of the harmful and wide-ranging effect of the coronavirus pandemic, many governments in the European territory and on the global framework rushed to introduce forms of financial support for those groups that are susceptible to be economically affected by the current situation in order to limit the economic fallout of the pandemic. Still, and in spite of the significant level of regulatory intervention, no major normative change addressed the area of consumer spending and consumer credit. Considering the expected length of the phenomenon, and its impact on the economy and employment, the management of consumers’ economic exposure and of unpaid debt is though meant to emerge as a major consequence during and after the expiration of the pandemic: in particular, significant risks are related to the consequences that the deterioration of consumers’ exposure caused by the COVID-19 and the pandemic-related factors is likely to have on their credit scores (and, subsequently, in terms of outcome of future creditworthiness assessments). Scores have a major impact on consumer credit landscape, and badly determined/distorted scores invest both consumers and credit operators: consumers are precluded from accessing credit in a moment of financial distress; banks and other institutions are not able to properly discriminate between the quality of potential borrowers, therefore facing risks of overexposure to losses and unprofitable operativity. In order to prevent a further worsening of the (already distressed) global economic health, it is therefore pivotal to promptly introduce harmonized corrective measures to mitigate the risk of unsought deviation in the credit scoring sector. The research investigates the impact of the events related to the COVID-19 infection on the scoring software’s functioning, considering how the pandemic is likely to impact on both macro and microeconomic factors related to consumer behavior and indebtedness. The analysis is then developed in order to provide a set of recommendations for interventions with the aim to preserve algorithms’ stability, accuracy and predictive power over the pandemic and during its aftermath.

Machine Learning, Big Data and the Regulation of Consumer Credit Markets: The Case of Algorithmic Credit Scoring

Recent advances in machine learning (ML) and Big Data techniques have facilitated the development of more sophisticated, automated consumer credit scoring models — a trend referred to as 'algorithmic credit scoring' in recognition of the increasing reliance on computer (particularly ML) algorithms for credit scoring. This chapter, which forms part of the 2018 collection of short essays 'Autonomous Systems and the Law', examines the rise of algorithmic credit scoring, and considers its implications for the regulation of consumer creditworthiness assessment and consumer credit markets more broadly. The chapter argues that algorithmic credit scoring, and the Big Data and ML technologies underlying it, offer both benefits and risks for consumer credit markets. On the one hand, it could increase allocative efficiency and distributional fairness in these markets, by widening access to, and lowering the cost of, credit, particularly for 'thin-file' and 'no-file' consumers. On the other hand, algorithmic credit scoring could undermine distributional fairness and efficiency, by perpetuating discrimination in lending against certain groups and by enabling the more effective exploitation of borrowers. The chapter considers how consumer financial regulation should respond to these risks, focusing on the UK/EU regulatory framework. As a general matter, it argues that the broadly principles and conduct-based approach of UK consumer credit regulation provides the flexibility necessary for regulators and market participants to respond dynamically to these risks. However, this approach could be enhanced through the introduction of more robust product oversight and governance requirements for firms in relation to their use of ML systems and processes. Supervisory authorities could also themselves make greater use of ML and Big Data techniques in order to strengthen the supervision of consumer credit firms. Finally, the chapter notes that cross-sectoral data protection regulation, recently updated in the EU under the GDPR, offers an important avenue to mitigate risks to consumers arising from the use of their personal data. However, further guidance is needed on the application and scope of this regime in the consumer financial context.

Comparing Two Novel Hybrid MRDM Approaches to Consumer Credit Scoring Under Uncertainty and Fuzzy Judgments

In the recent years, various statistical and computational intelligence or machine learning techniques have contributed to the progress of automation or semiautomation for measuring consumer credit scoring in the banking sector. However, most of the Taiwanese commercial banks still rely on seasoned staffs’ judgments on making the final approvals or rejections. To enhance the understanding and transparency of a decision support system (or model) that can assist bank staffs on making their consumer credit loan decisions—while uncertainty exist—is of high business value. One of the promising approaches is multiple rule-based decision-making (MRDM), a subfield of the hybrid multiple criteria decision-making that leverages the advantages of machine learning, soft computing, and decision methods (or techniques). The MRDM approach reveals comprehensible logics (rules or patterns) that can be justified and compared with the existing knowledge of veterans to reinforce the confidence of their judgments. Therefore, in the present study, we propose and compare two MRDM approaches in assisting decision makers on the consumer credit loan evaluations. A set of historical data from a commercial bank in Taiwan is analyzed for illustrating the plausible pros and cons of the two approaches with discussions.

Consumer credit scores as a novel tool for identifying health in urban U.S. neighborhoods

PurposeCredit scores may operate as a socioeconomic indicator of health: they represent cumulative financial history that directly influences ability to access financial and nonfinancial resources related to health. Yet, little is known about the relationship of credit score and health or to traditional measures of socioeconomic position (SEP). Our objectives were to (1) evaluate the association between area-level credit score and individual self-rated health and (2) compare credit score to traditional markers of area-level SEP in predicting self-rated health. MethodsEquifax estimates of average household credit score in 2015 among nine-digit zip code regions were combined with a representative survey of 2083 residents of Philadelphia to estimate the correlation with income, housing value, education, and occupational status and then predict the odds of self-rated health for credit score and each SEP measure. ResultsCredit score was moderately correlated with SEP markers (r = −0.78 to 0.49). After adjusting for area- and individual-level SEP and demographic factors, each SD increase in credit score is associated with 26% greater odds of better self-rated health (odds ratio = 1.26, 95% confidence interval: 1.09–1.46). Credit score had a larger effect size than other SEP markers. ConclusionsCredit score may be a useful complement to traditional measures of SEP in assessing health outcomes.

Nonlinear relationships in a logistic model of default for a high-default installment portfolio

This study uses data on consumer credit provided by a German retail and trading company along with generalized additive models to analyze nonlinear relationships and their effect on predicting the probability of default in the context of consumer credit scoring. In particular, this study examines which aspects of the contract and which characteristics of the debtor are nonlinearly related to the probability that specific debtors will default on the loans they are seeking to obtain. The findings of our analysis provide clear empirical evidence that certain contract and debtor characteristics have nonlinear and nonmonotonic effects on the probability that a specific debtor will default on their consumer credit. The results of our analysis and the forms that these nonlinear relationships take are interpreted with regard to their economic relevance. Specifically, the study shows that the nonlinear relationships identified are economically relevant because they take into account the costs of a range of misclassification errors. On the basis of this evidence, the study recommends that analysts include nonlinear relationships in the models they design to predict consumer default, although including such relationships increases the complexity of the models that are applied for this purpose.

Variable selection in classification model via quadratic programming

ABSTRACTVariable selection is an important decision process in consumer credit scoring. However, with the rapid growth in credit industry, especially, after the rising of e-commerce, a huge amount of information on customer behavior is available to provide more informative implication of consumer credit scoring. In this study, a hybrid quadratic programming model is proposed for consumer credit scoring problems by variable selection. The proposed model is then solved with a bisection method based on Tabu search algorithm (BMTS), and the solution of this model provides alternative subsets of variables in different sizes. The final subset of variables used in consumer credit scoring model is selected based on both the size (number of variables in a subset) and predictive (classification) accuracy rate. Simulation studies are used to measure the performances of the proposed model, illustrating its effectiveness for simultaneous variable selection as well as classification.

Public Policy Implementation to Reduce Credit Score Discrimination: Beyond Enacting Legislation to Reduce Socioeconomic Disparities

The credit reporting agencies are to consumers what credit rating agencies are to banks and other vendors. That is to say that, Transunion, Equifax and Experian are the credit reporting agencies (CRA’s) that collect data on consumers and runs risk assessments on the consumers that seek credit on the behalf of their costumers. The credit reporting agencies also allow vendors and lenders to report payment information to them as well. According to the Consumer Financial Protection Bureau (CFPB) and the Federal Trade Commission (FTC) there is often erroneous reporting that can occur in the process of reporting data, which negatively affects consumer credit scores. In comparison the Credit Rating Agencies for banks and federal loans Standard and Poor’s provides rating to government entities and banks. Consumer credit is accessed whenever a person seeks credit from a creditor. It is also accessed for some employers and rental properties.