IMF published a working paper on application of machine learning in assessing credit risk. This paper reviews the underlying challenges in assessing credit risk of particularly small borrowers and discusses the most prominent machine learning techniques applied in assessing credit risk of borrowers for a nontechnical audience.
The paper examines potential strengths and weaknesses of machine-learning-based credit assessment by presenting core ideas and the most common techniques in machine learning for the nontechnical audience; it also discusses the fundamental challenges in credit risk analysis. The paper presents the main elements of prudent lending in the context of the five Cs of credit and agency problems. The five Cs of credit are capacity, capital structure, coverage, character, and conditions. The paper also discusses main machine learning tools and techniques and examines the strengths and weaknesses of machine learning-based lending in contrast with the traditional methods.
The paper highlights that machine learning has certain weaknesses that should be taken into consideration when applying for credit risk assessment:
- Heavy reliance on learning from data, particularly in a context where the size of the sample is considerably larger than traditional ways of credit scoring, could result in noisy information playing a role in driving results of credit analysis and leading to financial exclusion of creditworthy applicants. The bias in the sample should be identified and avoided by analysts as much as possible to avoid digital financial exclusion.
- Machine learning may not capture structural changes in a timely manner, because arrival of informative data may be slow due to the lengthy process of default observation. This could negatively impact fintech lenders that rely on machine learning to assess borrowers without evaluating the relevance of data used for training the model for new applicants.
- Borrowers may realize and counterfeit certain indicators that drive their credit score, thus decreasing the relevance of those features for new applicants.
- Machine learning is exposed to some of the key concerns in econometrics, most importantly the endogeneity and selection bias problem. The analyst should check the sample to ensure proper treatment of these issues and avoid superfluous results. Nonetheless, proper choice of risk drivers makes these issues less of a concern for credit risk assessment and default outcome prediction.
Overall, the paper concludes that fintech credit has the potential to enhance financial inclusion and outperform traditional credit scoring by leveraging nontraditional data sources to improve the assessment of the borrower’s track record, appraising collateral value, forecasting income prospects, and predicting changes in general conditions. Nevertheless, the emerging market economies may face some challenges in reaping the benefits while ensuring that the development of fintech credit does not expose systemic risks to the financial system. Given the central role of data in machine-learning-based credit analysis, it should be legally and technologically possible to gather digitalized data reliably from various sources and avoid noisy and biased data as much as possible. As a complement for high-quality data availability, cyber-security measures should be in place because of the sensitivity of credit information. Moreover, the machine learning analysts should consistently review machine learning models for credit rating to avoid potential weaknesses of naive application of machine learning.
Related Link: Working Paper
Keywords: International, Banking, Financial Inclusion, Fintech, Machine Learning, Credit Risk, Big Data, Systemic Risk, IMF
OSFI has set out the near-term priorities for federally regulated financial institutions and federally regulated private pension plans for the coming months until March 31, 2022.
Under the Italian G20 Presidency, BIS Innovation Hub and the Italian central bank BDI launched the second edition of the G20 TechSprint on the lookout for innovative solutions to resolve operational problems in green and sustainable finance.
EBA proposed the regulatory technical standards on a central database on anti-money laundering and countering the financing of terrorism (AML/CFT) in EU.
ECB published its response to the targeted EC consultation on the review of the bank crisis management and deposit insurance framework in EU.
ACPR published Version 1.0.0 of the RUBA taxonomy, which will come into force from the decree of January 31, 2022.
BCBS, CPMI, and IOSCO (the Committees) are inviting entities that participate in market infrastructures and securities markets through an intermediary as well as non-bank intermediaries to complete voluntary surveys on the use of margin calls.
ECB published Decision 2021/752 to amend Decision 2019/1311 on the third series of targeted longer-term refinancing operations or TLTRO III.
The Central Bank of Ireland published Version 2.7 of the draft credit data template and rules for monthly AnaCredit reporting by banks.
OSFI proposed revisions to the Basel Capital Adequacy Reporting (BCAR) and leverage requirements returns for the 2023 reporting, with the comment period ending on July 09, 2021.
EBA published a discussion paper on review of the standardized nonperforming loans (NPL) transaction data templates, along with the proposed revised NPL data templates.