FinTech in Financial Inclusion: Machine Learning Applications in Assessing Credit Risk

IMF Working Papers, May 17, 2019

Source details

Canonical URL: FinTech in Financial Inclusion: Machine Learning Applications in Assessing Credit Risk

Recent advances in digital technology and big data have allowed FinTech lending to emerge as a potentially promising solution to reduce the cost of credit and increase financial inclusion.
The paper addresses the opacity of machine learning (ML) methods in FinTech credit for nontechnical audiences by:
- presenting core ideas and the most common techniques in ML for the nontechnical audience; and
- discussing the fundamental challenges in credit risk analysis.

FinTech credit can enhance financial inclusion and outperform traditional credit scoring by:
- leveraging nontraditional data sources to improve assessment of the borrower’s track record;
- appraising collateral value;
- forecasting income prospects;
- predicting changes in general conditions.

Because data are central to ML-based analysis, data relevance must be ensured, especially in:
- situations when a deep structural change occurs;
- situations when borrowers could counterfeit certain indicators;
- situations when agency problems arising from information asymmetry could not be resolved.
To avoid digital financial exclusion and redlining, variables that trigger discrimination should not be used to assess credit rating.
ML methods remain largely a black box for nontechnical audiences, raising transparency and interpretability concerns.

Subject: Credit, Credit ratings, Credit risk, Financial institutions, Financial regulation and supervision, Loans, Machine learning, Money, Technology
Keywords: bears risk, borrower default, capital structure, Credit, Credit ratings, Credit risk, Credit Risk Assessment, credit risk driver, credit scoring, Financial Inclusion, FinTech Credit, FinTech credit company, Global, Loans, Machine Learning, machine learning technique, ML analysis, ML analyst, ML evaluation, ML model, neural network, supervised machine learning model, WP