Financial Inclusion and Alternate Credit Scoring: Role of Big Data and Machine Learning in Fintech

Abstract

We use unique and proprietary data from a large Fintech lender to analyze whether
alternative data captured from an individual's mobile phone (mobile/social footprint) can
substitute for traditional credit bureau scores and improve financial inclusion. Variables
that measure a borrowers' digital presence, such as the number and types of apps installed,
measures of social connections and borrowers' "deep social footprints" based on call logs,
significantly improve default prediction and outperform the credit bureau score. Using machine
learning-based prediction counterfactual analysis, we find that alternate credit scoring
based on the mobile and social footprints can expand credit access for individuals who lack
credit scores without adversely impacting the default outcomes. The marginal
benefit of using alternative data for credit decisions are likely to be higher for borrowers with
low levels of income and education, as well as borrowers residing in regions with low levels
of financial inclusion.

Keywords: Fintech, Big data, Credit scores, Financial inclusion, Lending, Machine Learning, Mobile footprint, Prediction Counterfactual, Social footprint, Social capital

JEL Classification: G20, G21, G29