Algorithmic Discrimination in Digital Lending: Empirical Evidence from Emerging Markets

Digital lending platforms have extended credit to hundreds of millions of previously unbanked borrowers across emerging markets, fuelling hopes that algorithmic decision-making could eliminate the biases embedded in human lending. Yet a growing body of empirical work documents persistent discriminatory outcomes in algorithmic credit systems. This paper synthesises and extends this literature by integrating evidence from three convergent streams: (1) audit studies of proxy discrimination mechanisms in machine learning credit models; (2) cross-country empirical analyses of the fintech gender gap; and (3) regulatory mapping of algorithmic oversight frameworks across emerging market jurisdictions. Drawing on documented findings that only 21% of women use fintech services compared to 29% of men across 28 economies, SHAP-based audit evidence demonstrating that demographic proxies —including marital status, device type, and age—reconstruct protected gender attributes from non-sensitive features with ROC-AUC scores of up to 0.65, and the documented 12-percentage-point gender gap in bank access in Sub-Saharan Africa that has widened since 2011, we identify three recurring mechanisms through which discriminatory outcomes reproduce in algorithmic lending: redundant encoding of protected attributes through correlated proxies; training data that reflects pre-existing structural inequalities; and regulatory frameworks that have not been updated to address algorithmic decision-making. We document that no major emerging market jurisdiction reviewed mandates algorithmic impact assessment as a condition of digital lender licensing. We propose a three-tier governance framework—pre-deployment auditing, continuous monitoring, and mandatory third-party review—grounded in analogies from pharmaceutical and environmental regulation, and call for a reorientation of financial inclusion metrics from access coverage to equitable access conditions. Keywords: algorithmic discrimination; digital lending; machine learning fairness; emerging markets; proxy discrimination; financial inclusion; fintech gender gap; regulatory governance