Algorithmic Bias in Automated Credit Scoring Systems and Financial Exclusion in Nigeria: A Critical Empirical Investigation

Automated credit scoring systems powered by machine learning are increasingly deployed by Nigerian financial institutions and fintech companies to determine creditworthiness, yet the mechanisms through which these systems may perpetuate or amplify financial exclusion along demographic lines remain critically underexamined in the African context. This study conducts a critical empirical investigation of algorithmic bias in automated credit scoring systems and its contribution to financial exclusion in Nigeria, combining technical audit methods with socio-legal analysis. The research employs a mixed-methods design incorporating algorithmic fairness audits of credit scoring models used by three fintech lenders, conducted through collaboration with participating institutions under non-disclosure agreements, supplemented by a survey of 600 loan applicants in Lagos and Kano on rejection experiences. Technical analysis applies group fairness metrics including demographic parity, equalised odds, and calibration across gender, geographical zone, and informal sector employment status. Findings reveal statistically significant disparities in approval rates by geopolitical zone and formal employment status that are not fully explained by credit risk differentials, indicating potential indirect discrimination. The study develops an original Contextualised Algorithmic Fairness Framework for Sub-Saharan credit markets that accounts for data infrastructure limitations and informal economy dynamics absent from existing fairness frameworks developed in Global North contexts. Policy recommendations address the Nigeria Data Protection Commission's emerging algorithmic accountability regulatory agenda. Keywords: algorithmic bias, credit scoring, financial exclusion, Nigeria, fairness metrics.