Application of Logistic Regression in Predicting Loan Default Among Microfinance Clients in Abuja FCT

Loan default prediction is critical to the financial sustainability of microfinance institutions in Nigeria, where limited credit bureau infrastructure makes statistical default prediction models particularly valuable for risk-based lending decisions. This study applies logistic regression to develop a loan default prediction model for microfinance clients in the Federal Capital Territory (FCT), Abuja. Loan portfolio data from three microfinance banks covering 1,200 completed loan cycles were obtained, capturing borrower demographic characteristics, loan amount, loan tenure, repayment schedule type, prior loan history, and default status. The dataset was split 70:30 into training and validation sets. Logistic regression with forward stepwise selection was estimated. Model performance was assessed by area under the ROC curve, sensitivity, specificity, and Hosmer-Lemeshow goodness-of-fit test. Loan default rate in the sample was 21.4 percent. The final model (AUC = 0.81, Hosmer-Lemeshow p = 0.46) included six significant predictors: prior default history (OR 4.7), loan amount relative to income (OR 3.1), urban-rural location (OR 0.58 for urban), business type (OR 2.4 for trading), loan tenure exceeding 12 months (OR 2.1), and group-based lending participation (OR 0.51). Optimal classification threshold at 0.38 yielded 76 percent sensitivity and 72 percent specificity. The study provides a deployable statistical default risk scorecard for FCT microfinance institutions and recommends its integration into loan officer decision-support systems. Keywords: logistic regression, loan default, microfinance, Abuja FCT, credit risk