Mathematical Modelling of Credit Risk in Nigerian Deposit Money Banks Using Logistic Regression

This study applies logistic regression to the modelling and prediction of credit risk in Nigerian deposit money banks, developing a quantitative credit scoring model using borrower characteristic data from consumer and small business loan portfolios. Loan portfolio quality is a persistent challenge for Nigerian banks, with the ratio of non-performing loans to total loans consistently exceeding prudential benchmarks set by the Central Bank of Nigeria, and the absence of widely adopted quantitative credit assessment frameworks in mid-tier institutions contributes to suboptimal loan origination decisions. A dataset of 3,200 closed loan accounts from three mid-tier deposit money banks is assembled through research partnership agreements, with each observation containing borrower-level features including income level, employment type, monthly debt service ratio, collateral value, credit bureau score, loan purpose, and loan tenure. The dataset spans origination years 2019 to 2022, allowing adequate time for loan performance observation. Binary logistic regression is estimated with default status as the outcome variable, and predictor variable selection is conducted using a combination of univariate significance testing and variance inflation factor screening for multicollinearity. Model discrimination is assessed using the Gini coefficient and Kolmogorov-Smirnov statistic, and calibration is evaluated using the Hosmer-Lemeshow test. The developed scorecard achieves a Gini coefficient of 0.61 on the holdout sample. Keywords: credit risk, logistic regression, credit scoring, non-performing loans, deposit money banks