DOI: https://doi.org/10.36347/sjebm.2026.v13i09.004

Pages: 451-460

Credit risk is central to the financial sustainability of microfinance institutions (MFIs), yet borrower-level evidence that combines repayment history, financial capacity, institutional contact, loan purpose, and shock exposure remains limited in Bangladesh. This study examines the determinants of microfinance loan default and translates the empirical results into borrower-sensitive risk-mitigation strategies. The cross-sectional dataset contains 400 borrower-loan observations collected from 15 branches of four anonymized NGO-MFIs across seven districts of Bangladesh between June and August 2026. Default is defined as more than 30 days past due, with 60- and 90-day thresholds used in sensitivity analysis. The empirical strategy combines descriptive and bivariate analysis with multivariable logistic regression using HC3 robust standard errors, average marginal effects, multicollinearity diagnostics, nested model comparison, 10-fold out-of-fold validation, a probit specification, and alternative delinquency thresholds. Ninety-two loans (23.0%) met the primary default definition. Previous loan-cycle experience was associated with lower default odds (adjusted odds ratio [aOR] 0.787; p=0.020), whereas previous late payment (aOR 1.879; p=0.038), external shock exposure (aOR 1.719 per additional shock; p<0.001), and consumption/emergency-purpose borrowing (aOR 4.013; p=0.002) were associated with higher risk. The final model achieved an apparent AUC of 0.740 and a 10-fold out-of-fold AUC of 0.702. The findings support dynamic repayment-history screening, shock-sensitive assessment and restructuring, purpose-specific appraisal, affordability checks, and early intervention for watchful loans, while cautioning against automated exclusion based on a moderately discriminating model.