Multivariate markov chain model for loan default risk modeling in microfinance institutions: a case study of the Fundhouse finance
- Author
- Mahlahla, Lloyd
- Title
- Multivariate markov chain model for loan default risk modeling in microfinance institutions: a case study of the Fundhouse finance
- Abstract
- This study applies a Multivariate Markov Chain (MMC) model to analyze and predict loan default risk in microfinance institutions, focusing on FundHouse Finance in Zvishavane, Zimbabwe. With the microfinance sector facing rising default rates and traditional models proving inadequate, the research aims to model borrower behaviour dynamically across multiple loan states such as current, delinquent, pre-default, default, prepayment, and recovered. The study begins by outlining the problem and objectives, then reviews credit risk modelling literature, highlighting the limitations of structural and reduced form models. Using a positivist and quantitative approach, the researcher collected data from 2000 borrowers and incorporated macroeconomic indicators to estimate transition probabilities. Diagnostic tests, including the Chow test and Chapman-Kolmogorov test, validated the model's structure and predictive ability. The MMC model outperformed logistic regression, achieving a Brier score of 0.1774 and revealing key insights such as high default persistence (88.24%) and a 45.74% delinquency-to-default transition rate, which is useful for early risk detection. Though strong in overall accuracy (89%), the model underperformed in rare class prediction, suggesting a need for hybrid enhancements. The study concludes that MMC models offer superior accuracy, adaptability, and risk insight, and recommends integrating bias-mitigating algorithms and hybrid approaches to strengthen microfinance risk management.
- Date
- June 2025
- Publisher
- BUSE
- Keywords
- Multivariate Markov Chain (MMC)
- microfinance institutions
- loan default risk
- Supervisor
- Ms. Hlupo
