Le crédit scoring: une nouvelle voie pour réduire les problèmes de remboursement et améliorer la performance des IMFs

The aim of this research is to develop a scoring model using logistic regression applied to 496 individuals loans from Tunisian MFIs. The results show that the gender, the credit rationing, the house property, other sources of wealth, a fixed wage, and finally the age of the institution are negatively correlated with the probability of default. However, the marital status, the guarantor, the presence of other lending institution, contracting loan in order to implement a new project are positively correlated with the probability of default. The reject inference analysis has shown the consistency of the model’s prediction with the institution decision. Moreover, it made it possible to escape to the subjective judgments of the loan officers.