An Investigation of Default Prediction Models in the Taiwan Banking Sector

Abstract

This study investigates the determinants of bank default probability and the predictive performance of the logit and the hazard models. Using accounting and market price information in the Taiwan bank sector for the period 1999 to 2010, the result shows both models provides adequate default predicative performances. Furthermore, it finds banks’ default probabilities are negatively and significantly related to distance to default and stock returns, but positively and significantly related to loan loss provision ratios.

Keywords: bank, distance-to-default, hazard model, logit model

JEL Classification: E50, G18, G20