Assessing the credit risk of bank loans using an extended Markov chain model

In this paper, we adopted a continuous-time non-homogeneous mover-stayer model for the measurement of the credit risk associated with bank loans. This model is an extension of a Markov chain model. Furthermore, we extracted the time varying risk premium to convert the mover-stayer model to a risk-neutral mover-stayer model. This paper draws a number of conclusions and makes a number of important contributions. First, we determined that the mover-stayer model is better suited than the Markov chain model in estimating the credit risk of loans, according to likelihood ratio statistics. Second, we found that borrowers of investment grades are less likely to remain at their original rating. On the other hand, rating classes had a strong tendency to be downgraded, inferring the likelihood that downgrade momentum is an element of rating behavior. However, rating migration did not indicate the existence of upgrade momentum. Third, we estimated time-varying risk premium to transfer transition matrices to risk-neutral transition matrices. Fourth, estimated default probabilities match business cycle indicators particularly well. Finally, estimation procedures are easy to follow and implement. Consequently, the findings in this study have important implications for the management of risk assumed by financial institutions.