A Double mover-stayer model for credit ratings

Modeling the dynamics of credit ratings plays an important role in credit risk management and portfolio risk management. Both institutional and individual investors make use of credit ratings produced by some well-known international credit ratings agencies for analyzing the financial wealth of firms and other jurisdictions. In this paper, we develop a model for stochastic movements of ratings of a population of credit entities based on a double moverstayer (DMS) model with three independent discrete-time and discrete-state models. This provides a more flexible and general paradigm to incorporate the heterogeneity of rating behavior and to discriminate and classify a population of credit entities according to their ratings behavior. A two-stage procedure is adopted to estimate the DMS model. We first employ the maximum likelihood approach to uncover the dynamics of transitions of each credit rating sequence and to identify the relationship among ratings sequences in a population of credit entities. We then develop an efficient adaptive estimation method to estimate the model parameters. Numerical experiments are conducted to illustrate the practical implementation of the model.