A hidden Markov model of credit quality

This paper presents a hidden Markov model of credit quality dynamics, and highlights the use of filtering-based estimation methods for models of this kind. We suppose that the Markov chain governing the 'true' credit quality evolution is hidden in 'noisy' or incomplete observations represented by posted credit ratings. Parameters of the model, namely credit transition probabilities, are estimated using the EM algorithm. Filtering methods provide recursive updates of optimal estimates so the model is 'self-calibrating'. The estimation procedure is illustrated with an application to a data set of Standard & Poor's credit ratings.