Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
There is a large and growing literature on how to model the dynamics of the default-free term structure to fit the observed historical data. Much less is known about how best to model the dynamics of defaultable yield curves. This paper develops a class of defaultable term structure models that is tractable enough to be empirically implemented and flexible enough to capture some important behaviors of the credit spreads in the data. We compare two non-nested models within this class using a Bayesian estimation technique, which helps to solve the problem of latent state variables. The Bayesian approach also enables us to test the two non-nested models on the basis of the Bayes factor. The results strongly suggest that models with constant transition probabilities will not be able to fit the observed dynamics of inter-rating spreads.