Heterogeneity in Ratings Migration

We explore sources of heterogeneity in rating migration behavior using a continuous time Markov chain. Working in continuous time circumvents of the embedding problem, allows for arbitrary prediction horizons, mitigates the censoring effect, and facilitates term structure modeling. By adopting a Bayesian estimation procedure we are able to estimate for each issuer profile its own continuous time Markov chain generator. Using the Moodys corporate bond default database we identify significant country and industry effects on the determination of default intensity, rating migration volatility and conditional transition probabilities. We tabulate and compare these quantities for different issuer profiles to assess the heterogeneity in the sample. We compare the one year transition probability matrices for different profiles using Jaffry-Shuermann mobility metric. We show that other characteristics such as how long the issuer has been in existence, can also strongly affect the rating migration behavior. We therefore provide support and a tool for tailoring Markov chain generators to individual issuer profiles. The model may possibly be extended to incorporate other time-varying covariates such as age and momentum; this is the focus of ongoing work.