Credit Rating Transitions and Observable Covariates

Investors benefit from measuring and forecasting potential changes in the credit risk of securities. This paper examines the linkage between observable macroeconomic variables and credit rating transitions of U.S. residential mortgage-backed securities (RMBS). I specify a fully-parametric, heterogeneous credit rating intensity model, then use maximum likelihoodand forward selection methods to determine the significant covariates and corresponding rating transition intensities over time. These estimated intensities are converted to transition probabilities over intervals by numerically solving simple forward differential equations. I use a modified exponential distribution model to generate rating transition replicate records in order to study their sample distributions. My major findings are that RMBS with different levels of risk have different statistically significant covariates to which they are related and have quite different sample distributions. There is no evidence that rating agencies do not rate through the business cycle. RMBS investors should be alert to the fact that AAA securities of different RMBS classes have different transition probabilities over time. Also, the transition probabilities generated using covariate data from 2007 forecast a sharp increase in the probability of credit down-grades for Subprime RMBS after the 2007 crisis. Ratings data is from the Standard & Poor's CreditPro Structured Finance Database from 1978 to the second quarter of 2007.