Parnes, Dror, "Time Series Patterns in Credit Ratings", Finance Research Letters, Vol. 4, No. 4, (December 2007), pp. 217-226.
Abstract: This article offers a substitute setting to simulate credit rating migrations. The internal correlations model tracks time-series movements within credit rating entries, rather than cross-ratings correlations. The proposed nonhomogeneous process is authenticated through the likelihood ratio Dickey-Fuller test, and is found to be statistically and economically significant, by better fitting observed cumulative default rates. Several nonlinear regression models assist to better identify these time-related patterns. The economic structure underlying the time dependency often corresponds to changes in GDP, business cycles, and market risk. Furthermore, significant positive autocorrelation is detected mostly among downgrade probabilities.
Keywords: Credit ratings, Time-series patterns, Dickey-Fuller test, Homogeneous Markov chain, Serial correlation.