Credit default swaps and CreditGrades: Evidence from the Nordic markets

Abstract: This study examines and compares theoretical CDS spreads created by a structural framework with empirical CDS spreads. The model employed is the CreditGrades model based on the Merton framework from 1974 which calculate default probabilities and credit spreads from balance sheet and equity data. The aim is to measure how well the model can explain the observed CDS spreads and if it has any predictive ability. The model is tested for 22 companies in the Nordic market. Regression analysis is used to measure the explanatory power of the model. It is tested for the period between 2005 and 2009 and for two subperiods, 2005-2007 and 2007-2009. The model was found to have limited explanatory power with R-square value ranging from 0 to 21 percentages. Even though the explanatory value is low the CDS spreads obtained through CreditGrades are significant for 19 companies during 2005-2009 and 21 companies during 2007-2009. The predictive ability of the model is inconclusive with about a third yielding significant results for the one day lagged model and third of the companies’ CDS spreads are significantly autocorrelated with its lagged variable. The residuals were found to be highly cross-correlated. Principle component analysis reveals that 20-50 % of the variation in the residual can be explained by a systematic component not related to the company specific information. We propose the use of a counterparty risk index. With the inclusion of the index the R-square value is strengthened. The index is significant for 21 companies during the entire sample period and all of the companies during the second half of the sample.