We examine the correlation in credit risk using credit default swap (CDS) data. We find that the observable risk factors at the firm, industry, and market levels and the macroeconomic variables cannot fully explain the correlation in CDS spread changes, leaving at least 30 percent of the correlation unaccounted for. This finding suggests that contagion is not only statistically but also economically significant in causing correlation in credit risk. Thus, it is important to incorporate an unobservable risk factor into credit risk models in future research. We also find, consistent with some theoretical predictions, that the correlation is countercyclical and is higher among firms with low credit ratings than among firms with high credit ratings.