Using a Bootstrap Approach to Rate the Raters

This paper compares the accuracy of credit ratings of Moody's and Standard&Poor's based on 11,428 issuer ratings and 350 defaults in several datasets from 1999 to 2003. In order to improve the robustness of the results, the paper:

• Uses nonparametric bootstrap approaches;
• Has robustness checks to control for the impact of:
  • Watchlist entries,
  • Staleness of ratings,
  • Effects of unsolicited ratings on the results.

The paper finds evidence that Moody's performs slightly better than S&P.

• The overall result is driven primarily by the noticeable differences in the predictive power of defaults in favor of Moody's in the estimation periods 2001, and particularly 2002.

The authors conclude that, despite the fact that not all periods support the superiority of Moody's, the results of the base case are robust for variations of the mapping procedure, as well as for different adjustments to account for staleness in ratings and for unsolicited ratings.