Altman versus Merton - Are corporate credit rating changes new information?

Abstract: With the relative performance between accounting-based and option-based approaches for default prediction being a key subject in previous research, combined with criticism against the rating agencies’ timeliness in assigning credit ratings, surprisingly few attempts have been made to investigate the models’ usefulness in predicting corporate credit ratings. This thesis is the first to investigate the Merton and Altman’s Z-score models’ relative performance in predicting future credit rating changes. In addition, any asymmetry in the predictive power for downgrades versus upgrades is examined. The thesis is performed using data from 1,450 non- financial firms rated by Standard and Poor’s between 2002 and 2013. Using logit regressions, with rating changes as dependent variables and distances-to-default and Z-scores as independent variables, we find both models to have some predictive power for rating changes within one year but the goodness of fit is mediocre and the marginal effects are low. Although the Z-score shows slightly better results, in terms of percentage correctly predicted outcomes, it is concluded that no clear difference in the relative performance can be found. Meanwhile, it is concluded that the Merton model has more predictive power for downgrades than for upgrades while no such asymmetry can be found for the Z-score. Our results supports the notion of lagged credit ratings, which could be detrimental for the economy at large, and may work as a starting point for building more accurate prediction models to lessen the effect of rating announcements. It is further implied that credit rating agencies could be slower to assign lowered credit ratings, as compared to rating upgrades, which could possibly be explained by the interdependency between the rating agencies and the issuers.