The Lex team at the Financial Times introduced last week an article that intimates that Chinese rating agency Dagong’s credit ratings might be worth considering because certain of them are closely aligned, in the FT’s opinion, to the credit default swap (CDS) market spreads.
Fitch Ratings’ Grossman and Hansen pulled them up in a comment letter which explains that CDS spreads provide an imperfect measure when contemplating default probability:
“when using CDS spreads as default risk indicators, it is important to note that spreads can be driven by a number of factors not directly related to an entity’s fundamental creditworthiness, such as the leverage inherent in CDS trading, liquidity conditions, counterparty risk and the risk aversion of market participants.”
More important, in our opinion, is the examination of the superior measure and the environment in which it remains the most reliable measure. Certainly, each measure has its pros and cons, as Fitch points out. Nor will any measure be immune to the parties estimating it: within each measure, for example, each CDS trader has his profitable bets and his losers, and each rating agency has its successes and its failures. (If in general, traders of a certain company’s CDS tend to be poorly informed, a better-informed rating agency analyst might tend to be more accurate, even if it happened to be determined that CDS spreads typically outperform ratings from an accuracy perspective.)
In other words, both the rating agency and the market could be wrong: it therefore serves no purpose to consider a rating to be an informed opinion simply because it tends to agree with the market. In the worst case, a rating agency might actually use CDS spreads as a tool in coming up with the rating, in which case the ratings provided will be highly (if not perfectly) correlated to the CDS spreads. In such a scenario, the FT would doubtless find the rating to be entirely accurate. But what would be the usefulness of the rating?
Rather, we should hold the ratings and the market to a different accuracy standard, testing whether their opinions provided useful and meaningful predictive content, over time, of whatever they were estimating — in this case probability of default. Is there a strong correlation between CDS spreads and eventual default (over a certain tested time horizon); and is there a strong correlation between the default probability rating provided and the eventuality of default?
Posted on November 17, 2010
0