A lot of the theory of securities markets revolves around information: securities prices respond to changes in available information, you want to provide incentives for people to produce information, some kinds of information should be equally available to everyone, other kinds of information you should be able to trade on, etc. In the conventional model, rating agencies are information providers: they produce information that is useful to market participants, and thereby improve the functioning of the markets.
Well, forget all that. Nate Silver has the best article I've seen yet on S&P's sovereign debt ratings, and the summary is that it isn't pretty. Some of the things Silver finds, using some publicly available data and Stata, are:
- Debt-to-GDP ratio alone is a better predictor of default risk than an S&P rating (meaning that the rating subtracts information provided by the debt-to-GDP ratio).
- S&P ratings have almost no correlation with future default risk.
- S&P rates European countries higher than other countries, all other things being equal—and look where that got us.
- S&P ratings are serially correlated, which means they incorporate new information especially slowly.