I read yet another story on how the credit rating agencies (Standard and Poors, Moodys, Fitch, et al.) failed at their task of rating credit instruments (bonds, derivatives, etc.) to reflect the risk of those instruments.
The credit rating business works like this: Companies that want to issue instruments pay the agencies to rate them (AAA, BBb, etc.) for risk. The companies then sell instruments on the market to buyers who look at the rating when deciding how much to pay. The higher the rating, the higher the price that the companies will get, and the less risk the buyers think they are taking on.
Now there's a clear conflict of interest: Companies paying agencies want a high rating, agencies have an incentive to give high ratings (in exchange for bigger fees, more future business, etc.), but buyers depend on agency ratings.
Unfortunately, buyers cannot pay the agencies. They cannot pay in advance (they are not sure they will buy until AFTER ratings are done), and -- even if they did -- they would suffer from free-riding (if some buyers pay for the rating, other buyers can use that information without paying).
So, how do we fix the system (improving accuracy) while maintaining the current payment relationships? Change incentives in this way:
- Set a standard fee for rating that depends on the type of instrument, the size of the issue, etc. Such standardization would remove one obvious problem (negotiated fees) while giving agencies an incentive to turn down business that's too complicated to understand.
- Track the performance of all instruments rated by an agency in a given credit category (e.g., AA-) against all others in that category. Those that underperformed (price fell below the average) are probably riskier than the initial rating indicated, i.e., the agency was overoptimistic. Performance ratings should be weighted by the age of the instrument/rating, the size of the issue, etc.
- Adjust each agency's fees to equal some fraction of the standard fee based on that agency's performance relative to other agencies, e.g., 80% if ratings are too high (missing hidden risk) or 110% if the ratings are too low (seeing non-existing risk).
Note that this system depends on competition, so it's important to avoid cartels, oligopoly, etc. Even if standard prices are "set" (good place for a regulator), the competition will take place ex-post, when markets "reveal" the quality of the agencies' work.
Also note that #2 alone could do quite a lot to improve matters, but it's nice to link income to performance (#1 and #3).
Bottom Line: We need credit ratings, but we need to punish rating agencies that do a bad job, and the market is the best place for punishment.
Addendum: Cross-posted at AngryBear