Comments on: Subprime with a Ski Chalet: The Triumph of Data Over Common Sense http://www.concurringopinions.com/archives/2008/05/subprime_with_a.html The Law, the Universe, and Everything Sun, 22 Nov 2009 08:31:00 -0700 http://wordpress.org/?v=2.8.3 hourly 1 By: James Grimmelmann http://www.concurringopinions.com/archives/2008/05/subprime_with_a.html/comment-page-1#comment-49254 James Grimmelmann Wed, 14 May 2008 17:08:04 +0000 http://www.solove.org/archives/2008/05/subprime-with-a-ski-chalet-the-triumph-of-data-over-common-sense.html#comment-49254 There's another explanation for the failure of the rating agencies, one that doesn't suggest recklessness or fraud liability as the answer. <i>They were outsmarted</i>. The rating agency's goal is to find a metric that enables them to quantify the risk of default on an instrument. But as soon as they have a metric in place, the gaming begins. Loewenstein's article (which I found unsatisfying because it focuses on only one set of agency problems in a system utterly riddled with them) details how issuers were able to design instruments optimized against the rules used by the rating agencies. In other words, every time a rating agency is too harsh on a proposed instrument, it's shopped to another agency. Every time a rating agency is too lenient, it's presented with lots more that have been rigged in the same fashion. To me, the article is a story about how a loose collection of very smart mortgage-packagers found a mistake in the rating agencies' algorithms, and then systematically exploited it. This view suggests a few things. First, agencies do need to have their incentives properly aligned with those who <i>rely</i> on ratings. The current system, in which they're paid by issuers, certainly creates the risk of a structural skew in their outlook. Of course, since ratings are an information good, there's a standard public-goods problem in having them be paid by the client-side. The "skin in the game" proposals would help, but we could use something further, as well. Second, it's far from clear that greater transparency in rating would improve accuracy. If a rating agency is too up-front about the characteristics it's looking for, it's an open invitation to loopholing. Maybe we need more opacity, so that issuers would design their instruments based on financial fundamentals, not on the specific criteria used by rating agencies. Think of it as being like the problem of transparency in search algorithms--except that fewer people will shed tears for a financial engineer whose investment vehicle is misrated than would for a web site that's misranked. Third, there's something deeply messed-up about the quasi-official status the rating agencies enjoy. The actual letter-grades matter primarily because some institutions are legally bound to choose investments on the basis of their rating. The rating becomes a substitute for independent judgment. There’s another explanation for the failure of the rating agencies, one that doesn’t suggest recklessness or fraud liability as the answer.

They were outsmarted.

The rating agency’s goal is to find a metric that enables them to quantify the risk of default on an instrument. But as soon as they have a metric in place, the gaming begins. Loewenstein’s article (which I found unsatisfying because it focuses on only one set of agency problems in a system utterly riddled with them) details how issuers were able to design instruments optimized against the rules used by the rating agencies. In other words, every time a rating agency is too harsh on a proposed instrument, it’s shopped to another agency. Every time a rating agency is too lenient, it’s presented with lots more that have been rigged in the same fashion. To me, the article is a story about how a loose collection of very smart mortgage-packagers found a mistake in the rating agencies’ algorithms, and then systematically exploited it.

This view suggests a few things. First, agencies do need to have their incentives properly aligned with those who rely on ratings. The current system, in which they’re paid by issuers, certainly creates the risk of a structural skew in their outlook. Of course, since ratings are an information good, there’s a standard public-goods problem in having them be paid by the client-side. The “skin in the game” proposals would help, but we could use something further, as well.

Second, it’s far from clear that greater transparency in rating would improve accuracy. If a rating agency is too up-front about the characteristics it’s looking for, it’s an open invitation to loopholing. Maybe we need more opacity, so that issuers would design their instruments based on financial fundamentals, not on the specific criteria used by rating agencies. Think of it as being like the problem of transparency in search algorithms–except that fewer people will shed tears for a financial engineer whose investment vehicle is misrated than would for a web site that’s misranked.

Third, there’s something deeply messed-up about the quasi-official status the rating agencies enjoy. The actual letter-grades matter primarily because some institutions are legally bound to choose investments on the basis of their rating. The rating becomes a substitute for independent judgment.

]]>