« Paul Krugman: The Market Mystique | Main | links for 2009-03-28 »

Friday, March 27, 2009

Why Did Ratings Agencies Fail?

The failure of ratings agencies to properly price the risky securities at the heart of the financial crisis has been attributed to conflict of interest (being paid by the issuers of the assets they are rating) and shopping for the best rating (get more than one rating, then only make public the highest one). However, an objection to these explanations is that these incentives have always existed, yet the problems did not emerge until recently. Thus, any explanation relying upon these incentives must explain why they did not cause problems until recently.

This article says the answer can be found in the complexity of the assets that are being rated. When the assets are very simple, risk assessment is not very complicated and the dispersion of ratings across agencies is very low. Thus, there is no incentive to shop around. In addition, it is hard for the agencies to become beholden to asset issuers and inflate ratings because such behavior would be transparent enough so as to risk losing credibility. That is, people outside the agencies can independently check and verify the ratings easily so any manipulation of the ratings would be easy to discover, and the revelation that their ratings are inflated would damage their credibility and hence their business.

But all of this changes when the assets become more complex. First, because of the complexity the dispersion of ratings across agencies will increase. Thus, even if the mean rating does not change, the variance of the ratings make it worthwhile to pay for more than one rating and cherry pick the best of the lot, i.e. to shop around. (In numerical terms, suppose assets are rated from 1, which is the highest risk category, to 10 which is the safest. A non-complex asset might have a dispersion of, say, 7.95 to 8.05 among a fixed number of ratings agencies, while a complex asset might have ratings running from 6.50 to 9.50. In both cases, assuming a symmetric distribution, the mean is 8, but the rewards to shopping around are quite different).

Second, it is easier for the issuer to capture the rating agency, i.e. for the agency to produce the ratings the company is looking for, because the complexity makes such behavior harder to uncover. The ability of outside observers to uncover such behavior diminishes when the variance of the ratings goes up.

To be more precise about the incentive to shop around, there is a cost to obtaining one more rating, the fee the firm must pay (though the article below implies the firm can escape the fee it if doesn't like the rating it gets). The benefit is the chance that the new, incremental rating will be higher than the ratings already in hand, and this diminishes as more ratings are collected, i.e. there is a declining marginal benefit. If the fee is relatively low, it will be worthwhile to collect many ratings, and the expected rating outcome - the maximum of the ratings - will be higher as more ratings are collected. However, issuers do not necessarily collect ratings from every ratings firm since the expected benefit of an additional rating may not cover the cost. But if the fees are sufficiently low, if the assets are sufficiently complex, and if the number of firms is sufficiently small - a case that may describe the recent market fairly well - a corner solution will emerge, i.e. it always pays - in expected terms - to collect all the ratings available and then make only the best rating public.

The other side of this, though, is that the degree of distortion falls when the number of ratings agencies is small. That is, the expected maximum rating is increasing in the number of ratings collected (though the increase comes at a decreasing rate, that's why there is a declining marginal benefit to collecting another rating). It depends upon the nature of the underlying distributions, but it's possible - and I think likely - that the distortion from this factor was low due to the fact that there were only, effectively, Moody and Standard & Poor operating in these markets (do we count Fitch too?). If so, if the shopping around distortion is relatively minor because the number of firms is small (and that is highly speculative on my part, and based upon the quick reactions scribbled out above rather than days of careful thought), then the alternative explanation that the ratings agencies were beholden to asset issuers should be given more weight as the likely, predominant explanation for the problems in these markets.

In any case, here's the article:

The origin of bias in credit ratings, by Vasiliki Skreta and Laura Veldkamp, voxeu.org: Most market observers attribute the recent credit crunch to a confluence of factors – excess leverage, opacity, improperly estimated correlation between bundled assets, lax screening by mortgage originators, and market-distorting regulations. Credit rating agencies were supposed to create transparency, provide the basis for risk-management regulation, and discipline mortgage lenders and the creators of structured financial products by rating their assets. Understanding the origins of the crisis requires, at least in part, understanding the failures of the market for ratings. Proposed explanations for ratings bias have broadly fallen into three categories.

It was an honest mistake

New financial instruments were being traded, and rating agencies had no historical return data for these instruments on which to base their risk assessments. These new instruments had a degree of complexity that even financial professionals acknowledged was “far above that of traditional bonds" (Adelson, 2007) and “dizzying" (Zandi 2008). But complexity alone would generate independent errors in ratings, not ratings that were systematically upward-biased and subsequently downgraded in 2008. For this story to make sense, it must be that many raters made the same mistake. For example, they underestimated the correlation of defaults, particularly in residential mortgage-backed securities. This led them to underestimate the risk of a geographically diverse pool of mortgages and to assign such assets inflated ratings.

Agencies were beholden to asset issuers

A host of recent papers explore the conflict of interest that arises when rating agencies' fees are paid by asset issuers. Damiano, Li, and Suen (2008), Bolton, Freixas, and Shapiro (2008), Becker and Milbourn (2008), and Mathis, McAndrews, and Rochet (2008) investigate the extent to which reputation effects can discipline rating agencies who may feel compelled to deliberately inflate their ratings, either to maximise their consulting fees or because the issuer could be shopping for the highest rating.

Asset issuers shopped for ratings

Since, with few exceptions, an asset issuer decides which ratings will be published, he or she can choose to publish only the most favourable rating(s). Former chief of Moody's, Tom McGuire, explains, “The banks pay only if [the ratings agency] delivers the desired rating… If Moody's and a client bank don't see eye-to-eye, the bank can either tweak the numbers or try its luck with a competitor like S&P, a process known as ratings shopping."

Why the trouble emerged recently

While all three of these explanations likely played some role in creating ratings bias, only the first explains why an upward bias appeared recently. Asset issuers have been paying for credit ratings since the 1970s, and, until recently, ratings upgrades were more common than downgrades. Does this mean that the conflict of interest and ratings shopping were not possible sources of the ratings inflation of the last few years and should therefore not be the subject of new regulation?

Our research (Skreta and Veldkamp 2009) looks for a trigger that could explain why the incentive to shop for ratings might have remained dormant until recently. The trigger we identify is an increase in asset complexity. Suppose each rating agency issues an unbiased forecast of an asset's value but asset issuers can shop for ratings. If the announced rating is the maximum of all realised ratings, it will be a biased signal of the asset's true quality. The more ratings differ, the stronger are issuers' incentives to selectively disclose (shop for) ratings.

For simple assets, agencies issue nearly identical forecasts. Asset issuers then disclose all ratings because more information reduces investors' uncertainty and increases the price they are willing to pay for the asset. For complex assets, ratings may differ, creating an incentive to shop for the best rating. There is a threshold level of asset complexity at which shopping becomes optimal and ratings inflation emerges. Furthermore, the link between asset complexity and ratings shopping can work in both directions. An issuer who shops for ratings might want to issue an even more complex asset, to get a broader menu of ratings to choose from. This, in turn, makes shopping even more valuable.

A similar effect might have prompted a recent resurgence in asset issuers pressuring rating agencies to generate favourable ratings. If the guidelines for rating an asset are straightforward and all rating agencies must rate an asset the same way, then there is little pressure an issuer can exert. But if assets become more complex and there are now judgment calls to be made, the agency can legally come to many possible conclusions about what the rating should be. This creates the possibility for conflicts of interest that were previously not present or not so severe. Thus, an increase in asset complexity could have prompted rating shopping by asset issuers and manipulation by ratings agencies. The pattern of downgrades and defaults in the last few years confirms this relationship between asset complexity and over-optimistic ratings – complex CDOs had significantly higher default rates than simple corporate bonds with identical ratings. Similarly, mortgage-backed securities, whose underlying credit risk, correlation risk, and pre-payment risk are notoriously difficult to assess, experienced more widespread downgrades than assets based on other collateral types (Mason and Rosner 2007).2

What does the relationship between asset complexity and the incentives to bias ratings mean for future regulatory efforts? First, the conflict of interest that induces rating agencies to inflate ratings and the ability of asset issuers to shop for the best rating can each independently produce ratings bias. Dealing with one of these problems without addressing the other is unlikely to solve the problem. Second, just because these effects did not produce upward bias in ratings in the 1980s and ‘90s does not mean that the problems in the rating market structure are harmless. There is good reason to think that such incentives were latent and only emerged when assets were sufficiently complex that regulation was no longer detailed enough to keep them in check. Finally, the ability of ratings manipulation and shopping to affect asset prices only exists when the buyers of assets are unaware of the games being played by the issuer and rating agency. While that was likely the case for some buyers two years ago, today major market participants must have some awareness of the perils of relying on selectively disclosed ratings. If investors mentally discount ratings, then this problem has corrected itself. However, if we forego this opportunity to rethink how ratings are provided, the next bout of financial innovation could trigger another round of ratings inflation and the financial market turmoil that ensues.


1 On 26 January 2008, the New York Times quoted Moody's CEO saying “In hindsight, it is pretty clear that there was a failure in some key assumptions that were supporting our analytics and our models." He said that one reason for the failure was that the information quality given to Moody's, both the completeness and veracity, was deteriorating. See also page 10 of the Summary Report of Issues Identified in the Commission Staff's Examinations of Select Credit-rating Agencies, United States Securities and Exchange Commission, 8 July 2008.

2 Other collateral types that began to be securitised well after mortgages are far less complex. The first non-mortgage securitisation was equipment leases, followed by credit cards and auto loans, and, more recently, home equity, lease finance, manufactured housing, student loans, and synthetic structures. All of those types of collateral illustrate tranching structures that are measurably simpler than those for RMBS. They had correspondingly lower default rates for similarly-rated assets.


Adelson (2007), Director of structured finance research at Nomura Securities. Testimony before the Committee on Financial Services, US House of Representatives, September 27, 2007.

Becker, Bo and Todd Milbourn, “Reputation and Competition: Evidence from the Credit Rating Industry," 2008. HBS finance working paper 09-051.

Bolton, Patrick, Xavier Freixas, and Joel Shapiro, “The Credit Ratings Game," 2008. NBER Working Paper No. 14712

Damiano, E, H Li, and W Suen, “Credible Ratings," Theoretical Economics, 2008, 3, 325-365.

Mason, Joseph R. and Josh Rosner, ”Where Did the Risk Go? How Misapplied Bond Ratings Cause Mortgage Backed Securities and Collateralized Debt Obligation Market Disruptions," 2007. SSRN Working Paper #1027475.

Mathis, Jerome, Jamie Mc Andrews, and Jean Charles Rochet, “Rating the Raters," 2008. Toulouse Working Paper.

Skreta, Vasiliki and Laura Veldkamp, “Ratings Shopping and Asset Complexity: A Theory of Ratings Inflation," 2009. NBER working paper # 14761.

Zandi, Mark (2008) "Financial Shock," FT Press, July 2008.

    Posted by on Friday, March 27, 2009 at 02:34 PM in Economics, Market Failure, Regulation | Permalink  TrackBack (0)  Comments (55)


    TrackBack URL for this entry:

    Listed below are links to weblogs that reference Why Did Ratings Agencies Fail?:


    Feed You can follow this conversation by subscribing to the comment feed for this post.