The Rating Game

Whoever pays wins

Ever since AAA-rated subprime mortgage-backed securities (“MBS”) collapsed during the global financial crisis, sharp attention has been focused on the business model of the US credit rating agencies.

Long-term AAA grade of Credit Rating Agency

At issue is whether those inflated MBS ratings were based simply on mistaken assumptions about US housing prices (as the agencies argue), or whether they were the product of an inherently flawed payment system.

Since the early 1970s1, US credit rating agencies have generally been paid by the issuers of the debt instruments they rate. Of the ten nationally-recognized statistical rating organizations (“NRSROs”) currently authorized by the SEC to issue credit ratings2, nine employ the so-called “issuer pays” model3.

Unfortunately for the world financial system, the “issuer pays” model inherently favors higher ratings. To win business from corporate and government issuers and investment banks that package structured financial products such as MBS and other asset-backed securities (“ABS”), the rating agencies compete not only on the basis of credit expertise in the relevant industries, countries or asset classes, but on their willingness to elevate ratings so as to minimize issuers’ debt service costs4.

Among bond issuers, this phenomenon is known as ‘rating shopping ‘. Issuers use the promise of their current and future fees to attract favorable ratings from the agencies which may sacrifice the quality of their credit assessments for future business opportunities. The US credit rating business is big: NRSROs rate almost three million new debt securities every year and take in over $4 billion in fees. Apparently, exaggerated credit ratings are most easily disguised in ABS because of the complexity and opaqueness of their portfolios.

Biases in the “issuer pays” model have not been limited to residential MBS. The rating agencies missed the Enron (2001) and WorldCom (2002) bankruptcies and were still rating Lehman Brothers’ commercial paper AAA on the September morning in 2008 when the firm collapsed. The agencies contend that their ratings are just opinions as to the long-term risks of default and are not intended to address temporary credit conditions that may reverse themselves.

Biases in the “issuer pays” model have not been limited to residential MBS. The rating agencies missed the Enron (2001) and WorldCom (2002) bankruptcies and were still rating Lehman Brothers’ commercial paper AAA on the September morning in 2008 when the firm collapsed.

The agencies also defend their business model by pointing out that their ratings of credit-card and car loan ABS, for example, stood up well during the financial crisis, so they claim it’s not their payment model that failed but their housing forecasts (which are reportedly being fixed). More broadly, the agencies say that the widespread public dissemination of ratings enabled by the “issuer pays” model ultimately assures quality control, since NRSROs need to preserve their reputations for accuracy in order to sustain their businesses.

Because MBS ratings played such a pernicious role in the financial crisis, Dodd-Frank required the SEC to explore the feasibility of replacing the “issuer pays” model with a newly-created, public or private oversight board that would assign the issuance of credit ratings for new ABS to NRSROs on a rotating basis, fix the rating fees for the various debt instruments to be rated and then reward the agencies whose ratings turn out to be most accurate over time with greater shares of the ratings business. Senator Al Franken (D-Minn) champions this concept as “the easiest way to make sure there is no conflict of interest and to break the oligopoly.”

In December of last year, the SEC published its report to Congress describing the comments it had elicited from the marketplace regarding the proposed oversight board and a number of other credit rating models it had considered. The objections to all of the alternatives were compelling.

Under the control of an oversight board, NRSROs would be inclined to apply conservative rather than aggressive criteria to their credit ratings so as to ensure positive evaluations, which would improve their business prospects with the board but also increase debt service costs for issuers. For the same reason, NRSROs would be reluctant to downgrade their ratings, which would run counter to the government’s goal of ensuring the ongoing accuracy of ratings.

While an oversight board would eliminate ‘rating shopping’ by virtue of its independence from both issuers and agencies, it would also conflict with the government’s efforts following the financial crisis to eliminate ratings from its regulatory requirements for banks, broker-dealers, insurance companies and money market funds. It might also give the unintended but dangerous impression that credit ratings enjoy a government seal of approval.

Also concerning was the potentially depressing effect of the proposed board on competition and innovation in the already oligopolistic ratings industry. Currently, three firms – S&P, Moody’s and Fitch – control 91% of US and 96% of global ratings. Because of the substantial intellectual and technological resources needed to determine accurate credit ratings, newcomers would be more reluctant to make the required investment to enter a government-and-price-controlled market than to enter the current market in which they could at least compete freely with the entrenched incumbents.

The creation of such a board would also face potential constitutional challenges. The federal government would effectively be requiring one private party to do business with another private party of the government’s choosing, which would invoke Fifth Amendment concerns about a taking without just compensation. Government control over who can render a creditworthiness opinion would also raise First Amendment issues.

The federal government would effectively be requiring one private party to do business with another private party of the government’s choosing, which would invoke Fifth Amendment concerns about a taking without just compensation. Government control over who can render a creditworthiness opinion would also raise First Amendment issues.

Practical considerations pose even greater challenges to an oversight board. Think of the skills and other resources such a board would need to evaluate the NRSROs and their millions of credit ratings, and think also of how it would be funded. The costs of operating the board would have to come from taxpayers, the NRSROs or bond investors through transaction fees (or some combination thereof). The last of these would also require the creation of a world-wide system for determining and collecting the appropriate charges. At present, there are no accepted metrics for evaluating credit ratings over time, and the issuance of ratings would be slowed down by the additional layer of bureaucracy. On top of all that, the rotation system adopted by the board would deprive the rating process of the efficiencies of agency continuity.

In addition to an oversight board, the SEC reviewed the pros and cons of reverting to a “subscriber pays” model5, where bond investors rather than issuers would pay the rating agencies for their credit opinions on debt instruments. The two main pitfalls with this model are that it inherently favors the issuance of lower ratings by the agencies so that their subscribers can reap higher risk premiums (yields) on their investments (and continue their subscriptions) and it precludes the general publication of credit ratings because of the problem of ‘free-riding’. Publication of ratings is viewed as essential to the maintenance of a fair marketplace for all investors.

The SEC also elicited professional views of a privately-owned oversight board (financed and operated by large, sophisticated institutional investors) whose credit rating would be required in each case along with that of an NRSRO of the issuer’s choosing; a split-fee model in which issuers would select their NRSROs but the NRSROs would be paid from transaction fees imposed on both sellers and buyers in primary and secondary market transactions; and, finally, a so-called “designation model” under which all of the NRSROs would have the option of rating a new bond issue and investors in that issue would allocate fees to the agency or agencies on whose credit analyses they relied.

It would be too taxing to detail what it would take to implement any of these alternatives, including the government-controlled oversight board, so suffice it to say that there’s a good reason why the current “issuer pays” model has persisted for four decades and isn’t likely to be replaced anytime soon. Short of nationalizing the NRSROs, I don’t see any way of eliminating the inherent bias in favor of whoever pays an NRSRO for rating a publicly-traded debt instrument.

Does anyone have a better idea?

1 A brief history of US rating agencies can be found in Professor Lawrence White’s The Credit Rating Agencies in the Journal of Economic Perspectives (Spring 2010).

2 Under the Credit Rating Agency Reform Act of 2006, credit ratings on bonds can only be issued by SEC-certified rating agencies (NRSROs).

3 The one exception is Egan-Jones, the smallest of the US credit rating agencies. The other nine are Standard & Poor’s, Moody’s, Fitch, A.M. Best, DBRS, Japan Credit Rating, Morningstar, Kroll and HR Ratings de México.

4 Since 2008, SEC regulations have attempted to mitigate the “issuer pays” conflict of interest by, inter alia, prohibiting NRSROs from rating structured debt instruments they helped to design and insulating NRSRO credit analysts from fee negotiations with issuers.

5 Prior to the 1970s, credit ratings on bonds were published by rating agencies in manuals available only to paying subscribers.


  • Marc Joffe

    Thanks for this really well researched and well argued post. In response to your last question, yes I have a better idea. Credit ratings should be assigned by means of fully transparent open source models. I make the case for this idea at and This approach has also been discussed on NakedCapitalism.

    • Stephen Bornstein

      Very interesting, Marc, but how would we get the rating agencies to open their kimonos? Would open source require the adoption of the oversight board or do you think it could somehow be accomplished privately between the agencies and the issuers? If it could be accomplished either way, who would comment on agency ratings and how would they be compensated for their efforts/insights? And then, will the agencies be required to enter into dialogues with the commenters over differences of opinion or methodology? What happens to the ratings (and yields) in the meanwhile? And, lastly,why only governments? As you know, the whole ratings kerfuffle started with GSE and private-label MBS.

      • Marc Joffe

        Stephen: Thanks for your reply. I think open source rating analytics will need regulatory support in order to gain adoption. In a recent Guardian op-ed at, I discuss how an SEC-mandated certification body might oversee them.

        As for compensation, I don’t see a viable alternative to the issuer pays model, but with greater transparency, the embedded conflicts can be better managed. However, investors and other interested parties would have an incentive to pay quants to dive into these models in search of potential shortcomings.

        I don’t think the rating agency or whoever contributes to the open source model necessarily has a responsibility to debate every finding with third party users, but their models should face de-certification if enough unanswered valid criticisms are raised.

        With respect to the transition, instruments already trade on a lot of information other than ratings. I would not expect that a transition toward open source models would affect markets very much until one or more of the models gained credibility. Also, I don’t argue that legacy issuer-paid ratings should be banned; I proposed this concept to the SEC as a way of meeting the Dodd Frank mandate to take ratings out of regulation. An evolving model is a better alternative than the fixed algorithms that I have seen proposed to date.

        I started with governments because I thought that government bond ratings have the strongest public goods component and have been the least subject to modeling. There are already strong models for structured (had the CRAs only parameterized property) and for corporates.

        • Stephen Bornstein

          Marc, I read the article about your open source ratings proposal in the 5/23 FT and congratulate you on moving your daring idea into a broader public forum.