Rating Agency Failure: Conflicts, Models, and AAA Illusions
Why Did Rating Agencies Give AAA Ratings to Instruments That Were Near-Worthless?
In 2007 and 2008, Moody's and Standard & Poor's downgraded over $1.9 trillion in mortgage-related securities from investment-grade to junk in a series of mass rating actions. Many of the securities that were downgraded to junk had been rated AAA or AA six months earlier. The scale of the downgrades was unprecedented in rating agency history, and the speed at which "safe" instruments became worthless destroyed the credibility of a rating system on which the entire structured finance edifice had been built.
Quick definition: Rating agency failure in the 2008 crisis refers to the systematic assignment of investment-grade and AAA ratings to structured credit products whose true credit quality was far below what the ratings implied — driven by business model conflicts (issuers pay raters), flawed model assumptions (correlation underestimation), and competitive pressure among the three major agencies to maintain market share.
Key Takeaways
- The "issuer pays" model — in which the issuers of securities pay the agencies for ratings — creates a structural conflict that generates competitive pressure to assign higher ratings.
- The rating models used for structured products systematically underestimated default correlation across mortgages in the same geographic market or the same vintage.
- CDO-squared and synthetic CDO ratings compounded the correlation error at multiple structural levels, producing AAA ratings for instruments that had near-zero tolerance for adverse scenarios.
- Internal email evidence at both Moody's and S&P documented awareness of the model problems, but competitive pressure to maintain market share suppressed corrective action.
- The post-crisis regulatory response to rating agency failure included the Dodd-Frank requirement for increased transparency in rating methodologies, the SEC's registered rating organization oversight program, and the exploration of alternative rating models — though none resolved the fundamental issuer-pays conflict.
- Investors who relied exclusively on ratings for credit assessment — rather than performing independent analysis — suffered the largest losses.
The Issuer-Pays Business Model
The "issuer pays" model has existed in the rating agency industry since the early 1970s, when Moody's and S&P shifted from the previous investor-pays model in response to photocopier technology that allowed investment publications to be reproduced and distributed without payment to their publisher.
Under the issuer-pays model, the institution issuing a security pays the rating agency for a rating of that security. For structured finance products, the fee was typically a function of the notional size of the transaction — for a $1 billion CDO, the rating fee might be $500,000 to $1 million across the three agencies.
The conflict is clear: an issuer that wants a high rating for its product pays an agency whose revenue depends on winning the mandate. An agency that consistently assigns lower ratings than competitors loses mandates to the competitors. The competitive equilibrium that this dynamic produces is higher-than-warranted ratings for any category of security where all three agencies share the same systematic incentive.
For corporate bonds and sovereign debt — the traditional rating agency market — the issuer-pays conflict is partially mitigated by the availability of public financial information that allows investors to independently assess the rating's accuracy. For structured products, whose complexity made independent assessment by most investors prohibitively expensive, the conflict operated with fewer mitigating factors.
The Model Failure
The specific quantitative failure of rating agency models for structured products involved the treatment of default correlation.
In a mortgage pool, the correlation between two mortgages' probabilities of default determines how much protection the senior tranches receive from diversification. If defaults are independent (correlation = 0), a large pool produces highly predictable loss rates that the senior tranche can comfortably withstand. If defaults are perfectly correlated (correlation = 1), the entire pool either performs or fails together, and the senior tranche's protection is minimal.
The rating agencies' models for residential MBS used historical default data that implied low correlation because the data did not include a period of national home price decline. The United States had not experienced a national home price decline since the Great Depression, so the data available for model calibration did not contain the relevant stress scenario.
The Gaussian copula model — introduced by David Li in 2000 and widely adopted for CDO rating — expressed the correlation between mortgage defaults as a single parameter that could be inferred from the historical data. The calibrated correlation was low because the historical data was a biased sample (excluding the stress scenario). The resulting AAA ratings were too high because the models systematically underestimated the probability that many mortgages would default simultaneously.
When home prices fell nationally in 2006-2009, the realized default correlation was far above the model's assumption — approaching one in the worst-performing vintage pools. The AAA tranches of CDOs backed by subprime MBS pools experienced losses that the models had assigned probabilities of essentially zero.
Internal Evidence of Model Awareness
Subsequent investigations and legal proceedings revealed that employees within the rating agencies were aware of specific model limitations during the period when the flawed ratings were being issued.
A 2007 email by a Moody's analyst, disclosed in Congressional investigations, described a structured finance transaction as "financial engineering, nothing more, nothing less" and noted that the deal "could be structured by cows." A 2006 S&P internal email described the agency's model as producing ratings that were "out of range" for some CDO structures and acknowledged that the assumptions were inconsistent with the agency's own research on correlation.
These communications documented that specific analysts were aware of problems. They did not by themselves prove that the agencies intended to mislead investors — they are also consistent with the view that the problems were identified by some individuals but not elevated to produce model changes, due to competitive pressure or organizational dynamics.
The competitive pressure explanation is credible given the structure of the industry: an agency that unilaterally revised its models to produce lower ratings would immediately lose CDO rating mandates to competitors who maintained more favorable models. The equilibrium is a race to ratings laxity, not a race to accuracy — unless all agencies simultaneously revise models or regulators require common standards.
The Mass Downgrade Wave
The rating agencies' mass downgrades of structured products in 2007 and 2008 demonstrated the failure's scale. Moody's downgraded approximately $800 billion in structured products in the twelve months ending August 2008; S&P performed comparable actions. The downgrades affected instruments that had been held by money market funds, insurance companies, pension funds, and bank treasury departments on the strength of their investment-grade ratings.
For many institutional investors, the downgrades triggered mandatory selling — their investment mandates required them to hold investment-grade instruments. Forced selling from mandatory downgrades amplified the price declines that were already occurring from fundamental credit deterioration, creating a feedback loop between price declines, additional downgrades, and further forced selling.
Regulatory Response and Its Limits
The Dodd-Frank Act included several provisions specifically directed at rating agency reform. Section 932 required enhanced disclosure of rating methodologies, assumptions, and limitations. It required agencies to provide detailed information about their rating procedures and their track records. It created a new office within the SEC — the Office of Credit Ratings — specifically to oversee rating agencies' compliance with registration requirements and to examine their policies and procedures.
The Act also explored the creation of a government-appointed board to assign rating agencies to rate specific structured products, breaking the issuer-pays assignment mechanism. This provision was ultimately not implemented in its original form.
The fundamental issuer-pays conflict was not eliminated. The post-Dodd-Frank rating agency industry remained dominated by the same three major agencies with the same basic business model. The reforms improved transparency and created regulatory oversight but did not resolve the structural incentive problem.
The Rating Failure Mechanism
Common Mistakes When Analyzing Rating Agency Failures
Treating the failure as primarily about incompetence. The agencies employed competent quantitative analysts. The failure reflected systematic incentive misalignment, not simply bad modeling. Better models were available — they produced less favorable ratings, which were commercially unacceptable.
Assuming the post-crisis reforms were sufficient. The Dodd-Frank transparency requirements improved information disclosure but did not resolve the fundamental issuer-pays conflict. The three major agencies retained market dominance and the basic incentive structure persisted.
Ignoring the investor side of the failure. Institutional investors who relied exclusively on ratings without performing independent analysis shared responsibility for the losses. Rating agencies stated explicitly in their documentation that ratings were not investment recommendations and should not be the sole basis for credit assessment. Investors who treated AAA ratings as infallible verification of safety made their own analytical error.
Underestimating the model's role. The Gaussian copula model was not merely convenient — it provided mathematical rigor that gave the ratings an air of scientific precision. The model's deficiency was specific and documentable; addressing it would have required admitting publicly that historical data was insufficient for the rating task being performed.
Frequently Asked Questions
How did the three major agencies maintain oligopoly after the failures? The NRSRO (Nationally Recognized Statistical Rating Organization) designation — required for an agency's ratings to be recognized in regulatory capital calculations — was controlled by the SEC and limited to a small number of firms. New entrants faced a chicken-and-egg problem: issuers had little incentive to pay for ratings from agencies that regulators did not yet recognize. The post-crisis reforms created somewhat easier NRSRO registration but did not significantly expand competition.
Were rating agencies ever held legally responsible? Several rating agencies settled litigation arising from the crisis. S&P settled with the Department of Justice for $1.375 billion in 2015 without admitting wrongdoing. Moody's settled for $864 million in 2017. The settlements included elements related to the rating practices of the structured finance era.
Do ratings still matter for structured products? Yes — regulatory capital calculations for banks, insurance companies, and other institutional investors still reference external ratings for structured products in many jurisdictions. Despite the crisis, the ratings-based regulatory framework was not entirely displaced.
What alternative rating approaches have been proposed? The "investor pays" model — in which investors pay for ratings independently of issuers — has been proposed as the primary alternative. It eliminates the issuer-pays conflict but faces economic viability challenges (free-rider problems, since ratings information is difficult to prevent from spreading). A government-appointed assignment board would break the conflict differently but introduces regulatory capture risks.
Related Concepts
Summary
Rating agency failure in the 2008 crisis resulted from the interaction of three factors: the structural issuer-pays conflict that created competitive pressure for favorable ratings; model assumptions that systematically underestimated default correlation in structured products; and competitive dynamics that prevented any single agency from unilaterally correcting its models without losing market share. The result was over $1.9 trillion in structured product downgrades from investment-grade to junk in 2007-2008, triggering mandatory institutional selling that amplified price declines. The post-crisis regulatory response improved transparency but did not resolve the fundamental conflict. Rating agencies remain a systemically important element of the financial infrastructure, and the incentive structures that produced the crisis's rating failures persist in modified form.