Charles W. Calomiris is the Henry Kaufman Professor of Financial Institutions at Columbia Business School.
CREDIT RATING AGENCIES such as Moody’s, Standard and Poor’s, and Fitch used to be bit players in the dramas that attend financial crises. No more. Recent hearings at the House Committee for Oversight and Government Reform provided the latest in a series of surprising spectacles about the agencies, this time in accusatory exchanges among former and current executives of Moody’s. And new controversies are brewing: the engineering of a new set of complex transactions called “re-remics,” where rating agencies help repackage securities on a bank or an insurance company’s balance sheet to achieve an improvement of the ratings of the portfolios.
What is the evidence that rating agencies performed badly in measuring credit risk on the debts that they rate? Were rating agencies suborned, and, if so, by whom and to what purpose?
The evidence of rating agency failure shows up in inflated ratings and low-quality ratings. The inflation of ratings is the purposeful underestimation of default risk on rated debts. Low-quality ratings are ratings based on flawed measures of underlying risk. The recent collapse of subprime-related securitizations revealed both problems in the extreme.
What harm do these deficiencies do? Inflation subverts the intent of regulations that use ratings to control risk-taking, resulting in ineffectual prudential regulation. If rating inflation is accompanied by low-quality ratings, this causes deeper problems. Investors can “reverse engineer” a debt rating that is merely inflated and recover the true measure of risk; the revelation of severe flaws in risk modeling that usually occur in response to a financial shock leaves investors unsure of how to price the debts they are holding and unwilling to buy additional debts of similar securitizations, resulting in severe market disruption.
Evidence abounds that severe errors in subprime ratings were predictable. The two most important modeling errors relating to subprime risk were both assumptions that contradicted logic and experience, namely, that U.S. house prices could not decline, and that the underwriting of no-docs mortgages would not lead to a severe deterioration in borrower quality.
THE SOURCE OF THE PROBLEM
Who was behind these biased models? Many policy makers incorrectly believe that securitization sponsors are the constituency that control ratings. That is false. Ratings that exaggerated the quality of securitized debts were demanded by the buy side of the market (the institutional investors whose portfolio purchases are being regulated according to the ratings that are attached to those purchases) because inflated ratings benefited them. Why?
Ratings that understate risk are helpful to institutional investors because they (1) increase institutional investors’ flexibility in investing, (2) reduce the amount of capital institutions have to maintain against their investments (the objective of re-remics alchemy), and (3) increase their perceived risk-adjusted profitability in the eyes of less sophisticated ultimate investors (mutual fund, bank, and insurance company shareholders, pensioners, or policyholders) by making it appear that an AAA-rated investment is earning an AA-rated return.
If buyers wish rating agencies to inflate ratings to overcome regulatory hurdles and make them appear more favorably in the eyes of their ultimate investors, rating agencies can reap substantial profits from catering to buyers’ demands for inflated ratings. This has an important implication: rating inflation on securitized debts is done at the behest of the buy side.
Consider the case of the collateralized debt obligation (CDO) market. CDOs were constructed using unsold debts from other securitizations (often subprime mortgage-backed securities [MBS]). CDO issuance volume increased dramatically in the early 2000s, rising from $100–150 billion a year in the period 1998–2004 to $250 billion in 2005 and $500 billion in 2006.
Were institutional investors aware of the high risk of CDOs prior to the 2006 boom? Yes. Moody’s published data on the five-year probability of default, as of December 2005, for Baa CDO tranches of CDOs, which showed that these Baa debts had a 20 percent five-year probability of default, in contrast to the Baa corporate debts, which showed only a 2 percent five-year probability of default. Despite the rhetoric that rating agencies publish claiming to maintain uniformity in rating scales, institutional investors knew better: in 2005, CDO debts of a given rating were ten times as risky as similarly rated corporate debts.
Why did institutional investors play this game? Asset managers were placing someone else’s money at risk and earning huge salaries, bonuses, and management fees for being willing to pretend that these were reasonable investments. On one occasion, when one agency was uninvited by a sponsor from providing a rating (because the rating agency did not offer to approve as high a percentage limit for AAA debt as the other agencies), that agency warned a prominent institutional investor not to participate as a buyer but was rebuffed with the statement: “We have to put our money to work.”
Rating agencies gave legitimacy to this pretense, and were paid to do so. Investors may have reasoned that others were behaving similarly, and that all were protected by the biased models of risk. The script would be clear and would give plausible deniability to all involved. “Who knew? We all thought that the model gave the right loss assumption! That was what the rating agencies used.”
Strong evidence that buy-side investors encouraged the debasement of the ratings process comes from the phenomenon of “ratings shopping.” Before actually requesting that a rating agency rate something, sponsors ask rating agencies to tell them, hypothetically, how much AAA debt they would allow to be issued against a given pool of securities being put into the CDO portfolio. If a rating agency gives too conservative an answer relative to its competitors, the sponsor just uses another rating agency.
HOW TO STOP A RACE TO THE BOTTOM
It is crucial to recognize, however, that for ratings shopping to result in a race to the bottom in ratings, the race to the bottom must be welcomed by the buyers; if institutional investors punish the absence of a relatively good agency’s rating of an offering (by refusing to buy or paying a sufficiently lower price), then would-be ratings shoppers would have no incentive to exclude reputable rating agencies. Thus the evidence that ratings shopping tends to produce a race to the bottom implies that the buy side favors the low-quality, inflated ratings that result from the race to the bottom.
Under pressure from Fitch, Congress and the Securities and Exchange Commission (SEC) also played a role in encouraging the debasement of ratings of subprime MBS and related securities. Congress passed legislation in 2006 that prodded the SEC to propose “anti-notching” regulations that would have facilitated ratings shopping in the subprime MBS market. “Notching” arose when CDO sponsors brought a pool of securities to a rating agency to be rated, which included debts (often subprime MBS) not previously rated by that rating agency. When asked to rate the CDO that contained those subprime MBS, Moody’s, say, would offer either to rate the underlying MBS from scratch, or to notch (adjust by ratings downgrades) the ratings that had been given by, say, Fitch.
The new rules would have forced each rating agency to accept ratings of other agencies without adjustment when rating CDO pools. Even though the anti-notching regulations were still being considered at the time the crisis broke out, the legislation and proposed regulations sent a strong signal to the rating industry. It is quite possible (based on my discussions with insiders) that it had an effect on rating agencies’ practices in rating subprime MBS as early as 2006. This policy constituted an attack on any remaining conservatism within the ratings industry: trying to swim against the tide of ratings inflation would put a rating agency at risk of running afoul of its regulator!
WHAT WILL WORK
Once one recognizes that the core constituency for low-quality and inflated ratings is the buy side in the securitized debt market, that carries important implications for reform. Proposals that would require buy-side investors to pay for ratings, rather than the current practice of having securitization sponsors pay for ratings, would have no effect in improving ratings.
The elimination of the use of ratings for regulatory purposes would remove some of the incentive for ratings inflation, but, by itself, this would not solve the problem of inflated and low-quality ratings, since the buy-side agency problem would continue to generate a demand for inflated, low-quality ratings in the securitization markets, where incentive-conflicted institutional investors dominate.
Any solution to the problem must make it profitable for rating agencies to issue high-quality, non-inflated ratings, notwithstanding the demand for low-quality, inflated ratings by institutional investors. This can only be accomplished by objectifying the meaning of ratings and linking fees earned by rating agencies to their performance. If fees are linked to the quality of objectified ratings, then ratings agencies would find it unprofitable to cater to buy-side preferences for inflated, low-quality ratings. How could this be done?
All agencies wishing to qualify as Nationally Recognized Statistical Rating Organizations (NRSROs)—the rating agencies whose ratings are used in regulation—would be required to submit ratings for regulatory purposes that link letter grades to estimates of the probability of default and the expected loss given default. Once the ratings are presented as numbers, rating agencies could be held accountable for their ratings. For example, if an NRSRO’s ratings for a particular product were found to be persistently inflated, then it would face a penalty. That penalty could “claw back” fees the agency had earned on that product (enforced by requiring that agencies post some of their fees as a “bond” to draw upon). Alternatively, a rating agency found to have exaggerated its ratings could simply lose its NRSRO status for a brief time (say, several months), which would also provide powerful incentives not to inflate.
REFERENCES AND FURTHER READING
Bloomberg Markets magazine, July 2007.