One of the casualties of the financial crisis has been the reputation of the major credit rating agencies. To many, the problem with the credit ratings business seems obvious:
- The ‘issuer-pays’ market structure, in which the issuers pay the agencies to rate their debt instruments, distorts incentives.
The issuers want higher ratings to lower their cost of borrowing, and can shop among raters to get higher ratings (Pagano and Volpin 2010). Seems obvious, right?
Serious analysis and attention to history suggests that market structure is the wrong problem to focus on.
Our recent research argues that the main problem with the ratings industry is the government’s use of ratings for regulatory purposes (Cole and Cooley 2014).
Origin and use of ratings
Ever since John Moody (followed by Poor’s, Standard Statistics, and Fitch) began selling ratings of corporate bonds to investors, ratings have been viewed as useful information in the financial system. The ratings of corporate bonds and many other forms of debt have performed consistently well. This is true both in the era when investors primarily paid for ratings and in the period since the 1970s when issuers primarily paid for them.
Indeed, ratings were so useful that, as early as the 1930s, regulators began to rely on them to monitor risk-taking by banks. That reliance by regulators expanded significantly over the decades and across both eras as credit markets became deeper and broader. But something clearly went wrong in the run-up to the Great Recession, when it seems the ratings agencies systematically underestimated the risks associated with residential mortgage-backed securities.
Both the issuer-pays and the user-pays methods of financing ratings are inefficient, but in different ways.
- In the user-pays world, only users who pay for the rating get the information – even though it is essentially free after it is produced.
Moreover, users tend not to buy information about new and small firms – effectively keeping them out of credit markets.
- When the issuer pays for a rating, they can decide not to report their rating.
This selection problem distorts the information that is publicly available.
Although both market structures have inefficiencies, they both worked well enough for markets and regulators to rely on them. Why? Because of reputation. A simple model of reputation can explain why the ratings agencies had a strong incentive to produce reliable ratings. They did this in order to be able to sell ratings in the future, since if they were seen as having produced inaccurate or uninformed ratings when people wanted high-quality ratings, their ratings would not be believed. If the ratings were not believed, then their ratings would not affect a security’s price, and firms would have no incentive to pay for a rating under the issuer-pays system. Similarly, investors would not buy ratings in an investor-pays world if they were believed to be inaccurate.
Ratings and mortgage-backed securities in the subprime crisis
So how do we account for the failures with residential mortgage-backed securities in the subprime crisis? Aside from the obvious fact that virtually all market participants – including the Fed and the creators of these securities – underestimated the dangers of sub-prime mortgages and failed to anticipate a nationwide housing shock, what perturbed the reputational equilibrium that had prevailed for so long?
The best candidate is the confluence of the government’s use of the ratings for regulatory purposes and the opaque nature of these privately created ‘safe assets’. That distorted incentives in the market because it limited the ability of financial institutions to hold some securities based upon the agencies’ ratings, or forced them to hold higher reserves against them. Since these financial institutions were often well informed about the riskiness of these securities – in part because many of them helped create them – they ended up caring about the riskiness of the rating and not the information contained in the rating. This undid the reputation mechanism that had disciplined the market. It can explain both why the ratings of mortgage-backed securities were overly generous and why they paid higher yields.
What it means for reform
This explanation of the ratings breakdown during the Great Recession has important implications for many of the proposals for reform. The use of private ratings for regulatory purposes is inherently problematic because it can lead to these kinds of distortions. Further, having the ratings agencies publish all of the inputs to their ratings so that others may duplicate or redo their work (something no-one has done yet) probably will not help. It allows them to issue a public rating and still indicate the security’s actual riskiness to investors.
Forcing the issuers to accept a rating agency appointed by some outside panel (as proposed by Senator Al Franken) would undercut the reputational mechanism, since a ratings agency that was seen by investors as less accurate could still get orders from the panel.
What is to be done?
Constructing risk weights for regulatory purposes is inherently difficult, and there are no easy alternatives to private ratings. But using private ratings distorts incentives. Moreover, these ratings are forecasts of bankruptcy risk and don’t distinguish between aggregate risk – which cannot be diversified away – and idiosyncratic risk – which can. A better approach altogether may be to focus on capital requirements to provide safe risk buffers for the critical parts of the financial system.
There is no point in creating reforms that don’t address the root problem. Changing regulations just to do something usually ends up distorting credit markets in new and unanticipated ways.
Editor’s note: Professor Cooley was previously a member of the Board of Managers of Standard & Poor’s Ratings and is a consultant to the company.
References
Cole, Harold L and Cooley, Thomas F (2014), “Rating Agencies”, NBER Working Paper 19972.
Pagano, Marco and Paolo F Volpin (2010), “Credit Ratings Failures and Policy Options”, Economic Policy, 25(62): 401–431.
