In Search of Better Credit Assessments

July 21, 2014 was the fourth birthday of the Dodd–Frank Act (DFA). It is maturing faster than a human, but slower than a dog. Of the nearly 400 rules that DFA requires regulators to write, just over half have been completed. At the end of August, the SEC finished another one – regarding credit rating agencies (CRAs). The result makes us wonder what took so long. (A scoreboard tracking DFA rule-making progress is here.)

The Commission issued two specific rules, one on conflicts of interest and the other on transparency. The text itself runs more than 700 pages. We read the 4,000-plus word summary. Before getting to the specifics of the SEC’s rule, let’s back up to recall the central role CRAs played in the financial crisis of 2007-09.

Without the complicity of CRAs, it is hard to see how the lending that fed the housing boom could have been sustained. Their high ratings of mortgage-backed securities (MBS) were seen almost immediately as one of the villains in the drama. (For an overview of credit rating agencies, see here.)

During the years before the crisis, vast numbers of MBS pools where constructed, rated and sold. These typically included several thousand subprime mortgages. Each pool was cut up into pieces (tranches), with the cash flowing from the underlying mortgages allocated to the highest grade tranche first, and – if there was any left – to each successive tranche. This “waterfall” pattern – of filling up the top vessel first – was supposed to make the top tranche virtually risk free. The alchemy of transforming low-grade, high-risk mortgages into high-quality debt vastly increased the supply of housing credit and propelled housing prices upward, as MBS investors paid little attention to the quality of the underlying mortgages (information that was difficult and costly to obtain in any case).

What did the CRAs do? They blessed the alchemists, rewarding their MBS pools with extraordinarily high ratings, including (typically) a “super-senior AAA rating” for the top tranche. They did this based on statistical models calibrated using recent data from a period when nationwide housing prices had never fallen. (Compare that to the post-1890 history available here.)

Once housing prices tipped lower in 2007 and mortgages started to default in unison, CRAs slashed their ratings, fueling the bust just as high ratings had propelled the bubble. Taking all of the asset-backed securities (and collateralized debt obligations) rated AAA between 2005 and 2007, only 10% retained their original rating by the June 2009 (see chart). In fact, less than 20% were still investment grade!

Source:  IMF, Global Financial Stability Report, October 2009, Chapter 2, Figure 2.12.

Source:  IMF, Global Financial Stability Report, October 2009, Chapter 2, Figure 2.12.

By itself, the failure of expectations to materialize is not sufficient to demonstrate poor CRA behavior. But CRAs had strong incentives to pump up ratings. The most obvious was the concentration of their paying clients: in the half-dozen years before 2007, the top five MBS issuers accounted for 40% of the market, resulting in a large volume of repeat business. The resulting conflict of interest led to a documented bias toward high ratings.

This brings us to the two new SEC regulations intended to address the problems with bond ratings. The first one requires CRAs to establish various internal controls and provide certifications aimed at the conflict of interest arising from the “issuer pays” arrangements. The second is about transparency, and compels the CRAs to publish reams of information about the pools they are rating (anonymized to protect the individual mortgage borrowers).

Will this help? Probably not much. Take the transparency problem first. In an earlier post, we noted that the market for U.S. MBS that are not government insured has virtually collapsed. It seems unlikely that additional information – without a long data history – will improve the estimates of correlations within the mortgages in a pool. Unless investors can judge the diversification of the underlying mortgages, they won’t know how to price the private-label MBS. And they are likely to remain suspicious of CRA models, especially if they result in high ratings.

What about conflicts of interest? Here, we have several reactions. First, however much the people constructing the ratings are removed from the revenue gathering in a business, they will always be aware of the connection. They know that customers shop for ratings. And, they can judge if a customer is happy with their work.

Second, and more troubling, the desire for inflated ratings is not limited to issuers; many buyers want inflated ratings, too! Much less has been written about this, but the incentive problems are the same on both sides of this transaction.

Two examples make the point: asset managers and bankers. Asset managers’ performance is measured relative to a benchmark. When that benchmark includes bonds, it will be based on standard indexes of averages in particular ratings categories. If a manager selects bonds within the ratings category that are riskier than the rating suggests, this will provide an excess expected return relative to the benchmark (albeit, at a greater risk). Because funds that perform poorly are often simply shut down, the survival of the outperformers makes it look like their managers have superior skills, even when they are simply choosing riskier portfolios.

For a banker, the riskier the assets, the bigger the regulatory capital buffer required. Large banks are supposed to use their own internal risk models, but small banks employ a standardized approach based on credit ratings. Again, if a bond is riskier than its rating makes it appear, it will have a higher expected return. That allows the banker to take greater risk without adding capital.

Addressing all of the issues inherent in risk assessment – problems involving not only faulty models and insufficient data, but poor incentives and free riding – seems beyond anyone’s current capacity. But regulators can do better than the SEC has. For example, proposals to reduce ratings shopping existed before Dodd-Frank, but the new rules do not address this issue. More important, the rules do not eliminate the centrality of ratings in capital regulation. While various regulators have made some moves to reduce the reliance on CRAs, they have not gone far enough. In particular, systemic intermediaries should not be allowed to outsource credit evaluation. For banks, we see a straightforward solution: internal models that are calibrated using hypothetical portfolios. We wrote about this here, and it still strikes us as something regulators should try. Perhaps a comparable approach could work for large nonbanks.

To conclude the story, we hope that the SEC will bark louder in the future. And we will be watching (with considerable doubt) to see if their efforts to date have much bite.