This site has moved to
The posts below are backup copies from the new site.

December 5, 2007

Why Did Ratings Agencies Fail to Properly Assess Risk?

Charles Calomiris and Joseph Mason discuss the assessment of risk by ratings agencies, who is to blame for their failure, and possible improvements to the rating system:

We need a better way to judge risk, by Charles Calomiris and Joseph Mason, Commentary, Financial Times: Recent downgrades on residential mortgage-backed securities (RMBS) and collateralised debt obligations, and subsequent hedge fund failures and market turmoil, have led many to blame ratings agencies for the mortgage mess. ...

The agencies reply that accurately accounting for risk is not their job and that they are protected by the right of free speech. They point to disclaimers ... that make it clear that they are paid by the companies they rate and that ratings are statements of opinion, not recommendations. ...

[S]uch admonitions ring hollow. Ratings agencies do more than opine; they play an active role in structuring RMBS and CDOs. They also serve as important sources of information about securitisation performance and often enumerate measures that issuers must take to maintain ratings in troubled securitisations.

More importantly, unlike typical market actors, ratings agencies are more likely to be insulated from the standard market penalty for being wrong, namely the loss of business. Issuers must have ratings, even if investors do not find them very accurate. ... Portfolio regulations for banks, insurance companies and pension funds set minimum ratings on debts these intermediaries are permitted to purchase. Thus, government has transferred substantial regulatory power to ratings agencies, since they now effectively decide which securities are safe enough for regulated intermediaries to hold.

Giving ratings agencies more power actually reduces the value of their ratings by creating a strong incentive for grade inflation and making the meaning of ratings harder to discern. Regulated investors encourage ratings agencies to understate risk so that the menu of high-yielding securities available to them is larger. The regulatory use of ratings thus has changed the constituency demanding a rating from free-market investors interested in a conservative opinion to regulated investors looking for an inflated one.

Grade inflation has been concentrated particularly in securitised products, where the demand is especially driven by regulated intermediaries. In 1994, economists Richard Cantor and Frank Packer ... pointed out that grade inflation was ... driven by the least reputable ratings agencies: “...Most ‘third’ agencies ... assign significantly higher ratings on average than Moody’s and Standard & Poor’s.” ...

Although there is evidence that Moody’s and S&P remain relatively conservative..., it is clear that even Moody’s has allowed its ratings scale for securitised products to become inflated. Bloomberg Markets reported in July that: “Corporate bonds rated Baa, the lowest Moody’s investment grade rating, had an average 2.2 per cent default rate over five-year periods from 1983 to 2005... From 1993 to 2005, CDOs with the same Baa grade suffered five-year default rates of 24 per cent, Moody’s found.” In other words, long before the current crisis, Moody’s was aware that its Baa CDO securities were 10 times as risky as its Baa corporate bonds.

Given the different and shifting meanings of Baa and other ratings as measures of risk and given the high rate of financial innovation and the lack of transparency inherent in multi-layered structured finance deals, it is not surprising that investors underestimated risks so badly leading up to the recent crisis.

It is no use blaming the ratings agencies, which are simply responding to the incentives inherent in the regulatory use of ratings. The solution is for regulators to reclaim the power that has been transferred to ratings agencies...

How can regulatory power be reclaimed? ...[One] solution is to reform existing regulations to avoid the use of letter grades in setting standards for permissible investments by regulated institutions. In the absence of letter grades, banks and their regulators would look at the underlying risks of investments, not ratings. Indeed, ratings agencies sell tools to investors that permit exactly this sort of analysis. Full disclosure by regulated institutions of these new measures of portfolio risks and a greater reliance on market discipline to discourage excessive risk-taking would further improve the regulatory process.

Another solution I saw elsewhere is to assign the companies to the rating agencies randomly so that today's rating does not affect the probability of getting business from that firm in the future. But I'm not sure how to impose market discipline on the firms that issue poor ratings if the assignments are random. If I know I will get a certain percentage of business (by random draw) no matter what, then there is no penalty for being wrong. I like the idea of randomization because it breaks the link between inflated ratings and getting more business, but the percentage of business a firm is assigned should be connected to their success at rating companies, and there would need to be a mechanism for entry and exit. Thus, over time firms who do best would increase their share, and firms who do worse would lose out and eventually be dropped opening up a spot (by bid?) for a new firm to enter.

I'm interested in hearing other proposals to create incentives for ratings agencies to issue accurate assessments.

Update: In comments, two stories have emerged:

Were the signals bad? Even the article says that the tools are available to do more detailed analysis, so why didn't they if so much was at stake? Still, one has the sense risks weren't fully understood for some reason and bad signals are one possibility.

Or were the signals fine, but ignored due to some market failure or incentive incompatibility where, for example, those making decisions do not have a full stake in the losses and hence do not fully account for risk? I think this has merit, or at least warrants thinking harder about.

Or some of both?

No comments: