“We’re structure experts, we’re not underlying-asset experts.”
How exactly did the credit rating agencies assign ratings on collateralized debt obligations backed by mortgages? In “Triple-A Failure” in the April 27 issue of the New York Times Magazine, Roger Lowenstein explains in detail how Moodys rated one such issue, which Lowenstein calls “Subprime XYZ”. Lowenstein’s article is notable for the insight it provides into Moody’s ratings process. The ratings agencies were not the only culprit in this crisis, but they played an important role, and the Lowenstein article helps to elucidate that role.
The bombshell in the article is its depiction of the ratings process, which appears mechanistic and uninformed by serious economic analysis of the assets underlying the rated securities. Two quotes from the article are illustrative: “We’re structure experts. We’re not underlying-asset experts,” one employee said. Along similar lines, again from the article and from a different employee: “We aren’t loan officers. Our expertise is as statisticians on an aggregate basis. We want to know, of 1,000 individuals, based on historical performance [emphasis added], what percent will pay their loans?”
These are remarkable quotes. If you want to know what percentage of a mortgage pool will repay their loans, don’t you need to know what loan officers are doing? If you’re going to use historical default data, don’t you need to know whether loan officers are behaving the same way as in the past? Is history an appropriate indicator for a rapidly ballooning and relatively new asset class? Shouldn’t one consider external indicators of likely housing price changes? Moody’s knew that 43% of the 2,393 borrowers in the mortgage pool had provided no written verification of their income, but the article quotes one of the employees as saying that Moody’s was reassured by the fact that the mortgages were for primary residences: “When you get into a crunch, You’ll give up your ski chalet first.” (Doesn’t that count as an economic analysis of the underlying asset?) This is not to suggest that Moodys should have visited each house in the mortgage pool, but that they should have been thinking about mortgages as an industry, in this case, one that was in transition.
To top it all off, Moody’s had one day (yes, 24 hours) to rate Subprime XYZ.
It’s not fair to beat up on Moodys just because their ratings on a particular class of structures turned out to have been wrong. Unexpected bad things happen. And it is worth noting that Moody’s traditional bond ratings haven’t (so far) been as off-base as their CDO ratings. The important question is whether Moody’s did a reasonable job assigning the CDO ratings in the first place. After reading Lowenstein’s article it’s hard to conclude that they did.
A colleague who works for a government regulator pointed out that criticizing Moody’s for their “non-economic” default analysis echoes the criticism of audit firms for their failure earlier in this decade to detect corporate frauds. The auditing firms said that their job was to make sure that the numbers added up, not to determine whether the underlying numbers were false. Critics said that the auditing firms were being too mechanistic. The audit crisis catalyzed the Sarbanes-Oxley legislation, which among other things required CEOs to personally certify financial reports. One wonders what Congress and the regulators have in store for the credit rating firms. (If you wonder why Congress is involved at all, seven ratings agencies, including S&P, Fitch, and Moodys, have special status as “nationally recognized statistical ratings organizations” (NRSROs), and this designation gives their ratings regulatory status. If you’re curious, this report explains more.)
I haven’t said anything about the standard criticism of ratings agencies, which is that they are paid by the entities they rate. This creates an undeniable conflict of interest and criticism of this practice is completely justified. The link below to a Floyd Norris column discusses some forthcoming experiments, such as having investors pay the fees.
To think about what the ratings agencies should consider when assigning a rating, suppose that ratings agencies were at least partly paid in the securities they issue. For example, if a ratings agency gives a Aaa rating to an issue, the fee would be entirely payable in units of the issued security, which the ratings agency must then hold for at least 5 years. If the agency gives a C rating, the fee can be paid entirely in cash. For in-between ratings the fee is a combination of cash and securities. The fee would be paid in installments over time and the ratings agency would be permitted to hedge any risk associated with the security except credit risk.
This proposal has its own problems (the payments would have to be structured so that the agencies continued to rate fairly the securities they held), but the thought experiment is nevertheless suggestive. With payment in kind, the agencies almost certainly would find it worthwhile to give more serious thought to the underlying asset.