Wednesday, October 22, 2008

Rating Agencies Grilled by Congress

The rating agencies got a good grilling today by the Congress, and it was on CSPAN (video online here). My friend and colleague Jerry Fons testified, and his theme is that there is a conflict of interest inherent in rating agencies is a big problem that underlies the lack of due diligence by rating agencies. Indeed, it is a tough problem. An important point was a quote from the Moody's CEO, who noted in an internal discussion (leaked somehow) that issuers want 'high' ratings, and investors also want 'high and stable ratings'. Now, that suggests everyone got exactly what they wanted. Of course, the rating agencies then just played along, but if the sellers and buyers want AAA ratings, it seems inevitable that AAA ratings would be justified. How do you regulate out something where the seller and buyer are willing participants?

One congressman noted that while Baa rated corporate securities had a 2% default rate over 5 years, Baa-rated CDO securities had a 24% default rate over 5 years. What's up with that? The Moody's CEO mumbled something incoherent about being different in different time periods, though he didn't specify which ones. I worked on CDOs at Moody's, on the inputs for default rates of unrated securities (RiskCalc(TM)), and think the problem was primarily that they underestimated the correlation of defaults in bad times. The problem is still there, to my knowledge. But CDO defaults are not a big part of this current crisis.

Another congressman asked why the ratings for the mortgages started to fall? Why should I believe them today, given they missed this huge trend? No good answers from the CEOs of the Big Three (Fitch, Moody's, and S&P). The S&P CEO mentioned they did incorporate house price declines of 15% when they underwrote the structures, but I think that's total BS. That is, you can always say 'that assumption is in there', and with a complicated enough process, it is difficult to prove them wrong, but I would have followed up with: 'well, what were the loss assumptions for subprime versus prime mortgages? How would these have been altered with a 30% housing price decline? How did these assumptions change over the past 10 years as underwriting criteria changed? Where in your 'transparent methodologies' is there any record that gives an expected loss estimate for collateral on mortgage pools?'

A philosophical question from Eleanor Holmes Norton: if you rate a company AAA, and it defaulted, was it accurate or inaccurate? The Fitch CEO said, inaccurate. I think the question is incomplete. You are generating a rating that refers to a frequency times an expected loss estimate, and an observation contributes to that frequency estimate, but no single observation is determinative. However, given the AAA estimated default rate of around 1 in 10,000, a single default probably doubles the sample default rate, and so gives one great pause.

But, while everyone was frustrated, no one seemed to be converging on anything to fix the problem. I bet they'll put together some new structure to monitor the agencies, that will just be a waste of time, but also get rid of the Nationally Recognized Statistical Rating Agency, a designation that gave the Big Three more credibility than they deserve.

No comments: