Thursday, October 30, 2008

Bad Data, Bad Theory, No Surprise

So I'm reading a paper (Economic Catastrophe Bonds by Coval, Jurek and Staffard) on how CDOs tranches are mispriced. Big publication in the American Economic Review. One of their key assumptions is that empirically the yield spread over the expected loss increases over a default rate, that is, the total return is higher for junk bonds than for investment grade bonds; risk begets return and all that. But consider that the Merrill Lynch High Yield Index had a 6.75% return from 1987 through Sep 2008, whereas the Merrill Lynch Investment Grade Index had a 6.49% annualized return. That's 26 basis points difference in return, unadjusted for risk, in an index. Add the transaction costs, and the returns to the junk bond index are lower than for investment grade, yet academics seem unaware of this.

The authors cite two papers that note the difference between the AAA and BBB rate, and just assume that it extrapolates to junk bonds (err, no). They also cite a piece by Hull, Predescu and White (2005), who assert the B spread to Treasuries to be an average of 585 basis points. Bloomberg benchmark curves, and data I had at Moody's, would estimate a 405 basis points spread from 1991 to 2005, so I have no idea where they are getting their data. Hull et al mention Merrill, but I'm not sure which index. It's wrong.

Anyway, they have this wonderful model built on bad data, and a theory (the CAPM) that hasn't worked anywhere else, and shocker, it also does not match the data:

We show that losses on the most senior tranches referencing an index of investment grade credit default swaps are largely connected to the worst economic states, suggesting that they should trade at significantly higher yield spreads than single-name bonds with identical credit ratings. Surprisingly, this implication turns out not to be supported by the data.

Surprise? In the context of all that data supporting the Sharpe-Lintner CAPM (snark)?

But still, on an important point--what is the essence of a CDO?--their whirlwind tour of finance seminars never seemed to flag the absence of a risk premium in junk bonds, which makes their paper's premise moot. What are its potentially perverse incentives? They point out that AAA CDOs have more cyclicality than AAA for regular bonds, but the correlation of the AAA security is hardly as important or interesting as its default rate point estimate. They get an irrelevant answer, that people are arbitraging a risk-premium that doesn't exist. It's just wrong on so many levels: the risk premium does not exist in the collateral where it really matters in CDOs (junk), or the tranches; the marginal players in these deals are the equity owners of the deals, not the AAA investors; AAA ABS has higher returns than straight corporate AAA rated debt so the arb there is absent (especially as AAA ABS will take a generation to gain credibility as truly AAA again).

I don't mean to be so critical--it's a clear approach, plausible (though extrapolating 0.7 moneyness is not). Just wrong.

No comments: