I have argued that credit default swaps are fatally flawed because there is no natural seller. One commenter on this post asked, then,

Has Arnold gone into detail on why CDSs exist? I understand his point that they have no natural buyer, and thus may be a result of some weird incentives in the system somewhere. But which ones? What regulation causes this anamoly?

I should go into more detail.A B-rated bond has higher default risk than a AAA-rated bond. There are lots of institutions that could hold AAA-rated or AA-rated bonds but which are precluded from holding B-rated bonds. However, if it were not for regulation, those institutions might be able to do better with B-rated bonds than with AAA-rated bonds. Suppose that a B-rated bond has a default probability of .01, and a AAA-rated bond has a default probability of .0001 (note: those are strictly hypothetical figures, that I pulled out of the air–not based on the actual relationship between ratings and default probabilities.)

If two B-rated bonds are independent of one another, then the probability that both of them default is .01 times .01, or .0001, which is the same as the probability of a default on the AAA-rated bond. And if you have three independent B-rated bonds, the default probability gets even lower. Where this is leading is that with enough independence, a diversified portfolio of low-rated bonds can be created that has lower risk and a higher return than a AAA-rated bond.

The regulatory anomaly is that a bank or pension fund cannot take advantage of this. The regulators will see B-rated bonds in the institution’s portfolio and penalize the firm for taking risk. The regulators overlook the diversification.

Into the breach steps a AAA-rated insurance company, selling credit default swaps on B-rated bonds. The bank can buy a B-rated bond, protect it with a credit default swap from the insurance company, and have the regulator treat the bond as at least AA. The insurance company performs the diversification function, selling swaps on a boatload of different bonds. The insurance company gets to keep most of the extra return that is created by having the banks take on low-rated bonds.

It turns out that at least two things can go wrong here. First, the bonds may turn out to be more highly correlated than was thought. It could be that if some of them go bad, then a lot of them go bad. Second, even if defaults stay under control, an increase in the probability of default can force the insurance company to put up more margin (collateral), and this can put the insurance company in dire straits.

The regulatory issue is this: why are banks not allowed to take the risk of (a) holding a diversified portfolio of low-rated bonds, when they are allowed to take pretty much the same risk by (b) purchasing credit default swaps? If they do (a), then if the bonds turn out to be highly correlated, then the portfolio will not be so well diversified and the strategy will fail. However, if they do (b), then if the bonds turn out to be highly correlated, then the seller of CDS is unlikely to be able to perform as promised.

Regulators have to take a point of view on the issue of how highly correlated are bond defaults. If they take the view that they are not highly correlated, then banks should be allowed to take on diversified portfolios of low-rated bonds. If they take the view that bonds pretty highly correlated, then banks should not be allowed to claim that the purchase of credit defaults swaps provides insulation from risk.

The inconsistency in regulation is what accounts for the rise in the credit default swap market.

I think if you are going to regulate, you have to be really careful to trace through all of the consequences of regulation. If you see a financial innovation taking off like crazy, there is a good chance that it is being used to exploit a regulatory anomaly.

Effective regulation is really easy after something blows up. In real time, it strikes me as a darn hard problem.