So many teach-ins. The one at Yale School of Management builds up to an interesting argument in the last half hour over mark-to-market accounting. In all of the teach-ins I've seen, this one generates the most actual debate.
The link to the video can be found in a Wall Street Journal article on Gary Gorton (a participant in the Yale panel) and his role at AIG. An interesting question posed by the article (and ever-so-briefly hinted at during the panel) is what should happen to the seller of a credit default swap as a default becomes more probable. Suppose you have sold a default swap on XYZ corporation, and it has not yet defaulted. But things are looking a lot shakier. What happened to AIG is that its counterparties started asking for a lot more collateral. Presumably, if credit default swaps traded on an organized exchange, the exchange would be doing the same thing with margin calls.
Supposedly, Gorton's models say that AIG is not going to lose much on its default swaps. But the counterparties are worried that it might, so they ask for more collateral, and AIG has a hard time coming up withh that collateral. So what we are seeing, in effect, is a huge disagreement about the probability of XYZ corporation's default, with no clean way to resolve it. If you side with AIG and don't require them to post collateral, then counterparties will be afraid that their protection against XYA's default will not materialize when needed. If you side with the counterparties, then you create a run on AIG that makes it require a bailout, even if none of the defaults occur.
To me, this is just another flaw in the concept of a credit default swap. It is in some sense a deeply out-of-the-money option. There is a huge range of probability shifts (default risk going from, say, 0.2 percent to 0.8 percent, not to mention all the way up to 25 or 30 percent) that change the value of the swap while still keeping it far out of the money. It's not surprising that AIG got gummed up the way it did.