Knightian uncertainty generates a sort of double- (or more) counting problem, where scarce capital is wasted insuring against impossible events.
Read the whole thing. Of course, Mark Thoma found it.
The article provides a rationale for having the government insure toxic assets. By doing so, you might reduce liquidity preference by a large multiple.
For example, suppose that ABC holds a $100 mortgage bond enhanced by a credit default swap from XYZ. The bond is really shaky. Investors in ABC say, "You have to hold $100 in capital in case the bond defaults and the credit swap defaults." Investors in XYZ say, "You have to hold $100 in capital in case the bond defaults and you have to make good on your credit default swap." The overall risk is $100, but $200 in capital gets tied up, because investors in ABC want insurance against the "impossible" event of a default by XYZ (assuming XYZ puts up the $100 in capital.)
If government buys the mortgage bond (the original Paulson plan), the need for ABC to hold capital goes away, but I assume that XYZ's default swap stays in force. So XYZ's capital requirements stay in force. Instead, if government insures the bond (either directly selling a swap to XYZ or else buying the bond and canceling the swap), the entire $200 in capital gets freed up.
One could argue that this theory suggests a very counterintuitive view of how to address the panic. Most people think you should go after the "root" of the problem--buying the mortgage securities or, better yet, strengthening the underlying mortgages. Instead, what this theory suggests is that to free up the most capital government should take over the derivatives that are farthest from the problem--the bond insurance and the credit default swaps.
I think it is important to understand the theory, if for no other reason than to understand the limits of the "root of the problem" approach. But I still prefer shutting down unsound institutions to trying to come up with clever prop-ups.