OK - so I have a big mezzanine exposure in CDO's in which the underlying collateral is subprime mortgages. This sucks, but there isn't a lot to be done about that directly right now. If I think that the subprime mortgage world is going to be a complete rout, I should probably get rid of these things but it's going to cost me. Obviously, there are times and reasons for doing that. But it's a real short conversation - "Yep according to my model if the index goes down another 10 points, we would be looking at $50 M worth of losses and that's unacceptable. Since I now view that as reasonably possible, we have to bite the bullet and sell". What really should have happened is that as the risk became too big, I pared my position at better prices. There is no question that the risk of owning these things changes as defaults near attachment points, as credit reasons start making defaults past the attachement point more likely, or as correlation among the underlying collateral increases (btw - correlation isn't really the right measure of association here since correlation only exactly applies in normality). As the risk increased, I should have been on top of that and started doing something about the positions.
Recently, we had some discussion on AF about cutting positions as they lose money. There were a bunch of the usual objections. Of course, guys who deal with negative gamma in options books do that all the time and it is obvious to them. As the portfolio loses money, you still need to keep risk at some reasonable level. Since losing money is some indication that you are wrong, you might want to cut positions even more. In equity and fixed income investing, this is not normally the case. A stock that has dropped a lot contributes less risk to your portfolio than it did prior to the drop (you simply have a smaller proportion of your portfolio devoted to it). A CDO position that drops has lots more risk than it did before, but it's contained in esoteric models and obscured by manager marks and illiquidity. I wonder how many banks looked at BBB credit ratings and didn't realize they were taking on some negative gamma-like risk (like once it starts running it goes from BBB to default in no time).
Anyway, if you've got them, you probably have other credit risky positions. I'd be concerned that correlation among seemingly unrelated collateral will pick up in a subprime mortgage crisis. I think I can tell a story that says that a subprime mortgage crisis can spread to nearly any credit-risky security simply by starting a credit crunch that causes spreads to widen and available credit to dry up. How many credit risky securities were affected by Russia defaulting on bonds that were largely held by Russians? (ans: all of them)