My buddy Richard sent me this link today (6-18-2009) Tech Dirt analysis of a Yahoo analyst's analysis of a shambles.
Problem! "Systematic risk" is mischaracterizing risk by generalizing it, not to mention misquoting the phrase "systemic risk" which is what Duncan Watts actually said. The journalists are having a fantasy. The system *is* transparent enough to see what is happening, but it is obscured by distance and not-giving-a-shit.
The issue is understanding systems, in particular how several nonlinear variables that act together produce a solution. If you read Why Most Things Fail which is about cause and prediction and then The Black Swan which is about economic prediction specifically you can see what my point is. Ormerod is an economist and is sometimes on the outs with some of the community. They were in the middle of a pout when the bomb went off.
Measuring risk in a limited area of the economic system is a trivial part of the problem. Journalists are talking about risk and banks but the real problem is risk and the financial system. Libertarian economists say an individual institution should be allowed to fail. Techdirt is saying that there are individual institutions that are so large that toppling one will bring widespread harm. I don't know about that but I suspect that interlinking risk between institutions is what would be the big problem. I don't see people making any distinction between the individual elephant and the herd of elephants.
Which brings me to the issue I pointed out. Inventing a new tool to spread risk in an attractive way that makes it appear palatable may have contributed to the problem.
Watts' article in the Globe points out that institutions shouldn't be allowed to extend in certain ways. My way of saying it would be that if a system can't be predicted, it should be dismantled into components that can be predicted. Then we should create an institutional memory that explains why they can't be allowed to do a repeat performance.