Tuesday, July 29, 2003

Place Your Bets!

Much hoohah about DARPA's plan to try to predict terrorist activity by a "futures market". Supposedly, asking people to "bet" on things gives much more accurate results than simply asking them for an opinion.


It's an interesting question, and one I've noticed myself. Joe Blowhard will pontificate for hours, saying that, if we do "X", then "Y" is absolutely sure to happen. According to Joe, his result is as certain as the sun coming up in the West tomorrow morning. But just say "I've got twenty five cents, cash money, that says you're wrong. C'mon, put your money whare your mouth is. And no, I will not take your IOU. Cash. US currency." It's amazing how fast Joe will fade into the background.

What DARPA is assuming is that this "futures market" will somehow have a better result than other ways of using expert opinions. Personally, I doubt it. The problem depends on the nature of a futures market.

To those of the Libertarian persuasion, of course, markets are ironclad examples of Laws of Nature, not really different from the Law of Gravity. However, when you look at them more closely, markets are as artificial as the Infield Fly Rule. They are ruled by arbitrary social constructs and tradition that govern anything else. Futures markets, in particular, require a very delicate balance of information flow among the participants.

Let's look at the futures market in a commodity; say, coffee. Basically, what a futures trader is doing is making a bet that coffee will sell for a specific price at a specific time. Now, the actual price of coffee at the sale time is determined by all sorts of things like weather and local politics. The "futures price" is a guess. Now, as time goes on and the due date comes closer, the price of coffee starts to zero in on the final price, and the value of "futures" goes to zero. (The Chicago Board of Trade has a brief futures tutorial, if you want more info.) The gimmick for futures traders is that, the way margin requirements work, you can make (or lose!) a lot of money on very small price swings.

Now, look at the information required to make this work. Everybody has to know something about what's happening with the commodity, but there has to be some information that is confidential. If there was no information, or if there is full information, it reduces to a total game of chance. Professional brokers (and professional poker players) don't deal with games of chance. Keeping the appropriate amount of inside information makes the system work. (Technically, this is called "controlling transparency"). This is why, traditionally, the commodities markets are regarded as the fastest possible way to lose your money. The Big Guys have the information. You don't.

Now, how does this relate to terrorism? Durned if I know. In a commodity futures market like coffee, everybody knows when the due date is, when little pieces of paper get traded for sacks of beans. If we are trying to predict terrorism, there is nothing that has to happen. So what this will reduce to is simply a bunch of experts making their best guesses. How is that different from what we have now? Worse, I'd assume that they will all be working from the same intellegence assessments and listening to the same briefings.

So what we end up with is just another version of The Common Wisdom. But maybe the Blowhard Effect will kick in, and the experts will say what they really mean.

LATER -- Another little problem that I haven't seen discussed is that, to be useful, the "values" of the market have to be public. Otherwise, the whole thing degenerates into yet another Poll of the Experts, of which we already have a plentiful sufficiency. This means that the terrorists can track them too ...

STILL LATER -- WOOHO! Poindexter is going to resign over this mess. Couldn't happen to a nicer guy. However, he still has to go through the process of actually resigning. We'll see how honest he is -- my bet is he will either weasel out or find somebody worse to replace himself..

Weblog Commenting and Trackback by HaloScan.com