Web 2.0 again: Prediction markets

I just ran across an article in the Wikipedia on prediction markets, of which I've long been a fan. Prediction markets' basic idea is to turn predictions ("Obama will win" and "McCain will win," for example) into stocks, then have participants buy and sell the two stocks on an exchange.

I just ran across an article in the Wikipedia on prediction markets, of which I've long been a fan. Prediction markets' basic idea is to turn predictions ("Obama will win" and "McCain will win," for example) into stocks, then have participants buy and sell the two stocks on an exchange. The two prices will, if the market works its magic, reflect the relative share of the vote that'll accrue to each candidate. By watching the market, you should be able to learn the future. Simple. Ish.

So What?

This would be a very clever idea if it weren't so utterly insane. This was my reaction years ago when I first heard it, anyway. But it turns out that there's a fair amount of theoretical and concrete material to back up the claim. To your vast relief as well as mine, I'm going to skip the theoretical stuff in favor of the concrete.

The canonical example of a prediction market is the Iowa Electronic Markets (IEM), which for years (back even before the Web, if you can believe it) allowed you to trade on the outcome of presidential elections. Its record is remarkably good-- it consistently gets the share-of-vote for each candidate correct when election night rolls around. (Of course, it's possible that it's just following the--usually equally accurate--polls rather than being driven by traders' independent assessments. There's no way to know.)

The Hollywood Stock Exchange (HSX) has for years successfully predicted Oscar winners and movie revenues. (My wife played the HSX for several years. She did very, very well. Most of the players at that time were apparently men, and men tend to underestimate the draw of Julia Roberts, Hugh Grant and the rest of that crowd.) HP has used markets to predict sales with, apparently, some success. Google has over 100 prediction markets played by some 20 percent of its staff...it claims a 70 percent accuracy rate. Some companies have used them to predict software ship dates for large projects and have, again apparently, found that they're more accurate than official estimates. (This will be frustrating, by the way, since, as this article points out, you'll know that it's going to be late, but not why. So you're standing there, quivering like an attack dog, desperate to intervene, but with no idea how or where.)

What about predicting successful technology innovations? Good question. One company has used markets to decide which technologies to incubate and launch--at a predefined time, whichever technology has the highest price on the exchange is chosen and the people who held that stock are paid off. But that's really selection, not prediction. Prediction would defer the payoff until the technology was successful in the marketplace. The problem is, that could take years to determine. When the reward is that far out, it may be hard to get people to play the game.

And with that, we are out of time. Prediction markets seem to hold a lot of promise and if you happen to do something with them, I'd really appreciate hearing about it. Thank you.