Decision markets

I've always been impressed by James Surowiecki's lucid insights and clear writing. In this super piece he discusses how "decision markets", such as the current ones wagering on various aspects of the war in Iraq, are more accurate than experts and surveys.

He attributes this accuracy to 1) efficient aggregation of collective information, 2) openness to diverse viewpoints, and 3) rewards focused on being right, not being popular or influential within your crowd of choice. It strikes me that 1) and 2) are close to the sort of inclusive decentralization popular within "communitarian" or others Communist/Socialist thought, but the focus on "rightness" is what squeezes quality from diversity. While some people are aghast at the impersonal quality to speculating on what, in the Iraq situation, is life and death, I cannot imagine getting good predictions without some measure of detachment.

I also really liked his note on how HP used an internal market to predict sales, which turned out to be 75% more accurate than official, in house predictions. Sales forecasts are one of those things where everyone, from the optimistic sales rep to the factory foreman who wants to keep his people employed, has an incentive to lie, and I had never heard of an internal market, where presumably money was won and lost, for that. Neat.

But what about stock market bubbles? I hear you cry. When market participants start to think about what other market participants will do, instead of just independently trying to figure out the truth and put money where their mouth is, you get increased volatility and bubbles. But it's not like this sort of inter-market gamesmanship does not occur in centrally planned systems, it's just harder to pop and even more innacurate.