A "prediction market" lets participants buy and sell make-believe "shares" in future events. Who will win the next election? What will the dollar be worth a year from now? When the time arrives the shares in the market are paid off according to what actually happens. Investors whose predictions were right make a profit, so they have more resources to invest in other forecasts. It's the wisdom of crowds combined with a reward mechanism. As Cass Sunstein discusses in Infotopia, for some issues prediction markets are quite accurate.

They also, however, have inefficiencies. In recent months I've been involved in a prediction market experiment and several flaws are obvious. Questions start at arbitrary prices, far from equilibrium. For example, a yes-no question begins at 50% probability no matter how unlikely one of the alternatives is. Similarly, overnight or weekend news can move the market in an event wildly the next day. In both cases a few traders who come into the office early can make a big killing, regardless of their actual expertise.

This flaw wouldn't exist if there were limitations on individual trading, specified market hours outside of which no one could buy or sell shares, and rational pre-opening-trade price setting for new issues. And perhaps to a lesser degree, participants have a strong incentive to shoot the moon, make wild speculative gambles since they don't have real money to lose and the experiment isn't going to be running long enough for slow-and-steady wisdom to overcome lucky guessing.

A prediction market is no magic crystal ball, and it doesn't apply to many issues that require deep expertise. But as a source of data for analysis and further refinement it seems promising.

^z - 2011-03-09