In what ways do prediction markets fail? The paper provides some discouraging answers. First, they struggle when there is a high degree of insider information. On the question, "Will the mandate be struck down", for instance, only the Chief Justice himself could say for sure, and so the market was likely to be wrong. There must be information to aggregate. "Will there be WMDs in Iraq?" was the basis of one contract. But which Iraqi arms tattletale is trading predictions contracts? None, so the market was mistaken.
But if markets need information to predict accurately—as these two criteria entail—then so much is off limits. Who has information on next year's GDP? So much can change; so many things could happen. Sure, a merchant may have an idea, he may take a guess. But is that guess as reasonable as one's guess on WMDs? Is the future as foreign a realm as the far-off sands of Iraq? The thinking behind a market is that trading creates an incentive for players to develop the best possible information, to come up with new statistical models of the economy and place bets on their basis, for instance. The more dumb money in a market, the richer the pot for smart money, which should entice such money in and move the price in the right direction. But the very best processing of available information may still be wildly offbase where future events are concerned.
The third failure is a behavioural one. Individuals tend to overestimate low probabilities and underestimate high ones. (The former explains why so many play the lottery; the latter is just an inverse of the former.) But this means we have to rule out so many estimates as unreliable.
My friend Henry Manne would argue that one solution to the first and possibly third concerns would be to allow insider trading on the prediction market. I can see his point. If the prediction at issue is one that is either (a) known already by insiders or (b) uncertain but subject to much more accurate estimates by insiders, the contract will be far more accurately priced.
Henry discussed insider trading in prediction markets in his article Insider Trading: Hayek, Virtual Markets, and the Dog that Did Not Bark. Available at SSRN: http://ssrn.com/abstract=679662
Check it out.