Ko and Huanga, in the May Journal of Financial Economics:
In behavioral finance, overconfidence has been established as a prevalent psychological bias, which can make markets less efficient by creating mispricing in the form of excess volatility and return predictability. In this paper, we develop a model in which overconfidence causes investors to overinvest in information acquisition when this information could improve market efficiency by driving prices closer to true values. … Overconfidence generally improves market pricing provided the level of overconfidence is not too high.
In principle, prediction markets would require subsidies to get people to take risks trading on topics where they do not need to hedge risks. In practice, people are often willing to trade in zero-sum situations. This improves price accuracy overall, but also suggests that people who run markets where folks can now legally speculate will oppose legalizing new markets, for fear of diverting their arrogant traders.
Stuart, yes prediction markets can be far from perfect, and yes overconfidence may distort the choice of topics.
This post has made me aware of a personal bias: I am biased against anything that makes claims of perfection. I felt uncomfortable with prediction markets (and other similar, libertarian ideas) because of their claim to be the best possible solution in every possible case. Once I've seen the flaws in prediction markets (and other ideas) it somehow makes them more real, and I'm far more willing to accept them or compare them honestly with the alternatives.
Returning to the subject of the post:This also suggest that people who run prediction markets will be tempted to make the issues that are betted on more attractive to the overconfident. More convoluted, perhaps, or more random (so that people can imagine patterns where there are none).