If you have access to the Wall Street Journal Online, you may want to read this article. It talks of Prediction Markets and their fallacies:

So-called “prediction” markets such as Intrade and Iowa Electronic Markets are, like the stock and every other market, only as reliable as the information they process. They generally do a bang-up job guessing probabilities of outcomes, but can miss specific events badly if bad information gets in the mix.

One example was the 2004 presidential election, when Intrade futures were briefly skewed by misleading exit polls that suggested John Kerry was going to win. Critics of prediction markets often use this episode as a cudgel to pound on those markets’ credibility.

What needs to be understood is that, for elections, Prediction Markets are in direct competition with Polls. Polls have been used for decades in determining national sentiment, and it is only recently that people have been putting their faith in prediction markets.

During the 2004 presidential election, the prediction markets slated George W. Bush as the winner, yet the exit polls were saying John Kerry. Exit polls were considered the most precise form of polls, since voters were asked whom they voted for right after voting. It seems, rather, that people assumed the Prediction Markets must have been wrong and so they changed their predictions, thereby swaying the markets in Kerry’s favor as well.

Rather than record this as a fallacy of prediction markets, I would say it was one of the last times people felt they could have faith in polls. The faulty polls are what skewed the market. In fact, this incident was a huge impetus for realizing what prediction markets were capable of in their own right, and not as a complement to polls.