I might be getting involved in a prediction markets project soon. Here's some of my preliminary research.
The grand old lady of electronic prediction markets is Iowa Electronic Markets, a real-money market run by the business school at the University of Iowa, which claims to have provided better predictions of US presidential winners than opinion polls since 1988. There is a $500 deposit limit and a $500 limit per bet. The market structure is a continuous double auction.
Even in markets with binary outcomes, IEM has two contracts and a mechanism where they can be bought and sold in unit bundles at $1, thus splitting liquidity. In a wonderful series of posts, Daniel Davies offers some analysis and criticism of IEM. In this post he does some calculations on the Kerry '04 winner-takes-all contract and finds that in the Black-Scholes model, the Kerry contract is overvalued in such a way that there is no value of implied volatility for which the contract could be correctly priced.
The Foresight Exchange is one of the oldest prediction markets online and also runs a continuous double auction, but only has a single contract in each outcome that can be shorted. Its user interface is absolutely horrible, but it seems to be relatively popular because anyone can create new contracts.
NewsFutures is a prediction market with a slightly better design, run by a company that has a Java/PostgreSQL-based platform for prediction markets. They're the ones who set up the Yahoo! Buzz Game.
The Yahoo! Buzz Game does not use a strict continuous double auction, but rather a "dynamic pari-mutuel market" (pdf) mechanism, which is somewhat like a pari-mutuel market like the ones typically used for horse betting (more on this in a later post), but where there the price can reflect new information. (In a regular pari-mutuel market the price of each bet is fixed.) Unlike continuous double auctions, a dynamic pari-mutuel market can provide the infinite liquidity that is the chief advantage of pari-mutuel markets, and requires only a fixed subsidy by the market operator. (Continuous double auctions can have good liquidity if there is a market maker, but the market maker could incur substantial losses.)
Robin Hanson of George Mason University is one of the leading academic scholars working on prediction markets. He sits on the scientific advisory board of NewsFutures. He also had a role in the Pentagon's Policy Analysis Market.
I will probably blog more about this in the future.