18 Information ( in ) efficiency in prediction markets

This chapter examines a new class of markets at the intersection of trad· itional betting and traditional financial markets. We call these 'prediction markets'. Like both financial and betting markets, prediction markets focus on uncertain outcomes and involve trading in risks. Prices from these markets establish forecasts about the probabilities, mean and median outcomes, and correlations among future events. These prices have been used to accurately predict vote shares in elections, the box office success of Hollywood movies and the probability that Saddam Hussein would be deposed by a certain date. Other names for these markets include 'virtual stock markets', 'event futures', and 'information markets'. Financial economists have long known about the information-aggregating properties of markets. Indeed, the efficient markets hypothesis, a centre· piece of financial theory, can be stated simply as, 'market prices incorpo· rate all available information'. While financial instruments can be very complex, prediction markets tend to be analytically simple. Their current simplicity, however, belies their powerful potential future as a way to hedge against geopolitical and other forms of risk as envisioned by Athanasoulis, Shiller and van Wincoop (1999) and Shiller (2003). Currently, most prediction markets are quite small, with turnover rang· ing from a few thousand dollars on the early political markets run by the University of Iowa, to several million bet in the 2004 election cycle on TradeSports, to hundreds of millions bet on the announcement of economic indicators in Goldman Sachs and Deutsche Bank's 'Economic Derivatives' market. The most famous prediction market is the Iowa Electronic Market (IEM), which was started in 1988 to predict the vote share of the two major party presidential candidates. Since then, they have amassed a record of more accurate prediction than polls, all while limiting trading positions to a cap of $500.