1 Information Markets: Feasibility and Performance David Bray, Karen Croxson, William Dutton Information markets - also called ‘prediction markets’, ‘decision markets’, ‘event derivatives’, ‘event futures,’ and ‘idea futures’ - are markets designed specifically for the purpose of generating accurate information. Participants buy and sell assets whose payoffs are tied to the realization of future events. To illustrate, an information market might be used to forecast box office receipts for a new movie. A contract (or asset) might, for instance, pay $1 if ticket sales are above some pre-specified level by a pre-specified date and $0 otherwise. A market price of $0.43 would then be interpreted as a 43% chance of success. Probably the best known information markets are the Iowa political markets. 1 These markets were set up in 1988 by Iowa university academics to allow the public to “bet” on US presidential elections. Since their creation many more applications have emerged and now markets predict a wide range of events. Researchers at the University of Iowa have gone on to develop a market to forecast outbreaks of Avian Flu, 2 whilst their contemporaries at the University of Miami have established a Hurricane Futures Market. 3 Elsewhere, Hollywood play-money markets invite the public to predict opening weekend box office sales and the Oscars, 4 and, significantly, some corporates have begun to explore the potential of information markets to harness collective wisdom internally. Firms such as Hewlett Packard, Google, General Electric and Microsoft are leading the way in experimenting with business information markets. 5 Trading in corporate markets typically involves an internal group of “experts” with contracts written on such matters as whether a project deadline will be met or a sales target exceeded. Interest is spreading in the potential use of markets to generate conditional estimates (“Will our product ship on time if we take the following action?”). Decision-makers in many settings could benefit from the potential for conditional markets to provide neutral guidance for actions. In theory, as individuals trade on their private information, an asset’s price will move to incorporate all relevant news; the price, as a result, will constitute a collective prognosis which at any point in time is the best forecast available (Fama 1970, Hayek 1945). Markets therefore should outperform other Information Aggregation Mechanisms (IAMs) in terms of forecast accuracy, including the traditional alternatives of asking an expert for a forecast, obtaining the opinions of multiple experts and somehow pooling these, conducting a poll or survey, or leaving a group to deliberate. We discuss in this paper some of the features of these other IAMs that can undermine their prediction accuracy: • voting aggregates views but does not weight these by relevance (by contrast, in markets, misinformed traders will suffer heavy losses) and does not reward people for being right; 1 http://www.biz.uiowa.edu/iem/ 2 http://fluprediction.uiowa.edu/fluhome/Market_AvianInfluenza.html 3 http://hurricanefutures.miami.edu/ 4 http://www.hsx.com/ 5 http://en.wikipedia.org/wiki/Prediction_market