Information Systems Frontiers 5:1, 79–93, 2003 C  2003 Kluwer Academic Publishers. Manufactured in The Netherlands. Prediction Markets as Decision Support Systems Joyce E. Berg Department of Accounting, Henry B. Tippie College of Business, University of Iowa, Iowa City, Iowa 52242-1000, USA Thomas A. Rietz * Department of Finance, Henry B. Tippie College of Business, University of Iowa, Iowa City, Iowa 52242-1000, USA E-mail: Thomas-Rietz@uiowa.edu Abstract. Valuations from “prediction markets” reveal expec- tations about the likelihood of events. “Conditional prediction markets” reveal expectations conditional on other events occur- ring. For example, in 1996, the Iowa Electronic Markets (IEM) ran markets to predict the chances that different candidates would become the Republican Presidential nominee. Other concurrent IEM markets predicted the vote shares that each party would re- ceive conditional on the Republican nominee chosen. Here, using these markets as examples, we show how such markets could be used for decision support. In this example, Republicans could have inferred that Dole was a weak candidate and that his nomi- nation would result in a Clinton victory. This is only one example of the widespread potential for using specific decision support markets. Key Words. prediction markets, decision support, decision mar- kets, election stock markets, Iowa Electronic Markets, experi- mental economics, 1996 Presidential election 1. Introduction Berg, Nelson, and Rietz (2001) define “prediction mar- kets” as those run for the primary purpose of using the information content in market values to make predic- tions about specific future events. For example, since 1988, the Iowa Electronic Markets (IEM) have been running such markets, including markets designed to predict the outcomes of elections, box office receipts for movies, earnings reports, stock prices and returns, etc. In these markets, values of traded contracts depend directly on future outcomes and, hence, prices give in- formation about these outcomes. For example, in 1996, the IEM ran a market in which the value of traded con- tracts depended on the percentages of the vote taken by major candidates in the U.S. Presidential election that year. As Berg, Nelson, and Rietz (2001) show, values in such markets are efficient advance forecasts of the vote shares ultimately received. Depending on the contract payoff structure, market values can convey informa- tion about nearly any event that will be determined by a measurable future outcome. Here, we define “conditional prediction markets” as those run for the primary purpose of making predic- tions about future events conditional on other events. For example, in the 1996 Presidential election, the IEM ran a market with a set of conditional prediction con- tracts related to candidate vote shares. Contracts paid liquidating dividends of the form “$1 times the Demo- cratic nominee’s (two-party) vote share conditional on Lamar Alexander being the Republican nominee,” “$1 times the Republican nominee’s vote share conditional on Lamar Alexander being the Republican nominee,” “$1 times the Democratic nominee’s vote share condi- tional on Robert Dole being the Republican nominee,” “$1 times the Republican nominee’s vote share condi- tional on Robert Dole being the Republican nominee,” etc. As a result, values in these markets forecast the eventual election vote split conditional on the eventual Republican nominee. Hanson (1999) uses the concept of conditional prediction markets to illustrate his idea of “decision markets.” Decision markets are those designed primar- ily for the purpose of using the information in market values to make decisions. In such cases, markets become decision support systems. We argue that both prediction and conditional prediction markets can be used for decision support, either alone or in ∗ To whom correspondence should be addressed. 79