Prediction Markets: Betting on Risk Management

Risk managers looking for critical data needed for successful risk analysis and subsequent risk hedging would be well advised to consider prediction markets. Also known as decision markets, prediction markets have consistently proven themselves to be accurate predictors of a wide range of future events. This article examines these markets and discusses their considerable potential for use in risk management.

Prediction Markets Defined
Prediction markets are markets in which people bet on the outcomes of all types of events—political, economic, catastrophic, scientific, financial, cultural and so on. The most famous prediction market is the Iowa Electronic Market where anyone can bet up to $500 on U.S. politics. Established in 1988, this market has correctly predicted the outcome of every U.S. presidential election since its inception. Moreover, it typically predicts the percentages of votes garnered by the major candidates to within less than one percentage point, consistently producing more accurate results than even voter polls and expert opinion.

There are many other notable predication markets as well. Some deal in real money like the Iowa Exchange while others use artificial money that can be traded in for prizes. At the Foresight Exchange people can bet on the outcomes of nearly every conceivable event in categories that include news, politics, arts and entertainment, finance, religion, disasters, and science and technology. The NewsFutures Market is another comprehensive prediction market that offers betting on the likelihood of specific events occurring in the realms of politics, economics, science, natural disasters and more.

As its name implies, the Hollywood Stock Exchange is a prediction market that allows people to bet on the fortunes of Hollywood film and their movie stars. This market consistently forecasts the opening-weekend revenues of major movies more accurately than even the movie industry’s own official forecast. At the 2003 Hollywood Oscars, this market also correctly predicted 35 of the top 40 Oscar nominees.

The range of prediction markets is as varied as the public’s interests. TradeSports and the Athletic Stock Exchange specialize in sports and current events, while Long Bets is devoted to long-term predictions. There is even a market solely for German political and economic events called Wahl$treet.

Meanwhile, private companies have put their own versions of these markets to use in their own operations. Hewlett-Packard uses prediction markets to generate unofficial forecasts of its sales. Impressively, its prediction market has predicted H-P sales more accurately than the company’s official forecasts 15 out of 16 times.

According to a July 2003 Wall Street Journal article, Credit Suisse First Boston analyzed prediction markets in 2003 and found them to be “uncannily accurate” in predicting a broad scope of events. 

The Origin of Prediction Markets
Interestingly, prediction markets trace their origin to the 1600s, when scientific research was practiced quite differently than today. At that time, a few major universities held a virtual monopoly on research, recognizing only each other’s theories, and refusing to recognize the theories and research of “outsiders” who were not part of their inner circle. Many of these outsiders argued that their theories should be judged by empirical standards, and not by whether or not their results agreed with the prevailing wisdom of the inner circle. In an attempt to dismantle the teaching monopoly that a few British medical schools held on setting medical practices, chemical physicians in the 1600s—barred from teaching in the inner circle medical schools—offered the following wager in 1651, as cited by Allen Debus in his 1970 book, Science and Education in the Seventeenth Century:

“On ye Schooles…Let us take out of the hospitals, out of the camps, or from elsewhere, 200 or 500 poor people, that have fevers, pleurisies, etc. Let us divide them into halfes, let us cast lots, that one halfe of them may fall to my share, and the other halfe to yours;…we shall see how many funerals both of us shall have; But let the reward of the contention or wager, be 300 Florens, deposited on both sides: Here your business is decided.” 

For reasons that can only be conjectured, this concept lay dormant for more than three centuries until the late 1900s, when prediction markets suddenly flourished.

But How Do They Work?
The mechanics of these markets are essentially those of a horse race. Instead of entering a horse, a person enters a “claim” in a prediction market. A claim is simply a statement that something will happen by a certain time. 

For example, currently, one of the claims in the Iowa Electronic Market  concerns the outcome of the 2004 Democratic National Committee, which will determine the Democratic nominee for President of the United States. Claims can be made for John Kerry, John Edwards and the other major Democratic candidates.

In a typical prediction market, the holder of a winning claim (i.e., one that comes true) receives a dollar, while the holder of a losing claim (i.e., one that does not come true) receives nothing. For example, in March 2004, the cost of a claim on John Kerry in the Iowa Electronic Market was 89 cents. If a bettor bought this claim and John Kerry subsequently won the 2004 Democratic presidential nomination, then the holder would receive one dollar for the claim, thus earning a profit of 11 cents.

Because a one-dollar payoff in a prediction market represents 100 percent probability or certainty of an event coming true, and because zero payoff represents zero probability, the cost of a claim in a prediction market is actually the market’s consensus determination of the probability of a claim coming true. In the example above, the 89-cent cost of a claim on John Kerry means that the consensus of bettors in this market is that John Kerry has a 89 percent probability of being the 2004 Democratic candidate for president. By contrast, the cost of a claim on John Edwards was 10 cents at the beginning of March. Interestingly, a claim on Hillary Clinton would cost one cent even though she was not currently in the race, making her a 100 to 1 long shot. Of course, these probabilities change constantly as political events continue to unfold.

Just as with pari-mutuel horse racing, the estimated probabilities and the actual payoffs in a prediction market are set by the bettors. When more money is bet on a particular claim to win (i.e., come true), it means that bettors have attributed a higher probability to that claim so the cost increases and the payoff is less should it come true. With long shots, less money is bet on a claim so there is a higher payoff if that claim should  come true. Thus, prediction markets are just like betting on horses—if a heavily favored horse wins, its payoff is minimal, and vice versa if a long shot is the winner.

Rationale
The rationale underlying prediction markets is rather straightforward: they are able to flush out information that otherwise would not be available. Individuals around the world have different tidbits of inside information, and they know that using such information can enable them to earn a profit. Prediction markets are able to quickly and successfully aggregate such information as no other mechanism can. 

The probabilities existing in a prediction market thus reflect the collected wisdom of knowledgeable people everywhere, including those with inside information. Not surprisingly, academic studies have found that these markets are highly consistent (although not infallible) predictors of future events. Impressively, these markets consistently predict better than any other prediction tool, including expert opinion, quantitative modeling and computer simulations.

Thus, risk managers can look to prediction markets in order to obtain the world’s best and most informed probability estimates of all types of events happening, including crises and disasters. The probabilities in these markets continually reflect the collected wisdom of knowledgeable experts and savvy insiders. Moreover, because these markets exist only in cyberspace, the prices of claims in these markets are mostly immune to government regulation, political pressures, monopoly pricing and other factors that routinely affect and influence the prices in financial and commodity markets, keeping the probabilities in these markets pure.

Terrorism Futures
The ability of prediction markets to flush out inside information is what spawned the U.S. Pentagon’s recent Terrorism Futures Market, which was quickly terminated after the announcement of its existence in July of 2003 caused a sharp Congressional outcry. At first, such a market seems ghoulish until one realizes that it actually would have the ability to flush out inside information regarding terrorism. 

The Pentagon’s Terrorism Futures Market, officially known as the Policy Analysis Market (PAM), was designed to be a prediction market operated by the Pentagon and two private companies: the Economist Intelligence Unit (EIU), which is a division of Economist magazine; and Net Exchange, which was founded by several faculty members from the California Institute of Technology. The Net Exchange was created to conduct the actual trading of terrorism futures, with the EIU being the only registered member. 

By design, bettors would enter claims through EIU, and such claims would then trade on the Net Exchange. The Pentagon would not have had access to the bettors’ identities or funds. However, the Pentagon would have had access to all claims and their prices (i.e., probabilities). 

Claims were to be traded on all types of issues (politics, economics, terrorism, etc.) in the countries of the Middle East. PAM’s Web site gave examples of claims that included the assassination of Yasser Arafat and a biological-weapons attack on Israel. The Pentagon’s intention was to flush out inside information regarding events in the Middle East, including planned acts of terrorism. Presumably, the Pentagon intended to grow this market to include other areas of the world, especially the United States, which is a prime terrorist target. (In order to preempt the possibility of terrorists creating a claim and then profiting from it by making it come true, the Pentagon intended to limit the dollar amount of each claim so that the monetary incentive for particular acts of terrorism was minimal.) Highly priced claims in this market would have indicated high probabilities of the respective events coming true (barring any preemptive action), while rising claim prices would have suggested events that were currently being planned.

When existence of this market was made public in July 2003, the response by an angry Congress was swift and sharp. Denounced as “repugnant,” “morally wrong,” and “unbelievably stupid,” the market was quickly terminated by the Pentagon. Further, its creator, Admiral John Poindexter, was forced to tender his resignation. Quite possibly, a private concern will fill the Pentagon’s void and operate this market. Needless to say, the ability to uncover potential acts of terrorism and to thereafter preempt or, at least, hedge against such acts would have enormous social utility. 

Risk Management
Nevertheless, catastrophes are traded in existing exchanges. Under the category of “U.S. Catastrophes” in the Foresight Exchange, available claims include the likelihood of suicide bombers in the United States before 2005 or a large West Coast earthquake before 2010. Risk managers would be well advised to periodically examine the probabilities of such claims coming true, and thereafter implement hedging strategies should such probabilities become, in the manager’s opinion, too high.

Risk managers may also want to monitor other claims, as well. For example, political outcomes can affect government penalty risks, economic outcomes can influence profitability risks, and scientific outcomes can affect a wide range of risks. Prediction markets routinely reflect probabilities of stock prices, exchange rates, commodity prices, political events, natural catastrophes, acts of terrorism, scientific discoveries and cultural changes. 

Risk managers who operate in the world of finance will recognize that prediction markets are actually a type of “efficient” market called a “strong form” efficient market. In a strong-form efficient market, prices reflect all available information, including inside information. (By comparison, in a “semi-strong” efficient market, prices reflect only publicly available information.) In fact, since prediction markets are almost completely unregulated, they may be the most “efficient” markets in all of history.

Prediction markets can be handy tools for any risk manager. With their ability to consistently predict events, they contain critical data often needed for successful risk hedging.

Russ Ray, Ph.D. is a professor of finance at the University of Louisville in Louisville, Kentucky.