8 August 2003—The hue and cry over the Policy Analysis Market, a recently revealed and quickly canceled U.S. Department of Defense research plan to use a futures market to help it predict terrorism events, has obscured a couple of important facts. Similar markets have been used for years to predict everything from the next U.S. president to the toppling of Saddam Hussein, and generally they have had excellent track records.
The Policy Analysis Market, or PAM as it’s funding parent, the Defense Advanced Research Projects Agency (Darpa; Arlington, Va.) calls it, has been taken too seriously by some and perhaps not seriously enough by others. Quickly quashed after a drumfire of disdain from the U.S. Congress, the idea of a terrorism futures market reappeared in civilian form as the American Action Market [http://www.americanactionmarket.org]. But that Web site, despite being taken at face value by media outlets like Wired News, could be merely a parody of PAM. Evidence points to connection to the creators of a well-known political spoof site, gatt.org.
It’s too bad that the idea of futures markets is not given enough level-headed thought. The PAM project certainly had a creepy feel to it, studying as it did things like whether or not specific Middle Eastern prime ministers would be assassinated. But financial markets have always served an important role in the economy—their ability to aggregate opinion and forecast the future. For example, the futures market for oil is the best predictor economists have of that vital commodity’s future price.
On-line predictive markets like PAM have long been used to predict future prices, trends, and events such as elections. Months before PAM hit the news, the Irish betting site, Tradesports.com [http://tradesports.com/], had been taking wagers on an event of great interest to the Defense Department: when Saddam Hussein would be deposed.
The granddaddy of online predictive markets, one that has been accurately foretelling U.S. presidential election results since 1988, is the Iowa Electronic Markets (IEM), run by the University of Iowa’s Tippie College of Business in Iowa City [http://www.biz.uiowa.edu/iem/]. It consistently beats the polls at predicting election results. This shouldn’t be as surprising as it seems, since polls aren’t a test of who will win but, rather, a measure of how the electorate would vote at the time of the poll.
The IEM, on the other hand, is targeted at the actual election. There are two forms of wager. In the first, called a vote-share market, a contract pays off at US $1 multiplied by the relative share of the vote. Thus, a trader who thinks a candidate will get 40 percent of the vote will be willing to buy the contract at $0.39 or lower, but not $0.41 or higher.
The other type of contract, winner-take-all, gets a similar result by a different means. The contract will pay $1 or nothing at all, based on who wins the election. But traders can value the contract however they want. Those who think there’s a 40 percent chance of a particular candidate winning feel they are getting a bargain at any contract price up to and including $0.39. In the IEM, real money exchanges hands, though the amounts are $500 or less as part of an agreement with the Commodity Futures Trading Commission (Washington, D.C.).
Markets such as these ”act as complex, dynamic, interactive systems that incorporate information in new ways,” according to ”Accuracy and Forecast Standard Error of Prediction Markets,” a paper written by Joyce Berg, Forrest Nelson, and Thomas Rietz, three of the University of Iowa faculty members involved in IEM. Says Berg: ”There’s lots of research in experimental economics that says markets are good at aggregating information.”
The Iowa group and others have found that IEM traders with more information do well. Those without, like poorly skilled poker players, provide the profit that the better players walk away with. (As an old adage in poker puts it: ”If you can’t tell who at the table is the sucker, it’s you.”)
Berg and her colleagues also acted as consultants for other Darpa market experiments. ”There has to be information to aggregate,” she says. ”That’s the first question. So to run a market like PAM, you’d want to include policy analysts, experts in Mid-Eastern affairs. Specific people would be invited. On the other hand, you don’t always know who has useful information, so it might be a mistake to exclude anyone.”
The Department of Defense never said exactly how it would use the price-derived information culled from PAM, but Berg says it’s easy to see how such markets can serve government officials. For example, Neoteric Technologies (Huntsville Ala.) had already created similar, albeit more benign, markets, such as whether and when SARS would reappear. Berg says, ”A market like that might help us decide how to invest research money and other medical resources.”
Independent of its university affiliation, the Iowa group served as consultants to Neoteric. Though funded by Darpa under a so-called Small Business Innovation Research grant awarded in 2001, Neoteric did not work on PAM itself. [See a description of their work at .] Research on the PAM markets was done by Net Exchange (San Diego, Calif.), a group that includes faculty from the California Institute of Technology and George Mason University [ http://www.nex.com].)
Likewise, Berg notes that Tradesports, the Irish betting site, and her group in Iowa as well, have run threat-level markets. Referring to a U.S. government practice of assigning different colors to different levels of danger from terrorists, she says that one such contract might concern the number of days in a given month that are Level Red. ”That might help state and municipal governments decide how to budget, or allocate emergency personnel,” she says.
There just aren’t many tools out there for predicting events like coup d’etats and assassinations. The main one is called the Delphi method. Back in the early 1960s, Berg explains, the U.S. government was trying to find ways to aggregate expert opinion. Researchers at RAND (Santa Monica, Calif.) found that if you put a group of experts in a room and got them talking, the loudest voice’s opinion usually wins.
To avoid that, they tried using anonymous questionnaires. They would then report to the members of the group various statistics from the survey, such as the median and the mean, and then ran the survey again. The research says that after three or four iterations, the opinions converge and serve to aggregate the expert information.
Predictive markets, Berg says, are also forums for communicating where the loudest voice doesn’t win. A market transforms everything into a single statistic—the price of a contract—and conversations can start on whether that’s a good number.
Spoof or oracle?
Creepy as it was, PAM’s pedigree and potential value were lost in the debate following its disclosure. Further precluding any serious discussion, along came the American Action Market (AAM) . It says it will create much the same markets PAM would have. But despite AAM’s claim to have been founded by a group of academics at some of the top institutions in the United States, there are clues that it isn’t really a serious attempt to create futures markets.
For one thing, the examples of questions to be asked in the betting pool, which were given in AAM’s founding press release, don’t lend themselves to contracts. One was, ”Which corporation will be next to see its close relationship to the White House erupt in scandal?” Another asked, ”Was the invasion of Afghanistan planned from the start as a stepping-stone to an attack on Iraq?” isn’t even a question about the future. Futures contracts need to be easily settled. That’s why they ask crisp questions like ”Will at least 100 000 people die of SARS by the end of 2004?” rather than the vague ”Will SARS continue to be a threat?”
A look at the domain name system record for americanactionmarket.org is also revealing. (A public record of the owner is created when a domain name is registered.) The site’s owner is said to be Hedwig Ixtabal-Mono, a name that Google associates with Gatt.org, an organization known for creating spoof Web sites. While many such records have spurious data, it’s an odd coincidence.
Bob Ostertag, a San Francisco musician and freelance writer who is apparently AAM’s spokesperson and public relations director, was at a loss to explain the Hedwig Ixtabal-Mono name. When asked whether the organization—unincorporated, with no physical address, and a board of directors who refuse to reveal their own names—is a real market or a spoof, Ostertag said, ”When we launched, it wasn’t clear it wouldn’t be a goof, but it became clear that people were interested. It’s become more real every day.” He added, ”Hundreds and hundreds of individuals have sent e-mails wanting to become traders.”
The future of futures
Whether AAM is a hoax or not, online predictive markets are here to stay. Eric Zitzewitz, an assistant professor of economics at Stanford University’s Graduate School of Business (Palo Alto, Calif.), studied the Tradesports site, which, before the U.S.�Iraqi war, included a market for whether or not the war would be averted. ”Probably about $1.5 million traded hands” in the run-up to the war, he says.
Zitzewitz compared the Tradesports market to the Standard and Poor’s stock market index, since they both rose and fell during major events like U.S. Secretary of State Colin Powell’s United Nations presentation on Iraqi weapons of mass destruction. ”Tradesports incorporated the S&P information with lags of only a half a day or so. The S&P is incredibly efficient,” he says, ”but without the fluctuations in the market price concerning Saddam’s future [at Tradesports], you couldn’t separate the war news from the general corporate and economic news that affect the stock market.”
Predictive markets have also taken a particular hold in the entertainment world. The Hollywood Stock Exchange [http://www.hsx.com/] predicts the success of Hollywood movies, while at Celebdaq [http://www.bbc.co.uk/celebdaq/] one can bet on the popularity of Hollywood celebrities—what’s measured is the number of times in a given month the individual is named in the popular media. There’s also Fatdog Exchange [http://www.fatdogexchange.com/], where the notion of celebrity includes not just Britney Spears, but Bill Gates and Bill Clinton.
Zitzewitz thought the Policy Analysis Market might be useful, but only if it could attract enough people with information to be successful. He told me that having journalists participate would help. ”You could see journalists looking at a price, seeing a 10 percent chance of an event, saying that’s probably wrong, and buying a contract,” he said. At last, I thought, someone thinks I have the inside scoop. Plus, it could be a way to make a little cash on the side.