DOD's misguided bet on the future

System sought to use the market to help feds predict future terrorist attacks

On the face of it, the idea seemed "grotesque," as one senator called it. The Defense Department appeared to be betting on potential terrorist incidents.

The Defense Advanced Research Projects Agency's program to buy and sell futures contracts on potential terrorist events in the Middle East was publicized and promptly halted this month, but the economist behind the idea said it could be used to the government's benefit.

The Policy Analysis Market, which was set to launch Aug. 1, was seen by DARPA officials as a way of using the predictive ability of markets to anticipate terrorist events or other crises. Traders would have been able to place bets on myriad events, including whether Palestinian leader Yasser Arafat would be assassinated.

"The intent was to focus on big, aggregate [events] in the Middle East, like a whole country over a period of time and the average level of its economy or growth of the military, and how that would impact military casualties" in the United States, said Robin Hanson, assistant professor of economics at George Mason University.

Hanson said he created the basic concept and pulled the team together that responded to DARPA's original request for proposals (RFP) in 2001. That team included Net Exchange, a market technologies company, and later added the Economist Intelligence Unit, a division of the publisher of the Economist magazine.

"There was no intention of trading in individual terrorist attacks," and the Web site would not have accepted more than $100 on any particular bet, Hanson said.

The market was based on the concept of "combinatorial betting," where traders bet on various combinations based on a set of events. For example, the question is if the United States were to pull its troops out of Saudi Arabia, would that increase or decrease the number of terrorist-caused deaths in the United States next year, Hanson said.

Futures trading has proven more effective than surveys and expert opinions in predicting other events, including elections, said Erik Gartzke, associate professor of political science at Columbia University.

"The market is very effective at aggregating information. When someone is on to something in a market, people [rally behind] it even if they don't know what 'it' is," Gartzke said. "A decentralized information source — like which terrorist group is going to attack next — that's harder to do."

He said the government struggles where markets succeed because of all the bureaucratic hoops those trying to collect information must jump though in order to get senior leaders to act.

"Markets overcome that by forcing parties that make decisions to pay attention to frontline information" and quickly dispelling unimportant data, Gartzke said. "Nothing is perfectly random. The tricky part is getting through all the noise."

DARPA's piece of the market, the Futures Markets Applied to Prediction (FutureMAP) program, was granted approval by John Poindexter, director of the agency's Information Awareness Office.

Poindexter was not leading the DARPA office when Hanson and his team responded to the original futures market RFP, but the retired Navy admiral is responsible for granting approval to all programs under his purview.

DOD officials said this month that Poindexter is still a DARPA employee, but it has been widely reported that he will step down soon., a person-to-person trading "exchange," has listed a line on whether Poindexter will still be employed by Aug. 31, and a similar Web site lets users bet on whether Defense Secretary Donald Rumsfeld will still be in his office Oct. 1.

The Pentagon had planned to spend $8 million on the futures market through 2005, but that ended when DARPA's director, Anthony Tether, said that the agency's participation in the program had been withdrawn.

Hanson said the program was killed because of some "unfortunate examples used on the Web site," including the Arafat assassination, "which some senators spun to betting on terrorism."

Sens. Ron Wyden (D-Ore.) and Byron Dorgan (D-N.D.) held a July 28 news conference where each blasted the idea, with Dorgan saying it was "incredibly stupid [and] an appalling waste of taxpayers' money."

Hanson disagrees. "The combinatorial technology can be used for other things," including helping the government decide whether to approve a new drug by looking at whether its approval will increase of decrease the number of deaths over the next year, he said.

DARPA's handling of the program was "politically naive," said Gartzke, but the "failure politically to get [it] going is exactly the kind of problem the market can overcome [because] information doesn't get where it should be in a bureaucracy."