By Peter Coy
Suicide bombers kill 25 people in Saudi Arabia in May. A car bombing in early August in Indonesia kills at least 10 and injures 150. Where will terror strike next? The Pentagon's Defense Advanced Research Projects Agency hoped to gather fresh insights starting in October by launching a small-stakes betting parlor for speculation on geopolitical events. The idea was to use market forces to develop an expert consensus forecast on war- and terror-related topics. But no sooner was the experimental $1 million program publicized than it was killed on July 29. And within days, DARPA's Information Awareness Office chief, John M. Poindexter, resigned. Critics said it was immoral for Americans to run a pool to bet on death and destruction abroad.
The debate over the so-called Policy Analysis Market has been loud, but there have been few answers to basic questions: How was it supposed to work? And if it did work, would it have accomplished something useful? A closer look reveals a concept that's deeper and more interesting than either critics or backers have said.
The plan went through several revisions, including one in which the Pentagon would have allowed betting by the first 1,000 members of the public who applied. But the most acceptable version would have limited betting to invited experts from government, academia, and industry. They would have bet on events, initially in the Middle East only, such as whether the U.S. would pull its troops out of Saudi Arabia. Bets on assassinations -- the most controversial feature of the experiment -- could have been banned.
To prevent profiteering on bad news, the wagers would have been kept small. The maximum gain from any trade: $100. DARPA figured the thrill of being right would be enough to lure bettors. Each new wager would change the consensus, just as odds change at the horse track as people bet. The result: a constantly evolving collective forecast.
The Policy Analysis Market would have been more comprehensive -- and thus more useful -- than other online betting parlors. Forums such as Dublin-based TradeSports.com, which takes wagers on sports and current events, handle only bets where there's enough interest to keep the action lively. But thanks to a clever design by George Mason University economist Robin Hanson, the Policy Analysis Market could have taken action on literally millions of possible scenarios. That's because the Pentagon would have posted the original odds on each possible bet, giving people an opportunity to make money off the feds by making better guesses.
By giving odds to well-informed bettors, the Pentagon would lose money on average. But with bets limited to $100, it would have been a small price to pay for a snapshot of expert opinion. No one argued that the $1 million spent to launch the Policy Analysis Market could supplant conventional information-gathering and spying. But its forecasts might have become closely tracked -- not only by the Pentagon and the CIA but by multinationals on the lookout for news. After all, other current-events betting pools have worked well. Over the past four Presidential elections, Iowa Electronic Markets has forecasted outcomes better than polls.
Could businesses have used the Policy Analysis Market to hedge against events, the way airlines use energy futures to hedge fuel costs? No, because the money was too small. Even if a business successfully wagered that terror would increase in a country where it had operations, its gross winnings would be no more than a few thousand dollars -- not enough to offset its losses.
Yale University economist Robert J. Shiller calls the Policy Analysis Market blowup a "PR fiasco" for efforts to launch new markets. That's a shame, because a properly launched project could have made us smarter -- and maybe even saved lives. Coy is Economics Editor.