Business intelligence worth betting on
What if you could predict important trends that could greatly impact your company's fortunes?
For example, if you knew that there was a 90 percent chance that a rival company would develop a new product within the next two months that competed with one of your high margin products, you could react more swiftly and decisively. Or, wouldn't you like to know whether your sales forecasts are more or less accurate that what the sales people are projecting?
If this is the kind of insight you want, and your not planning a career in industrial espionage, then check out a discarded, dishonored project from Defense Advanced Research Projects Agency (DARPA). As you might have read, the Pentagon was far along the path of setting up commodity trading market system in which "investors" could bet on political and military events in the Middle East. A select group of investors could buy futures contracts on high value questions, such as whether the Jordanian monarchy would be overthrown during the U.S.-Iraq war or the length time the large contingent of U.S. troops would need to stay in Iraq.
The modest project---start-up funding was aimed at small business entrepreneurs---was scuttled by the Pentagon after Congress learned of its existence. DARPA withdrew its participation in the Futures Markets Applied to Prediction (FutureMAP) program, which was one of the sponsors of the Policy Analysis Market (PAM) Web site (now inaccessible) that drew attention to the Middle East trading market. Senate Minority Leader Tom Daschle of South Dakota called it a "plan to trade in death."
DARPA/PAM had planned to begin live trading on October 1, serving 1,000 traders, and offer three types of future contracts:
- Quarterly contracts based on data indices that track economic health, civil stability, military disposition, and U.S. economic and military involvement in Egypt, Iran, Iraq, Israel, Jordan, Saudi Arabia, Syria, and Turkey
- Quarterly contracts that track global economic and conflict indicators
- Specific possible events, such as U.S. recognition of Palestine in the first quarter of 2005
The DARPA plan was politically inept and subject to manipulation and obfuscation like any other trading market, but the underlying concepts are not without merit or precedence. I spoke with Robin Hanson , associate professor of economics at George Mason University in Fairfax County, Virginia, who came up with the concept of applying commodity-style trading markets and combinatorial betting technology to war gaming.
"DARPA [Michael Foster of the Terrorism Information Awareness program, formerly the Total Information Awareness program] put out a call for proposal on how to use information technology to help intelligence assessments," Hanson said. "We had to decide what we would apply market to, and we tried for the most high value questions to get the biggest benefit-to-cost ratio. We started out estimating what wars would happen everywhere in the world, but we have to scale back to the Middle East due to dollar constraints." DARPA awarded a $750,000 contract to Net Exchange, which is headed by John Ledyard, professor of economics and social sciences at the California Institute of Technology and Hanson's former faculty advisor.
The concept of using trading markets to predict the future isn't new. The University of Iowa Electronic Market (IEM) has been forecasting the outcome of U.S. elections since 1988 via on-line futures markets. Each market is tied to a specific event, such as a presidential primary or election, and contains a set of contracts with liquidation values pegged to the outcome of future events, according to Tom Rietz, associate professor of finance as the University of Iowa.
"Our average overall prediction error for all four Presidential elections since 1988 is 1.37 percent," Rietz said. "Other U.S. elections, such as Senate, House, and primaries, have had an average error of 3.43 percent, but that was driven by an extraordinary event in 1992 presidential primary when [Paul] Tsongas announced he had cancer." Compared to major polling organizations, IEM was more accurate 76 percent of the time in predicting the outcome of elections, Rietz said.
He noted that the trading markets aren't working with a random sample. With real money in play, participants are typically more informed and educated than in other types of markets.
Rietz and some of his colleagues from the University of Iowa who are associated with IEM formed Market Technology Systems LLC (MarTek) to commercialize prediction market technology. Until DARPA pulled the funding, the company had been partnering with Neoteric Technologies Inc. of Huntsville, Ala., on the PAM project. The two companies had also considered creating markets for predicting the outcome of events related to product development, such as when ships, aircraft or other pieces of military equipment might be completed.
Similar to IEM and PAM, TradeSports.com, a Dublin-based trading exchange, allows individuals to bet on everything from the outcome of sporting events to daily stock market closes. This week, TradeSports opened a trading contract on Admiral John Poindexter, the head of the Terrorism Information Awareness program, remaining on the DARPA payroll by August 31, 2003. (Shortly after TradeSports.com opened its Admiral Poindexter trading contract, the Pentagon announced that the controversial TIA head would be leaving his post in the next few weeks.)
The Hollywood Stock Exchange has developed technology for trading virtual shares of celebrities, movies and music with its own currency. The company then syndicates the data collected from the exchange as market research. The Foresight Exchange also provides a "fantasy" trading environment for predicting the outcome of events.
Hanson also believes that predictive markets have a role in formulating government policy. In a paper entitled "Futarchy: Vote Values, But Bet Beliefs," Hanson wrote, "A betting market can estimate whether a proposed policy would increase national welfare by comparing two conditional estimates: national welfare conditional on adopting the proposed policy, and national welfare conditional on not adopting the proposed policy. Betting markets can produce conditional estimates several ways, such as via "called-off bets," i.e., bets that are called off if a condition is not met.
Hanson cited internal corporate decision-making as another domain that would benefit from the technology. "Corporations agonize about whether to develop a product. These trading markets could be used internally to find out what people in the trenches think."
The question comes up as to whether terrorists or even product managers could manipulate the betting system to achieve a desired result. "We all understand that there are different ways in which people collect information and manipulate it," Hanson said. "The question is, is it better or worse than any other way? The comparison so a far is that the markets have beat other institutions. In situations where people want to manipulate things, any way we do it has the same problem--it's called counter intelligence."
The most practical application of an information market I could find was by scientists from the California Institute of Technology and Hewlett-Packard who wanted to see how a trading market fared against HP's official printer sales forecasts. The betters proved to be more accurate than the official forecasters six out of eight times.
While DARPA may have tanked the illicit project, I am sure the notion of information aggregation and predictive trading markets is finding applications within the Pentagon and other agencies. At the very least, the episode brought a lot of attention to the concept. Corporations looking to gain an edge and improve the quality of decision-making should look into creating pilot trading markets to test the concept. I predict that predictive information markets will become an important part of business intelligence activities among the most progressive companies by 2007. Does anyone want to bet on that?
Use TalkBack to let your fellow ZDNet readers know what you think. Or write to me at dan.farber@cnet.com. If you're looking for my commentaries on other IT topics, check the archives.