All seeing all knowing
Futures markets can help identify successful products, predict revenues, and even forecast whether Bush or Kerry will win in November
The superior performance of the Iowa market reflects one reason why these predictive exchanges work better than simple opinion polls. The people who trade are self-selected, political know-it-alls who are likely to follow the presidential race far more closely than the average voter who gets contacted by a pollster like Gallup or Zogby. In addition, says Malone, "the participants have a clear self-interest in predicting what will really happen rather than what they want to happen or what they want someone else to think will happen." In this case, there's money on the line, but benefits like reputation or personal satisfaction can also create powerful incentives. With a poll, the prediction is made without consulting anyone else. By looking at the current odds in a prediction market, traders are given another piece of information--they can examine their beliefs in the light of what others appear to believe. Each person sees the current "consensus" view and then makes a trade only if he thinks that view is wrong.
The skinny. In a world of accelerating technological change and global competition, anything that can help a CEO make a better forecast or smarter decision probably seems worth investigating. Pharmaceuticals giant Eli Lilly spends 20 cents of every sales dollar on research and development. That's up from 10 cents in the 1980s. And some two thirds of that money goes toward drugs that will never make it to market. So Lilly is experimenting with prediction markets to get a better handle on which drugs will be successful and which will win approval from the Food and Drug Administration.
Last year, Lilly ran an internal market experiment where instead of using the public as its pool of potential traders, the company tapped about 50 employees--chemists, biologists, salespeople--to trade contracts representing six hypothetical drugs. The object was to figure out which ones had the best chance of making it to the next phase of clinical trials. Participants used drug data from real compounds that the market's organizers would slowly leak out. As it happened, the three highest-trading drug candidates were, in real life, the three successful drugs, while the lowest-trading ones were trial failures. Asked if he was impressed by the outcomes, Alph Bingham, head of R&D at the company's entrepreneurial e.Lilly initiative, says, "Let me put it this way: We're not abandoning the experiments."
While Lilly is still figuring out how and where to apply internal prediction markets in the decision-making process, HP is already employing them for sales forecasts. Back in the mid-1990s, HP set up a series of experiments with California Institute of Technology economist Charles Plott to create a market around forecasts of future sales of its printers. A few dozen finance or sales people could buy or sell predictions about whether, say, August sales would be between 1,500 and 1,600 units. In six out of eight forecasts, the markets were significantly closer to the actual result than HP's official forecast. The company was so enthusiastic about the results--as well as those of some subsequent experiments--that one division is now using markets to predict monthly unit revenue. "The markets allow us to extract knowledge and put it to good use," says Bernardo Huberman, director of the Information Dynamics Lab at HP Laboratories.
advertisement
