Number 31, December 2001

Editorial

Prediction Markets: The Best Possible Forecast?

Nobel Prize winner Nils Bohr once remarked, “Prediction is difficult, especially of the future.” Recognizing the difficulties of prediction, what if there was a way to integrate all available information about the future into a single forecast that would instantaneously incorporate and make available new information? Sound too good to be true? Research in economics and recent developments in the private sector point to a novel way to think about forecasting in the earth and atmospheric sciences.

The basic premise of the approach lies in the efficient market theory from the field of economics and most closely associated with the work of Eugene Fama at the University of Chicago in the 1960s. The efficient market theory holds that the current price of a commodity in an exchange market reflects all available information. When you hear the phrase “you can’t beat the market” it is referring to the perspective that the “market” (usually the stock market) is efficient; i.e., if information were available that would allow someone to gain a trading advantage, this information would be reflected instantaneously in the price of the commodity through the actions of buyers and sellers in the marketplace. Whatever advantage the trader thought may have existed is absorbed into the market.

Economists have sought for many decades to prove or disprove the efficient market theory in the context of the stock market and other areas. In some cases research has shown that markets have inefficiencies (e.g., see the Research Highlight later in this issue which shows an “inefficient” relationship between the daily weather and stock market performance). But in many cases research has shown that markets are efficient to varying degrees. Consider the following examples:

• Researchers at the University of Iowa’s Tippie School of Business have created an “electronic exchange” where futures contracts are traded on political events like elections, but also phenomena such as box office returns for Harry Potter and Federal Reserve monetary policy. Although the Iowa Electronic Exchange is a non-profit enterprise, real money is traded in its exchanges. The exchanges provide a wealth of experience for understanding markets and also have proven to be highly skilled (but not perfect) predictors. The Iowa Electronic Markets appear to predict elections and other outcomes better than polls or models (see further reading below).
• Researchers have long studied sports betting and the apparent efficiency of the “point spread” between teams. In sports betting markets a “point spread” is the expected difference between two teams’ final scores. This spread moves in response to wagers placed by gamblers. Research suggests that the point spread is a more accurate predictor of sporting event outcomes than other methods such as statistics or power ratings (see further reading below).

What if a “prediction market” were created that would allow trading based on specific predicted outcomes such as the weather? Could such a market be created that operated efficiently and had sufficient participation to integrate all available information about the future? Here is a research hypothesis: An efficient “prediction market” generally will outperform all competing prediction methodologies in the earth and atmospheric sciences.

Recent developments in the private sector related to the securitization of risk suggest that “prediction markets” may not be so far off. In recent years a market has developed for “weather derivatives” to allow companies with weather-related risks to trade that risk in the marketplace, much like commodity futures such as gold and pork bellies are traded. Such markets typically focus on variables of direct relevance to particular industries such as heating degree days and cooling degree days, rather than the weather itself.

But this is changing. Recently a company called Aquila Energy announced a new weather derivative focused explicitly on the weather itself. Their “guaranteed forecast” promises to allow traders to “hedge their commodity position from changes in the forecast or a ‘busted’ forecast.” They are in effect creating a “point spread” for weather forecasts. One result of such a product, should it attract enough participation, is that Aquila (or perhaps its competitors) may wind up with more information than anyone else on the accuracy of weather forecasts, and this information itself may be of considerable value.

Several policy issues come to mind. First, there is an implication for the research community. It would appear to make sense to test the hypothesis about whether “prediction markets” can outperform other approaches to forecasting phenomena related to the earth and atmospheric sciences. Perhaps an equivalent to the Iowa Electronic Market could stimulate such research related to weather forecasting, climate forecasting, and, in principle, any area where predictions are made. Second, if the marketplace can provide skillful forecasts, then perhaps a mechanism might be created for this information to be provided systematically to consumers of forecasts through the public or private sectors, or through a partnership.

Make no mistake, prediction is difficult (especially about the future) and markets offer no silver bullet. But it is because prediction is difficult that innovative approaches to the integration and dissemination of knowledge should capture our attention. Perhaps four decades of experience in economics and its applications can contribute a fundamental innovation to forecasting applications in the earth and atmospheric sciences, maybe even leading us closer to the “best possible forecast.”

Roger Pielke Jr.
Center for Science and Technology Policy Research
University of Colorado
pielke@cires.colorado.edu

For further reading:

Forsythe, R., F. Nelson, G. Neumann and J. Wright, 1991. The Explanation and Prediction of Presidential Elections: A Market Alternative to Polls, pp. 69-112 in Laboratory Research in Political Economy, T. R. Palfrey, ed., University of Michigan Press

Stern, H. 1991. On the probability of winning a football game, The American Statistician, 45:179-184.

L. Zeng, 2000. Weather derivatives and weather insurance: Concept, application and analysis, Bulletin of the American Meteorological Society, 81:2075-82.