Home Help
 Academic Programs
 Medical Center
 Events Calendar
 Prospective Students
 Faculty & Staff
 About Columbia
 A–Z Index
 E-mail & Computing

Columbia News
Search Columbia News
Advanced Search
News Home | New York Stories | The Record | Archives | Submit Story Ideas | About | RSS Feed

Markets Show Greater Accuracy than Polls, Say Columbia Professors

Do you think Bush will win the White House again? Want to bet on it? Well, what you might be inclined to wager may be a more accurate indicator of a candidate's chances than traditional polling. And Steven Kou, associate professor in the Department of Industrial Engineering and Operations Research, and Michael Sobel, professor of sociology, have developed a theory to support that assertion.

Some real-money futures markets, in which self-selected participants trade shares of candidates at prices predictive of Election Day results, such as the Iowa Electronic Market (IEM), have performed better than polling data, the researchers found. In the summer issue of Political Analysis, the two professors published their theory in an article called "Forecasting the Vote: A Theoretical Comparison of Election Markets and Public Opinion Polls."

"We provide the mathematical justification -- a formal theory, and we show under what conditions that market will out perform the polls," Sobel said. "And not this one poll, but any combination of polls that you want. Our results show that as long as people have access to the polls that the market price will be a better predictor of the actual vote share than will the polls."

The way in which the IEM works is that a trader can buy a ticket containing contracts for each candidate for a dollar and then un-bundle the ticket and trade the individual contracts. What the trader holds after election gets paid out at whatever percentage each candidate gets. IEM, which was established in 1988, now allows traders to invest up to $500.

One indicator of the difference in which the markets and polls perform is the fluctuation in share prices based on a candidate's chances. For instance, after a party's national convention, a candidate's popularity often surges in the polls -- indicating a greater chance for the candidate's chances of being elected. Not so in the market. The shares at which the candidates are traded rarely fluctuate so dramatically at specific points, barring a major scandal or unforeseen development.

While urging that political scientists use market data in their analyses, Kou and Sobel envision the markets could be useful for other forecasting purposes as well. Along with any other U.S. elections of public interest, markets similar to the IEM predict everything from the outcome of Federal Reserve monetary policies to Hollywood box office revenues.

Related Links

Published: Oct 12, 2004
Last modified: Jan 10, 2005

Tell your friend about this story