The papers, which come in the wake of the disputed 2000 contest, offer pollsters, campaigns, and news organizations innovative ways of predicting which candidate will win the presidency in 2004.
Forecasting Tips from Pork Belly Traders "Election Forecasts from a Futures Market" is by Forrest Nelson, Joyce Berg, and Thomas A. Rietz of the University of Iowa. Prof. Nelson is speaking on Monday, November 5 from 8:15-9:30 AM in the Imperial I Room of the Fontainebleau Hilton Resort.
In a unique approach, the authors shun voter polls and turn instead to a predictor borrowed from the stock exchanges.
The authors point to weaknesses in the predictive power of traditional polls. Although polls quote a margin of error, they say, pollsters do not attempt, nor can they be expected, to measure the degree of uncertainty about the eventual popular vote based on polling numbers.
In contrast, they explain, futures markets have an advantage over traditional polls. Futures traders are constantly concerned with the concept of uncertainty about an event that has not yet taken place. Those who observe these markets cope by using mechanisms for measuring the degree of uncertainty about an eventual outcome.
The University of Iowa runs Iowa Electronic Markets, its own futures markets in which investors can buy futures shares. The market is structured in a way that lets observers infer predictions from contract prices.
IEM ran a presidential election market in 2000. From the middle of May on, says Prof. Rietz, "our market predicted a dead heat. At times that wasn't exciting news since the public wants you to predict the winner. In fact, predicting 'too close to call' was a much better prediction than the one from the polls, which were predicting large wins for Bush."
In general, he says, "We like the idea that we're accurate far in advance and relatively stable." The markets, he says, are not a random sample of voters. Typically investors are better educated, reflect higher income, and often include college students.
"But that doesn't make a difference in the ability to predict," he says. "You can probably make better predictions using a trader pool of well informed people. That's true in commodities, too. In a typical futures market, say the corn market, traders are well informed about corn. That's what makes it a good model for predicting the future price of corn."
Forecasting the Electoral Vote In the other paper researchers, noting the discrepancy between the popular totals favoring Al Gore and the Electoral College vote that chose George Bush as President, recommend new analytical methods that focus less on a candidate's share of the popular vote and more on the probable number of votes that the candidate will win in the Electoral College.
"A New Approach to Estimating the Probability of Winning the Presidency" is being presented by Edward H. Kaplan, Yale School of Management, and Arnold I. Barnett, Sloan School of Management, MIT. The authors are speaking on Tuesday, November 6 from 4-5:30 PM in Ballroom B of the Fontainebleau Hilton Resort.
Current polls focus almost exclusively on the popular vote, say the authors. The Kaplan-Barnett model, in contrast, converts state-by-state polling results into a probability distribution for a candidate's total number of electoral votes. The model, say the authors, may show a high probability that a specific candidate will take a state's electoral votes although popular vote totals suggest that the contest is too close to call.
"Would our model have changed anything last year?," asks Prof. Kaplan. "On the one hand I'd say, 'No.' I would still have predicted that Gore would have won Florida. On the other hand, I'd say, 'Maybe.' If the candidates had a clearer sense of where they were headed in electoral votes in March or June, they might have responded differently and the course of the campaign might have been different."
The annual convention of the Institute for Operations Research and the Management Sciences (INFORMS®) takes place in Miami Beach from Sunday, November 4 to Wednesday, November 7 at the Fontainebleau Hilton Resort. Operations researchers are little known but indispensable experts who use math and science to improve decision-making, management, and operations in a host of fields.
The convention includes sessions on topics applied to numerous fields, including air safety, the military, e-commerce, information technology, energy, transportation, marketing, telecommunications, and health care. More than 1,800 papers are scheduled to be delivered. Additional information about the conference is at http://www.informs.org/Conf/Miami2001 and http://www.informs.org/Press.
The Institute for Operations Research and the Management Sciences (INFORMS®) is an international scientific society with 10,000 members, including Nobel Prize laureates, dedicated to applying scientific methods to help improve decision-making, management, and operations. Members of INFORMS work in business, government, and academia. They are represented in fields as diverse as airlines, health care, law enforcement, the military, the stock market, and telecommunications. The INFORMS website is at http://www.informs.org.