NationalGeographic.com [an error occurred while processing this directive]


 

Online Extra
November 2004



<< Back to Home Page






Who'll Take the Prize?
Political Prognosticator Picks the President


Online Extra
Photographs courtesy STF/AFP/Getty Images


Facing off for America, Democratic presidential candidate John Kerry (left) and U.S. President George W. Bush (right) defend their positions during the first of three debates, held September 30, 2004, at the University of Miami.



By Shelley Sperry

Jim Campbell waited a little longer than usual this year to predict who would win the U.S. presidential election. With the Republican Party convention held just before Labor Day, the University at Buffalo, SUNY political scientist tweaked his model a bit to account for George W. Bush receiving a post-convention bounce in the Labor Day Gallup Poll, Campbell's most important factor in predicting the November vote. In the end, he created post-convention and pre-convention models, and both say that Bush will win. Campbell does not take into account the presidential debates. Using pre-convention polling numbers, Campbell says Bush wins 52.8 percent of the popular vote for the Republican and Democratic parties. Post-convention numbers predict he wins 53.8 percent. But what about the all-important electoral college vote? "If my forecast is close, within two points," Campbell declares, "Bush will clearly win the electoral vote as well."

Campbell created his current forecasting model in 1990 using two kinds of predictors: public opinion and economic growth. "The Labor Day Gallup Poll of likely voters accounts for about two-thirds of the model," he explains. That poll showed Bush ahead of Democrat John Kerry by 7 percentage points. But Campbell believes we have to read polls along with other factors to put them in context. "Historically, there is a relationship between the economy and people's voting patterns." So the second factor in the model is the gross domestic product (GDP) growth rate in the second quarter of the election year. This year's rate was 3.3 percent, and, according to Campbell, anything roughly over 3 percent favors the incumbent.

In early September, Campbell and colleagues who specialize in analyzing elections and public opinion took part in a lively roundtable in Chicago, where they discussed their 2004 election predictions. Of seven diverse models, six forecasted Bush would win in November—and the seventh saw another election "too close to call."

Although Democrats may see these as gloomy projections, they can take heart from a similar group of prognosticators who picked Al Gore as the winner of the 2000 election. That year Campbell came closest to predicting the popular vote outcome, saying Gore would win 52.8 percent. Gore did win the popular vote, with 50.3 percent, but lost the electoral college contest and the presidency. None of the pundits predicted the election would be a virtual tie, ultimately settled by the Supreme Court.

For more information about elections, polls, and predictions, Campbell recommends the sources listed below.

Related Links 
Update for The American Campaign
wings.buffalo.edu/polsci/faculty_and_research/campbell/campbell.htm
James E. Campbell provides an update to his book, The American Campaign, which extends through the 2000 presidential election.
 
Political Forecasting Special Interest Group
morris.wharton.upenn.edu/forecast/Political
Known as Pollyvote, this University of Pennsylvania site is packed with up-to-the-minute polls, a variety of forecasts, commentaries, and many other items of interest to political junkies.
 
PS Online
www.apsanet.org/PS
The October issue of PS, the American Political Science Association's journal, includes a full report on the panel of election forecasters chaired by Jim Campbell in September and links to more information about the forecasts. 
 
Gallup Poll: Election 2004
www.gallup.com/election2004
Read updated poll numbers, analyses, candidate profiles, and in-depth reports on the election in "showdown states." 
 
 You Can Predict the 2004 Presidential Election (Maybe!)
morris.wharton.upenn.edu/forecast/Political/PDFs/JonesEssayIndicatorsModels%
20_30Mar04_.pdf

For do-it-yourself pundits, Randall Jones, a professor at the University of Central Oklahoma, has assembled explanations of the major methods for predicting elections and links to sites where you'll find the raw material you need.

Top



E-mail this page to a friend.



© 2005 National Geographic Society. All rights reserved. Privacy Policy       Advertising Opportunities       Masthead

National Geographic Magazine Home Contact Us Forums Shop Subscribe