Professor-Created Models Predict Obama Will Barely Win 2012 Election

Professor-Created Models Predict Obama Will Barely Win 2012 Election

According to an average of 13 models created by political science professors across the U.S., President Barack Obama will win the election by .06 percent. (SHFWire photo by Jory Heckman) Of the 13 models created by political science professors across the U.S., five are in favor of Mitt Romney, five are in favor of Obama, and three are toss-ups. File photo by Tanya Parker.

By Emily Wilkins
Scripps Howard Foundation Wire
October 17, 2012

WASHINGTON — Fluctuating polls aren’t the only way to predict an election. Thirteen different models published in the October issue of Political Science and Politics (PS) give incumbent President Obama the win — but just barely.

The forecasts, which show the election to be much closer than the 2004 or 2008 elections, give Obama an average .06 percent lead ahead of Romney.

James E. Campbell, a political science professor at the University of Buffalo, helped select and edit the models for the October issue of PS. Campbell and three other professors who created forecasts spoke Oct. 16 at the National Press Club. …

This is the third election year the models have been collected into PS. In 2004 and 2008, a majority of the models correctly predicted the elections of George W. Bush and Obama.

This year, there’s less agreement. Of the 13 models, five predict Romney, five predict Obama, and three are toss-ups.

Because each model uses different factors in predicting the election, the results of each model are not the same. In nearly every report, changes in the economy played a role. Some forecasts used direct numbers from the economy, while others examined how the electorate has reacted to changes in the economy. Other factors included popularity of the current president, jobs and current polls.

When each forecast is taken individually, the highest Obama can get is 53.8 percent of the vote, and the lowest he can get is 46.9 percent.

“What we see every day in the newspaper and in the models is this conflict or confusion of all the numbers, and they’re not all going in the same direction,” said Michael S. Lewis-Beck, a professor of political science at the University of Iowa. “I think this is pointing to the tightening of this race and uncertainty of this race.”

When creating the models, professors seek to predict the correct winner rather than picking their favorite candidate, Campbell said. Although he is a Republican, Campbell predicts Obama will win with about 52 percent of the popular vote. He predicted Clinton’s victories in 1992 and 1996.

“That’s one of the nice things about a statistical model,” Campbell said. “If you have it in place before the election, the numbers speak for themselves.”

He added that he wasn’t very happy to predict Clinton – save the fact that he was right.

Thomas M. Holbrook, chair of the department of political science at the University of Wisconsin, correctly predicted George W. Bush and Obama’s wins in 2004 and 2008. Using a model that examines presidential approval ratings and personal finances of the electorate, he predicts a Romney victory.

Holbrook acknowledges that Obama was leading the polls for a while and was even being declared the victor by some groups. But according to the model, it just meant some event – in this case the first debate – would even things out.

“Campaign events serve as a corrective,” he said. “If a candidate is running way ahead of the polls, a campaign event favoring the other candidate can correct that overexuberance to bring public opinion back into line with the expected outcome.”

Reach reporter Emily Wilkins at or 202-326-9867. SHFWire stories are free to any news organization that gives the reporter a byline and credits the SHFWire.