The University of Iowa
The University of Iowa News Services Home News Releases UI in the News Subscribe to UI News Contact Us

University of Iowa News Release

July 19, 2004

UI Professor Says Election Too Close To Call, 2000 Replay Possible

The November presidential election could bring an eerie replay of the 2000 election, with one party winning the popular vote and the other winning the presidency with an electoral college victory.

That's the prediction from the University of Iowa 2004 Presidential Election Forecast. The statistical forecasting model created by Michael Lewis-Beck, UI professor and chair of political science, and Charles Tien of Hunter College, shows President George W. Bush and Sen. John Kerry (D-Mass.) in a virtual dead heat. The formula forecasts a final two-party popular vote split of 51 percent for Bush and 49 percent for Kerry.

"That narrow difference makes the race too close to call," said Lewis-Beck, who has been forecasting elections in the U.S. and France since 1982. "Even with a tiny margin of error, say one percentage point or so, this race could go either way."

The forecasting model, known as the Jobs Model, is based on a statistical analysis of political and economic trends associated with presidential elections over the post-World War II period. The final data are based on measures taken or available during summer of the election year.

Lewis-Beck notes that when the popular vote split is so narrow the chances are much greater for an outcome such as the 2000 election, when Bush won the electoral college vote and Democratic candidate Al Gore won the popular vote.

Weak popularity ratings and a poor record of job creation are the two primary factors explaining why Bush is not further ahead of his Democratic opponent going into the summer conventions, Lewis-Beck said.

"In the July Gallup poll, Bush's presidential approval stands at a mediocre 47 percent, and with respect to jobs, his presidential administration has seen less growth than any other, looking at the period 1952-2004," he said. "These modest scores, when plugged into the statistical model, offset his potential vote gains from other factors, such the advantages of incumbency and current economic growth."

The Jobs Model has an expected prediction error of about one and one-half percentage points, when used to forecast the incumbent share of two-party popular vote. It is a slight modification of Lewis-Beck's and Tien's Growth Model, which they used in 2000. That model did correctly forecast that Gore would win the majority of the popular vote, but exaggerated his final vote share. The Jobs Model attempts to adjust for mistakes made in the 2000 forecast.

STORY SOURCE: University of Iowa News Services, 300 Plaza Centre One, Suite 371, Iowa City, Iowa 52242-2500.

MEDIA CONTACT: Mary Geraghty Kenyon, 319-384-0011,