After the recount uproar in the 2000 election in Florida, several sources concluded that only 537 votes decided the election. With such small margins, no amount of polling can predict such a close election. Now President Obama and Governor Romney are neck-in-neck, as shown by the polls.
A Clarus Research Group poll Oct. 2 put Obama at 49 percent and Romney at 45 percent, a 4 percent difference. ABC and the Washington Post put Obama ahead Oct. 28-31 and Romney ahead Oct. 29–Nov. 1, both polls at 49 to 48 percent.
The marginal difference between the two candidates has remained between 0 and 7 percent during the month of October, sometimes favoring Obama and sometimes Romney, according to pollingreport.com.
“Polls are obviously a big part of the horse race,” said Dr. Greg Vonnahme, assistant professor in the Political Science Department. “They provide the bulk of media coverage of the election. News coverage of the ‘horse race’ is more abundant than any other aspect of the election.”
Vonnahme said he views most polls as equally trustworthy and untrustworthy, and this close to the election, he said they’re good at predicting the national popular vote and the state votes. He is, however, skeptical of polls with results that diverge too much from the others.
“The Gallup tracking poll, for example, is normally a pretty reliable poll but showed Romney having a huge advantage over Obama, which showed up in no other poll,” he said.
Gallup polls make up the vast majority of wide margins on pollingreport.com, with only one other October poll rising above a 5 percent margin. All of Gallup’s margins in October favor Romney, except two ties.
While Gallup seems to be the outlying poll, all public opinion polls are limited. Vonnahme listed the three biggest problems as: non-response, snapshot and turnout.
Most polls are completed via phone interviews, so non-response is an issue. Voters hang up, refuse to answer questions and/or only use a cell phone (which is not utilized in all polling).
Snapshot refers to the moment in time captured by a poll. Throughout the election process, attitudes are fluid. Polls can only reflect the moment in time a single voter was asked a series of questions; they cannot account for that voter’s next-day change of heart.
Turnout is also a factor. Polls can tell us which way people say they will vote, but they cannot tell us if those people will show up to cast their ballots.
Vonnahme said accuracy can depend on the length of the poll and whether the poll involves an automated system.
“People get fatigued with polls very quickly, and the questions that are asked first can affect responses given later,” he said. “Some think people are more candid and responsive talking to another person, others think the automated system is more neutral and reliable.”
Overall, Vonnahme said he is paying attention to the state polls rather than the national polls.
“There is a realistic possibility that the Electoral College vote and national popular vote could diverge,” he said. “If that happens, the Electoral College vote is binding and whoever wins that wins the presidency.”