Politics

DENVER (CBS) ― The general consensus among the media and political observers is that Barack Obama is favored to win the presidential election Nov. 4. A large part of that consensus is based on polling data. But can the polls be trusted? Could the race be closer than most assume?

An Associated Press-GfK poll out Wednesday showed John McCain and Obama essentially running even among likely voters in the election homestretch. The poll found Obama at 44 percent and McCain at 43 percent.

In sharp contrast, an ABC News/Washington Post poll out Thursday gave Obama an 11 point lead among likely voters (54 percent to 43 percent).

How can two polls released one day apart be so different?

"It depends on whether the poll is completely random," said Robert Hardaway, professor of election law at the University of Denver Sturm School of Law. "Even if you're trying to figure out what 100 million people's preference is, as few as 500 or 1,000 people polled can give a pretty accurate result if those 1,000 are chosen completely at random.

"Without complete randomness you're not going to get a completely accurate result."

Hardaway said the methodology of picking the sample is one of three factors that can affect the accuracy of a poll. For example, some polls might not accurately account for young voters who only have cell phones instead of landlines.

Another factor is the way questions in the poll are asked.

"We all know that how you ask the question can make all the difference in the world," Hardaway said.

The accuracy of polls becomes even trickier as Election Day draws near. News organizations rely on polling data, particularly exit polling, to help project the winners of the presidential vote in each state. If the exit polls are flawed in some way, as they were when John Kerry was projected to beat George W. Bush early on Election Day in 2004, then they could be unreliable.

"I was not surprised that the results of the exit polls did not match the actual results (in 2004)," Hardaway said.

National Election Pool (NEP) is the group the major national news organizations use for their exit poll data. NEP said it has changed its methodology to prevent similar problems in 2008 and get a more random sample of voters.

One challenge that is getting more attention this election is early voting. The ABC News/Washington Post poll found that 31 percent of likely voters plan to cast their ballots before Nov. 4.

"If you only do exit polling on the day of the election, you may only be polling a certain percentage of the electorate," Hardaway said. "So if you're not polling the early voting, then you're going to get a skewed result for that reason alone."

NEP said it plans to rely more heavily on phone polls conducted prior to Election Day to compensate for states with large early voter turnout.

Hardaway indicated there's at least one problem with basing media projections on exit polls that are inaccurate. The projections create an expectation, and if it turns out to be wrong, some might be inclined to believe there was wrongdoing.

"The fact that the exit polls don't agree with the actual results is probably the least important indicator of fraud," Hardaway said.

Hardaway made sure to point out that as a general rule, there should probably be less emphasis on polls by the media during an election.

"Most people know enough to take them with a grain a salt if for no other reason that they don't ever agree," Hardaway said.

Hardaway proposed an independent rating system for polls so people know how accurate polls are likely to be based on their history and methodology.

(© MMX, CBS Broadcasting Inc. All Rights Reserved.)