MADELEINE BRAND, host:
And we'll have more on these undecided voters in a moment with our political analyst, Juan Williams. First, though, the most hated of polls, exit polls. Those are the people with clipboards they're waiting for you after you vote, and they're supposed to help show who voted and why, while, as Day to Day's Alex Cohen reports, it's not for nothing that exit polls have a notorious reputation.
ALEX COHEN: Remember Election Day 2004 when early exit-poll numbers giving John Kerry the lead were leaked to the press? The effects were immediate. The Kerry campaign started celebrating. The markets on Wall Street dropped sharply. But as we all know now, those exit-poll numbers didn't match the election returns. That's why this year Edison Media Research and Mitofsky International, the groups conducting exit polls nationwide, are determined to stop any such leaks.
Dr. MICHAEL MCDONALD (International Affairs, George Mason University): They're going to have us under armed guard, and we're going to have our cell phones confiscated for about seven hours.
COHEN: Michael McDonald, an associate professor at George Mason University, is one of a select team of researchers and pollsters who will be the first to see the exit-poll results under quarantine.
Dr. MCDONALD: We don't want people to get this information before we've had a chance to put a fine comb through it.
COHEN: Early leaks aren't the only thing pollsters are watching out for. Andrew Kohut heads the Pew Research Center.
Mr. ANDREW KOHUT (President, Pew Research Center): Typically, exit polls go to the precincts and they hand out questionnaires to the voters as they leave, but in this election, the exit-polling organization and Pew and others are figuring that almost 30 percent of the electorate will vote early.
COHEN: There is a plan to deal with that, says Joe Lenski, executive vice president of Edison Media Research.
Mr. JOE LENSKI (Cofounder and Executive Vice President, Edison Media Research): We will be doing telephone surveys the week before the election in all the states that have significant numbers of early voting.
COHEN: But polling by phone has its own set of challenges, says Temple University's Michael Hagen.
Dr. MICHAEL G. HAGEN (American Politics, Temple University): Probably because of the use of cell phones, partly because cooperation rates are falling, and it just becomes more difficult to do interviews with random sample or representative sample of Americans.
COHEN: Even when done in person, polling a group of people that truly represents the entire voting population is tricky. You want to get the right mix of race, gender and age. On Tuesday, 1300 interviewers will be trying to do just that at polling places throughout the country. But bad weather and power outages can cause long delays in reporting polling results, and there's always the problem of human error, says George Mason University's Michael McDonald.
Dr. MCDONALD: It's not uncommon to see a precinct reporting results that are flipped, say, the Obama columns in the McCain column and the McCain column is in the Obama column. Those sorts of errors can throw off the modeling by quite a bit.
COHEN: McDonald says they can usually spot such errors by comparing results to previous election returns in a given precinct. But this year may be a bit harder to analyze if, as projected, an unusually large number of voters show up and a significant percentage of them are first-time voters. Temple University's Michael Hagen.
Dr. HAGEN: It's a different kind of election. There are lots of new people registered to vote across the country. I know here in Pennsylvania, there are 1.2 million more people registered now than were four years ago.
COHEN: And keep in mind, not everyone who wants to vote wants to fill out a questionnaire once they're done. Exit pollster Joe Lenski.
Mr. LENSKI: About half the people we approach decline to fill out the questionnaire for one reason or another. They don't have time or they're not interested. Sometimes polling goes late and it gets dark.
COHEN: But Lenski says exit pollsters have an advantage over those doing surveys by phone, in that they can see who's saying no to them.
Mr. LENSKI: We know if they're a man or a woman. We know if they're younger or older. We know if they're black or white. So, we're able to adjust our results to account for non-response by demographic.
COHEN: But Lenski admits there's one thing about a person you can't tell by looking, how they voted. Alex Cohen, NPR News.