Exit polling goes the way of changing media trends

• Exit pollsters are making phone calls to reach people who voted early
• Because of increased costs, fewer states will have full exit polls this year
• Networks have been more cautious in using exit poll data for projections since 2000

TV networks are marshaling a multimillion-dollar exit poll for Election Day, but they head into Tuesday knowing they may not be able to declare a winner in the biggest race: the presidency.

Exit polls – surveys of people who have already voted -- have been narrowed this year to focus on battleground states, in part to offset extra costs given the increase in early voting in states such as Ohio and Florida, which reduces the number of voters who can be surveyed at polling places on Election Day.

Networks hope the changes will provide an accurate indication of how the race will ultimately turn out, even though the likely closeness of the presidential race, among others, means they'll have to wait for more of the actual vote count before they project winners on the air.

The number of voters pollsters will contact by phone in swing states will be increased, to reach early voters who won't be at the polls Tuesday. The exit poll will also include more people reached by cellphone, given an increase in households without a land line. The total number of people interviewed for the exit polls will increase to 25,000 from 18,000 in prior years. To save money – because telephone polling is more expensive than traditional in-person exit polling – the media consortium will not run a full exit poll in 19 solid blue or red states.

"What we did was just what every journalistic organization does: We focused our resources on where the story is. The story appears to be in those states that are competitive," says Sheldon Gawiser, NBC's director of elections and head of the National Election Pool, a consortium of ABC, CBS, NBC, CNN, Fox News and the Associated Press. "We're working very hard to be as accurate as we can."

"We've just shifted our approach in order to keep up with the shifting way people vote,'' said Sam Feist, CNN's Washington bureau chief.

When television networks project a winner in a state, they are making a calculation based on a combination of exit polls, vote counts from representative precincts, the number of votes outstanding and historical voting data. Since the debacle of 2000, when TV networks incorrectly called the Bush-Gore race, networks have been cautious about relying solely on exit poll data to make projections.

"History has bred caution," says Tom Rosenstiel, director of the Project for Excellence in Journalism. "No one in television news wants to have a story written the next day about how they messed up the exit poll reporting or intruded on the process. And the exit poll professionals have also seen their vulnerabilities from 2000 and other years and tried to correct for them."

Networks have overhauled their data and their procedures since 2000, says Kathy Frankovic, former head of the CBS "decision desk,'' which projects winners for the network. "All of them reviewed their own internal decision-making processes and understood that they might not have been taking account of and paying attention to everything they needed to pay attention to.''

Networks still like to project clear state winners the minute the polls there close. But Feist says it is "better to be right than to be first and happy to be last if we're right. We're in no rush. We will not be making projections in competitive states based on exit polls, nor have we ever."

Exit polls, traditionally conducted in person at polling places, are valuable because they not only reveal how people voted but why they voted the way they did: which issues influenced voters' choices and which demographic groups voted for which candidate. This polling is efficient, because most everyone leaving the polling place has cast a vote, while phone pollsters make many unsuccessful calls before finding an early voter. Exit poll respondents are never undecided. And they're more willing to be candid about such things as party affiliation while filling out an anonymous survey form than talking to a phone interviewer.

But with more than 30% of voters now skipping the polling place on Election Day, the phone is the only way to reach early voters, says Joe Lenski of Edison Research, which conducts exit polls for the media consortium. By reducing the states being polled, "you're going to have less information about Idaho and Delaware and South Dakota -- (but) you're going to have a lot more information about Florida and Ohio and North Carolina."

Michael McDonald, a political scientist and exit poll expert at George Mason University, says that pre-election telephone surveys of early voters introduce a new uncertainty into exit polls.There can be differences in how people answer the two kinds of polls, he says. For one thing, some people lie, saying they voted when they haven't – not an issue when the exit pollster is standing outside the voting place.

For another, voters are much less likely to disclose their income level over the phone -- and voters who are shy about revealing their incomes tend to vote differently. In 2008, Obama beat Sen. John McCain by almost 9 points among those who provided their income but tied among those who declined to say.

Exit pollsters say they're aware of the problems and think they have a good handle on how to deal with them. "No, it's not easy," Lenski says, especially because states' voting procedures are different. For example, Oregon and Washington conduct elections entirely by mail, so exit polls are done solely through phone surveys.

The end of state exit polls in 19 states also will be a loss for political researchers, McDonald says. "It's unfortunate, because we used to have a very nice long-term historical profile of the electorate in the states based on the exit polls,'' he says "If we wanted to make comparisons … we're going to lose the capability of looking back at those elections.'' (The states losing exit polls are Alaska, Arkansas, Delaware, Georgia, Hawaii, Idaho, Kentucky, Louisiana, Nebraska, North Dakota, Oklahoma, Rhode Island, South Carolina, South Dakota, Tennessee, Texas, Utah, West Virginia and Wyoming.)

The networks that pay for the polling say that's less important than focusing on the competitive states. "I've never seen much conversation about the breakdown of the electorate in non-competitive states," Feist says.

The race for president appears so close, and the potential for prolonged vote counting in key states like Ohio so great, that even an all-nighter may not reveal who the next president will be. "There's a decent chance that we will not project the winner in at least one of the presidential battleground states,'' Feist says. "Election night is always exciting. … Election days that begin with the prospect of not actually knowing the outcome at the end of the day are even better."