"Don't believe the polls," Bob Dole warned voters in the final weeks of the campaign as he vowed an upset victory.
In the end, though, people were better off not believing Dole. The polls, generally, were right.
Eight national surveys showed President Clinton running well ahead in the week before the election. On Tuesday, Clinton won 49 percent of the vote, Dole 41 percent and Reform Party candidate Ross Perot 8 percent. And pollsters' expectations were thus largely fulfilled.
"I think, by and large, [pollsters] should be relatively pleased with their performance," says pollster Andrew Kohut, director of the Pew Research Center for the People and the Press.
Election Day, though, was not a unanimous victory for those who make their living gauging public opinion. All eight presidential polls underestimated voter support for Dole, and most gave Clinton a considerably wider margin of victory than the 8 percentage points he had in the end.
While polls generally predicted the outcome of gubernatorial and Senate races, several in Nebraska and New Jersey were well off the mark. One of election night's most embarrassing moments came when exit polls wrongly gave the New Hampshire Senate race to Democratic challenger Dick Swett instead of the eventual winner, Republican incumbent Robert C. Smith.
TV networks and radio stations announced a victory for Swett, only to correct it later. Sometime past midnight, National Public Radio's Robert Siegel said that after much "nibbling," it was time for him to "eat crow" and acknowledge the goof in New Hampshire.
Tuesday's results showed that political polling has become a reliable indicator of public opinion but that it also remains imprecise and, at times, an uneasy combination of art and science.
Questionable CBS-Times poll
The discrepancy between polls and Election Day tallies highlighted the difficulties firms continue to have in measuring the changing landscape of popular sentiment. And, as the famous "DEWEY DEFEATS TRUMAN" headline shows, people have sharper memories of when you get it wrong than when you get it right.
Of the presidential polls, the one that fell furthest from the mark was the survey by CBS and the New York Times. The survey gave Clinton an 18-point margin over Dole, with the president taking 53 percent, Dole 35 percent and Perot 9 percent.
As is often said, polls are merely a snapshot of opinion during a fixed period of time. When the lens is closed, you run the risk of missing the full picture.
Kathleen Frankovic, director of surveys for CBS News, said she stopped polling on the Saturday before the election. She thinks Dole's continued attacks on Democrats for accepting questionable foreign campaign contributions may have driven more people toward the Republican candidate in the final days.
"There was probably some movement going on that we missed," she says.
As for the polls' tendency to underestimate Dole's support, Frankovic says it is a common phenomenon. Elections are generally referendums on incumbents. Voters who don't make up their minds until Election Day probably aren't very happy with the incumbent and are more likely to vote for the challenger, she says.
While the CBS poll overestimated the margin of victory in the presidential race, New Jersey pollster Janice Ballou found herself in a different situation election night. Two days after the Newark Star-Ledger published her poll showing a dead heat in the Senate race between Democratic Rep. Robert G. Torricelli and Republican Rep. Dick Zimmer, Torricelli won, 53 percent to 43 percent.
Since the election, Ballou, director of polling at the Eagleton Institute of Politics at Rutgers University, has been trying to figure out what happened. She thinks the problem lies in the peculiarities of the state's politics and the challenge of gauging voter turnout.
New Jersey shares major media markets with New York and Philadelphia, but does not have a state-wide newspaper or TV station of its own. Therefore, Ballou says, there is often less news coverage of political races and an unusually high percentage of undecided voters going into Election Day.
"You have this really large bloc of voters that moves around and makes life hard. I constantly refer to them as the 'mushy middle.' "
The last Eagleton Institute poll, completed Friday, found Torricelli with 42 percent, Zimmer with 41 percent and 15 percent undecided. Over the weekend, though, a reported plan to intimidate urban Democrats from going to the polls became a rallying cry to turn out the vote in black communities. And on Election Day, Democrats and their friends in the labor movement aggressively worked phone banks and knocked on doors.
More Democrats voted
The result: Exit polls showed that far more Democrats came out Tuesday than Republicans, giving Torricelli an easy win.
Ballou says that figuring who is going to vote on Election Day is the single biggest challenge for pollsters. Survey firms try to screen people by asking if they are registered to vote, if they plan to vote and when they last voted. But, human nature being what it is, more people say they expect to vote than ever do.
"We depend on the numbers we get," she says.
Like polling itself, the question of whether polls influence voters is a matter of opinion. Kohut, of the Pew Research Center, doesn't think people factor poll results into their decisions. "Show me someone who is thinking about the polls when he walks into the polling booth and I will show you either a journalist or a political scientist," Kohut says.
Frankovic, though, thinks that polls allow some voters to cast their ballots strategically. She recalls a jewelry maker in California who didn't want Dole to win the state. He said he felt more comfortable voting for Green Party candidate Ralph Nader because polls showed Clinton would win California anyway.
So in some cases, polling "allows people to vote the way they feel," Frankovic said.
But given all the variables and all the races, some polls inevitably turn out to be wrong. The most glaring example Tuesday was not a pre-election poll, but the exit poll in New Hampshire.
Voter News Service, which is supported financially by the five major networks and the Associated Press, sent 1,500 people to interview about 150,000 voters at randomly selected precincts across the country. The service then weighted the information according to the size of the precinct, the number of people who voted there that day as well as the number of respondents.
First mistake since '88
After analyzing the data, the service called more than 100 contests correctly. But in New Hampshire, the service gave the race to the Democratic challenger -- its first mistake since 1988.
"For some reason, Democrats in New Hampshire were more willing to talk to us than Republicans," says Murray Edelman, the service's editorial director. "I don't have a good answer why."
Getting it wrong, he says, was "an awful feeling." But when he thinks about the service's track record over the years, he is gratified.
"Every once in a while, I'm amazed at how accurate it is," he says.
Pub Date: 11/08/96