Until Kamala Harris emerged as the Democrats’ presumptive nominee for president, talk of an election “landslide” for Donald Trump — however unlikely — was slowly building. In fact, now polls are reporting she’s slightly ahead. What gives?
A check on such speculation is the notable absence of a rout in U.S. presidential elections since 1984, when Ronald Reagan defeated Walter Mondale in a 49-state landslide. Reagan won the popular vote by 18 percentage points, the last presidential race decided by more than 10 points, which is a demarcation of an election “landslide.”
Although presidential campaigns have been decidedly closer in the 40 years since the Reagan-Mondale race, it is nonetheless instructive to look back at the last landslide; doing so tells us something about the abiding difficulties of getting it right in pre-election surveys. Even popular vote landslides, rare as they are, are not necessarily slam-dunk easy for pollsters, particularly when a sizable number of survey respondents declare themselves “undecided.”
In the closing days of the 1984 presidential campaign, Mondale took to assailing the polls as “dead wrong,” asserting that he had “an excellent chance of winning” the election.
In saying so, Mondale embraced the trappings of the epic polling failure of 1948, when pollsters, the press and the punditry all figured President Harry Truman was certain to lose, overwhelmingly, to Republican Thomas E. Dewey. Pollster Elmo Roper even stopped releasing poll results that September, saying he would turn his attention to other matters, given that Dewey was certain would win by “a heavy margin.”
Nothing better captured the misplaced certainty of 1948 better than the front-page banner headline in an early edition of the Chicago Tribune on the day after the election. “Dewey defeats Truman,” the headline proclaimed; a day later, Truman posed with the Tribune front page, holding it aloft and grinning broadly, in what made for perhaps the most memorable photograph in American politics.
In Chicago at the end of October 1984, Mondale declared at a torchlight parade and rally that the polls were failing to identify the movement of voters to his candidacy. To emphasize his point, Mondale held up a reproduction of the Tribune’s “Dewey defeats Truman” front page.
“The pollsters and the slick magazines are trying to tell us that the election is over,” Mondale said, invoking themes reminiscent of Truman’s 1948 campaign. “Well, I’ve got a little message for them: Public opinion polls don’t count at all. People vote, public opinion polls don’t vote, and we’re going to win!”
Mondale likely was putting on a brave front. Richard Cohen, a columnist for the Washington Post, wrote as the campaign ended that the polls had “spoken so unambiguously only a fool could think they’re wrong. Mondale is no fool.”
However, the polls were misleading in 1984, albeit not quite in the way Mondale argued. They were directionally accurate in pointing unanimously to Reagan’s easy triumph. But nearly all of them were too generous, or too parsimonious, in estimating the dimensions of his victory.
The variations were not trivial. The final poll for USAToday, for example, said Reagan would win by 25 points; NBC’s final poll pegged Reagan’s lead at 24 points. The New York Times/CBS News poll reported Reagan was ahead by 22 points.
Among the polls that underestimated Reagan’s support were those of ABC News/Washington Post, which said Reagan’s lead was 14 points; the Harris survey, which estimated Reagan’s end-of-campaign advantage at 12 points; and the Roper Poll, which said Reagan was ahead by 10 points.
Only the Gallup Organization, which in 1984 was still conducting election polls by in-person interviews, pegged the popular vote with spot-on accuracy, estimating that Reagan would win 59 percent to 41 percent.
“We did a pretty good job in this election,” said Andrew Kohut, Gallup’s then-president, in something of an understatement.
Burns Roper, an irreverent figure in a field that attracts few of them, assigned Gallup an “A-plus” for accuracy in 1984 while grading his own eponymous poll a “D.”
“I don’t know what happened,” said Roper about his poll’s underestimate of Reagan’s support. “I wish to hell I did.”
Roper, who took over the Roper Poll after the death of his father Elmo in 1971, also said in a post-election interview with the Wall Street Journal: “I’m very concerned. This raises real questions of whether this [polling] business is anywhere near a science.”
Gordon S. Black, who conducted USA Today’s surveys, attributed the newspaper’s off-target final poll to undecided voters making up their minds late in the campaign.
Black’s explanation spoke to an enduring problem for pollsters: What to do about respondents who say they are undecided for whom to vote? Are they best apportioned among the leading candidates? Should they be pressed to declare a preference? As data journalist and polling analyst Nate Silver has pointed out, “More undecided voters means more uncertainty” in election outcomes.
The impact of late-deciding voters was demonstrated more recently in the 2016 presidential election. According to a report for the American Association for Public Opinion Research, undecided voters in Michigan, Pennsylvania and Wisconsin swung mostly to Trump during the campaign’s final days, giving him the votes for narrow victories in those states — and, in turn, sufficient electoral votes to win the presidency.
Without those states, Trump would have lost the election.
W. Joseph Campbell is a professor emeritus of communication at American University and the author of seven books, including most recently “Lost in a Gallup: Polling Failure in U.S. Presidential Elections.”