Shock and surprise ran deep when Donald Trump won the presidency nearly four years ago in an upset that almost no pollster, pundit, or journalist anticipated.
The 2016 election was a case of polling failure in U.S. presidential elections — one of eight cases SOC Professor W. Joseph Campbell examines in his forthcoming book Lost in a Gallup, a narrative history to be published this summer by University of California Press. The book will be the focus of an online discussion August 4 at 5 p.m. in which American University experts address the polls, the media, the voters, and the presidential election.
Campbell — who has written six other, solo-authored books since joining the SOC faculty in 1997 — discusses how pre-election polls can go wrong in many ways. Just as no two elections are quite the same, he writes, no two polling failures are identical, he says. Lost in a Gallup notes that pollsters have anticipated tight elections when landslides occurred. They have signaled the wrong winner in closer elections. The work of well-known and venerable pollsters has been singularly and memorably in error. Exit polling has thrown Election Day into confusion. State polls have confounded expected national outcomes, which essentially was the story in 2016.
It wasn’t so much that national polls were off in 2016, Campbell notes. In aggregate, those polls anticipated Hillary Clinton’s popular vote victory. But polls notably failed in states decisive to the electoral vote outcome — Wisconsin, Pennsylvania, and Michigan, all of which Trump won narrowly. Had Trump lost those states, Clinton would have been elected.
Polling fiascoes such as the famous “Dewey defeats Truman” election of 1948 are rare, Campbell writes. Almost no one in 1948 gave President Harry Truman much chance of reelection. The polls signaled his certain defeat. Elmo Roper, one of the best-known pollsters of the day, declared in early September that the election was all but over, that Republican Thomas Dewey was as good as elected. Roper said he was so confident of that outcome that would release no further pre-election polls that year.
One of Roper’s contemporaries and rivals, George Gallup, also predicted Dewey’s victory and said that on Election Day, “the whole world will be able to see down to the last percentage point how good we are.”
Truman won the 1948 election by 4.5 percentage points in what Campbell calls an “epic polling failure.”
The cockiness that characterized polling in 1948 is a recurring affliction, Campbell writes. Misplaced cockiness emerged again in 2016. In one notable example, Samuel Wang, a poll-based forecaster at Princeton University, vowed to eat an insect on live television should Trump win more than 240 electoral votes. Trump received 304 electoral votes. Five days after the election, Wang went on a CNN program and brought with him a can of gourmet crickets. While on camera, he dug out a sample and ate it, redeeming his promise.
A key point of Campbell’s book is that polling failure is often journalistic failure.
Polls help fix media narratives about presidential elections. They are central to how journalists, and Americans at large, understand a campaign’s ebb and flow. Journalists, Campbell writes, invariably take their lead from polls in their reporting on presidential election campaigns. They are critical to shaping conventional wisdom about how competitive those races are.
So when polls fail or misfire, media narratives about the election often are in error as well. That was evident in 2016, Campbell notes, who argues that journalists conducted little detailed self-assessment as to why their analyses and expectations were so wrong.
“Why did journalists fail to alert the public to the prospect of a Trump victory?” writes Campbell, a former newspaper and wire service reporter. “The narrative-shaping influence of the polls and poll-based forecasts was partly to blame. But the explanations run deeper than that.
“Many journalists were afflicted by an ‘unthinkability bias,’ that there was no way such a repulsive and unsophisticated character could be elected. Trump’s winning the White House was simply unimaginable,” he writes. “Some journalists were quite open in saying so.” (“Unthinkability bias” was coined by Sean Trende, an elections analyst for the aggregation website RealClearPolitics.)
Campbell points out that poll-bashing, or severe criticism of polls and pollsters, used to be common among well-known American journalists. Some of them resented the presumption that polls could truly measure or interpret what the public was thinking.
Prominent poll-bashers among journalists included big-city newspaper columnists Mike Royko in Chicago and Jimmy Breslin in New York.
Royko once said a pollster was “a hired brain-picker trying to figure out what your personal fears, hopes or prejudices are, so that he can advise a politician how to more skillfully lie to you.” He encouraged people who were asked to participate in polls to lie in their responses, as a way to throw off polling results. It was a naïve view, Campbell points out, but indicative of Royko’s disdain of polling.
During the 2004 presidential campaign, Jimmy Breslin dismissed pre-election polls as “cheap, meaningless blatant lies” principally because they did not then reach the growing number of cell phone-only households (they do nowadays). “Anybody who believes these national political polls are giving you facts is a gullible fool,” Breslin wrote in his widely read column in Newsday.
Overt poll-bashing has mostly disappeared from American news media. An important factor has been the rise to prominence of data journalists and their statistics-based prediction models, Campbell argues.
The best-known of these data journalists is Nate Silver, whose election-forecast model, based on polling data, has made him a kind of celebrity. In the 2008 presidential election, Silver accurately predicted the outcomes in 49 states. In 2012, Silver accurately projected how all 50 states would vote.
He stumbled in 2016, however, estimating just a 10.5 percent chance of a split outcome — the popular vote differing from the electoral vote. But that’s what happened: Clinton won the popular vote, Trump won the electoral vote, and the presidency. Lost in a Gallup touches on the unfolding 2020 election, saying it’s not unreasonable to anticipate a polling surprise of some kind. Polls are not prophesies, at least not until late in the election campaign. They are not always in error but when they fail, they can fail in surprising ways, Campbell notes.
So what to expect in 2020 from the interplay of journalists, polls, and pollsters? Expect surprise, Campbell says, adding that whatever happens, whatever polling controversy arises, it may not be a rerun of 2016.
W. Joseph Campbell, PhD, is a professor in the SOC’s Communication Studies division. His other books include Getting It Wrong: Debunking the Greatest Myths in American Journalism and 1995: The Year the Future Began. His Twitter handle is @wjosephcampbell.