Why polling failure is often journalistic failure

It’s axiomatic that pre-election polls set the narrative for U.S. presidential campaigns. As we’ve seen this year, polls are central to shaping conventional wisdom about the competitiveness of the races. But when polls misfire, journalism can falter, too, leaving the news-consuming public feeling misled, ill-served, and even a bit contemptuous.

That polling failure can give rise to journalistic error has been evident for more than 70 years, at least since the “Dewey defeats Truman” election of 1948. That was when poll-based predictions gave President Harry Truman no chance of reelection, when Republican Thomas E. Dewey supposedly needed only to count the days to his presidential inauguration. 

Elmo Roper, a leading pollster of the day, stopped releasing survey results in early September 1948, saying Dewey was as good as elected and that he had better things to do. George Gallup declared without reservation that Dewey, whom he admired, would win “with a substantial majority of electoral votes.” Rival pollster Archibald Crossley likewise affirmed that Dewey “was assured of election.”

The polls failed epically in 1948. So, too, did the news media. Journalists took their lead from pollsters and were extraordinarily confident about Dewey’s prospects — so confident that Drew Pearson, one of Washington’s most aggressive syndicated writers, discussed the probable makeup of Dewey’s administration in a column sent to subscribers late on Election Day and tagged for publication the following day. “The men around Dewey who will take over the White House,” Pearson wrote, “… are an exciting, hard-working, close-knit clique who function with almost too much perfection… .” Pearson’s column appeared in the Washington Post and many other newspapers, as planned, on the day after the election — which Dewey lost in one of the most stunning outcomes in American political history.

The surprise of 1948 was profound, even more jarring perhaps than Donald Trump’s improbable win in 2016. Truman defeated Dewey by 4.5 percentage points and, in the election’s aftermath, some journalists conceded they had relied too enthusiastically on the pre-election polls. The press, declared Time magazine, “was guilty of laziness and wishful thinking; it had failed to do its own doorbell-ringing and bush-beating; it had delegated its journalist’s job to the pollsters.”

The “fatal flaw” for journalists and pundits, Marquis Childs, another prominent Washington columnist wrote in the aftermath of Truman’s victory, “was reliance on the public opinion polls. No amount of rationalization ever can explain away this mistake by Gallup, Roper & Co.”

Such assessments went beyond mere scapegoating. They spoke to a wider and enduring allure of polling data, which can exert a powerful attraction. Richard Morin, a former polling director for the Washington Post, once noted that “there’s something addictive about polls and poll numbers.” For journalists, this appeal is easily understood. Polling data offer a sense of assuredness and authority for a profession that deals unendingly with ambiguity, where the emphasis is on swift reporting that often comes off as provisional, even half-baked. Polls and poll-based forecast models can offer a promise of clarity and precision.

It’s small wonder that major news organizations have been conducting or commissioning polls since the mid-1970s and 1980s. Polling’s pedigree in the news business is even more extensive than that: historians have found newspaper reports about proto-straw polls in Pennsylvania and North Carolina as long ago as 1824, in the run-up to the bitter, four-way contest that John Quincy Adams won in the U.S. House of Representatives.

More recent U.S. election history is dotted with examples of polling error leading to journalistic failure — when polls set what proved to be a distorted narrative and news organizations failed to provide a reasonably clear sense about what was in store. The pre-election polls of 1980 almost unanimously signaled an eyelash-close race between President Jimmy Carter and Republican Ronald Reagan. The final Gallup poll gave Reagan a single-point lead, for example. So did the CBS News/New York Times poll. 

Reagan won by near-landslide proportions, defeating Carter by almost 10 percentage points while rolling up 489 electoral votes to the president’s 49. Recriminations and second-guessing came swiftly. Political columnist Michael Barone pointed out that “polls had not prepared most of us for what happened.” And the Washington Post’s ombudsman, Bill Green, wondered whether “the experience of 1980 will lead the press to turn its back on polls.”

That didn’t happen, of course. Election polling, if anything, has become more sophisticated and complex since 1980. Although the field of survey research is buffeted by declining response rates and characterized by experimentation with online methodologies and other techniques once considered far-fetched, polling and poll-based forecasts are perhaps more influential than ever in campaign reporting.

For confirmation, we need to look no further than 2016, when polls misfired in key states such as Michigan, Wisconsin and Pennsylvania. Narrow wins there propelled Trump to an unexpected Electoral College victory, despite his losing the national popular vote.

State polls were important to the forecast models of, among others, HuffPost, which declared just before the election that Clinton’s probability of victory was 98.2 percent, that she would “fairly easily hold onto Michigan, Wisconsin and Pennsylvania,” and that Trump had “essentially no path to an Electoral College victory.” 

In a brave and revealing mea culpa two days after the election, HuffPost’s polling director, Natalie Jackson, wrote: “The problem was that I placed way too much faith in polls. I assumed they would be right. … I kept looking at the consistency of the polls. They wavered in the exact margins, sure, but always showed Clinton winning in the key states that she needed to win. I saw no reason to question that the polls would be accurate overall. So I defended and stood by the numbers — as anyone who trusts their work does. That’s left me eating some crow.” 

The cases discussed here suggest that journalists — and the public — would do well to treat polls warily, especially polls taken well before the election that suggest a lopsided outcome lies ahead. Polls are not to be dismissed or scorned. But it is prudent to realize that they have had a checkered past, that even the best of them have inherent limitations, and that leaning on them too hard can produce mortifying reminders of journalistic fallibility. “Cover voters, not polls,” was advice gathered years ago by the now-defunct Committee of Concerned Journalists. Not covering polls is impracticable and doubtless unwise. But the sentiment is understandable.

W. Joseph Campbell is professor of communication at American University and the author of seven books, including, “Lost in a Gallup: Polling Failure in U.S. Presidential Elections.” Follow him on Twitter @wjosephcampbell.