This excerpt from Lost in Gallup: Polling Failure in U.S. Presidential Elections revisits the many confident expressions about Hillary Rodham Clinton’s candidacy in the final days and hours before the 2016 general election. A close outcome seemed possible, but her defeat was implausible. Clinton won the popular vote by 2.1 percentage points, but Republican Donald Trump handily won the electoral vote, and the presidency.

The final days also brought reminders that late-campaign developments — the proverbial October Surprises — can profoundly disrupt poll-driven expectations. While October Surprises materialize unpredictably, they have occurred often enough to be almost-anticipated features of presidential campaigns. The October Surprise of 2016 took the form of an announcement by the FBI director, James Comey. It may have cost Clinton the election.

As in the presidential election of 1948, vague hints emerged here and there that the dominant narrative [that Hillary Rodham Clinton was sure to defeat Donald Trump] was flawed. Final estimates posted at the RealClearPolitics site, which aggregates or combines poll results, indicated that Clinton led by only 272 electoral votes to 266, suggesting that an improbable victory for Trump was within reach. CNN’s final Electoral College map showed that Clinton was one state shy of an electoral majority. A report posted at Politico the day before the election declared, “Donald Trump is still in the hunt.”

But such contrarian notices seemed tentative. Ambiguous. A lengthy list of pre-election national polls, summarized and posted at RealClearPolitics, showed Clinton ahead by an average of 3.3 percentage points. The final poll of Monmouth University in New Jersey and the final joint poll by NBC News and the SurveyMonkey online tool were the most generous in their estimates, saying Clinton led by six points. Of the nineteen polls that completed field- work in November and included in the RealClearPolitics lineup, just one had Trump ahead in the popular vote—that of Investor’s Business Daily.

Persistently, if at times narrowly, Clinton had maintained a clear overall advantage in the national polls since early August. Her aggregate lead ebbed and flowed, swelling to 7.1 percentage points on October 18, after her third face-to-face debate with Trump. The Republican narrowed the gap to 1.9 percentage points on November 2, days after James B. Comey, the then–FBI director, announced that federal investigators were reopening an inquiry closed months earlier related to the private, nonsecure electronic mail system installed at her home and used while she was secretary of state. Two days before the election, Comey announced that the renewed inquiry was over. By then, Clinton’s advantage had begun to grow again.

So there seemed ample reason to believe that Clinton would win. The consistency of the pre-election polls seemed persuasive and maybe even conservative. “If the polls are wrong they are more likely to be underestimating Clinton’s support than overstating it,” Matthew Yglesias, a political reporter for, wrote on the eve of the election. He said that a “polling error that results in an unexpected landslide is a lot more likely than a polling error that results in a Trump win.”

If anything, the prospects of Clinton’s victory brightened during the campaign’s final hours. Silver’s FiveThirtyEight data blog estimated that her chances of winning had improved in the two days before the election, from about 65 percent to 71.4 percent. FiveThirtyEight projected that Clinton would win 302 electoral votes and carry important swing states such as Florida, Michigan, North Carolina, Pennsylvania, and Wisconsin. Silver’s forecast was more guarded and restrained than the Huffington Post’s poll-based projection, which on election eve pegged the probability of Clinton’s victory at 98.2 percent. The Upshot data blog at the New York Times gave her about an 85 percent chance of victory.

The Princeton Election Consortium, directed by the neuroscientist Samuel Wang, estimated Clinton’s probability of winning at 99 percent. In mid-October 2016, Wang declared on Twitter that the race was “totally over” and vowed to eat an insect on live television were Trump to win more than 240 electoral votes, or 30 votes short of the threshold for victory.

None of the poll-based prediction models anticipated or advertised the prospect of a split decision, with Trump’s winning the Electoral College while losing the popular vote. Silver figured the chances of that outcome were slightly better than 10 percent: not impossible, but far from likely. Even Trump figured he would lose. At a rally a month later, he said the polls on Election Day were not very promising and that he reckoned, “If I lose, I lose, and I’m going to have a nice easy life.”

Clinton and her entourage clearly thought she was bound for victory. Jeff Zeleny, who covered Clinton’s campaign for CNN, recalled that in the hours before dawn on Election Day, the candidate’s “blue and white campaign plane landed in White Plains, New York, after a final, midnight rally in North Carolina. “The Clinton staff wasn’t just counting their chickens before a single vote was tabulated,” Zeleny wrote, “they were popping the champagne. Jon Bon Jovi was aboard, holding court with Bill and Hillary Clinton. A couple dozen Clinton friends and top aides were too. As reporters watched from the back of the plane, their words were out of earshot, but their celebration was clear as they raised their glasses to what they over-confidently thought was a job well done.”

To celebrate on Election Night, Clinton’s campaign rented the Jacob K. Javits Convention Center, a sprawling, glass-enclosed structure in the Hell’s Kitchen neighborhood of Manhattan. The Javits Center was chosen not for its architecture, but for its glass. Shattering the metaphoric glass ceiling of American politics was a defining objective of Clinton’s campaign. Plans were to shower the party at the Javits Center with confetti in the shape of glass shards when the moment of victory came.

To say Hillary Clinton was going to win the presidency was neither outlandish nor extraordinary on November 8, 2016. As the day wore on, she seemed to be carrying battleground states such as Florida, Iowa, Ohio, and Wisconsin — at least according to poll-based, turnout-derived estimates by VoteCastr, a Silicon Valley data startup. Its estimates were posted periodically online at Slate and Vice News. VoteCastr’s data boosted stock markets on the afternoon of Election Day, the Wall Street Journal reported.

Exit polls, which were not released to subscribing news organizations until 5 p.m., Eastern time, also contributed to the sense that Clinton was ahead in key states like Florida and Wisconsin, and winning overall. When they came out, the first wave of exit polls “didn’t flat out say Clinton was going to win,” recalled Chris Wallace, an anchor and political commentator for Fox News. But “if you read it,” he added, “you had to think Clinton was going to win.” Kristen Soltis Anderson, a Republican pollster and commentator, was more emphatic. The exit-poll data “really pointed to a difficult night for Republicans,” she said.

Reflecting such convictions, David Remnick, the editor of the New Yorker, drafted an essay about the first female American president, incorporating references to Elizabeth Cady Stanton, Frederick Douglass, and the suffragettes. It was, Remnick recalled, “written in a mood of ‘at long last’ and, yes, celebration. The idea was to press ‘post’ on that piece, along with many other pieces by my colleagues at The New Yorker, the instant Clinton’s victory was declared on TV.”

Remnick conceded he had not written anything in the event of Trump’s victory, adding, “I don’t think our site had anything, or much of anything, ready in case Trump won.” After all, he said, the final pre-election polls “provided sufficiently promising news for Democrats in states like Pennsylvania, Michigan, North Carolina, and even Florida that there was every reason to think about celebrating . . . the election of the first woman to the White House.”

Omens of Trump’s imminent defeat seemed everywhere and inescapable in the run-up to the election. Trump’s sneering image was plastered in black and white on the cover of the election issue of New York magazine. Across the center of the image the magazine emblazoned the single word: “LOSER.” The cover was to evoke comparisons to the Chicago Tribune’s “Dewey Defeats Truman” post-election banner headline in 1948. And such comparisons were not far-fetched. Among pundits and journalists, certainty about Clinton’s victory was on the order of 1948.

The polls and poll-based statistical forecasts had set an election narrative that the news media embraced and locked into place. The final polling estimates showed little to challenge the dominant narrative. The election might be close, but an upset? That seemed implausible.

Learn more about Lost in a Gallup