Despite widespread reports to the contrary, most polls in the days before the election were actually right in finding that Hillary Clinton had an edge in the nationwide popular vote.
Most found a larger edge for Clinton than showed up in the results. Some state polls were off. And data analysts who interpreted polls to forecast an Electoral College win for Hillary Clinton clearly were wrong.
But most national polls leading into Election Day showed Clinton with a slight edge or a small lead in popular support. And while she lost the Electoral College and the presidency to Donald Trump, she does in fact have an edge in the nationwide popular vote.
As of Wednesday afternoon, she was ahead by about 200,000 votes out of 120 million cast.
The final McClatchy-Marist poll last weekend was closest to the popular vote total so far. It found Clinton with a 1-point edge. As of Wednesday, she led the real popular vote by 0.2 percentage point. That’s a difference of 0.8 percentage point.
Other national surveys did show a larger national Clinton lead. The Real Clear Politics average of public polls found a Clinton lead in the popular vote of 3.3 points. The final pre-election Bloomberg poll found her with a 3-point edge, or 2.8 points off of the popular vote total Wednesday.
Meanwhile, the final USC/L.A. Times Daybreak tracking poll, which found Trump in the lead more often than most of the competition, found him ahead by 3 percentage points on Tuesday. That was 3.2 points off from the popular vote as of Wednesday.
State polls were more varied.
The final pre-election average of public polls in the battleground states of Florida showed Trump with an edge of 0.2 percentage point. In North Carolina, Trump had an edge of 1 point. And he won the popular vote in both states.
The final average in Pennsylvania, though, found Clinton ahead by 1.9 points. She lost the state’s popular vote by 1.1 point. The final average in Wisconsin found her up by 6.5 points. She lost the popular vote there by 1 point.
“We saw in this election an avalanche of new polling methodologies with unproven track records . . . online polls, panel polls, internet polls,” said Lee Miringoff, the director of the Marist Institute for Public Opinion. “They can be quicker and cost less, but they’re unproven and they’re not successful.”
And it was the interpretation of the polls by some outlets that likely produced the real disparity by predicting the outcome of the Electoral College, and a Clinton win there.
“Mea culpa, mea culpa, mea maxima culpa,” said Larry Sabato, whose Crystal Ball website had forecast a Clinton win with 322 Electoral College votes. :We heard for months from many of you, saying that we were underestimating the size of a potential hidden Trump vote and his ability to win. We didn’t believe it, and we were wrong.”
“Some polls got it right; some polls didn’t,” said Republican pollster David Winston.
There’s a problem in the aggregators who just pile all the polls into some stew where all polls are treated equally. If the recipe is off, it’s not going to come out good.
Lee Miringoff, director of the Marist Institute of Public Opinion
He said that was because they all screened their voters differently, which made them hard to compare and get an accurate read on the race at any moment. They vary over age, party, ideology, education, gender and race.
“What happen is it turns into mush,” Winston said
Marc Farinella, a Democratic strategist who managed President Barack Obama’s successful campaign in North Carolina in 2008, said pollsters needed an accurate prediction of who would vote.
“I think many pollsters had the model wrong,” he said.
Chris Borick, director of the Muhlenberg College poll in Pennsylvania, said his model found Clinton ahead by 4 points and Sen. Pat Toomey R-Pa., ahead in his re-election contest by 1 point. Toomey did win by 1 point, and Clinton lost Pennsylvania by 1 point.
“It’s a weird performance year,” Borick said. “We didn’t anticipate the excessively high turnout in our model that propelled Trump.”
All along, much was made about how the GOP nominee was relying on white male voters, even as the electorate was diversifying, and on voters without college degrees.
“There was such an intensity voting in rural areas and some suburban areas, more than anyone expected,” said Bernie Porn, president of EPIC-MRA, a Michigan polling firm. “He was appealing to folks who were working-class white voters and getting a strong turnout, while Hillary Clinton was not getting the kind of turnout she needed from African-Americans.”
Trump won Michigan, a state Clinton had counted on, by 13,000 votes out of more than 4.5 million cast.
Fivethirtyeight.com threw some caution flags about its forecast of Clinton’s 71 percent chance of victory: A polling error could affect her lead, the level of undecided voters was especially high and her coalition was not tailored to provide a strong boost in the swing states.
Still, Miringoff said aggregating polls from here and there and developing a prediction could be fraught with error.
“There’s a problem in the aggregators who just pile all the polls into some stew where all polls are treated equally,” he said. “If the recipe is off, it’s not going to come out good.”