It will take months — maybe even years — to figure out why the polls were so wrong leading up to Election Day. But the autopsy is already well underway.
“The polls were a stinking pile of hot garbage," tweeted Sean Trende, the senior elections analyst for RealClearPolitics (RCP), expressing a view with which few pollsters would take issue.
"It is distressing," said Steve Koczela, president of The MassINC Polling Group, who has done dozens of political polls over the years, including many for WBUR. "It's a bummer to wake up after the 2020 election and see again that we're in a similar place as we were in 2016 as a profession."
Koczela pointed out that the failure in 2020 occurred even after pollsters thought they had learned the big lesson of 2016: they had underestimated support for President Trump by undercounting white voters without college degrees.
"We had the brightest lights in the polling industry shining on the states this time," Koczela said. "That was one of the things that 2016 didn't have. We all weighted by education. We all made our best attempt to solve all of the methodological issues that we identified after 2016."
And yet, the industry did even worse this year.
Four years ago, the national polls were actually pretty accurate. By the end of the 2016 campaign, Hilary Clinton led the RealClearPolitics average by three points; she ended up winning the popular vote by a little more than two points. This year, the RCP average had Biden ahead by more than seven points. Now, he's on track to win by just four points. But like 2016, the 2020 polls were even more off the mark in some key swing states.
"Certainly, we were underestimating — and I think most of the other pollsters were underestimating — Trump's support in the Midwest," said Spencer Kimball, director of polling at Emerson College, who spoke recently at a Zoom conference of pollsters from New England.
Consider the case of Wisconsin, where Emerson's final poll had Biden ahead of Trump by eight points; Reuters had him ahead by 10 points; and the New York Times-Sienna College poll, by 11 points — but Biden ended up carrying the state by just a fraction of a point.
"There may have been some Midwesterners who may have not wanted to say exactly who they were supporting, [which] created an undercount for Trump," Kimball said, referring to the theory of the "shy Trump voter" — the idea that lots of people didn't tell pollsters that they supported the president.
Andy Smith, director the University of New Hampshire Survey Center, has actually found evidence to support the theory.
"We [asked voters], do you agree or disagree [with the statement], 'I won't put a Trump sign in my yard because I'm worried my house might be vandalized,' " Smith said. "Sixty-eight percent of Trump supporters said that they agreed with that."
When Biden supporters were asked the same question, less than a third of them agreed with the statement, according to Smith.
It's easy to find anecdotal evidence supporting this idea. On Election Day, a Trump supporter named Claire from Littleton, Mass., held a Trump sign on the side of the road, and spoke reluctantly to a reporter, saying she was "afraid of the violent left."
"We put a [Trump] sign up this afternoon and someone stole it," she said. "There's not tolerance from the left. It's their opinion or no opinion."
That same day, David Niemela, another Trump supporter from Littleton, expressed deep distrust of the media.
"The mainstream media is totally against Trump," he said. "It's the truth."
Such entrenched and widespread attitudes represent new territory and new challenges for pollsters.
"Is the lack of trust in media and institutions, especially political institutions, going to persist past Trump being a candidate?"Jennifer Necci Dineen, director of the graduate program in survey research at the University of Connecticut
"There's suspicion of the media and there's this worry about the other side," said Dan Shea, a professor of political science and director of the Colby College poll.
"We've not seen such polarization in the electorate really since polling [began]," he said.
Shea tried to correct for that shy Trump voter, but he, along with several others pollsters, badly missed the mark on the much followed Maine Senate race. The pre-election Colby College poll had Democrat Sarah Gideon leading the Republican incumbent, Susan Collins, by four points. But Collins crushed Gideon by almost nine points. And she did it in a state that Joe Biden won easily, surprising pollsters like Shea that so many Mainers voted for Collins and for Biden.
"So our explanation is that in our zeal to find the shy Trump voter, we missed a whole bunch of voters [who split their tickets]," Shea said.
Koczela's biggest question about 2020 is: "Why did the polls miss in some places and not other places?"
For example, Koczela's polling of the hard-fought Massachusetts Senate primary between Ed Markey and Joe Kennedy III was pretty accurate, as was the presidential polling in neighboring New Hampshire. Koczela wants to know why, but says it will take months of studying the data to figure it out.
"And then it's years before we have another chance to see if we fixed the issue that we had," he said. "That makes it especially challenging — these [long] post-mortem periods — in figuring out what happened."
Another question facing pollsters: to what extent are the polling flaws of 2016 and 2020 related specifically to Trump?
"Is the lack of trust in media and institutions, especially political institutions, going to persist past Trump being a candidate?" asked Jennifer Necci Dineen, director of the graduate program in survey research at the University of Connecticut.
If it does persist, we might be dealing with the same kind of challenges with polling well into the future. And that might cause the public to give up on political polls altogether, if they haven’t already.
This segment aired on November 30, 2020.