2020’s polling miss? These Republicans say they saw it coming.
A misread of who is likely to vote. An overreliance on party registration. Failing to reach voters where they’re most likely to respond — via text and email — in favor of a phone call.
These are just some of the explanations for the great polling miss of the 2020 election, according to private political pollsters, who were still in their preliminary review of data of a race that still hangs in the balance Wednesday.
While Democrats are wringing their hands over how their numbers were off the mark, many Republicans claim that for the most part, they forecasted President Donald Trump in a much closer contest than the media and public believed.
“We had it a lot closer than the public polls. It showed pretty much what’s occurring right now,” said Brent Buchanan, a Republican pollster who helped conduct 62 surveys for the Trump campaign over the final six weeks before Election Day. “When you have good data, you shouldn’t be surprised.”
This year’s miss comes amid a turbulent period for the polling industry, which already saw trust in its surveys shaken after misjudging Trump’s support in 2016. Those errors appear to have been only magnified four years later, with the president significantly overperforming his pre-election numbers, even as he still trails Joe Biden nationally and in several key battleground states.
While many public pollsters rely on pushing live calls to Americans they think are likely to vote, Buchanan’s firm, Cygnal, hits their sample pool on multiple channels of communication — calls, texts, emails — in order to broaden their sample and reduce the growing non-response rate on phone calls.
“Live calls used to be the gold standard. Now I would call it the rust standard,” Buchanan said. “You get a call from a 902 area code, you’re like, ‘What’s that?’”
Trevor Smith, the research director for GOP-led WPA Intelligence, said his firm’s polling for a private client showed Trump down just 4 points in Michigan, 2 points in Wisconsin and up 2 points in Florida in early September. Given the margin of error, it meant the president was always within striking distance of a 2016 repeat.
“We get fired if we’re wrong. Our reputation gets destroyed and you won’t get hired unless you’re lucky,” Smith said. “I really think they need to do some soul-searching, the public people.”
Public polling averages found Biden holding 8-point advantages in both Michigan and Wisconsin and a 2.5 point lead in Florida. The final margins in Michigan and Wisconsin are likely to each be less than a single percentage point, while Trump carried Florida by more than 3 points. And in Wisconsin, the public pollster who came closest to the final result was the Trafalgar Group, a Republican-aligned outfit that was ridiculed by Democrats and professional forecasters alike.
Methodological concerns
One major problem Smith pinpointed was the neglect of the public voter file, which is a state-by-state listing of registered voters. Some public pollsters rely on calling adults and screening their sample to likely voters based on how respondents say they’ll ultimately act, rather than their actual history of voting. It’s much easier to tell a stranger you’ll vote than to actually follow through.
“I don’t know what I’m going to do next weekend, let alone next year. I feel like that’s true with most people,” Smith said. “People don’t know their behavior. A lot of people don’t even know if they’re registered to vote.”
Targeting people in a voter file will reveal who is registered and voted before, rather than simply making an assumption about it. The same goes for party registration.
Smith said simply identifying someone as a Democrat or Republican is less important than their previous actions during elections. In a random survey, many casual respondents will declare themselves to be a Democrat, but then vote as a Republican due to a personal fondness for a candidate or a recent news development.
Then there’s the issue of excluding certain groups that end up participating in greater numbers than anticipated. This helps explains missing a burgeoning Hispanic vote for Trump in South Florida and a rural boost in outstate Michigan and Wisconsin
At the heart of the industry’s problem is an upending of the methods top pollsters used to survey voters, using traditional landline telephones to call people at home. In the era of cell phones, voters are less likely to respond to potential questions, making it harder to complete surveys and introducing the possibility that those who do respond are not representative of their broader demographic.
Those problems become more acute for public pollsters, who often have fewer resources to correct for potential problems.
But in this election, Democrats say the problems affected their own internal surveys, the normally prized private data that inform big-spending party committees which races should receive the most attention and resources.
“The mood with a lot of people I’m talking to is despair,” said one Democratic strategist, granted anonymity to speak candidly. The strategist said Democrats’ errors included a broad range of individual surveys, tracking polls and the sophisticated analytics models.
“A lot of us feel very unsettled about the information we’re getting,” said the strategist, who predicted Democrats might be more reliant on gut feelings and simply listening to individual voters to develop opinions about the political climate in the future.
The return of the ‘shy Trump voter?’
Some Democratic pollsters urged caution about reaching definitive conclusions over the severity of the polling error before the final vote tallies arrived, saying that a late increase in Biden votes could reduce the discrepancy. Many of them also said they would need days and weeks ahead to truly understand how and why their polls were off.
Other pollsters still defend their methods as the best way to learn what voters think, but suggest they might have to revisit old assumptions about who does, or doesn’t, talk to pollsters.
At the top of the list, the previously dismissed notion among professional pollsters that so-called “shy Trump” voters refuse to answer surveys because they’re nervous about admitting their support for the president.
Glen Bolger, a Republican pollster who doesn’t think those voters skewed polls in 2016, now says he plans to investigate whether they were an issue in 2020. His firm, Public Opinion Strategies, conducted a post-election poll that found 19% of Trump voters said they kept their support for the president quiet, compared to just 8% of Biden voters.
All of the discussion in the media about Trump voters not wanting to share their opinion about the president, the pollster suggested, might have been a “self-fulfilling myth.”
“In other words, Trump voters heard they weren’t supposed to tell pollsters who they were supporting, and were told that if you live in a suburb, don’t tell neighbors who you voted for or you’ll be ostracized,” Bolger said.
Bolger said he was concerned that in the near future, other voter blocs could adopt the same mindset, potentially making polls even less reliable.
“Does this go away after Trump?” he said. “I don’t know.”
Still, there’s evidence that the internal numbers for both presidential campaigns were off. Trump was dispatched to campaign in Iowa on Sunday, even though he easily won the state by 8 points. Meanwhile, Biden chose to stop in Ohio on Monday, before going on to lose the state by 8 points.
Campaigns don’t allocate precious time that late in the game to states they don’t think are winnable, but Biden campaign manager Jen O’Malley Dillon had warned just weeks ago that the race was a margin-of-error fight to the finish.
“We had been abundantly clear that we thought that this could be a very close race in most of our states,” she said on Wednesday during a press call. “We worked hard to be very transparent about the fact that we did not believe that the race was in that wide margin that we were seeing [in] the national polling numbers.”
This story was originally published November 4, 2020 at 6:13 PM.
