A Look At Why Florida Presidential Polls Were Wrong
If the pre-election polls had been right, Joe Biden would have won Florida.
In the days ahead of the election, many of the nation's political survey organizations predicted Biden would narrowly win Florida in this week’s presidential election. In a poll published one day before election night, Quinnipiac University, a private school in Connecticut, projected Biden would win Florida by 5 percentage points and win Ohio by 4 points.
In fact, Trump won Florida by 3.4 percent of the vote and Ohio by 8.2 percent of the vote, according to unofficial returns.
The 2020 election – historic for its high turnout in the middle of a pandemic and extraordinary political divisiveness – might also be famous for the inaccuracy of prognosticators.
In a state like Florida – famous for narrow margins and challenging polling – the failures spurred President Donald Trump to spin a fanciful conspiracy theory, without evidence, that an entire polling industry of scientists, academics and others deliberately published false numbers to discourage voters from voting.
“The pollsters got it knowingly wrong,” Trump said Thursday at the White House, without presenting any proof of his theory. “I have to call them phoney polls, fake polls, designed to keep our voters at home.”
Preliminary estimates show turnout in 2020 might be higher than in any election in more than a century. Republicans in Florida, especially, turned out in high numbers in early in-person voting and on Election Day.
Trump’s criticism harkened back to 2016, when pollsters faced backlash for miscalculating his support among voters.
Inaccurate predictions are a problem nationwide, but especially in Florida. Quinnipiac, which calls itself the “the gold standard in national and state-wide polling,” predicted a victory for every major Democratic candidate in the state since 2016: Hillary Clinton, Andrew Gillum, Bill Nelson and Biden. All of them lost.
Quinnipiac’s poll in early October showed Biden ahead in Florida by 11 percentage points, which even Democrats here said was implausible.
“If you’re that bad at your job, maybe some of these prognosticators should find another line of work,” Republican Gov. Ron DeSantis said Wednesday.
The polling industry’s credibility is on the line again. Justin Sayfie, a lobbyist with Ballard Partners in Washington and publisher of a political newsletter in Florida, vowed to stop citing political polls.
“Quinnipiac has no credibility, and no respected news organization should ever report on a Quinnipiac poll again,” Sayfie said in an interview. “Clearly, they don’t know how to poll Florida.”
Quinnipiac’s polling director, Doug Schwartz, said he doesn’t yet know what went wrong. He said it will take months for the polling industry’s professional standards organization to study the 2020 results.
“After the 2016 election, it took six months for the American Association of Public Opinion Research to release their findings about polling errors,” he said. “I would expect a full evaluation of 2020 to take at least as long.”
Political scientists and pollsters cautioned against making hasty judgments.
“Do not bring out the torches and pitchforks for the pollsters,” said Casey Klofstad, a political scientist from the University of Miami. “This is not a crystal ball.”
Polling is a scientifically driven estimation, he said, and the expectation that it can predict exactly what the electorate is going to do is unrealistic and unfair. There will always be some degree of uncertainty to their findings, expressed in polls as a margin of error. Trump’s victories in Florida and Ohio fell outside some polls’ margins of error.
“We cannot divine a person's true and actual intention to vote,” Klofstad said. “We can estimate it, but we can't know it.”
Pollsters can’t reasonably survey all voters, so they survey a sample of people intended to accurately represent the electorate. The larger the sample, the more accurate the data – but this takes time and money that few polling firms have. Some polls survey as few as 400 or as many as thousands of likely voters, almost always by phone.
Another challenge for accurate polls: People aren’t answering their phones, said Hans Hassell, director of the Institute of Politics at Florida State University. He said the rate of people who agree to be surveyed by phone has dwindled to fewer than 1-in-10.
“People have caller ID on their phone, so they’re not picking up if they don’t recognize the number,” he said. “That’s a problem.”
When a poll’s respondents include too many Democrats or too many men or too few minorities, analysts adjust the results by weighting responses from under-represented groups. Hassell said another challenge is deciding how to weight responses and by how much.
With voter turnout fluctuating drastically in recent years, it’s also difficult to know who is likely to cast a ballot.
“Without them, you’re not actually getting the entire electorate’s opinion,” said David Burrell, CEO and co-founder of Wick, a market research company that predicted Trump would win Florida by 3 percentage points. “You’re just getting a representation of who’s-willing-to-take-polls’ opinion.”
Figuring out additional variables to control helps, Burrell said, but it didn’t result in accurate polling this year. He predicted there are still more challenges to help identify the uniqueness of Trump’s supporters.
“Pollsters are going to need to have a more agile approach to their polling,” he said. “Break it up into smaller data sets, and look at the data more often. I think we'll develop new and better methodologies, and we’ll have more accurate polling in the future.”
Burrell said between 2016 and 2020, there wasn’t enough incentive for polling firms to rethink their operations. A second terrible election cycle is the push they need.
“We’d seen pollsters get it wrong before, but everyone gets a mulligan,” he said. “But two times in a row? Two cycles in a row? You have to adjust.”
Some voters were reluctant to admit in polls they supported Trump, said Mario Canseco, president of Research Co., a public opinion research company. He called them “the shy Trump voter,” and said this isn’t unique to the United States. Conservatives polled poorly in the United Kingdom during the same election cycle that saw Boris Johnson elected prime minister.
“We had a little bit of this in Canada as well,” Canseco said. “People answer the phone, but they don’t want to provide an answer that might make people feel uncomfortable.”
Canseco said he doesn’t consider the 2016 or 2020 election cycles failures – no matter what DeSantis says.
“We're featured in the media, we have to be able to face that criticism,” he said. “That being said, when you get it right, nobody remembers.”
This story was produced by Fresh Take Florida, a news service of the University of Florida College of Journalism and Communications. The reporter can be reached at firstname.lastname@example.org