How wrong were Florida polls on Trump and Biden?

“This was a terrible night for the polling industry,” said one pollster.
A man carries a sign supporting President Donald Trump, outside a polling station that sits across the street from Trump International Golf Club, in West Palm Beach, Fla., Friday, Oct. 30, 2020.
A man carries a sign supporting President Donald Trump, outside a polling station that sits across the street from Trump International Golf Club, in West Palm Beach, Fla., Friday, Oct. 30, 2020. [ REBECCA BLACKWELL | AP ]
Published Nov. 6, 2020|Updated Nov. 6, 2020

In Florida, some pollsters are kicking themselves.

When the results of the state’s presidential election came Tuesday night, Donald Trump had beaten not just Joe Biden, not just his own 2016 performance, but also his standing in every major poll taken before Tuesday’s contest.

Asked how the state’s polling fared across the board, the head of the University of North Florida’s Public Opinion Research Lab did not equivocate.

“This comes from a guy who does this,” said Michael Binder. “I think the word is ‘bad,’ and I think that puts it generously.”

University of North Florida, whose polling Binder oversees, calculated Biden with a slight, one-point lead in its October 20 poll. Considering the margin of error, a close Trump win would be easily within range, as was the case for many polls. But Binder said that with his and most other prominent polls underestimating Trump’s performance by a few points or more, there is cause for major concern.

“Because all of these errors are biased in the same direction, that leads to an underlying problem,” he said Wednesday.

“This was a terrible last night for the polling industry.”

Of the 18 pollsters whose final surveys the Tampa Bay Times collected, none showed Trump ahead by more than two points, and most showed him trailing by multiple points. Still, applying the full margin of error to each candidate would make the average range stretch from a 9-point Trump loss to a 5-point win.

It was a close race, according to the polls, and polls are often a few points off from the actual election results.

Knowing that, outlets like FiveThirtyEight and The Economist don’t just produce polling averages. They also use historical data to guess how the race could shake out. If candidates are close, the frontrunner still has a decent chance of losing. In Florida, Biden was the frontrunner, and those outlets projected he would win by 2 or 3 points. But even with that lead, they estimated, Trump still had a chance to win. FiveThirtyEight gave Trump a 30 percent probability. The Economist gave him 20 percent.

Considering the averages taken by those two outlets and the New York Times, Trump’s ultimate result was better than his polls by about 5.5 points. That’s more than in other close states like Michigan, the Carolinas, Texas, Minnesota, New Hampshire and Georgia. Polls in Wisconsin, which were 8 points off in Biden’s favor, did worse than in Florida.

(Washington Post/ABC, which had the final poll in Florida that was closest to the actual result, overestimated Biden’s Wisconsin win by 16 points.)

St. Pete Polls owner Matt Florell said his Florida poll missed the mark. The organization’s Oct. 30 survey found Biden ahead by one point. Trump’s ultimate edge was outside their range.

Get insights into Florida politics

Get insights into Florida politics

Subscribe to our free Buzz newsletter

We’ll send you a rundown on local, state and national politics coverage every Thursday.

You’re all signed up!

Want more of our free, weekly newsletters in your inbox? Let’s get started.

Explore all your options

Florell said they nailed the breakdown of voters' party affiliations but that didn’t translate into accurate projections on the race.

“We predicted (Republicans by two percentage points), which is what it was, and we still had the final outcome wrong,” he said.

He said they don’t poll in Spanish, unlike many larger pollsters, and that may have done them in in a race where some heavily-Hispanic neighborhoods in Miami-Dade swung in Trump’s favor. But he was also concerned about local races — they were within one point in State House District 59, but off by 18 in District 60 — even though both are in Hillsborough.

Florell was uncertain whether his outlet would even try polling the governor’s race again in 2022.

“I’m not sure if it’s worth it, or if we need it, or if it can be accurate in any way, because turnout is up in the air. It’s really difficult to poll general elections in Florida,” he said.

Not all pollsters agreed that there was an issue.

Kevin Wagner, a research fellow with the Florida Atlantic University’s Business and Economics Polling Initiative, said it’s not yet clear if there was a problem with the polls. He emphasized that all polls should be read as ranges of outcomes, and that they’re snapshots in time rather than predictions of the future.

“I realize that people want a lot of answers very quickly,” he said. “I think it’s going to take some time.”

Florida Atlantic University’s poll finished Oct. 27 and found Biden ahead by two, with a margin of error that included Trump’s lead in the end.

Even in a perfect world, the same poll from the same pollster with the same process won’t always lead to the same result, Wagner said.

“Every draw from the population is going to get you a slightly different sample,” Wagner said. "We do have to do a better job of explaining what our polls mean and what they don’t mean.”

All pollsters take raw interview results and weight them by how they expect the electorate to look, using factors like age, race, party and education.

Binder, at University of North Florida, said once they get the public records of which Floridians in fact voted, they’ll re-run their numbers to match the group. If that weighting shifts their results closer to what happened in real life, it may show where they miscalculated.

After the results came in, he said he felt like a football coach who can’t explain why his team lost. He said he worries about trust in the industry eroding, and said that trust is important for giving regular people a voice on issues in government.

“Over time, and I believe this, credibility matters. You can’t just keep putting up nonsense, or what appears to be really-off results and have people still believe you.”

Binder said people assume polls are more precise than they are. He calls them butter knives, not scalpels, and he defends them when they might look wrong but actually showed a range of outcomes close to the final result.

“I can’t make that defense of these polls.”

Tampa Bay Times Politics Editor Steve Contorno contributed to this report.

Tampa Bay Times elections coverage

ELECTION RESULTS — FLORIDA AND TAMPA BAY: See all races, statewide and in Hillsborough, Pinellas and other Tampa Bay counties.

We’re working hard to bring you the latest news on the elections in Florida. This effort takes a lot of resources to gather and update. If you haven’t already subscribed, please consider buying a print or digital subscription. Or click here to make a donation to the Tampa Bay Times Journalism Fund.