We've all seen the polls. This candidate has pulled into a virtual tie with that candidate. That candidate has a big lead.
Only to find on election day they were dead wrong.
One of the most famous examples was pollsters overwhelmingly predicting Hillary Clinton's win over Donald Trump in the 2016 presidential election. She did win the popular vote, but lost the electoral college — which is what really counts.
ALSO READ: Polls are a snapshot, and other helpful things to remember during the midterm elections
"For 2016 we were polling that Clinton was had a chance of winning. She didn't have a chance of winning," said Matt Florell, president of St. Pete Polls, an independent polling service. "But it was turnout. You cannot predict turnout."
He said in Florida, the most recent poll that was completely wrong had Angie Nixon trailing Alexander Vindman by 10 percentage points in the Democratic primary race for U.S. Senate. She won by 12 percent.
"There were no polls," Florell said with a laugh.
By that, he means there were no statewide polls that were not commissioned by one of the candidates. Those are the most likely to be biased or have small sample sizes.
"We were never asked," he said. "Nobody even had us quote it because everybody thought it was a foregone conclusion Vindman was going to win. But surprises — surprises happen sometimes."
But why are these surprises happening?
"I would paint with a broad brush that since 2016 or so, I think most pollsters would tell you that it's harder to get accuracy, and I think there's a few reasons behind that," said Sean Freeder, who teaches political science and runs the University of North Florida's Public Opinion Research Lab.
He says polling was a lot easier — and more accurate — back in the day, when everyone had landlines and most numbers were listed in the phone book.
They still contact landline voters — but those with landlines are mostly senior citizens — who happen to be one of the most reliable blocs of voters.
You might think your cell phone number is unlisted, but that's only for marketers.
Florida publishes registered voter lists every month, which include phone numbers. So if you've given your cell phone number to the supervisor of elections at some point, and you haven't told them to remove it, it's available. Pollsters are allowed to text you with a survey link.
But results are all about how many potential voters are reached — and more importantly — what kind of voters they are. Freeder says it's called "non-response bias."
"Fifty years ago, you call out to people, and Democrats and Republicans have an equally likely chance of talking to you," he said. "Now we find — for at least the last decade or so — that Republicans tend to respond to polls at somewhat less of a rate than Democrats."
So they say they stop calling Democrats.
"You'd screen out people that say they're Democrats in the beginning to make sure that you're getting roughly proportionate numbers, so that's standard practice," Freeder said.
That may be because Republicans are bombarded by anti-media messaging.
"If Trump is on the ballot, those tend to be more difficult to predict elections," Freeder said. "It could involve maybe people are more likely than usual to not talk to pollsters if they're planning on voting for Trump."
Freeder says midterm polling in 2018 and 2022 —when President Trump was not on the ballot — were fairly accurate.
"The independents are coming out much more in favor of Democratic candidates two to one, whereas four years ago that was pretty much reversed. So this year is different, but also the landscape has changed. So who knows what will happen?"Matt Florell, St Pete Polls
And he said there's a lot more attention given to bad polls, rather than ones conducted by respected polling firms.
"I think people tend to hear the most saliently wrong polls first, and they're not necessarily paying attention to maybe a batch of polls that are maybe not perfect, but closer to what the actual results end up being," Freeder said.
"I tell my students whenever I go into the community and speak," he said, "that if you want an understanding of what the polls are actually saying, you don't get it from your local TV or radio sources. You don't get it from the internet. You get it from one source that models all of the polls, sort of averaged together."
So let the buyer beware, in essence.
ALSO READ: Click here for a guide to reliable polling from the American Association for Public Opinion Research
Now, St Pete Polls' Matt Florell says Republican registration has grown substantially in Florida, but independents are changing their tune.
"The independents are coming out much more in favor of Democratic candidates two to one, whereas four years ago that was pretty much reversed," he said. "So this year is different, but also the landscape has changed. So who knows what will happen?"
The latest poll from the Public Opinion Research lab of 1,400 likely voters shows Republican Byron Donalds and Democrat David Jolly are in a dead heat.
"Though Donalds enjoys a slight advantage, these numbers suggest a competitive gubernatorial race, similar to the race between Andrew Gillum and Ron DeSantis in 2018, which was won by less than a percentage point," Freeder said. "Given the margin of error, a win by either candidate would be unsurprising.”
And in the U.S. Senate race, Republican Ashley Moody's lead has narrowed to four points over Democrat Angie Nixon. That's down from an eight-point lead in July. Freeder said independent voters prefer Nixon by 15 percent.
But, as we all know, the only poll that counts is the one held on election day.
