Pollsters have messed up in Florida the past two election cycles — can we trust them this year?
Political dynamics in FL have shifted dramatically in only a few years
Early voting in the state capital on Oct. 24, 2020. Credit: Diane Rado
The dissatisfaction about public political polling in Florida reached its apotheosis on Election Night in 2020, when Donald Trump defeated Joe Biden by 3.3 points. That was even though at least five major pollsters had predicted in their last surveys before the election that Biden would win the Sunshine State by at least four points.
That prompted Justin Sayfie — a lobbyist now with Ballard Partners but best known for his “Sayfie Review” website, which has been aggregating the top political stories of the day in Florida for two decades — to announce that he would no longer publish any stories reporting poll results on his site.
“I think that they’re too unreliable,” Sayfie recently told the Phoenix (though the site’s “Blog Ticker” did post links to Orlando Sentinel stories on the Mason-Dixon poll last week looking at the governor and Senate races).
“I think that pollsters are having a harder time than ever figuring out who’s going to actually show up and vote.”
In the days leading up to the 2020 Florida presidential election, Emerson and Monmouth University separately had Joe Biden up by six points over President Trump, Quinnipiac had Biden up by 5 points, and Reuters/Ipsos and NBC News/Marist had Biden up by 4 points.
Well, anyone can have a bad night, right?
However, similar breakdowns happened before the 2018 Florida gubernatorial election: Quinnipiac’s last poll published the day before the Florida gubernatorial election had Andrew Gillum leading Ron DeSantis by 7 percentage points. Emerson Polling had Gillum up by 5 points; NRF-Harris had Gillum up by 3 points.
In both cycles, Republicans were obviously underrepresented in the polls. Opinions vary about why.
Longtime St. Petersburg-based political strategist and pollster Barry Edwards says that in part was because conservative voters don’t trust mainstream media sources and thus are more reluctant to answer questions from pollsters.
“There’s a perception by Republicans that if you express any Republican view that you’re a ‘MAGA’ or ‘Super MAGA’ and that you’re a fascist,” he says. “So, Republicans are very reticent about answering questions because they don’t want to be targeted for their views — so it’s taking on average three times longer to finish a poll.”
The theory gained resonance in the aftermath of the 2016 presidential election, following Donald Trump’s stunning presidential victory. It resurfaced before the 2020 election and gained force when an online study found Republicans and independents were twice as likely as Democrats to admit to withholding their true preferences when queried by surveyors.
The question leading into this fall’s election is whether those pollsters have made adequate adjustments to ensure that they’re more accurate this time. Some pollsters who spoke with the Phoenix say that it’s more difficult than ever to conduct surveys.
“Today fewer people have landlines,” says Frank Orlando, who conducts polls for St. Leo University. “Fewer people will answer. There’s caller ID. There’s widespread mistrust of polls.”
Veteran Florida Democratic political strategist Steve Schale adds that there’s a lot of “self-selection” when it comes to who is responding to polls, meaning it’s more difficult than ever to conduct a “random” survey. Not only are there fewer people with telephone landlines, but many people simply won’t answer a call or text on their cellphone from a number they don’t recognize.
And then there’s people who participate via internet panels.
“They’re not exactly your ‘random sample’ of voters,” Schale says. “They’re people that consume this stuff.”
Edwards says that, with just weeks before Election Day, pollsters should at this point only be sampling “likely” voters — not just registered voters.
“You need people who are going to participate in this midterm. If they only vote in the presidential — that’s one out of four voters or whatever — they’re not relevant to this turnout because they’re skewing the actual result that you’re going to have,” he says.
Those in the polling community agree that overall accuracy in polling has diminished in recent election cycles.
St. Pete Polls pollster Matt Florell attributes the problem to the fact that too many media outlets “publish a lot of the garbage polls” that he contends don’t come with enough caveats from reporters writing about them.
“For example, a Florida statewide poll with less than 300 respondents should be ignored by everybody, yet there were five of those this primary cycle in Florida that were picked up and published by multiple media outlets,” he says.
Florell has historically used “IVR” — interactive voice response telephone — polling using landlines since he began conducting political surveys in 2011 but, with increasing number of Floridians dropping landlines every year, he says he can’t do IVR polling at all in smaller districts where it worked a decade ago.
Brad Coker of Mason-Dixon Polling Strategy has been doing political public polling for nearly 40 years in Florida. He says that it costs more money to do polls these days because it requires more phone calls to reach people than it did 10 or 15 years ago. But he says that now that there are voter databases matched with cellphones, he doesn’t have many problems with landlines being phased out.
“I would say that 80 percent of my interviews are now done with people on their cellphones,” he says.
Coker acknowledges that there may be some hesitancy in certain quarters of the population to answer pollster’s calls, but believes the problem lies in lack of accountability for pollsters who get it wrong.
“If you’re some university think tank with a big endowment and a donor who wants to fund your operation, you can keep being wrong cycle after cycle, because there’s no consequence,” he says.
The dynamics are very different this year politically in Florida than even two years ago. Now, for the first time in the state’s history, there are more registered Republicans than Democrats — nearly 270,000, as of Aug. 31.
And, unlike the last two election cycles, there’s less big money being poured into the state from rich donors like Michael Bloomberg or Tom Steyer to stimulate Democratic voter turnout. That’s resulted in clear leads for Republicans going into Election Day.
The RealClearPolitics average shows DeSantis with a 7 percent lead over Democratic opponent Charlie Crist in the race for Florida governor, and Marco Rubio with a 4 percent lead over Democratic U.S. Rep. Val Demings in the U.S. Senate race.
Sayfie says that if the polls mostly turn out more accurately this cycle, he may reevaluate his philosophy on posting more stories on poll results on his site. But for now, “I just don’t have that confidence level.”
Our stories may be republished online or in print under Creative Commons license CC BY-NC-ND 4.0. We ask that you edit only for style or to shorten, provide proper attribution and link to our web site. Please see our republishing guidelines for use of photos and graphics.