Sun Sentinel Palm Beach Edition
Incessant stream of polls hits Florida
Here’s how to decipher the chatter and when you should ignore it
Depending on the day, Democrat Joe Biden is ahead of President Donald Trump by 5 points in Florida.
Two days later it’s 3 points, then 2 points, then 3 points again. In between, Trump is three points ahead of Biden, then 1 point. Or they’re tied.
Public opinion polls are multiplying, both nationally and in Florida, like the latest invasive species taking over the Everglades. They’re not only difficult to escape, but they’re difficult to decipher.
Although latest numbers provide fuel for frenzied chatter on cable TV and storms of 280-character analysis on Twitter, they don’t show who’s going to win Florida’s 29 electoral votes or win the presidency.
One thing the polls show with certainty: either Biden or Trump could win Florida. “If there’s anything to be taken away from the polls, it’s that the election is going to be close,” said Steve Vancore, a Florida Democratic strategist and president of the Tallahasseebased polling firm Clearview Research.
Also clear to analysts, including Vancore: The polling shows that almost everyone has made up their mind about Biden and Trump. “There are fewer and fewer undecided voters than ever before at this time of an election. The undecided vote becomes more and more microscopic with each passing day.”
How close is the contest? On Friday the FiveThirtyEight average of Florida polls had Biden 2
points ahead of Trump, 48% to 46%. The RealClearPolitics average also shows Biden 2 points ahead, 49% to 47%. (The two sites have different methodology in the way they collect and display data.)
Lots and lots of polls
Florida is a critically important state, and it’s always close. So, it’s one where top-rated pollsters and organizations that even political insiders have never heard of are conducting surveys, hoping to gain a bit of insight into the electorate. And, they hope, reap some publicity for having their results talked up by the news media and online.
In the first 17 days of the month, 15 polls have publicly released results for the presidential race in Florida. Of those, six have come from pollsters rated above-average by FiveThirtyEight, which specializes in data analysis. Some of the pollsters have released different versions of their surveys.
The sample
Besides arguing about the latest results, people wonder about — and sometimes vehemently disagree over — the notion that polls mean anything. Often people wonder how a sample of 428 or 631 or 1,235 can claim to show what Florida’s 13.9 million registered voters are thinking.
Pollsters use probability sampling, which means making sure everyone in the population has an equal chance of being selected, said Kevin Wagner, a Florida Atlantic University political scientist and research fellow at the university’s Business and Economics Polling Initiative.
“One doesn’t have to eat an entire pie to tell you whether the pie tastes good. They just have to make sure that the bite they eat has all the ingredients in it,” Wagner explained.
Vancore likened it to getting a blood test. It’s impossible to sample all a person’s blood, but a drop can reveal if a diabetic’s blood sugar is too high or a lab sample can reveal if someone has high cholesterol.
In a bygone era in which almost every household had a landline telephone and people answered their phones it was much easier to get a good sample by randomly calling from the pool of every phone number in use.
In the last two decades, many people have given up landlines, and those who still do have them tend to be older. People who rely exclusively on mobile phones tend to be younger.
The gold standard of polling this century has been considered live calls to landlines and telephones. But that’s expensive, and many people don’t answer unfamiliar numbers, so some pollsters are using combinations of live calls, automated calls to cellphones, and digital components.
Simply asking people to fill out a survey online isn’t valid, because then it’s not random; the sample would come from people who decided they want to participate and wouldn’t be representative. And an internet-only poll would miss many older and poorer people.
Vancore said he uses a system that sends text messages with links that respondents can use to complete the surveys from their phones.
Kathryn DePalo, a political scientist at Florida International University, said low response rates are a problem, and she wonders if that makes what’s supposed to be a random sample really random.
Two Florida polls released Tuesday used different methods — Monmouth University used live calls to landlines and cellphones and Florida Atlantic University used an online component and automated calls by computer — and reported similar results, a 3-point to 5-point lead for Biden over Trump.
Margin of error
Don’t make bets based on the percentage for Candidate A and the percentage for Candidate B.
Polls actually are showing a range, called the margin of error. If Candidate A has 54% and candidate B has 46% and the margin of error is plus or minus 5 percentage points, A and B could be tied at 50% each. Or it could be 58-42.
It’s especially problematic when trying to tease out results from smaller groups of people or regions of the state.
“People need to understand the limitations of these polls,” DePalo said by text message. “To only sample 400 people statewide in FL and make any claims dissecting this data by race or ethnicity is dangerous.”
The Monmouth Poll on Tuesday had a margin of error of plus or minus 5 percentage points. But pundits immediately honed in on two important voting blocs, Hispanics and voters age 65 and older.
But because they made up small slices of the overall numbers of people polled, they came with enormous margins of error — which sometimes weren’t acknowledged in online discussions and news accounts.
Monmouth said results for the 159 seniors had a margin of error of 8 points and 109 Hispanics was 9 points. That meant that the 49% for Biden among seniors could mean he has as much as 57% or as little as 41%. And the 47% for Trump meant he could have the support of 55% of seniors, or 39%, or somewhere between.
Are they accurate?
During election season, candidates who don’t like poll results typically say they don’t pay attention to polls. Trump has a different approach, with his campaign sending messages to supporters that “We can’t trust the Fake News Polls” and “The polls are LYING.”
Plenty of others criticize polling — especially when election results seem like a surprise, and a candidate who seemed to be leading ends up losing.
It’s easy to come up with examples. U.S. Sen. Rick Scott, R-Fla., referred to “all those inaccurate polls” after his 2018 election victory over then U.S. Sen. Bill Nelson, D-Fla. Most polls showed Nelson leading Scott, who was governor at the time.
The final RealClearPolitics polling average had Nelson winning 48.8% of the vote and Scott 46.4%, an advantage of 2.4 percentage points for Nelson. Scott ended up with 50.1%, winning the race with 0.2% more than Nelson.
Also in 2018, virtually every poll from the August primary until the November election showed Democrat Andrew Gillum ahead in the governor’s race. The final polling average in the governor’s race was 49.4% for Gillum and 45.8% for Republican Ron DeSantis, a difference of 3.6%. DeSantis won with 49.6% of the vote, 0.4% ahead of Gillum.
It was similar in the 2016 presidential race. National polling averages had Democrat Hillary Clinton winning the national popular vote by three percentage points. She won by two points, but lost the Electoral College.
There were some big state polling errors, largely because of faulty assumptions on how different voting blocs would participate. A major issue in 2016, Wagner said, was that many pollsters underestimated the number of white voters without college education who would vote. More turned out than expected, and they overwhelmingly supported Trump.
Also in 2016, people who reported themselves as undecided at the end broke largely in Trump’s favor.
In Florida, the polling average was close to the mark. The final RealClearPolitics average of Florida polls had Trump at 47%, which was 0.4% ahead of Democrat Hillary Clinton. He ended up with 49% of the vote, which was 1.2 percentage points more than Clinton.
Polling averages
It’s a mistake to make too much out of one single poll and the specific numbers.
“One poll doesn’t mean anything,” said Jim Kane, a South Florida pollster and political strategist, who taught graduate-level survey research classes at the University of Florida for 15 years.
The best way to look at polls is through averages of multiple polls, such as the ones compiled by FiveThirtyEight and RealClearPolitics.
When looking at an individual poll, it’s a mistake to compare it with the most recent poll from another polling outfit. Each organization can have slightly different wording on questions, or make different assumptions about who’s going to vote.
“You can always have one or two polls that can swing fairly widely,” Wagner said. “Watch the averages to see whether or not there is a trend, an observable, measurable trend.”
Good for what?
Miramar Mayor Wayne Messam wasn’t the biggest fan of polls in 2019, when he spent most of his time seeking the Democratic nomination for president. He dropped out in November after he didn’t gain any traction — and didn’t register at all in many public opinion polls.
When a September 2019 Florida Atlantic University survey found he had 3% support among Florida Democrats, Wagner said it could have been a sampling quirk. Messam countered on Twitter with a positive spin, suggesting the number “shows that many voters still want something new and fresh.”
Messam said polls have a place in people’s information diets. If it’s from a credible organization, he said, polling “gives a pulse of those that are sampled.”
But he cautioned people not to put so much faith in polls that they don’t take in a broader reality. “I think [polls] can be used as information, but not as the gospel truth and the absolute.”