Do you trust surveys?

Pollsters explain what they do and how it’s changed

Each time Susan MacManus gives a speech on politics, and she averages two a day this election season, she’s inevitably asked one question.

“Everywhere I go,” the retired University of North Florida professor said, “everyone wants to know about the accuracy of polls.”

Just two years after polls seemed to indicate a win for Hillary Clinton over Donald Trump, people have become more critical and skeptical of voter surveys – much to the pollsters’ chagrin.

The directors of many of the state’s most prominent polling organizations say they’ve done everything they can to track changes in the use of cell phones versus land lines and establish unbiased ways of reaching out to not just likely voters, but lastminute ones who show up on

Election Day.

And, they explained, even in 2016, they weren’t really too far off — Clinton’s popular vote margin was close to what pollsters predicted nationally and in Florida.

“The common misconception is that there’s a gold standard for

polling, one way that’s really, really good way and every other way is bad,” said Florida Atlantic University pollster Kevin Wagner. “But every method does things well.”

The difference between polling methods, said Michael Binder, Public Opinion Research Lab Director of the University of North Florida, “is monumental. A lot of people see ’48 to 47’ and don’t look behind the curtains at all.”

So far in 2018, the polls in Florida have brought mixed results.

One race, for the U.S. Senate between incumbent Bill Nelson and Gov. Rick Scott, has gone back and forth with both holding small to decent leads while averaging out to a statistical tie.

Another, the Democratic primary for governor, saw the eventual winner Andrew Gillum running in third or fourth place for much of the race.

“I think part of the problem is that polls are not a prediction of what’s going to happen in the election,” Wagner said of the primary polls failing to catch Gillum’s rise in the last week before the primary – after many pollsters had released their final surveys.

“Polls show a moment at a specific period of time,” Wagner said. “You can’t capture something that’s not there.”

With primaries, Binder said, “There’s no party cue to help you. Often people know little about a candidate and make up their mind at the last minute. Maybe they caught him at one debate, or one interview. It’s easier for voters to change their mind.”

Brad Coker, director of Mason-Dixon Polling & Strategy in Jacksonville, said his company’s last Democratic primary poll was July 27, a month before the vote, so it couldn’t catch any late-breakers.

As for the polls themselves, the two biggest decisions pollsters make are who to call and how to reach them.

The rise of cell phones, Binder said, “certainly has changed polling dramatically over the last 20 years. … Half of the people in Florida are cell only. You have to call people on cell phones.”

Since it’s mostly illegal to call cell phones on auto-dial,

Binder said UNF uses live callers who dial by hand. One-third of their respondents are on landlines, two-thirds are on cell phones.

But, he added, the problem they always face is that today, “if you get a call on your cell phone and don’t know the number, you’re not inclined to answer it.”

Mason-Dixon uses live callers as well.

“In Florida, you’ve got to use bilingual interviewers,” Coker said. “Do robocalls to press one for English or 2 for Spanish work? I’m not sure.”

With live callers, “You can turn on a dime depending on who you talk to. If they answer ‘hello,’ you answer in English. If they answer ‘Hola,’ you answer in Spanish.”

Wagner, though, said FAU’s method of using a combination of robocalls combined with online polling has benefits over live callers.

“Live calling has all kinds of [variables] – who’s calling, whether it’s a man or woman … In IVR, everyone’s questioned the same way. There’s no bias there.”

Then there’s the question of who pollsters call.

“‘They didn’t talk to me, therefore the poll’s no good!’” as Binder described the complaints they receive. “You’ve got to understand, there are 13 million registered voters in the state of Florida. A large poll for us is 800 people.”

Some surveys ask voters how intensely they support a candidate. Others prescreen voter lists to find people who have a history of voting, which could also miss new voters such as the young college students Gillum cited as a key to his victory.

Timing also matters. In the governor’s race, FAU had Gillum up 2 points and UNF had him up by 4 points in mid-September. MasonDixon had Gillum only up by 1 during the last week of September – the week of the Brett Kavanaugh hearings, which some polls said fired up GOP voters. But newer polls since his confirmation have shown enthusiasm up among Democrats.

“The only way to judge accuracy is the last poll versus the actual vote count,” Coker said.