PANAMA CITY — There’s good news and bad news for anyone interested in the race to represent Florida’s 2nd Congressional District.
If you’re rooting for Rep. Steve Southerland to keep the seat he’s held for two terms, the good news is Southerland is leading his Democratic challenger Gwen Graham 45 percent to 39 percent, according to figures his campaign released Monday.
But Graham supporters need not despair — Graham on Monday released poll numbers that show her with a 48 percent to 45 percent lead.
With so many moving pieces in a political poll — from sample size to the way a question is phrased to the type of phone pollsters call —it’s impossible to explain the difference without more information than the campaigns released.
The purpose of a poll is to determine how people who are actually going to vote will vote, said University of Florida political science professor Stephen Craig, who studies public opinion. But each of those moving pieces within a public opinion poll has the potential to introduce error into the results, and there is no such thing as an error-free poll, he said.
“You would be stunned how much a tiny little wording change can change the results,” Craig said.
Neither Craig nor Florida State University political science professor Charles Barrilleaux were familiar with Southerland’s and Graham’s polls, so they spoke generally about the methodology they look for to determine if a poll is likely to produce reliable results.
What are the questions — word for word — pollsters asked, and who did they ask? There’s a difference between registered voters, likely voters and actual voters, but actual voters can’t be identified before an election.
How often did pollsters call a number without success? The more callbacks, the more randomness in the sample, which leads to more reliable results, Craig said.
Here’s what the campaigns have released about the methodology to the surveys:
Both camps surveyed likely voters in the 2nd District, which stretches from Panama City to the Big Ben region. Southerland’s poll had a margin of error of plus-or-minus 4.9 percent, while Graham’s margin of error was 4.4 percent. Both had 95 percent confidence levels.
The Graham survey respondents were selected at random, and the campaign didn’t provide demographic details about who responded.
Southerland’s pollsters broke down their sample by demographic: Fifty percent of respondents were registered Democrats, 34 percent were Republicans, 16 percent were independent, unaffiliated or other, and 30 percent were “cellphone only” voters.
Hitting that cellphone-only crowd is critical; Craig said he would be skeptical of a poll that didn’t includes cellphones because those voters as a demographic tend to be younger and more likely to vote for a Democrat —even more so than a young person with a landline, Craig said. A survey that only includes landline phones is going to get responses from older voters.
Graham’s poll had a greater sample size, 500 phone interviews, which generally makes for a more representative sample. Southerland’s sample size was 400. Craig would want a sample size of at least 600 if he were designing a poll for District 2, though 1,000 interviews would be preferable.
Southerland’s poll was conducted Oct. 1-2, while Graham’s was conducted between Sept. 21 and 24. Those two weeks saw a lot of airtime filled with ads on both sides, as well as the first of two debates between the candidates.
The time period might explain a difference in results, said Craig, but that would be more likely in a primary fight than the general election. Primary polls can vary more widely because there is only one party in each fight; parties tend to anchor voters to the candidate of their party, Craig said.
“Generally, numbers usually don’t move a lot in a short period of time” in general elections, Craig said.
Why release it?
Because they are internal polls, it doesn’t make a lot of sense to release them, Craig said, unless the intent is to send a message to voters.
“I don’t know why they’re releasing these numbers. That’s the first thing that makes this feel fishy,” Craig said. “It’s not that I think they’re making up numbers; it’s just there’s not much purpose except to show you’re a viable candidate.”
Graham and Southerland, R-Panama City, are engaged in one of the most closely watched and hardest fought congressional races in the nation this year, so the viability of both candidates has been clear for some time.
Craig described what he called the bandwagon effect. Certain voters — particularly voters who are not informed about or attentive to the race — might see the results and get on board with a candidate they think will win.
“There are people out there who can be influenced by the poll result ... but again those people aren’t paying attention, so it’s not going to work,” he said.
There’s also a phenomenon known as confirmation bias, Barrilleaux said. Confirmation bias is why Coca-Cola and Pepsi spend so much money on advertising — not to convince cola drinkers to change brands, but to convince them they already have chosen the right brand.
“If I hear things that lead me to believe my choice is right, it really reinforces my choice. … We know that people are biased to believe what they already believe,” Barrilleaux said.
So what should a voter do with these seemingly contradictory poll results? Take them with a grain of salt, Barrilleaux said.
“This is really just campaign fodder,” Barrilleaux said. “I would say this race is too close to call.”