For weeks, Alabamians watched as presumed front-runners Tim James and Bradley Byrne engaged in a steel-cage match for the Republican nomination for governor. And when voters cast their ballots on Tuesday, it was Robert Bentley -– the Tuscaloosa dermatologist who consistently placed fourth in pre-election polls –- who emerged as a sur-prise contender.
And in the Democratic race, Agriculture Commissioner Ron Sparks trounced U.S. Rep. Artur Davis by a 24-point margin. Pretty good for a candidate in a race that most polls were calling a toss-up.
The political meanings of the Bentley surge and the Sparks rout will probably be debated for weeks. But in the short term, many voters are wondering why no one saw this coming. How could the polls be so wrong?
Patrick Cotter, a retired University of Alabama political science professor who studies political polling, wasn’t so surprised. Primaries, he said, are notoriously hard to predict.
“As a general rule, pre-election surveys in primaries are difficult,” Cotter said in an e-mail. “Compared to general elections, primaries are generally less visible and the amount of information available to voters is less. As a consequence, voters' preferences may be less stable.”
Ask experts in the field, and they’ll give you several reasons why Alabama’s pre-primary polls should have been taken with a grain of salt. First and foremost, there’s the problem of finding a true “likely voter” in an off-year election with a crowd of candidates and a flurry of negative ads.
“It’s easy to ask people, ‘do you intend to show up at the polls,’ but it’s hard to ask that question in a way that gives you a reliable answer,” said Sam Fisher, a University of South Alabama political science professor and former head of the university’s polling group. “People may not really know whether they intend to vote, or they may change their minds after you speak to them.”
They also may not know who they’ll vote for until they get into the voting booth. Republican voters had seven can-didates vying for their attention this year. Democrats had only two choices – but according to former U.S. Congressman Glen Browder, Democrats were also hard to track in a year that presented voters with their first African-American contender for the nomination.
First, Browder said, the issue of race makes voters uncomfortable when talking with pollsters. Second, many African-American voters in the South remain undecided until the last few days of an election. Browder said they often wait until churches or organizations within their community offer an endorsement of a particular candidate.
“Those polls do not reflect this traditional process in the Southern black electorate,” he said.
Browder, a political science professor at Jacksonville State University, said several issues contributed to the difference between Bentley’s poll numbers and his actual votes.
Bentley spent much of his advertising budget during the last days of the primary campaign. Browder said this tactic helped to sway many voters who were either undecided or frustrated with the more popular candidates. Also, the more popular candidates focused their negative advertisements on one another, allowing Bentley to come out of the campaign reasonably unscathed, Browder said.
There may be a larger trend at work here –- the end of polling as we know it. Technological change and economic stresses are changing the quality of the polling information available to the average voter.
One reason: the cell phone. Polling truly took off when the telephone became a household appliance –- because the phone made it relatively cheap and easy to quiz a large swath of the population. The rise of cell phones –- and in particular, the rise of cell-only families –- has added a new wrinkle to that method.
“Now that numbers are transferable, I can move to Wisconsin and keep my 251 (cell) phone number,” Fisher said. “It’s increasingly hard to find phone numbers that are representative of a geographical area.”
Americans have historically assumed that polls are conducted by nonpartisan groups -– nonprofit agencies, universi-ties or large newspapers –- who don’t have a dog in the fight, or at least have an interest in providing accurate numbers. Increasingly, that’s no longer true.
“The economy has hit newspapers pretty hard,” said Fisher, who used to conduct polls for The Mobile Press-Register. “Putting the paper out is getting more and more expensive, and papers are looking at ways to cut costs. Polling is usually something they feel they can cut.”
Meanwhile, the candidates themselves appear to be spending plenty of cash on polls. Most of those polls involve reputable polling companies that use good research methods, Fisher said, but the candidates aren’t necessarily sharing all their information with the public.
“If I’m a candidate, I want everyone to think that everyone else is behind me,” Fisher said. “So I’m going to release the numbers when they make me look strong.”
In many ways, that’s the story of polling in this campaign. Early in the race, Republican Roy Moore held a press conference touting himself as front-runner claiming that other organizations were hiding polling numbers that showed him in the lead. That set the tone for a season of breathless press releases, with each candidate claiming the new poll showed him on top.
By the time Robert Bentley released his final polling numbers, showing him with a realistic chance of getting into the primary, the poll looked like just another self-serving press release. Bentley's numbers turned out to be pretty much on the money.
In the age of candidate-funded polls, Fisher said, it’s more important than ever for voters know how to read polls critically –- to get under the hood, as it were, and to kick the tires.
“The first thing to ask is who did the poll, and was it done for a candidate,” he said. “That alone will tell you a lot about how to interpret the results.”
Voters should find out whether a poll involved a probability sample –- in other words, were the polls conducted using a random, representative selection of voters. And sample size matters. Fisher said that when he was in the business of conducting statewide polls, he would usually use a sample size of about 400-600 people, which gives a 4 percent to 5 percent margin of error.
Finally, he said, it’s important to know exactly what question was asked in the poll, because the wording of the question can affect the results. Fisher says most reputable research firms stay away from “push-polling,” but he notes that “if they’re really hurting for business, some firms might change their policies.”