AUGUSTA – Energized by a heated five-way gubernatorial race and competitive congressional campaigns, 2010 was one of the most polled elections in Maine history.

But the polls — which rarely failed to garner front-page attention — failed to pick up late-breaking trends in two contests.

“This was a really historic, really interesting, really challenging election,” said MaryEllen FitzGerald, president and founder of Critical Insights, a Portland-based polling firm. “I’ve never seen an election that has had so much in flux and that so many people were not paying attention, weren’t engaged and didn’t make up their minds until the last minute.”

FitzGerald defended her firm’s work — which included four Maine Polls commissioned by MaineToday Media — saying its final poll pegged candidate Paul LePage with 40 percent of the vote. He was elected with 38 percent.

Democrat Libby Mitchell was predicted to get 21 percent versus the 19 percent she actually received.

And independent Eliot Cutler, who benefited from a last-minute endorsement from former Gov. Angus King and an e-mail from Equality Maine encouraging supporters to “vote for the pro-equality candidate with the best chance of winning,” earned 37 percent of the vote, though he only had 21 percent in the final Maine Poll released days before the election.

“We didn’t pick up the Cutler surge, but that was after we were out of the field,” FitzGerald said. “The methodology was pretty good, we just didn’t get the surge.”

While all agree methodologies are the key to good polling, different polling firms use different methods.

National firms such as right-leaning Rasmussen Reports — which has received criticism in the national press for showing a GOP bias in its results — and left-leaning Public Policy Polling offered periodic surveys using large samples and automatic polling methods. comparison, the Maine Poll used live interviewers.

Pan Atlantic SMS Group and the Maine Center for Public Opinion, commissioned by the blog Pine Tree Politics, also released polling data this cycle.

One poll that drew attention was a Maine Poll released Oct. 29 that showed U.S. Rep. Chellie Pingree, D-1st District, trailing Republican Dean Scontras by about 4 points. It was the only poll that showed her behind. She eventually trounced Scontras, 57 percent to 43 percent.

“We predicted that he was going to be at 45 percent. He came in at 43 percent,” FitzGerald said. “So, that was right on target. What happened in that race, you had about 14 percent undecided. Those people had been undecided and tended to buoy Pingree. There was also a lot of negative advertising about Scontras that came out at the end.”

Scrutiny of polling methodology tended to focus on whether a firm called landlines, cell phones or both. FitzGerald said although her firm has surveyed cell phone-only users in the past, it did not this year.

That decision is “largely driven by resources, to be honest with you,” FitzGerald said. “In this particular election, in the polling that we did, I don’t think that including a cell phone-only cohort would have made a difference because I think so much was breaking at the last minute.”

Nationally, about 25 percent of adults rely solely on cell phones but in Maine it’s about 12 or 13 percent, Fitzgerald said.

The controversy about whether or not to poll cell phone users stems from demographics. Some say not doing so tends to skew results, since cell phone-only users tend to be younger.

“Increasingly, people will be including cell phone-only components in their telephone surveys. The challenge is, it’s not just that easy because you don’t have simply cell phone and landline,” she said.

Some people still have both, she said.

“What makes a poll predictive is that each member of your defined target audience has an equal probability of being sampled. So if you have someone who has a cell phone and a landline, they are going to have more than one opportunity to participate in the survey,” FitzGerald said. “There are some real technological challenges to doing that, but I think the industry itself is evolving in that direction.”

Michael Franz, a political science professor at Bowdoin College, said polling firms have been able to get by without using cell phones because earlier studies have shown it did not make much of a difference.

“But now that you are seeing these differences, they are just going to get more pronounced over time, so it really probably will require pollsters in the future to include cell phones in their sample,” he said.

Experts said that, overall, the polling was pretty accurate, but said numbers for the congressional district races would have been better if the sample sizes had been larger.

“One thing that I did find frustrating was the reliance on half the sample to make inferences about the 1st or 2nd Congressional Districts,” Franz said. “You really do need more than 300 people to make that kind of an inference.

“To do a solid estimate of where the race stands in the congressional districts, you need more. The numbers can jump around a lot, just with a different set of 300 people and it’ll settle down when you have a bigger sample.”

Mark Brewer, political science professor at the University of Maine, agreed.

“The criticism of the polls, at least what I have been hearing, isn’t really directed toward the gubernatorial race, it’s more along the lines of, what the hell happened in the 1st Congressional District?” he said. “Without seeing the methodology it’s hard to say, but a sample size of 300, I would not be terribly comfortable with that. For something like that, I would want to go for 400 at least.”

Maine is one of the most difficult states to poll accurately because of its low barriers to voting, said Brian Duff, a political science professor at the University of New England.

“It’s harder to predict who is going to vote in Maine, simply because we’re a pretty unique state when it comes to voting laws,” he said. “The barriers to voting that affect states all over the country are just a little lower here in Maine.”

Maine allows same-day registration and makes it easier to maintain your registration when you move than other states, Duff said.

“So it’s actually a state where, even though it’s a little older, young people and basically the working class votes a little more than in most other states,” he said.

And polling is far from an exact science, according to Duff, with most polls consisting of calling a bunch of landlines, getting a response rate of about 15 percent and writing down what is said.

“That is really cruddy social science — and we wouldn’t get away with it in peer-reviewed journals,” he said. “But it becomes a national obsession every two years.”

 

Copy the Story Link