Methodologies, demographics key factors in confidence with poll results

Polls taken within days of each other place the Arkansas governor's race as either a competitive race between Democrat Mike Ross and Republican Asa Hutchinson or a far lead for Hutchinson with the race in its final two months. But with two national polling firms having such different results in the race, which one should observers of politics and polling trust?

Dr. Jay Barth, a politics professor at Hendrix College in Conway and a pollster for The City Wire content partner Talk Business & Politics, said it is important to look at a collection of polls to get a sense of where an election stands versus looking at a single poll.

"Once you have looked at a number of polls, you can have confidence in trends that are shown," he explained.

While outliers like a poll by Marist College and NBC News – which has Hutchinson up by nine points against Ross – can shift the average slightly, he said looking at "an aggregation" of polls gives a better idea of not only where the election is going but where it has been. An easy and quick way to view an aggregation of polling data is by visiting RealClearPolitics.com, a site that tracks polling in most of the top political races across the country including the race for Arkansas governor.

Dr. Janine Parry, a political science professor at the University of Arkansas in Fayetteville and the director of the Arkansas Poll, said many times there is little movement from poll to poll.

"There are a good many reasons why these polls are far apart since the vast majority of polls confirm the previous polls," she said. "Typically, there's very little movement, so what happened here in an outlier."

Even though focus has been paid to the NBC News/Marist poll with Hutchinson besting Ross by nine points, a recent poll by right-leaning Rasmussen Reports taken just six days before the Marist poll had Ross up by two. Previous polling typically had Hutchinson up by only five or six points with a margin of error around four points. In the case of the NBC Marist poll, the margin of error was 3.9%, while the Rasmussen was 4%.

"In terms of why these are different, sometimes no matter how well a poll is conducted, there's always room for error," she said. "That's why good pollsters and good reporters report the margin of error. Even if the margin is tight … at any given time, we're only 95% sure that the result we arrive at is (the right) number. There's some noise possible in any poll. It could be that one of these polls falls into this category. Or it could be that both do, though it's unlikely."

Parry said political scientists and pollsters like herself and Barth would always look at methodologies when it comes to polling, noting that sampling of likely voters to registered voters impacts results. So does the inclusion of a select number of cell phones and the number of females versus males in a given poll.

"Those are all places you have to look," Parry said. "If those things aren't reported, any polling number should be suspicious. What I emphasize (when I discuss polling information with civic groups) is the more information they give you, the more confident you should feel about results. Samples, margin of error, questions asked, etc.”

Barth noted that polling nearly always sees a large difference in polls that screen for simply registered voters versus likely voters, with some of the pollsters asking specifically if the respondent plans to vote while others go based on voter data to see how many of the last several elections respondents voted in.

Barth and Parry said it is important for political observers to focus on the overall picture instead of depending on one specific poll, with Barth noting that voter turnout is likely to play the biggest roll in the election regardless of what polling data may be showing.

"The big thing I'm seeing in the polls is that if it's more (based on) all registered voters, the Democrats are advantaged. If it's a very narrow likely voter screen, then it really skews Republican. So this means it's really all about turnout. If turnout is especially high for the mid-term and if the electorate is going to show up and vote, Democrats are advantaged. If that does happen (and voter turnout is low), Republicans are advantaged significantly.”

The state of the races in Arkansas, Barth added, varies. In the Senate race, he said it would stay close until the end with polling data barely moving for either Democratic incumbent U.S. Sen. Mark Pryor or his Republican opponent, U.S. Rep. Tom Cotton.

In the race for governor, he said the last two months still provide room for movement for both candidates.

"The governor's race has a little more malleability in it. I think the candidates are still a little less known," he said, adding that as both campaigns continuing pouring money into television and other advertising avenues, polls could start to see more movement.

In the races for Congress, Barth described those races as more "fluid." The state legislative races are also described as fluid, he said, adding that individual voters' partisanship could ultimately decide many of the races.

Parry said whatever happens, political observers and the public wanting to have an idea of what is happening in the races should dive into specifics of polls released in the coming weeks instead of just rely on the talking heads, campaigns and bloggers to dictate the facts.

"But the public should probably pay more attention to evaluating the candidates instead of who is ahead," she added.