Polls say a lot about the 8th District congressional race between four-term Democratic incumbent Jim Marshall and Republican challenger Austin Scott, but they don’t all say the same things — or address exactly the same questions.

Three firms have released polls on the race recently: The Mellman Group, hired by Marshall’s campaign; independent but Republican-leaning pollster Landmark Communications; and Penn Schoen Berland, at the behest of the congressional newspaper The Hill.

Two of the polls show Scott with a strong lead, while Mellman’s says Marshall has a lead that’s within the poll’s margin of error.

Scott spokesman Sam Ray says the Scott campaign has its own numbers too.

“We’re doing internal polls every week, but they’re not public,” he said. The last poll the campaign released, a month ago, showed Scott leading by 8 points, and there hasn’t been much change since then, Ray said.

Just as interesting as the final tally, however, is how the polls were done and who they reached.

The Hill’s phone survey of 400 “definite” and “probable” voters gives Scott a 13-point lead over Marshall, and it indicates that national trends against incumbents and Democrats are having an impact on the 8th District race.

Scott has worked hard to tie Marshall to President Barack Obama and House Democratic leaders, and The Hill’s poll indicates that it has had some impact. Marshall is not personally unpopular among voters, according to The Hill’s poll, but 41 percent of respondents said they’re less likely to vote for him because he’s an incumbent.

Respondents disapproved of Congress’ performance by more than 3-1. They also disapproved of Obama by 64 percent to 36 percent, with three-quarters saying that strongly impacted their congressional choice.

The most extensive poll — surveying nearly three times as many people as Mellman or The Hill’s poll — was released this week by Landmark Communications of Duluth.

“Only respondents who said they ‘definitely plan’ to vote ‘somewhat likely to vote’ are included in the final report,” according to Landmark’s methodology.

The most recent Landmark poll said voters were going for Scott about 53 percent to 39 percent, leaving the remaining undecided voters too few to change the outcome. That’s not a real shift from the previous week, though both candidates have now cut further into the pool of undecided voters.

A previous Landmark poll, released Oct. 21, said Scott’s margin held up across all age groups. In this week’s poll, however, Marshall had evened out the figures among voters under age 35. He also gained a little more support from male voters, but they still favored Scott by a considerable margin.

Marshall spokesman Doug Moore said Landmark’s automated “robopolling” is far less likely to get good results than interviews by live callers.

“People hear a recorded call of any sort … and they just hang up the phone,” he said. “I don’t consider that to be worth anything.”

Moore, perhaps not surprisingly, said he considers the Mellman poll most reliable.

The Mellman poll said Marshall is well liked by about 48 percent of voters and disliked by about 34 percent, despite representing a district that’s an estimated 45 percent Republican.

Scott, less well known than the eight-year incumbent, has a similar proportion of people who actively like or dislike him, but a bigger chunk undecided, the Mellman poll said.

Marshall’s campaign is counting on get-out-the-vote efforts by national and gubernatorial campaigns to help close the poll gap, along with predominantly younger voters who use cell phones exclusively and are therefore underrepresented in polls based on land line phones, Moore said.

Charles Bullock, head of the University of Georgia’s political science department, agrees with Moore on some poll mechanics — but not necessarily on results.

He wonders if that poll is assigning substantially more weight to expected black voters than the other polls, he said.

Major national pollsters tend to agree that the 8th District is “probably going to elect a Republican,” Bullock said.

While there is concern about missing people who rely on cell phones, polls for this year’s primary election tended to predict the results fairly well, he said.

Despite problems, poll results can’t be dismissed out of hand, Bullock said. They’re the best evidence anyone will have until election results are in, he said, and both campaigns pay them close attention.

“The candidate who’s getting unfavorable results out of polling will say that the polls are bad,” Bullock said.

To contact writer Jim Gaines, call 744-4489.