That's not the most notable thing about this data, though. Since Perdue won the GOP nomination on July 22, six pollsters have released numbers for the Senate race, including Landmark. Landmark, however, is the only outfit to show Nunn ahead. Polls from a variety of other firms have all found Perdue up anywhere from 6 to 9 points. Almost all of these companies surveyed the gubernatorial race as well, and have usually found Carter performing a bit better than Nunn. However, he, too, has generally been trailing or neck and neck with Deal; only Landmark has found him leading by more than a hair.
So Landmark is clearly the odd pollster out on the Senate race and also shows the best numbers for Carter in the gubernatorial contest. The good news for Democrats is that none of the pollsters who've been in the field here represents any sort of gold standard. Rasmussen has been awful for years; InsiderAdvantage was terrible in 2008 and has produced very few pre-election polls since then; and Vox Populi has only released one pre-election poll in its entire existence, a primary poll for the Eric Cantor race that it blew chunks. HEG does not appear to have any track record at all: We have no polls from them in our 2010 or 2012 pollster databases.
Of this batch, SurveyUSA (the most pessimistic pollster for Democrats this time around) is the only one that does have a decent track record. However, they have demonstrated some erratic behavior in the past, often releasing weird results and unexplained swings in the middle of the cycle, before coming down to earth at the end. Georgia Democrats can take some comfort in knowing that the polls showing them down aren't exactly of the highest quality (and in some cases are awful).
The bad news is that Landmark doesn't have much of a recent track record either. Head below the fold to learn more about how they've done.
Nate Silver didn't include Landmark in his 2008 pollster ratings, meaning they did not publish any polls within three weeks of the 2008 election (and nor did frequent partner Rosetta Stone). We also can find no late polls from either firm in 2012. Landmark did, however, conduct a few surveys in Georgia in the last week of the 2010 race:
• GA-Sen: Landmark: Isakson (R) 56-35; Actual: Isakson (R) 58-39; error: +2 RIn 2010, Landmark's average error was 2.4 percent in the Republican direction. Landmark underestimated the eventual winner (regardless of party) by an average of 4.2 percent. On balance, they were pretty good.
• GA-Gov: Landmark: Deal (R) 49-42; Actual: Deal (R) 53-43; error: +3 D
• GA-LG: Landmark: Cagle (R) 50-34; Actual: Cagle (R) 55-42; error: +3 R
• GA-AG: Landmark: Olens (R) 47-40; Actual: Olens (R) 53-44; error: +2 D
• GA-School Sup't: Landmark: Barge (R) 48-36; Actual: Barge (R) 54-42; error: +0
• GA-Ins. Comm'r: Landmark: Hudgens (R) 54-28; Actual: Hudgens (R) 54-42; error: +14 R
• GA-Ag. Comm'r: Landmark: Black (R) 48-32; Actual: Black (R) 56-40; error: +0
• GA-02: Landmark: Keown (R) 50-46; Actual: Bishop (D) 51-49; error: +6 R
• GA-07: Landmark: Woodall (R) 60-30; Actual Woodall (R) 67-33; error: +4 D
• GA-08: Landmark: Scott (R) 53-39; Actual: Scott (R) 53-47; error: +8 R
However, don't have much to go on beyond this 2010 data set. Earlier this year, they polled shortly before the Republican Senate primary runoff, where they gave Rep. Jack Kingston a 48-41 lead over Perdue; Perdue ended up winning 51-49, a miss of nine points. Other firms also found Kingston up a week before the runoff, but by a smaller amount. Republican runoffs are different animals than general elections, but it is worth noting Landmark missed here.
Ultimately, this newest Landmark poll leaves us with more questions than answers. Just as Democrats shouldn't have despaired when they were on the wrong end of the last few polls, they shouldn't celebrate right now. As frustrating as it is, the best thing we can probably do right now is wait for more numbers to see if anyone backs up Landmark or if they continue to be the only ones to show Nunn up. To paraphrase the West Wing's Leo McGarry, with Landmark showing Nunn up 7 and SurveyUSA showing Perdue up 9, we can at least be confident of one thing: A few months from now, at least one of them is going to look pretty stupid.