GA: Survey USA: Perdue re-takes lead

[Recalcuate]

I know SUSA crosstabs are always wacky, but is it realistic that Perdue is winning the early vote 54/44?

It probably depends on where the early vote is dispersed. I don't think it's out of the realm of possibility if the redder parts of the state have higher totals than the bluer ones.

As always with cross-tabs on a sub-sample though, the MOE will be horrendous, so much so to be practically meaningless.

[Recalcuate]

This is an outlier from the other polls, as Nunn has lead in three polls straight from other pollsters.

Disagree. This is noise around the same MOE for a race that is basically tied and likely headed to a runoff.

[Recalcuate]

This is an outlier from the other polls, as Nunn has lead in three polls straight from other pollsters.

Last 8 polls from RCP.

SurveyUSA                                   Perdue +3
CBS News/NYT/YouGov           Perdue +3
Atlanta Journal-Constitution        Perdue +2
InsiderAdvantage                   Nunn +2
WSB-TV/Landmark                   Tie
CNN/Opinion Research           Nunn +3
SurveyUSA    (superseded)            Nunn +2
WRBL/Ledger-Enquirer/PMB   Nunn +1

If anything, the trend is Perdue's friend, but neither candidate has shown the ability to reach 50. They'd probably need a margin of around mid 3s to 4 to do so.

[Recalcuate]

To follow-up, I did three calculations based on historic preference by race, what the polls have consistently showed by gender, and the likely party breakdowns available via voter file's early vote results posted above. All three were quite close to one another, so I'm confident in this estimate:

Race: Nunn 50.8
Gender: Nunn 49.5
Party: Nunn 50.1

Considering early vote in GA is usually only a point or so more "Democratic" by race than it is on Election Day, Nunn is likely looking at 49-50% of the vote when this is all said and done (sigh).

It's very hard to lock down party affiliation in Georgia because the primaries are basically open primaries. Voter affiliation can change from year-to-year based on which primary voters want to vote in.

As to your other point, statistically, it's very plausible that Perdue holds a 54-44 edge with early voters, just as much as it's plausible that Nunn holds a 54-44 edge. The sample size of about 103 voters (17% of 611 LV) is too small to make any reliable statistical determination given the MOE is between 9-10%.