SurveyUSA has a new poll out about the impending runoff in Georgia between Saxby Chambliss and Jim Martin. For some reason, they didn't poll the trial heat question (e.g. "who are you going to vote for?"), but there are a number of other items of interest.
In particular, 87 percent of registered Georgia voters claim that they are going to participate in the runoff on December 2nd -- this in a state where only 67 percent of registered voters turned out for the general election last Tuesday! People may have the best intentions, certainly, and there is probably some response bias in the sort of person who might answer a series of a pollster's questions about the runoff election. Nevertheless, a lot of these people are pretty much ... lying. They're not going to show up. In 1992, when the late Paul Coverdell won his senate seat against Wyche Fowler in a runoff, participation declined from 2,251,587 Georgians in the first go-around to 1,253,991 in the runoff, roughly a 50 percent drop.
The tricky thing for pollsters will be in figuring out just which of these people are lying about their intent to participate and which of them aren't. Pollsters like to root their models in recent precedent, but things like runoffs and special elections happen so rarely that there's just not very much to key oneself off of. The point is ... if the polls going into December 2nd say that Saxby Chambliss is going to win the runoff by 7 points, you shouldn't be a but surprised if Jim Martin actually wins instead. And you also shouldn't be surprised if Chambliss wins by 20. This will be a return to the high margins of uncertainty that we saw in the primaries.