Polling the Georgia race

Polling a special election is hard. You usually don't have a lot of precedent to base turnout projections on, and that makes it difficult to figure out ahead of time what the electorate is going to look like.

Polling a special election over a holiday weekend is even harder. People are out of town, at the mall, and generally just harder to reach than usual.

That's the confluence of circumstances folks polling the Georgia runoff are having to deal with.

So here's one conundrum. On our poll over the weekend the unweighted percentage of respondents who were African American was higher than it was for the Presidential election. Does that mean Republicans and/or conservative leaning white Democrats are planning to stay home on Tuesday, or does it mean those demographics were more likely to be visiting family out of state or in Athens for the football game over the weekend?

At the same time African American participation in early voting was under 23%, considerably less than what it was for the Presidential race. I can absolutely guarantee you that if that number doesn't increase when those who turn out tomorrow are factored in that Saxby Chambliss will win by double digits.

So we have data from the poll this weekend suggesting black turnout could be as high as 33-34% and turnout from early voting suggesting it could be as low as 23-24%. Our poll the previous weekend, not over the holiday, looked like blacks would make up 27-29% of the electorate and we're sticking with that for our projection. But because we've seen evidence to the contrary in either direction really nothing between a 2 point Martin victory and a 16 point Chambliss victory would absolutely shock me. That's just the nature of the uncertainly with an election like this.