By popular demand we'll have polls on the Georgia Senate runoff out each of the next two Mondays.
There are some tough choices to make in terms of sampling and weighting with a weird election like this, and the fact that we're going to have to conduct the final poll over Thanksgiving weekend just makes it that more challenging.
We're going to use the same sampling criteria we did for November 4th: folks who voted in the 2004 general, 2006 general, 2008 Presidential primary, or 2008 state primary. We gave some thought to narrowing the sampling criteria, and think it's highly unlikely turnout will match or even come that close to matching what it was for the general election. But one of the nice things about IVR is that people who aren't going to vote or don't care about the election can weed themselves out by just hanging up on us, instead of possibly feeling a compulsion to tell a live interviewer that they do intend to vote. So we're going to go with the broader sample and assume that anyone willing to take the time to answer the poll is also going to be willing to take the time to go out and vote.
As far as weighting, we went into our set of 16 final state polls with a pretty clear idea of how we were going to set up the polls by race and age regardless of what the composition of those who responded to the polls was. I think we did a very good job of projecting demographics, but this election is a whole different bird so we're going to take the composition of who is actually responding to the polls into account a lot more than we usually would over the next two weeks. I've had several questions about what we project black turnout for the runoff to be and the answer at this point is we really just don't know. We'll look at how our polls come out and decide based on that.
In addition to looking at Jim Martin and Saxby Chambliss our first Georgia poll will peak forward to 2010...and the Georgia/Georgia Tech football game. Stay tuned.