Thursday, October 30, 2008

Will undecideds break overwhelmingly for McCain?

The assumption that they will, is one of the things that keeps many diehard McCain supporters from jumping out of windows, given the pretty bleak poll numbers at this late stage. Some of you may remember going through this same dance as John Kerry supporters four years ago; convincing yourself that those last few undecided voters were going to get into the booth and decide to kick out the incumbent in large numbers. It was a reasonable theory, with some historical support, but it obviously didn't turn out that way. Despite the fact that this late breaker theory didn't work out so well last time, many(Republican and Democrat alike) think it very well could happen this time... why? Well, I don't know if you've noticed this, but Barack Obama is a black man. A black man with the middle name Hussein... Just sayin'. You'll also notice that many of the undecideds lack a high school education and are uhm... "downscale", as it were.

So anyway, there are some reasons, that nobody is laughing the idea off as desperate straw grabbing by the McCain camp. But what can do other than wait and worry with our fingers crossed?

Enter Charles Franklin at Pollster. He took the Diageo/Hotline tracking poll data from October 3-11 and fit a statistical model to those respondents who had expressed a preference in Presidential candidate. Using such a wide swath of polling means that he could assemble enough data sets of undecided voters to actually do some predictions on these voters. In a typical poll you are only going to have a couple of dozen undecided voters, but by using eight days of a daily tracking poll he was able to examine 267. Pretty good. All of the people who responded to the survey were asked a slew of attitude and demographic questions, beyond just preference for President, so Professor Franklin was able to fit a model to including those characteristics:
...including favorability to Obama and McCain, party id, marital status, kids at home, education, race, age, sex, church attendance, region and urban, suburban or rural residence.

He even included data on the following question:
I'd like you to tell me whether you strongly agree, somewhat agree, somewhat disagree or strongly disagree with the following statement. ... African Americans often use race as an excuse to justify wrongdoing

Which he has seen to have "a statistically significant effect on vote choice, even after controlling for other political attitudes and demographics." Proffesor Franklin wouldn't do this, but let's call that the "unenlightened racial attitutes question."

Now that he had fit this model to the people who picked either Obama or McCain, he could then use the same parameters to predict how undecided voters were going to vote based on their demographics and attitudes. My one quibble with his approach, not seeing a detailed write-up of his methods, is that he didn't divvy up the group he fit his model to into a "training set" and "validation set" - I could then be more certain that his model wasn't over-fit or the fit artefactual, though I don't suspect that is the case. Regardless, what were his findings?
Bottom line: Undecided and refuse to say voters are estimated to break 50% for McCain and 50% for Obama. As even as it gets. There is no evidence here of a large bias towards McCain that is hidden within the undecided respondents.

And:
Finally, what happens if we ignore racial attitude and predict vote among the undecided without it? The split is 52% Obama to 48% McCain. So at most the impact of incorporating racial attitude in the model is a rise of 2% for McCain among undecided. Given the sample sizes involved, that is well within the margin of error. And if we take out candidate favorability from the model we get estimates of 52-48 without racial attitude and 53-47 with racial attitude.

Pretty interesting stuff, and a really nice approach to a tough question. It also highlights one of the reasons pollsters are so excited by all these daily tracking polls... soooooo much data. Professor Franklin and his colleagues should be able to start examining some very interesting issues as they sift through it all.

So, good news for Obama fans, but it's by no means definitive... in the end we will just end up having to wait and see as the votes are counted. Stock up on the pepto.