Thursday, November 1, 2012

Is it Gut Check time on the Presidential Polls?

An interesting phenomenon is occurring this election cycle. Both sides are pointing at polls to say their guy will win. That is all fine and dandy, except often they use the exact same poll. As the saying goes, a good pollster can make his guy look good given the correct set of questions and sample. So it was with some excitement that I noted challenges to the conventional wisdom that the big pollsters get it right. After all, I clearly remember all of the media outlets having to walk back their predictions in 2000 when the actual vote count failed to follow the polling model.

Still, I never really understood the two big issues as to why polls, and, more to the point, subsequent predictions based on them, vary by so much. A piece by Dan McLaughlin at Red State http://www.redstate.com/2012/10/31/on-polling-models-skewed-unskewed  brought it into sharp focus.) Read the whole thing at the link, but in short it's because the Math Model people (typically the Democrats and big polling houses) believe the science of the poll, which can and quite often is flawed, whereas the other bunch (generally the Republicans who believe polls are skewed through over or under sampling) has a Historical Perspective, which I equate to the "gut check model."

Everyone, or at least the intellectually honest parts of everyone, agrees that for the majority of the current polls showing the President with a huge lead (or any lead in some cases) to be correct the Democrats need to turn out in at least the same percentage this year as in 2008. Using that math, the polls, which typically over sample likely Democrat voters to reflect the inflated turnout number, would be right. Where the Math Model fails is when the Historical Perspective gut check comes in and we realize that no sitting president has had the same level of turnout from their base  when running for re-election. It is therefore not likely to happen for the first time this year, and certainly not where such a large percentage of the President’s previous vote appeared to come from new, first time or previously disaffected voters. I won’t even go into what some refer to as the "white guilt" vote, which also is unlikely to be repeated.

So how does that affect this election? No idea. The only poll that really matters happens on the "Tuesday following the first Monday in the month of November," which we will all watch unfold together. For the record, I’m in the Historical Perspective gut check camp on the polls, I just don’t know if the gut check completely negates the President’s lead in some of these polls, or just dials it back a notch.