Before I became an obscure blogger and show business peon, I was an aspiring academic. I have a B.S. in Cognitive Psychology and have studied many statistical and data modeling techniques for dealing with large multivariate data sets. I’ve tried to stay fresh in these skills over the years.
As such, I’ve been following the election polls closely. And playing around a bit with some of the polling data; looking into the methods and models of the various polling outlets. I’ve discovered a pattern in a few battleground state polls, particularly Ohio, which I can’t find anyone else noticing, at least anyone you can google. Pollsters are almost uniformly under-weighing a couple unsung but key variables in “likely voter” data. In my opinion, the correctly weighted model points to a clearer Barack Obama electoral victory than anyone in politics or the media anticipates. To explain what I mean, look at the latest “poll of polls” from Ohio via the Huffington Post:
Which shows debate-driven fluctuations, and a slight Obama lead, but essentially a race that could go either way. I’ve taken what raw data I could find and re-weighted it a bit according to my hypothesis, into a new graph below:
As you can see, the trend in the data becomes MUCH more clear. With this new picture in place we can extrapolate patterns in the next two weeks heading up to the election:
Pointing to a spurt in Obama’s numbers, before a tightening of the race that is too little too late. If the election were six weeks away, the poll numbers would diverge even further to show the expectant face or ass of America, which would herald an even steeper Romney decline.
So relax, my fellow liberals. This one’s in the bag.