It’s been a while since I updated this, but I figured this would be fun. I made my primary prediction for Michigan late last night and here’s the breakdown I went with:
Santorum – 40%
Romney – 39%
Paul – 11%
Gingrich – 9%
Others – 1%
Then today, I decided to look at the recent polls for Michigan. The first poll I went with was starting on February 12, since anything before this wouldn’t factor in Santorum’s triple win in Missouri, Colorado, and Minnesota. I made a scatter plot of Santorum – Romney:
I then added 2 trendlines. The first one (red) was a linear trendline which I knew wouldn’t be reflective of Santorum’s momentum recently. I plotted the second (green) second order polynomial trendline, which did a decent job at predicting his momentum, but I noticed that the relatively low lead he had right after his triple win was skewing the results a bit lower. Those polls are so old that they don’t correctly account for what’s going on now with just a simple second order polynomial. So I put in the third (purple) trendline, which was a third order polynomial. This correctly factored in his slow start to gain momentum, but looks like it drastically overestimates it at the end.
So I thought, how can I better examine Santorum’s momentum. I took out the polls right after Santorum’s triple-win, which I thought weren’t representative of the momentum he gained, then lost, and regained, and got this plot, with the red representing a linear trendline, and the green representing a second order polynomial:
This plot matched up almost perfectly with my predictions (which I wasn’t expecting it to turn out just like that), but I think accurately portrays the general momentum Santorum has had since his triple win.
Ultimately, gauging the momentum a candidate has is hard to do, but I think the second order polynomial in the second graph accurately portrays Santorum’s momentum. Ultimately, both of these models could end up being wrong, and there’s no perfect algorithm to predict voters, but the results of my second model, I believe, accurately portray Santorum’s momentum, and cement my confidence that Santorum will have a close victory over Mitt Romney.