Five thirty eight got it almost 100% right, pending the uncalled, only Indiana is wrong. They had it lean McCain and it went Obama.
But they’re the only ones I see with this side by side comparison. How did other polls and pollsters do in predicting the many races?
I’m expecting that FiveThirtyEight will do a breakdown of the final polls versus the election results sometime in the near future.
They also got the Alaska senate race wrong, big time. Nate was predicting a 100% win for Begich. As screwy as Alaska is, that’s worth noting.
It’s worth noting that 538’s algorithm is just based on polling and to a lesser degree demographic data. It doesn’t factor in who has the better “get out the vote” operation, which is what they say accounts for Obama’s Indiana victory. Their commentary, in contrast, has been touting Obama’s get out the vote operation for months now.
Incidentally, when I looked at the various poll-of-polls sites (fivethirtyeight, pollster.com, electoral-vote.com, realclearpolitics, etc.) the day before the election they all seemed to have nearly the same electoral map, with some differences on the states that ended up being really close like North Carolina and Missouri. I’m not sure if there were more differences over House and Sentate races.
Yeah. From his comments, it looked like Nate just threw out all previous data, and assumed that Begich’s post-conviction gains were real and growing. Not a terrible idea given how much of a game chager a conviction could have been, but it bit him on the ass.