The final RCP average for national polls was 52.1% Obama, 44.5% McCain. The current popular vote count is 53% Obama, 47% McCain. That implies no significant Shy Tory effect, and undecideds splitting 4:1 in McCain’s favor. It’s also consistant with a 2% Shy Tory Factor and a 3:2 split of undecideds, or a 3% Shy Tory Factor and an even split of undecideds.
The state polls tell a different story. State-by-state, things are all over the map. There’s states like Arizona, Iowa, Oregon, and Utah, where McCain outperformed by a large margin. And there’s states like Massachusets, Indiana, and Missouri where Obama outperformed. And there’s states like Virginia and North Carolina where the state polls were spot-on.
But overall, the states polls show an even split of undecideds and a 0.1% Shy Tory Factor in Obama’s favor.
I decided to apply a Shy Tory adjustment based on looking at Party ID weightings in the national polls. And it turns out there was a skew in the national polls. Where I went wrong was assuming the state polls had the same skew; they didn’t, at least not on any consistant basis. I also went wrong in assuming an uneven split in undecideds. It happens in many races, but it didn’t in this one. I’ve heard a theory that undecideds break evenly in high-profile races where both candidates are very well known by the time of the election; that increasingly looks like it’s the case in Presidential races (I believe the last Presidential election with a large uneven break of undecideds was 1996).
{ 1 trackback }
{ 2 comments }
Whoa! You have too much time on your hands.
I’m almost afraid what you will get up to while waiting for the next election. Go to Vegas. Count cards. Do something constructive.
It took me about 15 minutes with a spreadsheet to do the analysis. It takes longer than 15 minutes to go to Vegas, and most casinos frown on spreadsheets.
Comments on this entry are closed.