Before tonight's election, I compiled the most recent predictions by several of the most prominent polling firms for outcomes in the battleground states where the election would be decided. I then compared what they predicted with the actual result. I looked at
- PPP
- SurveyUSA
- NBC/Wall Street Journal
- CNN
- Rasmussen
I did this mainly because Rasmussen polling has gotten a reputation for having a GOP bias. I wanted to test that hypothesis. There were only two firms that published polls just before the election with errors of +-4%, Rasmussen and NBC/Wall Street Journal. These errors were in six battleground states.
- Colorado
- Iowa
- Nevada
- Ohio
- Virginia
- Wisconsin
Of these six errors, FIVE were from Rasmussen. More importantly, in EVERY case the error was biased towards the GOP. Thus Rasmussen managed to simultaneously be both the least accurate of the major polling firms AND the most biased. A twofer!
Here's the raw data.
Positive Error numbers lean towards Romney and negative error numbers lean towards Obama. Another point is evident looking at this data. Other than the sizable 3.67 GOP-bias seen with Rasmussen, the remaining polls were remarkably unbiased, showing a slight bias towards Romney. PPP which is often thought of as a Pro-Democratic firm showed an insignificant 0.2 bias towards Obama.
No comments:
Post a Comment