One of the interesting things I wanted to see in this cycle was how Rasmussen covered their ass as the election got close. Obviously missing the result by 5 points would be really bad for a company that makes money on accurate results, and I consider Rasmussen nothing more than a propaganda mill for Republicans, therefore I expected to see their polling come into line with the other agencies leading up to the election.
Did it work that way? For the most part yes. Here is a snapshot of polls taken in July-August. Rasmussen is off by 5-13% from everyone else. They are either the only polling agency in America that has a good statistical sampling model, or they are grossly biased and trying to give Newscorp fodder for their misinformation campaign.
Now here is a snapshot from this morning. Miraculously Rasmussen is within 1% of almost everyone else. The outlier is Gallup which is an interesting case in itself and will either be the ONLY polling company that got it right or will be reading an apology statement on November 7th:
This analysis wouldn't be complete without a third chart showing that Rasmussen sometimes DID conform to other polls although it was generally for a short period of time. As I see it, there are only two explanations for this. Either Rasmussen is simply a bad polling company that doesn't understand statistical sampling OR (and I consider this more likely) Rasmussen manipulates their polling sample based on contemporary news stories and exposure.
Sure enough August was right before the Republican convention, offering Fox a potential nice story about how Republicans got a bump in the polls.
Comments 1-5 of 5 Last »
That does look pretty suspicious; we'll see what happens in the next few days. Gallup is still offline from the hurricane; it will be interesting to see if they too are singing a new tune. My inclination is to assume pollsters are dutifully following their own procedures and publishing whatever they come up with; they may be 'biased' in the statistical sense but are not cooking the books. That assumption may be strained if all of a sudden Rassmussen starts saying that, yep, Obama is indeed up 2-4% in Ohio...
Gallup has Romney by 1 today. Still an outlier but considerably less so than before Sandy. Make of it what you will.
Now the number are in, how did the polling organizations do? Here's a breakdown:
No surprise, but in case anyone wasn't paying attention Rasmussen and Gallup were way biased toward the republicans. PPP, YouGov, and Ipsos/Reuters were the best. The last of these was a surprise to me because Ipsos did a lot of online polling.
My premise in this thread was that Rasmussen cleaned up its numbers right before the election to avoid being horribly wrong.
Even after they did that, they still failed miserably getting almost every swing state wrong.
Gallup is the one that has a lot of explaining to do. They put a huge reputation on the line and got smacked hard.
I was surprised by Gallup though-they were wrong almost throughout. Rasmussen is a joke - no one takes them seriously well except for Dick Morris and hardcore republicans.
Then there was that Suffolk pollster who quit pollingh FL, VA and NC because they felt it was a lock for Romney. I wonder if they were actually trying to shift the narrative-quite a lot of polls put VA as Obama's and FL a toss up. NC was always close, but leaning Romney. But this one just publicly announced that all three states were locked up for Romney and so they stopped polling. Just seemed strange for a polling org to do that during the election cycle-so close to the election date.
Watch comments by email