Huh?
Peter Robinson ·
Sep 7, 2010 at 2:11pm
Last night I expressed my elation at Rasmussen's latest poll, which showed Republicans with a 12-point lead in generic polling, saying that any more good news might kill me. Apparently, I've been reprieved.
But there's something undeniably screwy going on here. Today Gallup may be telling us that the GOP and the Democrats are tied, but just a week ago Gallup had Republicans up by ten. One poll is an anomaly or outlier. But which one?
- Comment (9)
- · Quote
- · UnfollowFollow (1)
Comments :
Feb '10
Re: Huh?
How about we split the difference, and call it a 5-point lead, as the folks at FiveThirtyEight do?
Edited on Sep 7, 2010 at 2:25pmMay '10
Re: Huh?
Remember that all polls have a margin of error. Here's how a 4% error margin works:
Whatever the unknown, actual fraction of Americans is who support the GOP generic ballot, let's call this the true value. If you performed the same poll 20 times using a different random sample for each poll, the result will fall within 4% of that true value in 19 out of 20 polls. So in 1 out of every 20 polls the error in the result actually exceeds the margin of error. The error can be on either side of the true value.
It could be that the 10% spread in the previous poll was an anomaly, and the 47/44 from August and 46/46 from last week are closer to reality.
Jun '10
Re: Huh?
Not disputing anything in your post Mark, but could there also be error due to an over sampling of Dems versus GOP. Having read Gallup's meethodology statement nothing jumped out at me as causing bias, but I admit I don't know what constitutes "registered voters" in the US. Is it possible that based on the last election cycle there is a bias toward Dems among the registered votors? I'm not saying there is, because I just don't know, perhaps, you have an answer?.
Sep '10
Re: Huh?
Stick with Rasmussen. They are genuinely neutral and much more reliable than Gallup (whom I suspect of having a slight pro-liberal bias). During the Brown / Coakley battle in Massachusetts, Rasmussen consistently produced the most accurate polling numbers. They just strike me as being more efficient & professional.
It is still early, with two long months to go before the elections. About a week before Halloween, start monitoring the polls very closely. At some point, the numbers for the Democrats will go into free-fall. The bottom will simply drop out. I predict that, about three days before the elections, you will see Republicans take a 15-20 point lead on the generic ballot.
The Democrats are going to get crushed – that is a given. The only question is: How lumpy will the applesauce be?
May '10
Re: Huh?
Ahh, polls. Wake me when we're there.
One thing polls can't show is how many votes Democrats will secure by Chicago-style voter fraud.
Jul '10
Re: Huh?
It was the "Mom-jeans on a girlie bike" photo that did the trick.
Metrosexuals and Safety Moms ate it up.
May '10
Re: Huh?
Zing!
That one made me chuckle Kenneth :D
May '10
Re: Huh?
Aaron, I once saw a poll on a cable news channel that asked, "Will bird flu become a national pandemic?" Something like 75% of Americans answered "likely" or "very likely", but the number is not important to my point.
I said, "Humbug! That's just a poll! It has no influence on the reality of whether bird flu will become a pandemic. Get back to me when we have a real election on this question!"
Re: Huh?
Even the history of that Gallup poll is all over the place. Are other polls showing such violent lurches back and forth?