Skip to comments.(Rasmussen) Summary of Party Affiliation (Sep 30: R-36.8; D-34.2; Other-29.0)
Posted on 10/10/2012 8:12:40 AM PDT by xzins
click here to read article
Rasmussen’s internals *are* posted in tatown’s link to my post.
There is no way the split can be R+2 and independents going 16 points for Romney, and the race would still be only at a 2 point advantage. There is a significant D bias.
He is not using the party affliation numbers for his national polling. I don’t know why, but he’s simply not.
You are 100% correct!
That’s because you’re confusing a base model with an adjustment model.
If the base model of a car dealer says he needs 5 Mustang Shelby GTs, but his local adjustment model says he needs 2.7 more, then how many Shelbys does he need?
Correct. Romney leads in indep voters by 10 points, translates to about 3 pt lead if indies are 1/3rd of the voters.
If we indeed get R-D of 2.5 pts, then a 5-6 pt lead for Romney total.
IBD/TIPP today have Romney up 5 points. We may be on track to that kind of result. So Romney will get around 310 EVs.
And I disagree at the point that it says he doesn’t use his own data. He can’t help using it.
If I have, for example, a base model of R 3 D 5 and O 2 = 10
and I combine it with my adjustment model of R 5 D 3 and 0 2 = 10
Then I end up with R 8 D 8 0 4 = 20 = 20/2 = 10
It is then from that adjusted base that I begin interpreting my polls with any other special sauce that I think is important (like attentiveness, likeability, job approval, etc.)
And I can have some percentage of dems dismayed more than reps, and I can have some percentage of indies more rep than dem, and NOT blow the numbers out of the water because it is a difference I’m dealing with and not a raw number. You can add, for example, a .05 difference between O support of R or D, to the Rep candidate and bump his number just a tad by “2*.05 = .1” to 5.1 and cut the 0 to 1.9. (It seems you could also split the difference bumping up the R and lowering the D, eacgh by 1/2 the difference, but I’m not sure which makes more sense. In any case, you’ve effectively placed leaning O’s with their candidate, and lowered the numbers of the Other category.)
So, what I understand Rasmussen to be doing is presenting a history adjusted by a contemporary that might be further adjusted by considerations he thinks are important.
If we survey 100 people and ask “Will you vote for Romney or Obama?” and we get 66% who say Obama and and 33% who say Romney and 1% who can’t make up their minds, we don’t accept that as an accurate predictor of a vote.
I think we’d toss out first those who aren’t going to vote as best as we can tell. If that leaves us with 75 people, then we’ve got a bit better picture, but it’s still not a balanced view of America Then we’d want to make sure we had enough women, enough minorities, enough old people, enough rural dwellers, etc. Say that process knocks us down to 65 people.
If we now impose a pre-determined, adjusted formula on that body that is, for example, 40/40/20, then to reach that we might go down to 55 people, but we’d have an n = 22/22/11 when we finally starting counting Obama, Romney votes.
If we end up with Romney 49% and Obama 48% and 3% O/U, have we not already made adjustments with our pre-determined, all-inclusive formula?
I’m no professional pollster, but this is the way I think it works from what I’ve read and studied over the years.
I want to say I’m open for correction, and I am, but I’m refusing to believe that Rasmussen does not use those Party affiliation numbers he spends so much money and time and resources gathering. A 3 month running sample of 15000 people is not easy, and especially when updated about every 3 weeks.
It doesn’t matter what you believe, the fact is, is that he doesn’t use his own data. Scott has said so himself. He believes that democrats will outnumber republicans by 2%-5% on election day.
Again you don’t have to be a ‘pollster’ to understand second grade math:
If R is winning a larger percentage of the “D” vote than O is winning of the “GOP” vote AND R is winning the “Ind” vote by 10 points then it is IMPOSSIBLE for Rasmussen to be oversampling the “GOP” and still have the race tied. If you cannot understand that extremely basic premise then we are wasting our time trying to educate you. I would have an easier time trying to teach my cat how to play the guitar.
Like a lot of other pollsters, Rasmussen tinkered with his likely voter "model," making it progressively tougher as the campaign progressed (adding other questions to the mix such as political interest). The different models allowed the ultimate party mix to vary from the initial weight target. As Rasmussen explains it:
By Election Day , our baseline was still 35-39-26 but our Likely Voters sample had just over 36% R and just under 38% D. If we [had] adjusted to 37-37-26, we [would have] nailed the actual election results even more closely (our final projection before Election 2004 was within half a point of each candidate's actual total). http://www.mysterypollster.com/main/2006/04/index.html
So, this says very clearly in Rasmussen's own words that he adjusts the base model by some formulaic considerations.
By the way, it’s not 2d grade math. I’ve already shown that placing the differences in a formula applied to the base model after the data is balanced for general demographic concerns removes your concern.
Lets say you had a dozen previous elections with an average of what percent voted Republican, Democrat, and other. It represented the history of these things.
Look at it like the baseball card of presidential elections.
Now we have THIS season. And this Season your guy is batting .315 but his baseball card says he is a lifetime .287 hitter. Based on that what kind of hitter would you say he is? And if you had to arrive at a number, how would you do it? Would you just average it in or would you add other considerations such as history of past injuries, steroid use, new team with new hitting coach, etc.
You are right, it’s not second grade math. It’s probably closer to first grade. Yes, I’m being serious.
Without a doubt I’ve wasted enough time on this subject matter. You obvious cannot understand the basic premise of modeling or statistics. Rasmussen, just like every single pollster that is currently publishing polls for the 2012 GE is oversampling democrats. FACT!
BTW, I’m an engineer so math comes pretty darned easy for me.
"By Election Day , our baseline was still 35-39-26 but our Likely Voters sample had just over 36% R and just under 38% D. If we [had] adjusted to 37-37-26, we [would have] nailed the actual election."
Rasmussen is SAYING that he adjusts his baseline with his LV sample. What more do you really need than Rasmussen's own words?