Skip to comments.Survey of Polls 10/2 Romney increasing lead
Posted on 10/02/2012 2:07:13 PM PDT by Leto
Party ID and the weighting of poll samples has been a big topic of discussion in the blogs and even in the some of the old media recently. It is a good thing to shine light on the shoddy and indefensible polling practices, but the light has to get even brighter before it will cause a shift with the pollsters to become more than another front in the election campaign. They need to be systematically held up to ridicule when they engage in practices that cannot stand up to intellectual scrutiny. There should be something in polling akin to what a high school sophomore or junior has to learn to do when writing an argument with justification and supporting footnotes or when a math teacher requires a student to show their work. Writing "81" is not an answer. In the same way that Obama 48, Romney 46 is not an acceptable "result" without fairly good documentation and justification for any judgment calls. I wish RealClearPolitics had some clear standards for polls and more systematic analysis of the polls they use.
However it's not just party ID that needs more light, but also other dimensions of the internal structuring of polls. Party ID and other internal categories are for all practical purposes, independent of each other. Actually , they are all mutually independent variables because a person can change one response without changing anything else. Changing the Party ID will cause a change in the outcome, but by itself does not cause a shift in the other internals. When a pollste decides to change the ID% of D's nothing needs to happen as far as the other internals of a poll. The % of D's voting for the D candidate will be influenced by other factors and not the absolute number of D's. That is why there are two different sets of "skew" numbers above. You can see this effect in the changes in the ID skew and the polling skew - ID can change without changing the polling skew and vice versa. I use the term "polling skew" to lump all other variables together as one category separate from Party ID.
Party ID is only part of the problem and getting the ID right may not completely fix a bias problem. Theoretically they can can both be very large in opposite directions and offset each other. Focusing solely on correcting for Party ID alone does not give a complete picture and may give an inaccurate picture of what is going on within a poll. However the nature of public polling today is that polls are biased in one direction and the focus is on party ID because it is so easily "weighted" and the egregious use of it to push polls in a preferred direction, but there are other more subtle ways as well.
Looking at the result from 11,000+ respondents to all these polls:
demRATS: 90.54% Obama, 7.07% Romney, 2.4% others/undecided
Republicans: 4.14% Obama, 93.01% Romney, 2.86% Undecided/Other
Independents: 37.97 Obama, 44.17% Romney, 17.86% Undecided/Other
The latest party affiliation numbers (updates today) show a +.13% R affiliation relative to the dems.
When all these polls are corrected to the current party affiliation the numbers are:
Romney 48.51% Obama 44.86% !!! Undecided/Other 6.63%
Some observations about these numbers:
Things to note Romney has a 6.2% lead among independents. The % of undecided/other among independents is 17.86%.
Looking at the undecideds/Other, typically 1-2% of voters go to 3rd parties. If we take the 6.63%. lets 'allocate' 1.63% to third parties that leaves 5% up for grabs. If the voters break as they typically do they should go 4% to 1% for Romney that would leave is with Romney winning 52.5% to 46%.
If Romney and Ryan show they are an acceptable alternative this is the result I would expect. I don't think the recent revelations about Libya and the administration lies will not help Obama.
Leto, many thanks for posting threads that are informative and relevant and a good alternative to the “romney sucks and is behind by 15 points” nonsense.
Wow, very close to what I predicted by rebalancing the Quinnipiac poll this morning.
Maybe I should go into business!
I don't know if that is a result of his methodology, or if it is actually validating his methodology. I need to study it more closely.
I know: Dead horse beaten and rotting at this point... But WHO in their right mind could be “undecided” at this point. Good grief...
This is the most NO BRAINER election we have had since... Well, since Carter... (Another whack at the decaying horse...)
Go Mitt...Go, Mitt; Go Debunk the liberal shtuff !!
That was another election where the claim was the incumbent was going to win and I knew otherwise.
The polls are very consistent with each other each candidate get 90% of their base and Romney leads in ALL these polls among independents.
That is the dirty little secret the only real variable among these polls is the distrubution of D/R/I in their sample.
THAT”S IT!!!!!!!!! NOTHING MORE.
One thing this blogger does that unskewed polls doesn’t do is update his party affiliation numbers when new data becomes available.
I would note that Romney went from 55+ after the pub convention to as low as a +0.86% lead on 9/26. So the number is bouncing up currently.
OOPs meant 5.5%+ not 55 of course.
I just did the Glenn Beck Show with GOP pollster Kellyanne Conway. She explained many of the weaknesses of these polls beyond merely party sampling. Among other things, public (as opposed to private internal pollsters like here) polling firms tend to give up easily, after one call, on a demographic (whites, for ex) and that they ar all in urban areas and hugely OVERSAMPLE city dwellers ( Ds, even if they say they are Indies)
Agreed LS, I would also think that one of the things that drive me crazy here at FR is people attacking the Romney campaign for not doing this or that due to the polls that say Obama has a lead.
They don’t consider that Romney might be behind and not ahead and conducting his campaign on that basis.
As you point out the internal polls are very different than the Obamedia polls.