Election Polls

[landrun]

I've spent some time tonight reading up on the 2012 election polls.

In short, it seems republicans this year are saying the polls are slanted to the democrats because democrats are being over sampled in the polls. Sometimes by over 10%.
Democrats complained about this in the previous elections. Now its the Republicans' turn.

I'm not sure how this works - or should work. But here's the scoop...

It seems that the vast majority of polls right now are using models from the 2008 election - where 39-41% of the population identified themselves as democrats while 32-34% identified themselves as republicans. Hence, the higher number of democrats in polls we see right now.

But, Repubicans say to be accurate you have to use how people currently identify themselves.
Rasmussen keeps a monthly tracking of this on their site.
here --> http://www.rasmussenreports.com/public_content/archive/mood_of_america_archive/partisan_trends/summary_of_party_affiliation

In Nov 2008, it was Dems at 41% and Repubs at 33%. +8 for the dems

Today it is Dems at 33% and Repubs at 37%. +4 for the repubs

In 2010, where Repubs had historic gains, it was Dems 34% and Repubs at 36% -- only +2 for the repubs

So, since 2008, the dems have dropped 8% and the Repubs have gained 4%

And the repubs are even better now than they were in 2010.

In 2008

Rasmussen found that the partisan make-up was about 40% Democrat, 33% Republican, and 26% Independent. That was very close to the actual turnout of 39% Democrat, 32% Republican, and 29% Independent. (See link below on about.com)

So, the argument is that if anything, the polls should over sample the republicans by 4% or so rather than the dems by 8% or so.

Sounds logical I guess.

But... to me, you should just decide to call XXX number of people (once they determine their needed sample size) and then the number of dems/repubs just fall where they may. But it doesn't look like this is how it is done. I'm no statistician so I can't say how it should be done. But it does seem a little off to me that you would intentionally sample more democrats then republicans.

Here's a couple of interesting links I've found regarding this:
http://nationaljournal.com/2012-presidential-campaign/republicans-to-pollsters-too-many-democrats-in-your-surveys-20120925
http://usconservatives.about.com/od/campaignselections/a/A-Look-At-Voter-Turnout-And-Enthusiasm.htm

So, my question for my fellow Sooner fans, left and right included, is what do you think? Are the poll a little off right now in favor of the dems? :pop:

[SouthCarolinaSooner]

But... to me, you should just decide to call XXX number of people (once they determine their needed sample size) and then the number of dems/repubs just fall where they may. But it doesn't look like this is how it is done. I'm no statistician so I can't say how it should be done. But it does seem a little off to me that you would intentionally sample more democrats then republicans.

Here's a couple of interesting links I've found regarding this:
http://nationaljournal.com/2012-presidential-campaign/republicans-to-pollsters-too-many-democrats-in-your-surveys-20120925
http://usconservatives.about.com/od/campaignselections/a/A-Look-At-Voter-Turnout-And-Enthusiasm.htm

So, my question for my fellow Sooner fans, left and right included, is what do you think? Are the poll a little off right now in favor of the dems? :pop:
From someone with a bit of experience in polling, I personally don't agree with weighting by party identification, and the poll I've worked for does not weight by party identification. I think the biggest reason is that party identification can be affected by question ordering, not to mention it doesn't account for those who may cross party lines. Its an unnecessary weight for those reasons, at least from what I've been taught in survey research. Things like race, gender and age need to be weighted in polling...not party identification or partisanship.

[diverdog]

I've spent some time tonight reading up on the 2012 election polls.

In short, it seems republicans this year are saying the polls are slanted to the democrats because democrats are being over sampled in the polls. Sometimes by over 10%.
Democrats complained about this in the previous elections. Now its the Republicans' turn.

I'm not sure how this works - or should work. But here's the scoop...

It seems that the vast majority of polls right now are using models from the 2008 election - where 39-41% of the population identified themselves as democrats while 32-34% identified themselves as republicans. Hence, the higher number of democrats in polls we see right now.

But, Repubicans say to be accurate you have to use how people currently identify themselves.
Rasmussen keeps a monthly tracking of this on their site.
here --> http://www.rasmussenreports.com/public_content/archive/mood_of_america_archive/partisan_trends/summary_of_party_affiliation

In Nov 2008, it was Dems at 41% and Repubs at 33%. +8 for the dems

Today it is Dems at 33% and Repubs at 37%. +4 for the repubs

In 2010, where Repubs had historic gains, it was Dems 34% and Repubs at 36% -- only +2 for the repubs

So, since 2008, the dems have dropped 8% and the Repubs have gained 4%

And the repubs are even better now than they were in 2010.

In 2008


So, the argument is that if anything, the polls should over sample the republicans by 4% or so rather than the dems by 8% or so.

Sounds logical I guess.

But... to me, you should just decide to call XXX number of people (once they determine their needed sample size) and then the number of dems/repubs just fall where they may. But it doesn't look like this is how it is done. I'm no statistician so I can't say how it should be done. But it does seem a little off to me that you would intentionally sample more democrats then republicans.

Here's a couple of interesting links I've found regarding this:
http://nationaljournal.com/2012-presidential-campaign/republicans-to-pollsters-too-many-democrats-in-your-surveys-20120925
http://usconservatives.about.com/od/campaignselections/a/A-Look-At-Voter-Turnout-And-Enthusiasm.htm

So, my question for my fellow Sooner fans, left and right included, is what do you think? Are the poll a little off right now in favor of the dems? :pop:

Landrun:

I look at the weighted average of all the polls. So far it has been pretty spot on in predicting election results. If you look at one or two polls then your are going to get slightly slanted results.

[rock on sooner]

I follow polls pretty closely. It is interesting that national polls
have the race closer than the battleground states. Does anyone
know if national polls are weighted differently or conducted differently
than the battleground states polls?