Monday, October 20, 2008

Yahoo, Indeed

Last week, the Associated Press once again proved they are cheerleaders for Obama, rather than honest reporters of fact. This is illustrated in this case by the AP/Yahoo poll released, touted as evidence of Obama’s continued dominance in the election, when in fact the numbers from that poll confirm the tightening race as it nears the end. Of course, you have to go looking for the link to the actual poll detail, and then you have to dig through and find the relevant results. But there are some interesting details indeed in this poll, details which should be considered in weighing the claims made by the Associated P.

Starting with the big question, who folks will vote for, the poll in question VOT3B tells us that Obama gets 42% of the support, and McCain gets 39%. Another 4 percent want someone else and 15% say they are undecided, which tracks with what I have been seeing in other polls. This tells us two very important things. First, 42 percent is not going to win, so both Obama and McCain have a lot of work to do, no matter what the media says. And with only 3 points separating them, the 15% undecided pool means that either candidate could reach a clear majority, or fail miserably. If you want to know why Obama is stressing, I bet it’s because his own private polls are warning him about similar conditions.

Now, I generally don’t like polls which press ‘leaners’ to pick a candidate, since we cannot say how sure they are about someone if their first statement was ‘I don’t know’. But the AP/Yahoo poll notes that when pressed, more went to McCain than to Obama (question VOT3BA).

The poll gets hinky when AP/Yahoo asks again who the respondents prefer in question PRES. The response from “all adults” is 44-42 Obama, but the “likely voters” are posted as 49-44 Obama. The problem here, is that nowhere in the poll does the group explain how they defined “likely voters”, nor are there any questions which clearly define the sub-group, nor do we even know how many respondents are classified as “likely voters”. The top of the poll identifies “1,769 adults; 1,528 registered voters. 873 Democrats; 650 Republicans”, but that’s it. Since we have seen a 3 point separation in the base question, and we know undecideds broke for McCain, it gets a little weird for AP/Yahoo to go from there to expand Obama’s lead to 5, and since they never explain how they got there, the reader would do well to ask why they hid that part of the methodology.


LV5 asks if the person is registered to vote. 86% say yes, so that tells you that unless the question was only asked of registered voters, one of every seven answers is irrelevant.

LV6 asks how often the person votes. 74% say always or nearly always, so another one in eight respondents (and one in seven registered voters) are not sure if they will vote this year, on the basis of their admitted practice.

LV31. goes on to ask if the respondent is “certain to vote” using a 10-point scale. Only 66%, about two-thirds of the respondents, say they are certain to vote. So we must be even more skeptical about the quality of this respondent pool.

VOT3BB asks if the respondent might change their mind, and 14% say they may change their mind.

PID1 notes that the party affiliation split for the poll was 40% Democrat, 27% Republican, 21% Independent, and 12% undecided or other, which actually creates a DRI split of 40/27/33, as usual for this year far out of balance with historical norms.

PARTYID goes further, noting that 49% of the respondents consider themselves Democrats, 37% Republicans, and just 14% other or Independent, a contrast to PID1 which is not explained.

Looking at the end-poll demographics, more information is revealed which helps us see the bias. 22% of respondents are identified as 18-29 in age, versus 17% in that age group’s actual voting in 2004. In this poll, 13% of the respondents did not complete High School, versus only 4% of voters in 2004 who did not complete High School. 31% of the poll respondents have a High School diploma as their highest education, versus 22% of voters in 2004 in that category. The overweight is obvious.

Whites in the poll made up 69% of the pool, versus 77% of voters in 2004. Hispanics in the pool counted for 13%, versus 7% who voted in 2004.

In the poll, 84% of respondents live in urban areas, versus 30% of actual voters in 2004.

In the poll, a staggering 41% of respondents do not have a job.

And finally, 58% of respondents to the poll make $50,000 a year or less, versus 45% of the actual voters in 2004.

This poll is biased in six distinctly invalid ways relative to known demographics, and even then the AP/Yahoo poll admits the race is close, though they do everything they can to hide that conclusion in the press release.

Critics may claim that this is just one poll, to which I invite you to suggest another poll with complete internal demographic data. I will indeed be looking for just such polls, and you can count on a full dissection and report. For here, I consider this poll to be an intriguing example of how spin and deceit are hiding the true state of the race --- that is, unless you expect all of the behaviors of past elections to be completely abandoned this year.

12 comments:

Lea said...

Wow, they really skewed those numbers. Excellent analysis, as always. Thanks so much.

David Marcoe said...

Interesting. I would appreciate your thoughts on Politico's article: "Obama leads among white independents"; politico.com/news/stories/1008/14683.html

DJ Drummond said...

david, the first thing that occurred to me when I read that article, is "white independents WHERE?" The second is how OLD? You get the idea

One problem with some of the parsing, is that context gets lost. I was initially surprised to see how different this year's polls are, from those of the past in terms of methodology. Even Gallup has abandoned the known historical values in favor of something, well, invalid in statistical terms. I'm sure you noticed that they re-introduced the historical measure from likely voters, calling it 'traditional', because they have had to admit - to themselves, though never in public - that the 'expanded model' they created for this year is dysfunctional.

The polls appear to have bought into a foolish circular logic - Obama is a "game changer", so that means the old weights do not apply, so they over-weight younger voters, urban voters, and assume that newly registered voters from the primary season will vote in much greater proportion than the known historical norms. So when they over-weight these vectors, the polls show big gains for Obama and that makes the pollsters believe they were right.

Until ...

Until someone tests the models for collinearity and finds out they do not work in a controlled setting.

And that's where we are now. The polls are invalid and they know it, but they have to scramble these last two weeks to make the final numbers come out close to the actual election, but they cannot get a handle on where that will actually end up, because of the undecideds.

David Marcoe said...

And yet, McCain is still steadily gaining (though some of that is probably just poll corrections), even with the shenanigans. Throw in B.E. (which, given widespread anecdotal info--as good as anything this year--appears very real, though not race-based), compounded with the PUMA defection, and McCain may actually be ahead. At least, that's what I hope.

Anonymous said...

DJ,

What are your thoughts about how

1. inclement weather can affect voting on election day (first look at the WAY extended forecast has a significant snow storm for the midwest (OH, MI, PA, WI))

2. state initiatives and referendums can increase the conservative base?
Florida - sanctity of marriage
Missouri - immigration/affirmative action
Colorado - abortion rights
Nevada - property tax cap

DJ Drummond said...

jeff, those are good questions which deserve their own post. That, and I have to go get my daughter, help her with her homework, then feed the dogs, take out the trash, then finish and submit a case that's due today for school.

Joshua said...

DJ,

I've quoted and linked you at GateWay Pundit OT: under his Obama - Grandmother post.

http://gatewaypundit.blogspot.com/2008/10/barack-obama-will-visit-white.html

Great work. I know it takes time to go thru the bogus junk. I've done this in the past. And you end up getting very disgusted with the Deception that is taking place.

This is why We the People all Must Vote.

Thanks for this work and post. Keeping them honest! I think big Media companies just don't get it. The American people are going around them now. But we have to work together to spread the truth.

commoncents said...

Great post!

Would you like a Link Exchange with our new blog COMMON CENTS where we blog about the issues of the day...

http://www.commoncts.blogspot.com

Eugene J. Mascoli said...

I wonder if the pollsters are basing some of their conclusions on the primaries. I remember Operation Chaos had a sizable number of Republicans switching to Democrats to vote for Hillary. Any chance then that pollsters are overcounting Democrats for their models?

Also Barone seems to think recently that the wide gap in party affiliation is real. Any thoughts?

Anonymous said...

dj,

Good job with the internals.

Does anyone know where the Dems' reputed 10,000,000 new registered voters came from and when?

I'm assuming ACORN has been in the game for the last four years but all we hear about is the number of voters they registered this year.

Second it would be good to know, if possible, how many of those 10,000,000 voters were motivated to vote because of the possibility of a woman in the White House--Hillary--as opposed to a generic male democrat, black or white?

Anonymous said...

"In the poll, 84% of respondents live in urban areas, versus 30% of actual voters in 2004."

Well, that's completely ludicrous. Forget everything else. Throw out the bitter God-n-gun clingers and of course Obama leads.

Biased Girl said...

I have been beating my head against a wall trying to calm down my friends every time one of these polls comes out. I will share your analysis as further proof that the only poll that matters is on Novemeber 4th.