« Clinton Support (?) for Obama | Main | Breaking: John McCain Suspends Campaign to go to Washington for Bailout Debate »

Obama's Weight Problem

The ABC News/Washington Post released a poll today, which claims that Barack Obama is leading John McCain for President by 9 points. If that is so, just why is Obama so angry and short-tempered these days? A number of possibilities come to mind, but the most likely answer for me is that he knows those numbers are bunk. As Kim Priestap noted, the people running the poll weighted the responses by a 38-28 margin, with 29% listed as independent(question 901). But when they ask whether folks lean more towards the republican or democrat parties, it becomes a 54-38-07 weight (question 904). This is critical to the poll's meaning, because the 52-43 Obama lead (Question 3) was extrapolated from the "net leaned vote"; in other words 54% of the response used was from democrats, versus 38% from republicans, and only 7% from independents.

The ABC/WP poll also did not break down support by party ideology, depriving the curious a simple way to see what folks really though within internal demographics. They also neglected to break down support by any other demographics, something which further degrades any claim to validity in the poll. I used the internal supports from the Gallup and Pew polls to reverse the calculations and found an interesting revision which shows what I consider a clearer picture, but for here I will just say that the ABC News/Washington Post poll shows what kind of problem Barack Obama really faces; the reality does not match the poll projections. Obama has a serious weight problem, and I don't mean his cholesterol levels.

Every opinion poll taken this year has been weighted to show more democrats than republicans. There is some historical justification for that, but only to granting the democrats about 3 more percent of any respondent group. The weighting in polls produced by media sponsors, however, has at times been much heavier in democrat proportions, the ABC/WP poll being an egregious example. There is simply no rational basis for presuming that one party would be overwhelming in its representation. As an example, in 1984 Ronald Reagan absolutely crushed Walter Mondale, yet the republicans only represented 35% of the voters in that election. There have been minor fluctuations, but for more than three decades, party affiliation has been largely stable and predictable. There is simply no rational basis for claiming that either major party would have more than 39% of voters, or less than 32%.

Barack Obama is certainly capable of winning this election. He has a ton of money, the media is a team of Obama cheerleaders, and the public mood is desperate. But the polls released to the public do not reflect the actual level of Obama's or McCain's support, nor do they show an accurate pattern of support growth over time, or properly relay the time and source of momentum shifts and changes in support. When the polls are reverse-engineered and nominal weighting restored, it becomes clear that Barack Obama is depending on a three-tier plan for the election:

1. Barack Obama must collect at least 90% support from democrats nationwide; or
2. Democrats must make up at least 43% of the voters with present aligned support levels; or
3. Obama must collect at least 55% of independents' support with present aligned support levels.

At this time, there is no evidence that any of the three conditions exist in fact. This is a serious problem for Obama with just the election a little over a month away.

For John McCain, the following conditions may result in victory, using the Obama plan as a template:

1. John McCain must collect at least 92% support from republicans nationwide; or
2. Republicans must make up at least 37% of the voters with present aligned support levels; or
3. McCain must collect at least 55% of independents' support with present aligned support levels.

From the internal support reported by the polls, Barack Obama is presently 6 points below the 90% support level in his party, while John McCain is presently 3 points below the 92% support level in his party. Democrats historically represent 38-39% (4 below what Obama needs) of the voter pool, while republicans represent between 35-36% (1 below what McCain needs) of the voter pool. And depending on the poll, Obama enjoys between 42 and 48 percent support among independents (7 to 13 below his need), while McCain enjoys between 45 and 56 percent support among independents (he may need as much as 10 percent, or he may have what he needs right now).

There are several significant events between now and the election. The debates between Obama and McCain are yet to come, and the fallout from the financial crisis could well affect voter opinion. But right now, Obama has a weight problem, in that the election proportion of democrats in the actual election is not going to nearly match what the media is pretending it will be. While some believe the spin on these polls is meant to dismay McCain supporters and encourage Obama supporters, in the actual case this distortion could well come back and hurt Obama's campaign.


TrackBack

TrackBack URL for this entry:
/cgi-bin/mt-tb.cgi/31833.

Comments (31)

Talk about trying to make a... (Below threshold)
JFO:

Talk about trying to make a silk purse out of a sow's ear. Or to put it another way: when you put lipstick on a pig you still have a pig.

An excellent article on wishful thinking DJ.

Really, JFO? I have number... (Below threshold)
DJ Drummond:

Really, JFO? I have numbers to back up my opinion. All you have is a failed, juvenile insult as part of yours.

JFO: Could you point out e... (Below threshold)
Paul Duffau:

JFO: Could you point out exactly where DJ's reasoning is flawed?

It appears that the different weight given in the multiptude of polss do significantly affect the outcome. Knowing that, it is very possible to engineer results that you want.

Party weighting after factoring leaners at Gallup for example show 5-10 points of separation R-D with spikes of 3-5 percent in responents rates to the winner after the election.

The ABC/WaPo has a separation larger than any of the Gallup historical records I could access.

The other problem is that it looks as though they've used regitered voters rather than likely voters which iwll skew the results (historically to the left.)

Good analysis. Ano... (Below threshold)
xaix:

Good analysis.

Another point: The media sponsored polls seem to have much more in the way of wild gyrations than independents, like Rasmussen.

Bias may be one reason, but I also think the media likes the drama, so they fail to use methods which will result in a more stable horserace. Week to week, Rasmussen barely moves, regardless of what happens in the campaign. Kind of boring, but probably a truer picture of where the race stands.

Since the polls this year c... (Below threshold)
mantis:

Since the polls this year consistently show that party affiliation has stayed at the historical levels for Democrats while a good deal has shifted from Republican to independent, one wonders (again), why you think the polls are over-representing Democrats. A more logical reading would be that they are under-representing Republicans (though such a reading would still be wrong).

Read Gallup's recent article on the shift in party affiliation. According to their tracking polls, an average of 34% of Americans have identified as Democrats and 26% as Republicans, for the entire year. The most recent tracking poll breaks down the numbers at 35% of Americans identify as Democrats, 26% Republicans, and 33% as independents, only 1 point difference from the average for the year.This shows that survey respondents have consistently affiliated themselves less with both parties, but with the Republican party in lower numbers. Do you have a reason to doubt this other than historical precedent?

While it's entirely possible that as we get closer to the election (and in the election itself) the numbers will move closer to their historical norms in terms of party affiliation, the unpopularity of the President, combined with the economy (which is historically blamed on the incumbent party) give pollsters good reason to believe that the Republican affiliation numbers this year will be considerably lower than in recent elections.

Remember also, self-reported affiliation and actual voter registration are two different things.

DJ,JFO's just doin... (Below threshold)
Sheik Yur Bouty:

DJ,

JFO's just doing his thing. He gets paid by the comment, with bonuses for each insult.

mantis: Even the article t... (Below threshold)
Paul Duffau:

mantis: Even the article that you link to uses the population as a whole to get a 10 point advantage for Democrats, tightening slightly when only registered voters are considered. These are historically consistent with Gallups numbers through the years and thus more likely support DJ's argument than proof of a 16 point gap provided by ABC/WaPo.

The numbers typically change substantially when likely voters are considered.

In JFO's world: actu... (Below threshold)
Son Of The Godfather:

In JFO's world:
actual logic = "wishful thinking".

Also, JFO, I think your camp is trying to get away from the whole "calling Sarah Palin a pig" debacle. It seems it turned even more women away from The Empty Suit.

But by all means, please continue.

Btw, just discovered that M... (Below threshold)
mantis:

Btw, just discovered that Mark Blumenthal (Mystery Pollster) wrote a good piece on party affiliation last week covering a good deal of what I've been talking about here.

mantis, I have read Blument... (Below threshold)
DJ Drummond:

mantis, I have read Blumenthal's work (he's been doing it for years, after all, as have I). I simply do not agree with his contentions, not least because he has never even tried to explain why

1. polls with weights disparate from historical norms are more likely to be at variance with polls which use nominal proportions - that is, to be statistical outliers; and

2. just why it is that the historical proportions in actual elections (by definition the 'most likely' of voters, since these are people who are known to have voted) are so consistent with other years, if the choas Blumenthal embraces is otherwise in effect.

I think he - and you - confuse the issues of poll group motives for attention and client share, voter turnout, and popularity of individual candidates with the actual party identification by voters.

Very interesting analysis. ... (Below threshold)
Therese:

Very interesting analysis. This does explain the seeming "panic" in Obama's campaign. In your analysis, you point out the following:

"From the internal support reported by the polls, Barack Obama is presently 6 points below the 90% support level in his party"

This, to me, helps explain why there has been all of this media attention lately about racism. They are trying to "shame" Democrats into supporting Obama by using racial guilt as a reason. I don't believe that he is down in support in his party because of race. I really believe that it is his liberal and socialist agenda complemented by a string of questionable associations - Bill Ayers, Rev. Wright, Tony Rezko and even his wife.

Please keep up the good work!

polls with weights dispa... (Below threshold)
mantis:

polls with weights disparate from historical norms are more likely to be at variance with polls which use nominal proportions - that is, to be statistical outliers; and

That's because you apparently believe that all polls weight by party ID. Most do not. That includes Gallup and, unless they have changed things since I last checked, ABC/WaPo.

You've been arguing that Gallup should weight by party ID using historical data. They don't weight by party ID at all. You argue that Mark should explain why weights that aren't used are used. Are you sure you've read his work?

just why it is that the historical proportions in actual elections (by definition the 'most likely' of voters, since these are people who are known to have voted) are so consistent with other years, if the choas Blumenthal embraces is otherwise in effect.

Not sure what "chaos" Mark embraces, but as I've already explained, exit polls are self-reporting just like polls before the election (though of course you do know that the people actually voted), and because they are historical, pollsters would be completely distorting the data by weighting according to exit polls if there were a significant shift in party affiliation, as there seems to be this year.

Go back and look at the polls from before the 2006 or 2004 elections. Did you see the kind of disparity in party affiliation among respondents in those years that we are seeing now? You did not, and the exit poll numbers stayed in line with previous elections. You are seeing it this year. Should the pollsters ignore it? Do you have a good reason why? I certainly have reasons explaining why we're seeing it this year.

I think he - and you - confuse the issues of poll group motives for attention and client share, voter turnout, and popularity of individual candidates with the actual party identification by voters.

Huh?

mantis, you continue to ign... (Below threshold)
DJ Drummond:

mantis, you continue to ignore what I said, I am not sure becuase you do not understand, or because you try to ignore what causes you problems.

Almost all polls do weight by party affiliation. Gallup certainly does. You have either forgotten or are trying to dismiss the evidence of Gallup's own tracking poll. As I showed, only manipulation of party affiliation weighting can cause the results shown for September 7 and 14. Your denial does not improve your credibility.

IN fact, even polls which previously did not weight for party affiliation in 2004, now do so, most notably Rasmussen and Fox News.

And not least, but certainly most significant, I am surprised that you would continue to try to support the clearly circular reasoning employed in things like "dynamic weighting", used by CBS, NBC, Fox News, and others, where the party weighting is driven by the raw return from the previous month; it makes a number of assumptions, not least that a subjective guess using internal data is acceptable as an objective control. The NCPP never approved anything remotely like that practice, you know.

DJ said:"Bu... (Below threshold)
JFO:

DJ said:


"But right now, Obama has a weight problem, in that the election proportion of democrats in the actual election is not going to nearly match what the media is pretending it will be.

This is the pig and wishful thinking reference.

Got any numbers to support that assertion DJ? Or will you respond with one of your usual juvenile insults when challenged to defend a statement?

Almost all polls do weig... (Below threshold)
mantis:

Almost all polls do weight by party affiliation. Gallup certainly does.

Well, if you're going to openly lie than what good is debating? Gallup does not weight by party ID. Most public pollsters don't. Rasmussen does, which is why their results are more stable than others week to week.

As I showed, only manipulation of party affiliation weighting can cause the results shown for September 7 and 14.

You didn't show anything other than your misunderstanding of the numbers, and your propensity for dishonesty.

Your denial does not improve your credibility.

That's rich. You are willing to lie about poll methods and you talk about my credibility for calling you on it? You are hilarious.

And not least, but certainly most significant, I am surprised that you would continue to try to support the clearly circular reasoning employed in things like "dynamic weighting"

And you're willing to lie about what I say as well. I never wrote one word in support of dynamic weighting.
depp=true

bye-bye mantis. I let you ... (Below threshold)
DJ Drummond:

bye-bye mantis. I let you have your say, but your lies and arrogance are over the line.

I caught you in several lies, but gave you the chance to step back and/or correct your statement. Instead, you not only repeated the lies but also made false accusations to try to muddy the waters. I should have expected that from a Lefty, but it's classless thuggery, and I will not permit it on a thread I can address.

The rest of the article led... (Below threshold)
DJ Drummond:

The rest of the article led up to that conclusion, JFO. And I find it ironic, that you of all people would dare to demand support for an opinion, when your opinion lacks any semblance of reason, to say nothing of support in evidence.

I take your umbrage, however, as implicit confirmation that I hit the target. You and mantis have become quite exercised by the fact that Obama is not quite what he advertises, in substance nor even in support.

Great perspective and insig... (Below threshold)
Knightbrigade:

Great perspective and insight into the details of polls once again DJ.

To me National polls mean nothing, it comes down to a game of RED state-BLUE state.

Now finding what STATE polls are more objective/accurate for analysis would be very helpful.
Where this election stands on EV's at this time seems to be in flux, but at a slight Obama lead.

DJ is now disemvoweling my ... (Below threshold)
mantis:

DJ is now disemvoweling my comments for pointing out that he is blatantly lying by saying that Gallup weights by party ID, which they do not. Rather than admit his error, as an honest person would do, he is eliminating criticism in order to cover it up. We'll see what Jay Tea has to say about this.
depp=true

For the audience, mantis is... (Below threshold)
DJ Drummond:

For the audience, mantis is angry that he cannot continue to spread false accusations and lies in my thread.

I guess when he was younger, his schools did not believe in 'time out' or detention, so he's learning about it now.

2FB.

mantis - "We'll see wha... (Below threshold)
marc:

mantis - "We'll see what Jay Tea has to say about this."

Look-out JT, ready your spam filter.

Greedy for punishment aren't you mantis?

*ZAPP!*Another lefty... (Below threshold)
Son Of The Godfather:

*ZAPP!*
Another lefty made impotent through the nefarious disemvoweler. heh

Another effort by the no-lo... (Below threshold)
doug in colorado:

Another effort by the no-longer-mainstream media to discourage potential republican voters (Why should I bother to vote when we're gonna lose) and influence the weak minded (I wanna have bragging rights, and vote for the winner)...

And the lefty trolls here clearly lack even a high-school-civics-class understanding of how polls work and why weighting makes a difference...they must have had all Democratic teachers in school.

My biggest concern right no... (Below threshold)
doug in colorado:

My biggest concern right now is that with all the celebrities saying "if Obama doesn't win it's because we're racist" that when (not if) Obama actually doesn't win, there will be race riots in major cities, a la the Rodney King verdict...and this supposedly post-racial politico will actually have set race-relations in this country back about a hundred years...
Shame be upon him and his cronies for that.

doug in coloradoI'... (Below threshold)
Sheik Yur Bouty:

doug in colorado

I'm afraid Obama and his minions have set race relations back decades regardless of who wins.

OK, who's Beetlejuicing me?... (Below threshold)

OK, who's Beetlejuicing me?

(Putting on my "Main Page Editor" hat -- which NEVER fits comfortably)

I HATE when things like this happen. I like both DJ and mantis, and in fact am halfway through a rather lengthy piece that owes a great deal to mantis.

I believe that it lies within my power -- at least technically -- to reverse DJ's "disemvoweling" and restore mantis' comments. And looking at the whole thread, it's not the decision I would have made.

But I'm not going to intervene. This is DJ's article, and it is his call on how he wishes to moderate the discussion. If he had asked my opinion, I would have advised against it, and had I seen the comments before he acted, I would not have done so.

Part of the reason is the topic at hand. Both DJ and mantis have studied polling extensively, and have very strong opinions based on their studies. Personally, I don't give a rat's ass about the subject, and think they tend to cause more harm to the political process than whatever negligible good they might provide.

But that is part of the reason why, I think, Wizbang is so successful. We, as a staff, are generally complementary, but not clones.

What is in my purview is whether or not mantis will be banned from commenting at Wizbang. That will not happen, unless Kevin or I choose to do so.

Sheesh, you two... I'd love to take you both by the scruffs of your necks and whack your heads together.

Then, maybe, you'd have the same kind of headache as I get whenever I have to put on my "Main Page Editor Hat."

J.

Here's the thing. I have a... (Below threshold)
DJ Drummond:

Here's the thing. I have a reputation for using the hammer at times, and there's a reason for it.

No, it's not that I can't handle an argument. Mantis and I have been back and forth on this for days now.

No, it's not that I cannot defend my position. If you read the two previous posts on this topic, you a lot of evidence from me versus a lot less from mantis.

But ... when I write a post, the discussion should be on the topic of that post, not rehashing things from other posts. Mantis does not agree that almost all publicly-released polls weight their results by party affiliation just as they do for other demographics. Fair enough, but that discussion was not today's discussion. Today's discussion examined the ABC/WP poll in the context of its very obvious weighting, and the resulting significance on the election as various aspects of that weighting are examined. Mantis first tried to rehash old arguments, while ignoring the present topic. That's strike one and two, because he never did address the ABC poll.

But worse, when I explained to mantis why I did not accept Blumenthal's contentions, he returned to the old argument he knew (or certainly should have) was going nowhere and had nothing to do with the topic. More, he made two statements I knew to be false. I corrected him and reminded him that party affiliation weight ing is circular logic by definition (yet another fact mantis has ignored because it causes him problems in his arguments), after which he went on the attack and left all hope of addressing the topic. At that point, mantis stoppped being even a marginal contributor and instead took on the role of troll.

So, if you disagree with me, that's one thing. Make your case or vent your opinion, it;s no big whoop. But try to derail the thread, and I will have your vowels.

Besides, I am sure there are venues for spiteful rants against neocons like myself. Mantis should have no trouble finding one to suit his mood.

DJ,I remember 2002... (Below threshold)
Baggi:

DJ,

I remember 2002, 2004, and 2006 pretty vividly. It seemed like the Democrats were winning in all three elections the entire time and as always, the media was pumping for them.

So I thought in 2006 that all the polls were just as wrong as the previous years. But, like most people, I don't keep a bunch of polls sitting around my house to see how wrong they all were.

Then, after the elections are over, I see all these polls come out from before that elections showing just how accurate the pollsters were.

I guess what i'm trying to say is, do you have anything for us to show how in previous years you are right?

Put another way (I know i'm not too clear) can you show how the media has released polls that weren't weighted properly and therefore, they missed the turnout?

2004 would probably be best, since that's when Kerry was supposed to win. I even remember hearing Hugh Hewitt telling us not to panic becuase exit polls had Kerry drilling Bush. Then in 2006 Hugh Hewitt said pretty much the same thing, but where he was right in 2004, he was wrong in 2006.

Get what i'm saying? I'd love to see you do some historical comparisons here.

DJ, it look like our observ... (Below threshold)
arrowhead:

DJ, it look like our observation from the Gallup internals that independents are heavily trending towards McCain was also observed in today's LA Times Poll:
http://www.latimes.com/media/acrobat/2008-09/42527865.pdf

Obama is in trouble big time.

DJ, I looked at the Fox Pol... (Below threshold)
Pam:

DJ, I looked at the Fox Poll and thought it was pretty good, though it is of registered voters. Would you say that Obama most likely has about a three to four point lead? I still think McCain will win at the end, but it will be close; too close and I'm sick of close elections.

Barack is screwed! M... (Below threshold)
hyperbolist:

Barack is screwed! McCain wants to postpone the debates; the American people think that's a shitty idea; ergo, uh oh for Obama. Did I miss anything?

Come January 2009, your loudmouth minority will finally be sidelined and America can reclaim its stature as the leader of the world. Progress!




Advertisements









rightads.gif

beltwaybloggers.gif

insiderslogo.jpg

mba_blue.gif

Follow Wizbang

Follow Wizbang on FacebookFollow Wizbang on TwitterSubscribe to Wizbang feedWizbang Mobile

Contact

Send e-mail tips to us:

[email protected]

Fresh Links

Credits

Section Editor: Maggie Whitton

Editors: Jay Tea, Lorie Byrd, Kim Priestap, DJ Drummond, Michael Laprarie, Baron Von Ottomatic, Shawn Mallow, Rick, Dan Karipides, Michael Avitablile, Charlie Quidnunc, Steve Schippert

Emeritus: Paul, Mary Katherine Ham, Jim Addison, Alexander K. McClure, Cassy Fiano, Bill Jempty, John Stansbury, Rob Port

In Memorium: HughS

All original content copyright © 2003-2010 by Wizbang®, LLC. All rights reserved. Wizbang® is a registered service mark.

Powered by Movable Type Pro 4.361

Hosting by ServInt

Ratings on this site are powered by the Ajax Ratings Pro plugin for Movable Type.

Search on this site is powered by the FastSearch plugin for Movable Type.

Blogrolls on this site are powered by the MT-Blogroll.

Temporary site design is based on Cutline and Cutline for MT. Graphics by Apothegm Designs.

Author Login



Terms Of Service

DCMA Compliance Notice

Privacy Policy