Zogby: "My polls use a party weight of 39% Dem, 35% Rep, and 26% Ind"

[Hegemon]

Zogby reveals an ingredient of his secret sauce:

http://zogby.com/news/ReadNews.dbm?ID=859

Same weighting as Rasmussen, and very simliar results this year.

[Bandit3 the Worker]

Democrats outnumber Republicans by more than that.

[ATFFL]

Democrats outnumber Republicans by more than that.

Not in terms of numbers who vote and not according to Gallup and ABC, hwo both believe the Democratic advantage is down to 1% or less.

[Bandit3 the Worker]

Not in terms of numbers who vote and not according to Gallup and ABC, hwo both believe the Democratic advantage is down to 1% or less.

Gallup and ABC are wrong if they think Democrats outnumber Republicans by only 1%. In my own county Democrats outnumber Republicans by a good margin, yet Bush managed to win it by just as much.

Among people whose votes actually get counted, I have no doubt there's just as many Republicans.

[Lunar]

To put it in some context: Bandit believes Democrats outnumber Republicans in self-identity by more than +10.  In response to nationwide polls of tens of thousands showing it about even, he said it was because polling agencies are part of some right-wing conspiracy determined to get Bush elected.

And yes, those numbers are indicative of some Zogby bias.

[The Vorlon]

Same weighting as Rasmussen, and very simliar results this year.

Not true

Prior to his most recent poll, Zogby NEVER had bush ahead in any poll.

Repeat after me now...

Zogby

20 of his last 45 polls have be wrong by more than the margin of error....

Since 2000, his average reported error has be 5.71%

According the the National Council on Public Polling, in 2002 Zogby had the highest average error and lowest % of correctly forcast races of any company.

Zogby is trying to use a hard weight by party id (as well as race, age, horoscoope sign, favorite flavor of Ice Cream, etc) to save a catestrophically flawed sample design.

Take all Zogby polls and burn them

[agcatter]

Vorlon, in the article today in which Zogby explained his partisan weighting, he ended the article by reminding us once again how he nailed the 2000 election.  He specifically said he was closer than any other 2000 poster.  He also said Gallup has been "all over the place" on partisan breakout.

Other comments by Zogby

1) Bush had a good convention (dah)

2) Bush is still having negative re-elect numbers

3) Bush still has a negative job approval number.

Kerry ought to post Zogby's comments on his campaign website as a morale booster for his troops.

[The Vorlon]

Vorlon, in the article today in which Zogby explained his partisan weighting, he ended the article by reminding us once again how he nailed the 2000 election.  He specifically said he was closer than any other 2000 poster.  He also said Gallup has been "all over the place" on partisan breakout.

Other comments by Zogby

1) Bush had a good convention (dah)

2) Bush is still having negative re-elect numbers

3) Bush still has a negative job approval number.

Kerry ought to post Zogby's comments on his campaign website as a morale booster for his troops.

Point 1:

His claim he was the most accurate in 2000 is just a plain old bald faced lie.

Here is a link showing, depending if your count the Nader result, 3 polls that did better than Zogby (if you don't count Nader) or 7 polls (if you do count the Nader result) that did better than Zobgy in 2000

The link is to the National Council on Public Polling, which I assume will be deemed and acceptable source.

Zogby ranked 7th in 2000 by NCPP

In addition, Opinion Dymanics had the race tied, Celinda Lake had Gore +1, as did Greenburg Quinlan Rosen.

If the first link showing 7 firms doing better than Zogby is not enough to establish that this is just plain simply a Zogby lie, I am sure I can find links to the other 3 firms to provide 9 firms that did better.  (there are a few more too over and above the 9, DDC/Yerxa had it tied, POS had Gore +1, ...etc....)  

Zogby most accurate in 2000...?  It just ain't so

He did do well, but the most accurate..?   nope and repeating the lie don't make it truth.

Point 2:

Job approval:

Zogby askes a different question than the other pollsters, it is not approve/disapprove, it is Excellent, Good, Fair. Poor.

A lot of folks interpret "fair" as a middle value (neither approve/disapprove) hence a systemically lower result than other questions.

This question wording, also used +/- by Harris, has shown for literally decades a presidential approval rating lower than the question asked by other pollsters.

He askes an entirely different question, one that for decades has yielded a result 5% or so lower in favorability.

Zogby's Jop Approval for Bush is 48% Good/Excellent with the balance Fair or poor

Ask a different question, get a different answer

BTW, his current internet madness has bush +11 in Ohio, while down in Florida. (Bush apparently is again ahead in Tennessee however)

I have posted Zogby's track record since 2000 a few time, it speaks for it's self.

Based on this track record, judge his comments as you see fit

I will not drive up Dave's bandwidth bill by posting it again...

[agcatter]

What exactly did he mean when he said Gallup has been all over the place on his partisan breakout.  Does Gallup even use party weighting?

[Smash255]

Most of Zogby's National polls seem decent and close to what most other pollsters have, his state polls are utter & complete garbage.  Gore did win the popu;ar vote by .5%.  So you can make the argument that CBS (which had Gore 1) , Fox & harris were all tied for the mosta ccurate off by .5% (CBS .5% in Gore's favor, Gox & Harris  .5% in bush's favor)   Zogby off by 1.5% in kerry's favpr  Gallup, Pew, TIPP & ICR all off 2.5% in Bush's favor, ABC & NBC off 3.5% in Bush's favor, battleground off 5.5% in Bush's favor, rasmussen off 9.5% in bush
Bushs favor

[The Vorlon]

What exactly did he mean when he said Gallup has been all over the place on his partisan breakout.  Does Gallup even use party weighting?

Gallup uses very very minimal weights for everything.

They do not weight for party ID for example.

I agree with this approach of NOT weighting

Party ID does change slowly over time, and is also impacted by the party's get out the vote efforts, as well as recent events

To use a recent example, during the week long Reagan funeral, I think it is virtually certain that more than a few Independants and even democrats would have self identified as Republicans in a poll.

If you did a poll during Reagan week, and weighted back to a "normal" party ID, you would have been way out.

A second actual example is the 2002 midterm elections.  The GOP had a good Get out the vote effort in 2002, and in 2002 GOP voters actually outnumbered Dems at the polls by about 2%.

Zogby in 2002 continued to use a formula that assumed the Dems would have about a 4% advantage in voter turn out.

This is what happened in 2002:



If you do a hard weight to party ID, you live and/or die by it.

If you get it right, you polls should be bang on.

If you get it wrong you look like, well, zogby in 2002.

Gallup as a deliberate choice uses virtually no weighting.  The price you pay for this is you get a bit more RANDOM error (ie you naturally drift a few % either way from the truth due to statistical noise) but the Benifit you get is you do not have the possibility for STRUCTURAL error.

In 2002 Gallup has the lowest average error of any polling company because, in addition to being a damn good firm, also did not weight away the extra republicans that showed up on Election day.

IN summary,

In 2002 Zogby did his polls and likely found that in his samples that among the folks likely to vote, there were about 2% more Republicans then Dems.

Zogby then said - ah... this is wrong, I will adjust my data to the true result which is 4% more Dems - he thus shifted all his polls 5% to the Dem side.

Gallup looked at it's polls and said, Hmmm... Looks like the GOP base is pretty energized... and did nothing...

Hence Gallup's 2002 polls were dead on, and Zogby's were a joke.

[mddem2004]

Zogby's weighting he states is the same as the 2000 and 1996 election turn out by party ID.

Though he may have had a bad polling year in 2002, I for one won't completely discredit the guy. His state polls may be a bit off as well but for the national polling I don't really see him nearly as far off as say Time, Newsweek, LA Times, CBS, etc.

After all he did pick up on the over the weekend desertion that occured to Bush 3 days before the 2000 election, and there were very few firms that did catch that.

I would listen to his polls on the national level, state polling less so......

[Floridude]

What exactly was the turnout of actual voters in the 2000 presidential election, and how has registration changed since then?  

Tis a shame that polling is so difficult