Zogby says Newsweek poll was skewed

[KurtNYC]

I have not yet been able to get the details of Time's methodology but I have checked out Newsweek's poll. Their sample of registered voters includes 38% Republican, 31% Democrat and 31% Independent voters. If we look at the three last Presidential elections, the spread was 34% Democrats, 34% Republicans and 33% Independents (in 1992 with Ross Perot in the race); 39% Democrats, 34% Republicans, and 27% Independents in 1996; and 39% Democrats, 35% Republicans and 26% Independents in 2000. While party identification can indeed change within the electorate, there is no evidence anywhere to suggest that Democrats will only represent 31% of the total vote this year. In fact, other competitors have gone in the opposite direction. The Los Angeles Times released a poll in June of this year with 38% Democrats and only 25% Republicans. And Gallup's party identification figures have been all over the place.

http://zogby.com/news/ReadNews859.html

That's how skewing is done. Over-represent one side in the sample and don't do anything to correct for it. Seems like they wanted to validate the abhorrant tactics of the RNC. Beware - many polls are thinly disguised propaganda.

[Frodo]

1. Not "propaganda" , just unadjusted.

Most pollsters weight by party affiliation to avoid these types of swings. Others choose not to. Your post implies they had different raw data and then balanced it in Republicans' favor. Instead they just gave you their raw figures.

This isn't INTENTIONALLY over-representing Republicans any more than they were bending things the other way when Kerry was way ahead. They simply don't re balance their figures to represent expected turnout.

That isn't necessarily a bad thing as long as they report their internals by party so you can balance it yourself (as Ziggy did to show they had essentially the same results he did).

[Raya]

2. Can we confirm that Time, Newsweek NOT adjusted by weighting

Do they use a simple, unweighted, random population sample?????

[Frodo]

5. They did not weight by party affiliation

I don't know if the weight by other demographic data (gender, race, age, etc.)

Why? Well, some pollsters disagree with weighting by party because party affiliation DOES shift and turnout obviously does as well. And any weighting you pick would by nature be arbitrary.

On the other hand - NOT weighting means that polling taken when one party is more likely than another to answer a survey (like right after their convention?) can throw things off.

Obviously the "solution" that would most please me is just to give me the rawest data and let me make up my own mind what it says. But pollsters are very picky about that.

[zaj]

13. I'm looking for a source on that too...

... I'm being challenged to prove that they didn't reweight it.

[KurtNYC]

3. Of course it IS propaganda!

And what better way to cover for election theft come November??

This is in theory a democracy. Mis-representing the views of the populace to further one's political ends is manipulation of elections through media -- eg. propaganda.

And why is Frank Luntz suddenly being used by CBS, NBC and Fox even though he is a discredited RW pollster who never released his methodology for stating that 60% of the US population was in favor of the Gingrinch 1994 Contract on America?

[jab105]

6. Frodo...in this case the polls were all over the news...it was characteriz

ed as Bush running away with it because of the 11 point lead...

But, if the skew wasnt accurate, then the reporting which was all over the place, wasn't accurate either...

From what I've read, it should be balanced out, if you are at all interested in accurate polling...I could use 99% Republican and 1% democratic...and say that it shows that Bush is 85% ahead...but that wouldnt be accurate either...

[Frodo]

8. That's not what I'm saying.

If you take 1000 calls an re-weight the numbers to over represent Republicans then you ARE creating "inaccurate polling" as you say.

But if you DON'T reweigh the numbers at all and there just HAPPEN to be a different number claiming to be Democrats than you expected THAT could simply be changing dynamics of the race. You can't possibly know.

Some say (and I agree) that balancing to a presumes turnout is the correct methodology. But we're just as likely to be wrong. Because sure as I'm sitting here, I'm likely to be wrong about the actual turnout.

People keep talking about the "correct" weighting. Except that previous years have been well off of that. Democrats claim (and with good reason based on the ABB feelings) that THEY will have better turnout than in 2000. Republicans claim (also with good reason based on 2002 turnout and their "72 hr project") that THEY will have better turnout this year. What SCIENTIFIC basis can there be for making a prediction? Those who do NOT re weight are just saying they are letting the polling dictate where that weighting will be (perhaps by their "likely voter" determination).

In any case. There's nothing "wrong" with it. As Rassmussen said, almost ALL of the polling data lately says the same thing. Just in different ways.

[jab105]

11. Fine...but its questionable at best considering

what the percentages have been in recent presidential elections...

That's where my problem is, because this 11 point lead was all over the news, and there was no asterick for the weighting of more republicans...if there had been, I would not have had a problem...

In America, it is all about perception, and right now, the perception is that Kerry is in an 11 point hole based on accurate polling...AMerica likes a "winner", and perception is everything...

[The Magistrate]

14. Still, Mr. Frodo

In most polls, there are questions asked of the potential respondent before the poll questions are asked, to determine if the person is within the demographic desired: in a marketing poll intended to learn about motivations of people who order beer at the bar, for instance, you will be asked your drinking habits, and you will be bid farewell if you respond that you purchase mostly from liquor stores. It is not as if the random sample is simply the first thousand people who answer the telephone and express a willingness to answer the question. Thus, the building of such a skewed sample as that present in this poll suggests that a predetermined quota of party affiliations was guiding the assembly of the cohort of persons to be asked the poll questions.

"Figures lie, and liars figure."

"LET'S GO GET THOSE BUSH BASTARDS!"

[Frodo]

15. Good afternoon Mr Magistrate!

True. But a person's demographic information does not change (well, "age" changes, but not enough to affect the polling - same with marital status).

So I'm sure they DO weight by that. But party affiliation can certainly change.


Just search back for the posts about the LATimes poll that showed Kerry so far up. The posts were not "nuts, that's really no gain at all since the poll includes far more Democrats than last time." They were all about how it was a reflection of the changing electorate. That when people decided to vote for Kerry instead of Bush they start considering themselves Democrats instead of Republicans.

And it's an interesting question to ponder. With Bush being a little more popular with Republicans than Kerry is with Democrats... and with a virtual tie among independants... the election really comes down (as it almost always does) to turnout. HOW should we measure this critical factor if the figure that the polls are being "normed" to is predetermined? Wouldn't the ONE thing you would NOT want to adjust be likely turnout?

I don't know what's going to happen, I just know that taking the 2000 turnout as gospel is a mistake. My calendar doesn't SAY "2000".

[The Magistrate]

16. That Is True Enough, Sir

In dealing with polls, people really ought to adopt something like the average scoring in those quasi-sports events like figure skating, where both the one highest and the one lowest scores given by panel members are tossed out before the average is figured.

My own view, as an old card-shark and nothing more, is that the result of this rate will be about a five point margin of victory for Sen. Kerry, and that there has been, and will be, very little movement among the people from the start of this campaign back last winter, through to its end this fall, on the question.

There does seem to me, though, a real indicator of higher than normal Democratic turn-out this year, namely, the tremendously increased number of small donations this year to the Party's Presidential campaign, and to other groups supporting same. People sufficiently excited to part with a small amount of cash will certainly vote, and if there is a greater number of persons excited to that pitch, it seems reasonable to suppose there is greater number of persons excited to a greater level of intensity than usual. The Republican Party, on the other hand, is not getting as many small contributions as it has usually received this year.

"Democracy is a form of government based on the belief the people know what they want and ought to get it, good and hard."

"LET'S GO GET THOSE BUSH BASTARDS!"

[Frodo]

17. I agree. With ONE correction.

"My own view, as an old card-shark and nothing more, is that the result of this rate will be about a five point margin of victory for Sen. Kerry, and that there has been, and will be, very little movement among the people from the start of this campaign back last winter, through to its end this fall, on the question."


That would be President Kerry then. :-)


Have a good evening!

[WiseMen]

4. Methodology of Time, Newsweek poll is extremely important. Very hard

to believe they don't use statistical weighting to adjust for
a lack of representation in the sample.

It is impossible to use a pure random sample when wants to obtain
larger subsamples for reliable results from subgroup analysis --
such as military families.

It they used a simple random sample, all kinds of bias related
to the interview days and times can be postulated.

[Snellius]

7. The suspicious stat on that poll: almost half were "military families"

Something like 48%. That particular poll was so skewed and came at such a politically critical time, there should be an investigation into whether it was fixed.

[hurley]

9. hmmmmm

[Nicholas_J]

10. The media is

PLaying this one wonderfully. They have papers to sell, and month after month of neck and neck polls just isnt news. In fact it seems odd that the polls done by or media outlets have Bush with double digit leads, while more independent pollsters have the results you would expect. A small boost for Bush after his convention.

Now the same media outlets have just released "New Records" regarding Bush's military duty. Lets see if a month from now, we dont see a sudden spike for Kerry, leading by not a nose, but two lengths.

This all is vaguely reminiscent of the media polls of Howard Deans lead on all of the other candidates, and the very ,very large soapbox Dean was given for several months before the primaries. Polls done by independent pollsters did not show anything like the leads Dean was being given in polls done directly by media outlets, or for them.
And in the last few weeks of the race, the polls started changing slightly, and in the last week, changing a great deal more.

This seems to me to be a case of the news media creating its own public anxiety, tension. Polls done by more independent pollsters, like Zogby, ARG, Rasmussen, seem to show far less of a lead by euther of the candidates duyring this entire campaign. Probably most accurate, most dull, and most un-newsworthy.

It seems they are creating their own horserace to keep people watching tv, and buying magazines and newspapers.

[sofa king]

12. Thank you, DU.

As of late, I've been whining a lot about these polls and how virtually everyone is bushitting us with that "likely voter" tripe.

As I have said before, I think that both the press and the pollsters have a vested interest in keeping up the appearance of close-ness.

Since Newsweek is permitted to blindly speculate, I figure I am, too. You want to know my guess as to why that sample was skewed so heavily in favor of "likely" Republican voters? (I mean above and beyond the interest of the magazine in selling something new and interesting and challenging.)

Because the Republicans are slinging as much shit as possible in order to keep Americans away from the polls. Negativity turns off undecided and independent voters, while it minimally affects their own base.

I think that Newsweek, having viewed the vitriol of the Convention and knowing it's a mere taste of the vile tactics to come, is going on the premise that non-Bush supporters are going to stay away in droves on election day thanks to a flood of Republican smear-and-fear.

That's not how I see things where I live, but it's a big country.

[Generator]

18. Can we have this bookmarked for

the next time DU goes apeshit over a poll that everyone's gut knows is WRONG?

Thinly disguised propaganda. Well there's a surprise. The media wouldn't want to taint an election or anything.

[yardwork]

19. Take all polls with a huge cube of salt!

Polls are not real news. They are efforts to create news. Even the best-designed and conducted polls are essentially an educated guess at creating a representative model.

Please stop freaking out or celebrating everytime a new poll comes out. It's pointless. Please

[a new day]

20. poll adjustments

In 2000, in the actual election results, the breakdown of registered voters was : 39% Democrats, 35% Republicans and 26% independent. I.e., these were the actual voters, as in not only likely to vote, but DID.

So how do the corporate pollsters get away with their new "formulas" for "adjusting" poll results in which the number of "likely" voters is adjusted DOWN by several points for Democrats?!