Webb By 8 - I Don't Buy It

News7 has a poll out today claiming Jim Webb has an 8 point lead over Senator George Allen in Virginia.

Here's why I'm not buying it. In fact, someone should look into these ridiculous polls being conducted so close to an election. Another recent poll by the same station and Survey USA had Allen by 3. If that link goes to today's poll, use the drop down window at left on the page and select 10/25/2006 Election Poll #10547.The key item to look at in the internals is Party ID for that poll.

R - 42%       D - 35%        I - 21%

Fast forward to the internals for today's poll:

R - 35%       D - 38%        I - 24%

There's no integrity to this poll and no one is going to convince me that party ID in Virginia changed dramatically in a week or so. They're just dialing the phone until they get 700 or so people without any real control to provide a sample representative of the voters on the ground.

Now go back to Virginia in 2004 and here is the Party ID for Virginia voters from exit polling - you know, where people actually show up at a voting booth.

R - 39%       D - 35%        I - 26%

All things being equal, if voter trends hold from 2004 Allen is possibly up by a few, at worst the race is a dead heat. You get that simply by adjusting the party ID close to something which has been shown to be relevant to an actual election ... instead of an attention grabbing headline declaring an alleged Webb surge.

Comments

[dc]

Dan,

I had already wadded that one up and thrown it in the trash. The breakdown of the internals puts it in context. Thanks.

[jaydawg]

Party affiliation is not a static number. People tend to associate with the party they are voting for in a particular election. Hence the trouble for Dixie's own George Allen. The poll's legit....

[John]

Another problem with the poll is that its sample is 53% female. Or is someone going to claim that gender is not static, either?

[Dan Riehl]

Party affiliation is not a static number.

Maybe, maybe not - depends on the respondent. Fact is there is damn little science involved in most of these media polls. Your argument would hold that it's a dependent variable, clearly it isn't. SO, if it's meaningless, why publish it at all?

It could show a skewed poll, particuarly as it was a weekender, as a trend for Webb. That renders it worthless as predictive, especially when three other major polls contradict it. But, hey, enjoy the kool aide. It's a close race and that's about all anyone knows.

[mehes]

The poll does overweight Democrats, but that just makes it not 100% accurate. It doesn't make it meaningless. Take a look at the breakdown of votes within each party affiliation. If the poll did not overweight Democrats, it wouldn't show a tie, let alone Allen leading by a few. It would show Webb leading by 5, which still makes him a huge favorite.

[Chris S.]

The poll is questionable as it polled 1000 'adults', of which nearly 20% cannot vote. Not particularly valuable regardless of your side of the isle.

[Chris S.]

or aisle...

[longtom]

yup, the old "polls are wrong" excuse.

[mehes]

You can find discrepancies between every poll and the representative population. That's obvious. Samples can never be completely representative of a whole population. That's why they're just samples. The only question that matters though is if the discrepancies are meaningful enough to alter the poll's predicted outcome. In the case of the Virginia poll, they're not. More Republicans are voting for Webb than Democrats are voting for Allen, and a huge majority of independents are voting for Webb. That's just more important overall than the poll oversampling Democrats. If you want to see a poll that is skewed in favor of Democrats in a statistically meaningful way, look at the Missouri poll SurveyUSA just released giving McCaskill a 9-point lead. If that poll broke down Missouri Republicans and Democrats properly, it would probably show McCaskill with a statistically insignificant 2-point lead.

[Alex]

All polls showing surging Republicans are excellent; all polls showing Dems surging are all lefty MSM propaganda (except when a Fox poll shows that - then we should just ignore it). Why don't you guys just stop blogging about polls for 24 hours and you'll know ??!!

[larry]

the polls are blogworthy not so much because of what they say....they're blogworthy because we think they're wrong. and wrong on purpose.

in 2002 & 2004, the dems screamed that they'd exact hideous revenge for the "stolen election of 2000". the polls dutifully agreed. nothing of the sort came to pass. not to mention the polls absolute failure to even hint at the gop house takeover of 1994. also, every year that passes sees poll 'internals' skewing more and more left-wing. it's not random, and it's not accidental.

polls are worthy of commentary because it's becoming clear they're just another tool used by the MSM/psycho left to try & shape events via their favorite tactic: lying.

[mehes]

Larry, that's completely wrong. Take the time to look at the actual data: http://www.surveyusa.com/Scorecards/2004PresGovSenOnly.xls. SurveyUSA polls predicted the 2004 elections with remarkable accuracy, be it in presidential, senate or governorship contests. They correctly predicted the winner in all but 2 contests, and their average margin of victory error was only 1.5 points. So, all of a sudden, they're 9 points off on Virginia? Maybe, but unlikely, especially when you look at the crosstabs. You're mixing up bad exit polls during the 2004 election, which involved people interviewing voters as they left their booths, with SurveyUSA's automated telephone polling.

[Alex]

Larry : On this blog's home page, the Pew poll and the Gallup poll are being cheered (and assumed to be accurate) ; the CNN and News7 polls are being jeered (and being accused on being biased - "tools of the MSM") and the Fox poll is being ignored. It is just magnificently coincidental that the first two are showing Repubs rising and the second two are showing Dems rising. Lastly, when it's convenient Fox is "fair and balanced" and when it is not, it's ignored.

[larry]

sorry, guys, i still don't buy it. i have distinct memories of '02 &'04: at no time did those polls ever tell a story of dems not winning. in '02, especially, the polls and the dems were as one: cold would be their revenge for 2000, and heaven help the republicans who'd be paying the price for dem rage. the media, along with their democrat soulmates, were rubbing their hands in anticipatory glee! just like flounder used to say, they'd crow to anyone dumb enough to listen, "ohhh, boy! this is gonna be GREAT!"

result: in '02, gop picked up seats. gains that the media had to reluctantly admit were "historic". '04 was same song, verse 2. and now here we are today.

maybe i'm full of it; maybe the polls are right. we'll see soon enough. but until i see results that more or less match what the polls have been telling us, i have no reason to believe them. their credibility is no more than the 'times', or 'newsweek'. which is to say, nonexistent.

[Dan Riehl]

On this blog's home page

The FOX poll wasn't out when I made that post. If it was, I hadn't seen it. But it is a weekender, don't forget that. I haven't been "cheering" a lot of polls - especially generic vote. I look at the polls for specific races. When Allen and Webb are tracking plus 1 or 2 for days and suddenly a 7 or 8 comes along, the first assumption is that it's an outlier. And frankly, I don't trust anything labeled CNN. There's no reason why any reasonable person would, unless they are telling you what you want to hear.