Would the Kerry campaign, the rest of the Dems, and their shills in the MSM be acting as crazy as they have since the convention if the race was tied a la Pew or even if Bush was up 3-4 a la Rasm'n?

They're acting like they're down big-time, to reference Cheney.

It may not be 14, but I think must be at least 9-10.

For some reason this doesn't sound right to me... a 12 point swing overnight???

The latest national survey of 1,972 registered voters by the Pew Research Center for the People & the Press, conducted in two waves over a seven-day period, finds that the president's large margin of support in the initial period (Sept. 8-10) dissipated in the polling conducted Sept. 11-14. Among all registered voters Bush initially led Kerry by 52%-40%. However, the second wave of interviewing shows the race even among registered voters, at 46%-46%. When the sample is narrowed to likely voters, Bush holds a statistically insignificant lead of 47%-46% in the second wave, down from a huge 54%-38% advantage he held in the first wave of interviews.

http://people-press.org/reports/display.php3?ReportID=224

Sorry thats 6 points... I need a drink.

Efrain,

Now do you see why I stick with the tracking polls?

Stephen,

Whenever I've discussed poll tracking with people I've always described them as an extension of the Heisenberg Uncertainty Principle. I may have to adopt your photography analogy, I'd be sure to get a lot fewer blank stares.

Michael

Ever notice how when Bush is doing well, violence picks up in Iraq.

Aaron, I think you mean, that the MSM steps up the Henny Penny, the sky is falling news articles.

I would think that after the learning experience many had in 2000--Gore won the popular vote; Bush won the Electoral College--that following any national poll, tracking or otherwise, would be dangerous when trying to predict the Presidential race. [Note to self: Gotta work on those run-on sentences.]

I did a quick little linear regression going back to 1936 on Gallup's poll numbers midway through September.

The best fit line to the incumbent party's share of the two-way popular vote is:

y = .613x + .212

Where y = the eventual incumbent party's share in the two-way popular vote and x = the incumbent party's share in the two-way Gallup numbers.

There's a relatively high correlation between Gallup's mid september numbers and the actual result (correlation coefficient = .79). What that means for non-math folks is that historically Gallup's numbers have been a fairly good indicator of the actual results.

Bush's predicted final popular vote share of the two-party vote is 56%. Only twice in the 17 years has the difference between the prediction and the actual results been large enough to overturn that lead (1980 when Carter was ahead and Reagan wound up winning easily and 1936 when Roosevelt was ahead and eventually won in a huge landslide). Other than those two, the line fits fairly well.

That is to say, if Gallup's historical accuracy holds, this is very good news for Bush.

AZlibertarian,

People keep saying this but it wasn't like Gore ran away with the popular vote. He "won" it in the sense that he had more votes but it was 50.3% to 49.7% of the two-way vote. A virtual tie.

If either candidate gets 53% of the two-way vote, they're going to win pure and simple. If either gets 52% of the two-way vote, they likely win.

Only when the popular vote is very, very close would you expect the electoral to shift opposite of popular.

Steve,
You make good points about tracking polls -- but even a poll that shows someone ahead 49-45 with 5-6% undecided, when you factor in the margin of error -- then you can have ANY result.

Face it, head-to-head polls are fun, but useless. The key figure from the Gallup survey is that Bush's approval rating is holding at 52% - and that's good. Historically, that means he should win with something like 52-47 or 51-47 (depending how much Nader peels off). As Mr. Vee correctly states, the EV race will not be a contest with popular votes like this.

The Pew poll was done over a weekend - from Saturday to Tuesday.

The Galluo poll was done only on weekdays - from Monday to Weds.

The Pew poll likely had far fewer respondents who were observant religious folks (because they were at synagogue on Saturday and /or church on Sunday)than the Gallup poll did.

In other words: by polling on a weekend Pew likely under-polled a important categpory of voters.

"Religiously observant folks" is a category of voters that OVERWHELMINGLY votes for W.

Therefore: IF relgiously observant folks are in fact likely voters, then the Gallup poll is likely far more accurate.

The problem that I see here is that Pew provided enough information to treat it, to some degree, like a mini-tracking poll. For what is not getting reported but was the case is that the result showing Bush up by just one was only part of the poll.

They actually polled from Sept 8-14th. During the first half of the poll, Bush was up huge. Gallup-like.

During the second half, it was essentially a tie.

Consider it to be two data points from a tracking poll. They showed about 10 points of movement within a handful of days-- movement that the one tracking poll going right now did not see, movement that other polls have not seen. The other polls showing the race tied, showed the race tied before. The other poll (Gallup) showing the race big in Bush's direction, show Bush up big.

So the question I have for you-- do you really think that the race changed 10 points between Sept 8-10, and Sept 11-14?

If you doubt that, then you have to doubt at least part of Pew's results. And which do you doubt? Is there a reason to believe the front part over the last part, or vice-versa?

The overall Pew survey, conducted from Sept 8-14 has Bush up 49-43 among registered voters and up 51-42 among likely voters. Unless someone can give me a really good reason for believing the second half of that survey (where it was a 1 point race) over the first half (where it was a 16 point one), I'll take the whole thing with a grain of salt, but put more stock in the results from the whole survey than just a portion of it.

Some of the media are breaking ranks, but some of them are sticking it out with CBS. But I think the Kerry campaign has been hurt at the margins. And it is at the margins that this presidential campaign will be won. 0.1% here, 0.1% there, and pretty soon, you're focusing on your concession speech. But I wouldn't get too worked up as yet. The debates are yet to come. Save the bubbly for Nov 5.

Pew, like Zogby, adjusts their results to factor in the same percentage of Democrats and Republicans as reflected in the 2000 vote. Party ID is not a demographic since people self-identify and when they support Bush, a portion tends to view themselves as Republicans as well. The effect of this adjustment is to remove Bush supporters, when they have changed parties. I'm surprized that these polls show any Bush lead. Because they do, Kerry is in real trouble.