Polling truth

October 26, 2004 4:03 AM

With one week to go, Americans are being inundated by polls. At least 112 have been published for the presidential contest in the last week alone. Democratic pollster Mark Blumenthal maintains that, in the campaign's last hours, we tend to see 'undecided' voters 'break' for the challenger. Testing this theory, blogger Chris Bowers examined presidential poll results since 1976, and calculated that undecided voters broke for the challenger 86% of the time. So, is this really how it's going to turn out? Are the Republicans' attempts to 'steal' another election going to bear any fruit?
posted by acrobat (11 comments total)

Sorry, it now appears the second link is not coming up, even though it came up fine when I linked it. The link also appears in the last paragraph of the first linked article, but it seems to have gone dead there also. I really don't know what to do about it. Sorry again.
posted by acrobat at 4:09 AM on October 26, 2004

Well, I think it does now.
posted by acrobat at 4:13 AM on October 26, 2004

It's my biggest worry--and it makes you realize why Rove spent all his time playing to their base instead of wooing swing voters or moderates or new voters. They've been planning this stuff all along.

and how cool is it that Kos is writing for the Guardian now?
posted by amberglow at 5:27 AM on October 26, 2004

"These people are cheating. When you tamper with it, you cheat the system. And cheating in elections is the worst form of cancer because it's uncontrollable."

Exporting democracy indeed!
posted by acrobat at 6:17 AM on October 26, 2004

and how cool is it that Kos is writing for the Guardian now?

sure is good to keep all the loonies in one bin.
posted by Krrrlson at 12:38 PM on October 26, 2004

Didn't 2000's final Gallup poll, that showed Bush up by, what was it 10 or 13 points, demonstrate that such polling is worthless in such a close race?

Personally, I'd be glad never to see another f#*$!%& poll in my life. Do we have any pools going on this yet? My money is on Kerry winning the EV by a fairly wide margin, but losing the popular vote by 2-3%.
posted by psmealey at 1:05 PM on October 26, 2004

Didn't 2000's final Gallup poll, that showed Bush up by, what was it 10 or 13 points, demonstrate that such polling is worthless in such a close race?

Polling might well be worthless in a close race, but that can't be demonstrated solely by one poll conducted by one organization for one election.

Do we have any pools going on this yet?

See here.
posted by DevilsAdvocate at 2:27 PM on October 26, 2004

Thanks, DA. Yeah, I misspoke. There were several polls that had Bush up by significant margins over extended periods four years ago, the Gallup was just the most egregious example I could think of.
posted by psmealey at 3:03 PM on October 26, 2004

How about Americans are inundated by polling. I've been called by two pollsters already in the last month, apparently from the Democratic party or its operatives. they asked questions about hypothetical matchups in 2006 or about the local assembly race after the proverbial Bush or Kerry question. And I don't live in a swing state, and my phone number is not reported to the registrar of voters (to prevent being harrassed by the phonebankers I once was).
posted by calwatch at 9:36 PM on October 26, 2004

Gallup had two big problems in 2000: their "daily" tracking poll and their method of likely voter determination. The first they've dropped for this election, the second they've kept. But both together are why their polling for 2000 is so ridiculed.

As discussed many places, including here, their "likely voter" determination method is why we've got some really weird and misleading results from them in the last month, showing Bush ahead by 6+ points. Just discard those out-of-hand. They're wrong.

This week many polls are showing Kerry closing whatever small gap there was (which, mostly, was within statistical error anyway) or pulling ahead. Kerry's ahead in the majority of the swing states which will determine the election, not taking into account that independent voters, as mentioned above, overwhelmingly move to the challenger on election day.

All this put together paints a picture pretty much exactly what I've been saying for almost ten months: a small but significant Kerry win in the popular election with a much larger margin in the electoral college.

And yet there's still many people who think Bush has a lock on the election. It's really weird and has been very educational for me. This year's the first Presidential election that I've watched as closely as I have—in the past I've had good general intuition and called two of the last three elections long before (I wasn't sure how 2000 was going to turn out, but thought that Gore might win)—but this is the first where I've watched all the polling data and learned about polling method and related stuff very closely. And the result? I've learned to deeply distrust how the major media presents "state of the horse race" info and analysis to the general public. It's very misleading in two different ways: it's very uninformed and uncritical of/about polling methodology; and it's built around the major media's need to invent new and interesting horse race stories to keep people interested. But most people only see the headlines or the TV news glosses, and the result is that they've thought and still think that it's more likely than not that Bush will win.

It's been a very educational election season for me.
posted by Ethereal Bligh at 9:45 PM on October 26, 2004

I've been doing some number-crunching on the data available at race2004.net. I haven't done the statistical analysis yet to prove it, but it looks to me like Gallup is skewing significantly more towards Bush than many of the other polls. In several swing states, Gallup has Bush up by 4-6%, while other polls have the two candidates within a percent of each other. If that's so, it might indicate the uselessness of Gallup polls specifically (as well as any news reports or meta-analysis that relies too heavily on Gallup), but not of polling overall.

I've already thrown out the Strategic Vision polls based on my analysis showing that they're skewed towards Bush (relative to all the other polls combined) with p<.002.
posted by DevilsAdvocate at 8:48 AM on October 27, 2004