Exit Polls

[Matabiri]

Anyone got any idea how the exit polls got it so wrong? Inaccurate weightings? Bad model? People saying they voted Democrat when they actually voted Republican? Or possibly dodgy counting?

(Oh yeah: damn you, CNN. I don't want breakdowns, I just want the overall totals. Why is that so hard?)

 

[Anathema]

Well, the exit polls measured exactly what they claimed - the self-reports of people willing to be surveyed. The big skew likely came from Kerry supporters feeling more insecure - making them more likely to get all gushy when approached by an surveyor. I imagine more Bush supporters simply waved the surveyors away, and went on their way.

 

[BobK]

Probably some of the error can be laid off to the long lines early in the day causing business people to not vote until after work when they have more time.

I wouldn't doubt that some people might have simply felt it was safer to say they voted for Kerry even though they didn't. Although not exclusively one-sided, newspaper coverage of vandalism and generally boorish behavior seemed to indicate more unhinged people on the Kerry side of the equation.

It would only take 4-5 people out of each 100 to skew the result. Some of the people may have been unsure of who was asking them, and what sort of response they would get.

It's easy to lie to a stranger to avoid any hassle.

 

[Matabiri]

I wouldn't doubt that some people might have simply felt it was safer to say they voted for Kerry even though they didn't. Although not exclusively one-sided, newspaper coverage of vandalism and generally boorish behavior seemed to indicate more unhinged people on the Kerry side of the equation.

Ooh, I dunno if that's fair - there's been plenty of unhinged behaviour on both sides. I haven't seen any evidence of clustering on one side or the other.

I can certainly see people in some places being ashamed to admit voting for Bush in Dem areas, more so than being ashamed to admit voting for Kerry in Rep areas, though.

 

[Rob Lister]

Probably some of the error can be laid off to the long lines early in the day causing business people to not vote until after work when they have more time.

Even though many (if not most) on the right are claiming some underhanded conspiracy, I tend to agree with your assessment. I think it VERY likely that more Kerry voters voted early (for more reasons than the single reason you stated) and thus skewed the numbers.

There might be some plot but it is my understanding that the exit-pollers tell their clients (the MSM) that election-result predictions are not what exit polls are designed for and should not be relied upon for that; rather, exit polls are meant to reflect the reasons people vote they way they do.

 

[BobK]

Ooh, I dunno if that's fair - there's been plenty of unhinged behaviour on both sides. I haven't seen any evidence of clustering on one side or the other.

I can certainly see people in some places being ashamed to admit voting for Bush in Dem areas, more so than being ashamed to admit voting for Kerry in Rep areas, though.

Kerry voters tend to be some what younger. Therefore, more physically fit and likely to be willing to defend themselves. The older you get the less hassle you want out of life.

 

[crimresearch]

Well, the exit polls measured exactly what they claimed - the self-reports of people willing to be surveyed....
<SNIP>.

Very astute... one of the senior faculty where I was, also owned one of the country's leading polling firms...

I asked him how it was that they had called Virginia's governor's race a landslide for Wilder, when it turned out to be a squeaker, and he mentioned that they can only measure what allows itself to be measured..if people want to assume that as fact, they might be mistaken.

 

[Matabiri]

Kerry voters tend to be some what younger. Therefore, more physically fit and likely to be willing to defend themselves. The older you get the less hassle you want out of life.

The polls show greater support - and not by a huge margin - for Kerry only below the age of 30, when most people are still pretty fit, so I don't think that really holds up.

 

[Snide]

There might be some plot but it is my understanding that the exit-pollers tell their clients (the MSM) that election-result predictions are not what exit polls are designed for and should not be relied upon for that; rather, exit polls are meant to reflect the reasons people vote they way they do.

That seems reasonable. If they wanted to "predict" who won, they wouldn't ask a zillion questions, which only serves to make people hesitant to respond. Just ask which presidential candidate did you vote for, and I would suspect nearly everyone would comply, and honestly so.

 

[BobK]

I read an article a while ago that said the early polling was about 59% female. Maybe they didn't weight Kerry's popularity with women correctly. Or maybe they didn't bother to weight the respondents at all.

Another possibility that one journalist mentioned, is that maybe Kerry operatives each went to several polling places that were being surveyed and pretended they just voted. Then answered the poll in an effort to get some sort of ground-swell started by showing Kerry with an early commanding lead.

 

[Segnosaur]

I think it VERY likely that more Kerry voters voted early (for more reasons than the single reason you stated) and thus skewed the numbers.

Heard something along those lines on the radio.

Kerry tended to have a higher level of support from females, many of whom are 'stay-at-home' moms, who had the time to vote for Kerry during the day. Bush had a little more support from the male population, but many of them had to wait until after work to vote.

Not sure how accurate that is. Anyone have any stats to proove or disprove it?

 

[BPSCG]

Democratic Underground's blog is spinning dark conspiracy theories about this.

 

[patnray]

I think the pollsters include all the above factors in their models, but they cannot predict who will eventually turn out to vote. The Republicans did an outstanding job getting out their voters, but it appears most came to the polls later in the day. By election day they have accurate and detailed models of the demographics, so the exit polls are mostly about determining who is actually voting.

 

[aerocontrols]

America did vote for John Kerry - before we voted against him.

 

[Snide]

America did vote for John Kerry - before we voted against him.

That is excellent!

 

[crimresearch]

America did vote for John Kerry - before we voted against him.

 

[NoZed Avenger]

Women were oversampled, possibly because of the scheduling mentioned above, and women traditionally vote more heavily democratic.

I would also speculate that the polls, especially for the networks, concentrated on urban areas -- simply because it is easier to cover than running out to multple polling spots in rural areas. That would also skew the results towards the democratic side.

IIRC, similar problems happened in 2000 and 2002 in some races -- Colorado being called a toss-up when it turned out to be Bush's by 10 percent or more.

I was surprised that none of the coverage seemed to remember those problems, much less tried to compensate for them or warn the viewers about them. If they did, I missed it -- but then I didn't see a whole lot of the coverage, either. The parts I saw made no mention of the previous problems, though.

 

[Kevin_Lowe]

Democratic Underground's blog is spinning dark conspiracy theories about this.

Okay, I'll be the resident alfoil-head.

That site has linked images which claim to be the CNN exit polls, showing that between 12:21am and 1:41 am the exit poll results somehow reversed themselves, from showing a Dem win to a Rep win.

How the heck does that happen after midnight?

Has anyone managed to track down an actual side-by-side comparison of (unaltered) exit-poll predictions and final results on a state-by-state basis?

Some netizens are claiming that the exit polls line up with the final results within experimental error everywhere but Ohio and Florida, and they're blaming Diebold. However no site I've seen yet has the numbers set out to show this relationship exists, and I'm not going to cry foul until I see the numbers.

It's possible that upset partisans are just making stuff up. On the other hand, Diebold is certainly as dodgy as hell and any result generated by their machines has to be held as suspect.

 

[corplinx]

However no site I've seen yet has the numbers set out to show this relationship exists, and I'm not going to cry foul until I see the numbers.

Yes, I showed this "chatter" in my post election WOO thread.

The emerging meme from the people in denial seems to fit this template:
exit polls show kerry won
diebold gave election to bush

There is no evidence for this theory. They people in denial are using a quote from a bigwig at diebold who is a GOP supporter and once said he would help Bush win. The context was "i am involved with grassroots campaigning and im going to bust my butt for a bush win". The kook version is "i am so dumb that i am publicly admitting i will rig my machines for Bush".

 

[Kevin_Lowe]

The emerging meme from the people in denial seems to fit this template:
exit polls show kerry won
diebold gave election to bush

There is no evidence for this theory.

Easy on the partisan politics there. If fraud took place, it's far more serious than the outcome of any one election between marginally distinguishable candidates.

The simple fact is that there is ample evidence that the Diebold voting machines were built to make fraud possible. The security holes are totally unnecessary (as proven by the Australian and Brazilian systems) and built in. The journos at Blackboxvoting.org demonstrated how a trained chimp could delete a Diebold machine's logs.

Given that fraud was possible, is there any evidence at all it happened? Well, the way you would generally measure fraud in a foreign election is to compare the exit polls to the counted vote. So the mere fact of a serious discrepancy is evidence, although far from conclusive evidence.

Take off your Republican hat for a bit and put on your skeptic hat. The claimant has arranged to be tested under conditions that allow cheating, and has pulled off a surprising result. Is it really "denial" to want to take a second look at what happened?

There are a lot of claims floating around that need the 48 hour rule applied to them. Allegations of serious electoral fraud are absolutely in that category. Nonetheless I can't see any rational reason to dismiss them or accept them yet. I think they are very much on the table.