Slashdot Mirror

That is not even close to a true statement. I know you really want to believe it because it is good ammo that the election must have been stolen "this time" but it is simply not true. Go to one of the referenced studies. Flip to page 32. Starting there - polling data is compared with actual returns from every state from 1988-2004. The negative numbers are when the polls overestimated Democratic turnout. There are a lot of negative numbers. Skip down to page 34 (second table). In every election the democratic position has been overstated. It just so happens that in 2004 it was overstated by the largest margin (6.5% error). However, the error in 2000 was only 1.5%. Interstingly, the next largest bias in the polling after 2004 was in 1992 at 5% (so you will obviously say it runs in the family...) which doesn't make the 6.5% value such a stretch. If you skip down another page they show the correlation between voter "paying attention to the election" and poll bias. Interestingly, the more interest the voting public has in the election, the greater the turnuout and the bias against the republicans in the polls.

Unless you are a statistician, I'd have to say that you're just coming up with either excuses or rationalizations.

Statistician live to do this stuff. When they have uncertainties, they quantify it and label it a "margin of error"

Don't you think all your points have already occurred to them? Do you really believe they didn't try to adjust for it?

For cryin' out loud, it's their job & they've been doing it quite well for years.

Even loosening up their assumptions beyond what could reasonably be expected (I'd say 'liberal assumptions' but I don't want to confuse the issue) didn't change the polling data enough to match what was coming out of certain states.

So instead of doing a proper audit of the elections, we got weeks of rationalizations... all of which are being rehashed today.

Do you have any facts to back this up?

What part of "scientifically selected sample" don't you understand?

Actually, we can never validate one way or the other if these elections are valid because you can't do recounts in Diebold elections.

All I meant with my post was that polls show Republicans should get trounced this November.

If they don't, and if exit polls are wrong for the 3RD election in a row, then our election system is officially rigged and we have no easy way to stop it.

This is not a problem for a survey. In exit polls, voters are chosen in a random way, (say every 5th that comes out the door), it's not up to them. And if someone diclines to answer, it is noted, and counted in the statistical error: from 100 ppl, 40 voted for A, 30 voted for B, 20 for C and 10 refused, so you have a 40% +- 10% for A

• They screwed up.
• The early numbers released were inaccurate due to bad gender participation weighting factors. (the end-of-day results were actually much closer to the actuals than most people realize)
• There was no difference in exit poll errors between touch screen and other methods."Some have suggested that the exit poll data could be used as evidence of voter fraud in the 2004 Election by showing error rates were higher in precincts with touch screen and optical scan voting equipment. Our evaluation does not support this hypothesis. In our exit poll sample overall, precincts with touch screen and optical scan voting have essentially the same error rates as those using punch card systems. In the larger urban areas these systems had lower WPEs than punch card precincts."
• Kerry supporters were more likely to participate and complete an exit poll
• strong correllation between the age of the poll volunteer and the pollee's willingness to participate

You can speculate all you want, or you can read the 77 page report from the company who did the exit polls.

> Exit polls have been used the world over to predict election results for
> decades. [link to wikipedia snipped]

With generally good, but less than perfect accuracy. Wikipedia is not the best source of information on a subject this hotly debated.

> So we have a single event where the long-working exit polls (which are
> normally accurate) are suddenly and significantly different from the final
> official tally. This could be written off as a statistical fluke, but the
> Diebold and ES&S machines are already suspected of widespread insecurity
> and/or deliberate tampering, and then when it all hits the media the
> administration announces it won't be conducting exit polls any more?

Tha's somewhat oversimplified, There are -without doubt- problems. The immediate leap to deliberate tampering is not the only explanation, especially in light of historical incompetence of election workers. The reality is that finding, training, and supervising the volunteers who do the work is a large task filled with problems.

Many of the inaccuracies cited in Florida and Ohio in particular are not evidence of election tampering as much as evidence of the fact that election jobs are handed out to party hacks who aren't competent to do any real work.

Many pollsters are changing their methodologies to deal with problems that have been affecting accuracy for some time.

> Why, when they've been used for decades without problem, are exit polls
> suddenly considered dangerous or misleading? Apart from, that is, their
> potential to provide an indication of election-tampering?

The problem is that they can only be used to support a -charge- that the election was botched. They can provide no value in determining the cause of inaccuracies, or determining culpability.

To actually audit the elections, we need an audit trail, not a secondary source of information. I'm not arguing against most of your points. You made some good ones. But you let it fall into simple two-valued logic which I feel weakened the value of your post.

A couple of links people might find useful:

http://www.exit-poll.net/faq.html

http://www.mysterypollster.com/main/2004/12/what_i s_the_sam.html

The one at mystery pollster has some nice links to other sources. I found it very informative.

I think this does a pretty good job of explaining why exit polls resulted in such a poor estimate of election results.

By the way, I have read the Mitofsky-Edison report to which I assume you are referring regarding the exit polls. I do find it interesting that there are other statisticians who conclude differently.

Plain and simple, repeat after me: Exit polls showed Bush winning .

The only exit polls which showed Kerry winning were early results from the National Exit Pool, and were only reported on sites like DrudgeReport. The problem was that taking a sliver of the NEP is completely inaccurate. There's a reason major news organizations are very conservative with the data. It's just not 100% accurate. Less so early in the day.

Of course someone will say that late at night Bush's exit poll numbers suddenly jumped up. This too, was caused by a server failure in the NEP which hadn't updated the exit poll information.

Bush did not win by mass conspiracy. There is no cover up.

Either way, the system needs to be changed.

Here is the final exit poll data bush by 1 which puts bushes victory in the margin of error for the exit poll (+/- 3). Also how accurate can an exit poll be if only 13,047 voters made up the exit poll in an election of well over 100,000,000 people. have a nice day, your average anonymous coward

Why don't you do yourself a favor and read up a bit on How Exit Polls Work

Yea what are the odds that a poll would have a mirgin of Error or 3%, I mean that would be unheard of amount of error...

Oh Wait What is the Margin of Error for an exit poll? Every number estimated from a sample may depart from the official vote count. The difference between a sample result and the number one would get if everyone who cast a vote was interviewed in exactly the same way is called the sampling error. That does not mean the sample result is wrong. Instead, it refers to the potential error due to sampling. The margin of error for a 95% confidence interval is about +/- 3% for a typical characteristic from the national exit poll and +/-4% for a typical state exit poll. Characteristics that are more concentrated in a few polling places, such as race, have larger sampling errors. Other nonsampling factors may increase the total error.