New U. of Penn Study -- 622,000 to 1 Odds against Vote being Fair

[JMDEM]

Dr. Freeman has updated his report, originally released in early November, about the statistical odds against the exit polls being so wrong in Florida, Pennsylvania, and Ohio. This time around, he takes the apologists head-on in their arguments that the exit polls were fundamentally flawed. He also discusses how exit polls are so very important for ensuring fair elections in other countries.

http://www.appliedresearch.us/sf/Documents/ExitPoll.pdf

[Woodstock Liberal]

1. Thanks for the post!

[m.standridge]

4. Sounds like the odds against error are dropping, though

But the key states are still questionable as to outcome. Those are the ones Professor Freeman focussed on in his first report, anyway.

[understandinglife]

2. Thank you and, everyone, please note the following:

"The Center for Organizational Dynamics operates within the
University of Pennsylvania’s School of Arts and Sciences,
Graduate Division, conducting research and scholarship
relevant to organizations, public affairs, and policy.

Copyrights remain with the authors and/or their publishers.
Reproduction, posting to web pages, electronic bulletin
boards or other electronic archives is prohibited without
consent of the copyright holders. For additional information,
please email [e-mail] or call
[phone]"

Peace.

[StephanieMarie]

6. Just emailed them

And asked them to email their report to every U.S. Senator and ask the senators to contest the electoral vote on Jan. 6.

[Wilms]

3. Thanks JMDEM

Consider posting this on the daily thread.

<http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=226125&mesg_id=226125>

[JMDEM]

8. Posting to Daily Thread

I'm a newbie. Please feel free to repost to anywhere more appropriate. You know the landscape of DU better than I do. Thanks!

[Wilms]

10. Done, JM

The pleasure was mine. Thanks again.

[TruthIsAll]

5. Freeman should analyze ALL the states. His odds are conservative.

Its much less likely then 1 out of 662,000.

The odds of at least 16 states exceeding the Exit Poll MOE for BUsh is

******** 1 out of 13.5 trillion ********

This had never been refuted. Because it can't be.

[SicTransit]

7. Why not 12.5 zillion?

Every time you post the "1 out of zillion" stuff you conveniently forget to state your assumptions. Like "If it is assumed that each deviation is an independent event". And "If it is assumed that the methodology of the exit polls was correct". And "If it is assumed that raw exit polls can be relied upon to make predictions about election results, in spite of the guy who invented exit polls saying otherwise."

[JMDEM]

9. Did you read the article?

A lot of these variables are addressed in the article. Perhaps if you pick out specific examples in the article to dispute, this would be clearer.

[mgr]

11. yes, I did

One of his caveats is with the sampling methods and design.

Mike

[SicTransit]

16. You will note that I was commenting on TIA's methods, not

on Freeman's, but - about the article.

1. Freeman admits that pollsters say that the polls were not designed to verify election results, and should not be used for that. In fact, the person whom Freeman credits with inventing exit polls - Mitofsky - said that exit polls cannot be used to prove fraud - yet Freeman attempts to do it anyway. Does he know more about this subject than Mitofsky does? It is a bit presumptuous.

2. He says that the data he is using is correct because "NEP would < not > ever distribute unweighted data to anyone" - but he really has no idea whether it is or not, no confirmation from the source of data, and no way of knowing whether that's right or not - because the raw data has not been released yet. Any scientist who bases his research on something that nebulous should be ashamed of himself.

3. He compares exit polls in US to exit polls in Germany, blithely ignoring the differences in methodology of collecting data in US and in Germany. See http://www.mysterypollster.com/main/2004/12/what_about_thos.html

He also ignores studies that have shown that exit polls were inaccurate to the point of complete uselessness in other places in Europe.

4. He admits he has no access to either the exit poll data or methodology, and claims that his analysis is based on "reasonable approximation". Well - "reasonable" is in the eye of the beholder. See my point (2) above.

5. Key phrase of one of his assumptions: "assuming independent state polls with no systematic bias". So if there was systematic bias in the methodology or the data, his conclusions are wrong. The polls are also clearly not completely independent - they are, after all, conducted by same people with same methodology. That's a big assumption to make - especially on data that has no proven provenance.

6. He examines pollsters' explanations for the discrepancies and dismisses them - on his say-so. I am not sure what makes him more of an expert on polling that Mitofsky. Do you know?

(edit for typo)

[Yupster]

17. Very interesting article

Taking Ohio as the most important example, his appendix said Kerry should have won 51-49 %, but Bush actually won 51-49 %. So how to explain the discrepency?

Besides the oddity of making such a big deal out of a poll being 2 % off, I think he answers his own questions on pages 15 and 16 of his report.

Let's say that every so often the exit pollers which I think are mostly younger people approached an angry old white guy who got out of a white pick-up truck with the Starry Cross displayed over his NRA sticker right by his W icon. The guy gets his news from Rush Limbaugh and hates the "liberal" media. They asked him if he'd mind taking an exit poll and he used his best ninth grade schooling to cleverly tell the exit poller to fuck off and go back to Red China.

The exit pollster noted that he/she needed another older white man. The next older white male came in wearing a vest and dress slacks as he is a creative writing professor at the local junior college. Would he take a minute to take the exit poll? Of course he chuckled. Democracy demands a minute of our time once in a while.

The exit pollster is happy that he/she has now evened out his/her demographic sample.

Twenty minutes later, a woman was stopped by the same exit pollster. Wearing a frock and a bonnett, she seemed skittish when asked if she'd take an exit poll. She looked at her husband who shook his head "no," and she sheepishly declined and hurried away with her six kids in tow. The pollster noted that she needed a white woman. No problem as a white woman came out soon enough. She was carrying a briefcase in one hand, a starbucks in the other, and was dressed in business attire. Would she take the poll? "I've got to get back to work, she said, but I can spare a minute on election day.

Again the pollster was happy that her demographics were back in balance.

Did this happen? The author said it's a hypothesis and left it at that. I agree. It's just a hypothesis to explain the 2 % poll variance, but I'd call it a much more likely hypothesis than a massive conspiracy all over the country stealing millions of votes.

The author also mentions the fact that early voting is a problem for exit polls, but left it at that. In Florida didn't 1/3 of all voters vote early? If that's at all accurate then yeah, early voting would be a big problem for exit polls.

The author also said respected pollster John Zogby respeccted the figures so much that he called the election for Kerry. LOL. Zogby called the election for Kerry back in May. Talk about your early results.

Anyway, there are apparently a number of hypothesis as to why the exit polls and final vote didn't add up. I guess everyone can choose which hypothesis they find most likely.

For those who think the exit polls were accurate though, I'd like some explanation of New Hampshire where the exit poll said Kerry would win by something like 12 %. Now obviously that exit poll was wrong. No one expected Kerry to win New Hampshire by double digits. They expected a very close race and that's what it was last time and that's what it was this time too.

[Carolab]

18. Why would it be any MORE likely that the second male and second female

would answer the pollsters in THIS election over ANY OTHER?

That is a simple-minded assertion that is completely without merit.

Furthermore, have you ever listened to a Rushie? They LOVE to spout off about * and Red China.

You are SPECULATING to an unreasonable degree about who did or did not answer pollsters' questions on Nov. 2, 2004, and that the scenarios would be significantly different from those affecting any other election ever held in the past.

[Yupster]

19. I think in general

left-leaning people are more friendly and trusting and cooperative than right-leaning people. That's why the Republican always seems to do better than the exit polls say he will do.

Anyway, that's a hypothesis. Is it the correct one?

I'm pretty sure it's correct to some degree. Whether it's worth .25 of 1 % of the poll difference or 2 % of it is subject to argument, but I'm pretty sure it accounts for some part of the difference.

But then again, I'm speculating and prone to make simple minded wild assertions completely without merit, or so I've been told.

[Igel]

12. Yeah, he stated his assumptions.

That the samples were random, there was no systematic bias or error, and that Mitofsky's too professional to make a mistake. The usual assumptions you have to make, but the ones that would have to be wrong if the exit polls turned out skewed.

I still like when he miscalled the NJ race 15 or so years ago. He had three forms going, to see if he could get the same info from a short form as a long form, with his usual form being the "middle" one for length. A lot of people, it turned out, were (he concluded) annoyed or offended by the short form, and people patiently waded through the long form. This much was obvious early on. What wasn't obvious is that those offended from predominantly from one political party (which one it was escapes me now), and so his sample was far from random.

[JMDEM]

14. 6 errors in 2200 exit polls

That is Mitofsky's record until 2000. 6 erroneous polls in 2200 tries. 99.7% accurate.

Then suddenly -- 10 states wrong in one election.

Now what are the odds on that?

[JMDEM]

13. in spite of the guy who invented exit polls saying otherwise

Now why would Mitofsky, who has made his living for the last 30 years doing exit polls, advertise to the world that his exit polls were screwed?

"Don't buy MY product, it is JUNK! I have no other way to make a living, but don't buy MY product!"

Go figure...

[Joe Chi Minh]

15. Some of the points made are

so pungently expressed, albeit with due academic rigour and solemnity (indeed, partly at least because of it), that they make extremely comical reading.

The forensic annihilation of the ever more outlandish hypotheses (not explanations!!!!!) and "undocumented"!!!! assertions, of the "hear-no-evil, see-no-evil, speak-no-evil" apologists for the election, is so comprehensive and meticulous, that, well, frankly, it can, induce an unhealthy access of frivolity in the democratic/Democratic heart. A plight, no doubt, that must have befallen many who heard the young lad call out that the Emperor had no clothes.