Exit poll report is self-contradictory

[L. Coyote]

It states: "There is no significant difference between the completion rates and the precinct partisanship" on page 37. This table then follows:



The refusal rate does not agree with the "probable" findings.

In the 32 states with Senate races, the exit poll margin errors (population mean) are:

Presidential = 4.990%
Senate = 3.597%

Difference in Error Margins = 1.394%

The greatest error for US Senate races was in Ohio. The Republican won with a 15.10% greater margin than the exit poll predicted. Non-responder percentages cannot ever come anywhere close to explaining this one.

Updated Exit Poll spreadsheet

New Presidential Results spreadsheet.
.

[rzemanfl]

1. The exit poll report was a bullshit coverup. n/t

[jkd]

2. Hi L. Coyote

I haven't seen any posts from you for awhile. What's your opinion about the Mitofsky report. If the problem wasn't non-responders, where do we look? Could it be the interviewers or the weighting of the polls. What's your opinion?

[L. Coyote]

5. What's my opinion... There best defense is "bad election."

Well, my first impression is that a business organization is trying to make the best of a very bad performance.

If they get sued by the people who paid them all that money, we might see some more interesting data! Meanwhile, they are just trying to hang in there and justify their work result. It they end up on the "defense" (in court) instead of on the "defensive" (money in the bank), they too will be focusing on election problems instead of polling problems.

One consideration I have not seen expressed yet it the possibility that the report actually supports the view that the elections were flawed. Since the factors they point to as problem areas cannot adequately account for the margins of error, it is logical to look elsewhere. They are showing how much of the margin can be attributed to their own problems and methods. They are also showing that their own problems cannot account for all of the margin of error. You fill in the blanks.....

There best defense is "bad election."

[anaxarchos]

3. More...

An average 1.4% differential BY ITSELF is significant enough to make WPE through non-response virtually impossible, especially as you move to the two ends of the curve...

It's complicated though because you have to go through the individual Senate races (crossover votes, etc.) to prove that the issues outlined in the report would have to apply to the senate races as well.

[L. Coyote]

4. The Senate data is added to the Exit poll spreadsheet

so you can do that now with ease.

I agree that the Senate races are more complicated. For example, Dorgan (Dem) in ND. He wins with ease and Bush sweeps the state.
The ND exit polls overrate Kerry by 7.0% and underrate Dorgan by 6.0%
The report certainly does not account for this sort of "utter failure."

I am surprised by the inaccuracy of some Senate exit polls. The standard deviations of these 32 states are:

Pres. Exit Margin Error stdev = 5.256
Senate Exit Margin Error stdev = 5.838

while "on average" the Senate polls are more accurate.

[anaxarchos]

18. Take a look at Alabama... I was initially looking for a proof of 1.

In Alabama, the Presidential and Senate races are pretty close:

http://www.cnn.com/ELECTION/2004/pages/results/states/AL/P/00/index.html

http://www.cnn.com/ELECTION/2004/pages/results/states/AL/S/01/index.html

Both the county and demographic results are similar for presidential and senate races... But the exit poll differential is nearly 12 points (-10.163 for president versus 1.56 for senate)!

There are many others that I am going to look closely at.

[bleever]

6. Excellent. I expect they will now make more mistakes, and more

obvious ones, as they try to explain this away.

They blew it election night. If Edison-Mitofsky had started bastardizing the exit polls early enough, they might have gotten away with it.

At this point, the damage control gets asymptotically harder.

It's the cover-up that gets 'em, every time.

Thanks, L. Coyote!

:thumbsup:

[L. Coyote]

8. You write, "It's the cover-up that gets 'em, every time." I wonder. ???

I'm not so sure they are doing that. They are laying out what they know. And, they are trying to appease the people who sign their paychecks, so they can keep the money. And, in the end, they seem to admit that the best they can do is provide a weak "probably."

I think they are trying to stay in business. And since they admit they cannot fully account for the margin of rerror, what are they really saying about the elcetion?

[Carolab]

13. That it was stolen. n/t.

[jkd]

11. Different kind of cover-up

The only thing Mitifsky has to cover-up is the inadequacy of his poll to predict this election. Samples were too small, and interviewers were unsupervised. I had the BYU exit poll interviewers at my precinct this election. They had Mitofsky interviewers outnumbered ten to one. They were a check upon each other. I also believe they had better cooperation with election officials. They were within the walls of the polling place. They were all young, but well trained and highly motivated. They polled six times the number of precincts and conducted over ten times the number of interviews. I don’t believe their WPE has been published, but they only missed the actual state presidential election results for the state by .5%. NEP’s exit polls were off by 4.5% in Utah.

Mitofsky is conducting debriefings with many of his interviewers. The WPE was based upon the data supplied by these interviewers. Perhaps their cover-up will be exposed. Do you think Mitofsky should ask that their wages be returned?

[L. Coyote]

7. The "Reluctant Bush Responder" Theory Refuted? by Mark Blumenthal

Check out this article on the same topic:

http://www.mysterypollster.com/main/2005/01/the_reluctant_b.html

[anaxarchos]

9. This report is beginning to come apart in less than 48 hours...

Even Blumenthal and company are now grasping at "random" non-random error or even further out with false-response.

Mitofsky would have been better off just punting... "gee, it's a complicated business and 'something' happened but we don't know what but it couldn't have been election fraud...", quack, quack, quack.

This is something to watch...

[anaxarchos]

10. L. Coyote, it's time to start kicking this thread of yours... n/t

[intensitymedia]

12. big kick from here - n/t

[TruthIsAll]

14. Here's a truth kick. Good to see you guys carrying the ball. n/t

.

[davidgmills]

15. He's been shilling the no-fraud mantra for some time

And now it looks like he has moved from the bogus non-responder hypothesis to the incompetency of the poll takers.

Seems that it's hard work asking someone if they want to take an exit poll and if they say no to go back to your chair to record their non-response. Apparently this is above the competency level of the average college student.

[anaxarchos]

17. When in doubt, blame those who are doing the sweating...

...that hasn't changed.

Too bad it doesn't hold together.

[anaxarchos]

16. One more from me... n/t

[chorti]

19. maybe people were lying

OK, how about this theory. People were so ashamed to tell pollsters that they voted for * that they lied in mass to the exit poll interviewers. To me it is more plausible than interviewer error - that the interviewers intentionally picked Democratic-looking people to interview. But based on the MEP report, it seems clear that the shy-Republican theory has been busted. So let's look at the lying, ashamed Republican theory.

If it is true that a bunch of Bush supporters, primarily Republicans lied to exit pollsters by saying that they had voted for Kerry then it would also be true that a slightly smaller group lied to exit pollsters by saying that they were Democrats when in reality they were Republicans. In addition, we have already established that another bunch of Bush supporters lied by saying that they voted for Bush in 2000 when in reality they did not vote in 2000. (Percentage of respondents who said they voted for Bush in 2000 would result in Bush-2000 vote total that is 104 percent of Bush's actual 2000 vote tally.) In other words, some Republican Bush supporters were too ashamed to say that they voted for Bush in 2004 so they lied to exit pollsters and said they were Democrats who voted for Kerry but another group of Bush supporters were so enthusiastic about Bush that they lied to exit pollsters by saying they voted for * in 2000 but in reality had sat that one out. Make any sense to you? Not to me either.

If this is true then why were they only ashamed and only liers in certain states? Why were they ashamed/liers/avoided pollsters in Nebraska but not in Kansas? Why were they ashamed/liers/avoided pollsters in Alabama but not in Georgia?

[anaxarchos]

20. I think you made your point.... Kerry won. n/t