The Reluctant Gore Responder (RGR)

[TruthIsAll]

To quote Febble:
"E-M made the hypothesis that Bush voters were more reluctant to respond. A prediction flowing from this is that where there was more opportunity to avoid being polled, the error should be greater. This was so, and thus their hypothesis was supported".

This hypothesis was contradicted by E-M in their FINAL NATIONAL EXIT POLL.

Fact 1
Gore won the popular vote with 50.999 million to Bush's 50.456.

Fact 2
The FINAL 13660 National Exit Poll states that Bush 2000 voters comprised 52.57 million (43%) of the total 122.26 million who voted in 2004. That is 2.11 million more than the 50.456 million who actually voted for him.

Fact 3
Gore voters comprised just 37% (45.23 million) of the 2004 vote.

Question 1
If Bush voters were reluctant to respond, how come the Final 13660 National Exit Poll indicates that 104% of Bush 2000 voters came to the polls while only 89% of Gore voters did?

Fact 4
The National Exit Poll at 13047 respondents had the split at 41%/39%.

Fact 5
The Bush 2000 weight could not have been 41% at 13047 respondents, much less 43% at 13660.

Fact 6
The maximum possible Bush 2000 voter turnout/2004 vote was 39.82%, after reducing his turnout by the approximate 1.77 million Bush voters who have died since 2000.

Fact 7
Even 39.82% is too high, since at least some Bush 2000 voters stayed home this time.

Fact 8
The 52.57 million Bush turnout is overstated by 3.88 million, when you add the 1.77 million resurrected Bush 2000 voters to the 2.11 million phantoms.

Hypothesis 1
These 3.88 million Bush voters must have been very motivated, indeed.

Fact 9
According to the NEP, only 45.23 million of Gore's 2004 voters participated in 2004, since .37*122.26 = 45.23 million.

Fact 10
If the 37% Gore weight is accurate, then you must believe that 5.77 million Gore voters (11.31%) were reluctant to vote in 2004 and stayed home.

Fact 11
To accept as accurate the 37% Gore weight, you must also accept that the 43% Bush weight is also accurate.

Fact 12
But the 43% Bush weight is impossible - at least 3.2% too high. Therefore, the 37% Gore weight must be impossible - and at least 3.2% too low.

Question 2
Was there a RGR (Reluctant Gore Responder)?

Fact 13
Kerry was the 51-48% winner at 13047 respondents.

Question 3
What happened to change the 41/39% to 43%/37% over the final 613 respondents? Bush won the FINAL by 51-48%.

Fact 14
The IMPOSSIBLE 43% Exuberant Bush Responder invalidates the FINAL EXIT POLL, which matched to the recorded the vote for a Bush 51-48% win.

Fact 15
The IMPOSSIBLE 43% weight indicates an Exuberant Bush 2000 Responder (at least in the Exit Poll) which is at variance with the Reluctant Bush Responder hypothesis.

Fact 16
The IMPOSSIBLE 43% weight is a necessary condition for Bush to win.

Fact 17
The 41% Bush 2000 voter weight, though EXTREMELY IMPLAUSIBLE, still showed that Kerry won the first 13047 repondents by 51-48%.

Fact 18
If the weights were a more realistic 40% Gore/39% Bush, then Kerry wins in a landslide - by at least 7 million votes.

Fact 19
The final 613 respondents in the National Exit Poll had to have voted 100% for Bush to overcome Kerry's lead among the first 13047.

Hypothesis 2
Bush voters must have been EXTREMELY EXUBERANT to do that.

[kansasblue]

1. Is this getting back to DailyKos and Febble?

[TruthIsAll]

2. Do you think it should be? If so, you can e-mail it.

This was a reply to Febble in another thread.

I thought it should stand alone.

[Febble]

9. Well, I'm new to this board

and I never know where you've got to.

Dunno about 2000. The WPEs showed a lot less over an over-estimate of Gore's vote, although by my calcs it was still significant.

I will say this one last time. My paper does not say the evidence supports rBr (although I think it may). It says that USCV's refutation of rBr is not justified on the bases of their analyses. Other things may refute rBr but not that.

It would help if you read my paper to the end. Here is the end:

"Mathematically, the observed pattern could arise from widespread fraud as well as from widespread response bias; differential vote spoilage rates for Kerry votes or “ballot stuffing” of Bush votes would produce results indistinguishable from “reluctant Bush responders.” However, this is not the inference currently drawn from the data by USCV in their report. "

[TruthIsAll]

10. The question is not the UCSV analysis. It is the facts in this post.

"I will say this one last time. My paper does not say the evidence supports rBr (although I think it may). It says that USCV's refutation of rBr is not justified on the bases of their analyses. Other things may refute rBr but not that".

The question I pose to you is: Do these facts refute RBR?

[Febble]

11. Well, to be honest, I don't think they do

But don't put words in my mouth, keyboard or whatever. If you are going to critique my paper, critique it. Not some straw person you imagine me to be.

I think the evidence advanced by USCV supports widespread error, not error concentrated in high Bush precincts, which is what they infer. To the extent that they base their vote-corruption argument on this inference, I think they are wrong.

I think I have demonstrated that the error was far more widespread. I think you agree. I think we disagree on whether it was rBr.

I think it was, because I find the arguments in the E-M report somewhat persuasive. But a) they used a confounded variable (which I have fixed for them) and b) they don't specify their statistical tests.

Until they fix those things I am not going to commit myself to rBr, merely to widespread error. Which could just as easily, in principle, be in the vote count.

[TruthIsAll]

12. You are avoiding the factual isues I raise in this post.

Is it too much to ask for a detailed response?

These facts are very relevant to RBR.

They go to the heart of the issue.

For respondents 1 to 13047 it was Kerry, 51-48
For respondents 13048-13660 it was all RBR.

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x363938

				National Exit Poll Timeline					
	4:00pm	7:33pm	12:22am	1:25pm		4:00pm	7:33pm	12:22am	1:25pm
	8349	11027	13047	13660		8349	11027	13047	13660
		Category Weighting	       Kerry percentage		
VOTED IN 2000 FOR
NoVote	15	17	17	17		62	59	57	54
Gore	39	38	39	37		91	91	91	90
Bush	42	41	41	43		9	9	10	9
Other	4	4	3	3		61	65	71	71

[Febble]

13. Because they reweighted the data.

as new information (not simply new responses) came in. We know they did. It is all in the E-M report.

The interesting question is why these initial weights were wrong. We know it wasn't precinct sampling, because E-M tell us that was pretty good (actually a slight Bush over-estimate), so it must have been that the voter-sampling didn't match the vote counts.

The fact that they reweighted the data later tells us nothing about why the first weights were out. All we know is that they were.

It does not go to the heart of the rBr issue. It's an epiphenomenon.

What goes to the heart of the rBr issue are questions as to what precinct characteristics were associated with what degree of mismatch (I'm choosing my words carefully here, as you don't like "wrong".)

Did precincts in swing states have greater errors? Did precincts with touch-screens have greater errors? Did precincts in Ohio have greater errors?

Did precincts with inexperienced interviewers have greater errors? Did precincts where the pollsters have to stand a long way from the precinct have greater errors? Did precincts in shared polling places have greater errors?

We have some of these answers in the E-M report, but they are all potentially confounded by the unresolved problem with the WPE.

But if the answers to the first set of questions are "yes", will you infer fraud?

And if the answers to the second set of questions are "yes" will you infer sampling bias?

Or will you cherry pick?

[TruthIsAll]

16. The first weights were "out"? What do you mean by that? They were wrong?

"The fact that they reweighted the data later tells us nothing about why the first weights were out. All we know is that they were".

No, we don't know that they were "out". Only you seem to feel that way. The first weights were essentially correct. The FINAL weights were out.

You say the first weights were "out". But which ones? The time-line of respondents: 4pm: 8349, 7:33pm: 11027, 12:22am 13047? Were they also "out" for the confirming state exit polls as of 12:22am?

What about the FINAL at 1:25pm?
We know the 43%/37% was "out? I proved it.

Why do you say the early weights were "out"?
What is the basis for that statement?

Is that just a corollary of your careless wording of "the exit polls were wrong" even if that's not what you really meant to say?

[Febble]

19. I said OUT because you don't like WRONG

They didn't match! So they were reweighted!

[TruthIsAll]

24. They didn't match WHAT? THE VOTES? WHAT IF THE VOTE COUNT WAS WRONG?

.

[TruthIsAll]

17. When I reweighted the state data to the votes, Kerry won..

And the sum of the state re-weights matched the national exit poll to within a SIX one-hundredths of a percent.

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x362208

[Febble]

20. Well in that case

you were using an inappropriate number of significant digits.

[TruthIsAll]

23. You are flailing, Febble. I show the results to the nearest .01%

because that is what they are.

ARE THE CONFIRMING RESULTS TOO ACCURATE FOR YOU?

[Febble]

26. Yes, they are too accurate.

you need to give a confidence interval.

[TruthIsAll]

28. Dupe delete

.

[TruthIsAll]

29. Is the standard 95% C.L. good enough for you...

or would you insist on the totally inappropriate 99%, like MP did?

National Exit poll (Gender demographic)
MoE = 1/sqrt(13047) =0.86%

How Voted in 2000 demographic (3168 respondents):
MoE = 1 / sqrt (3168) = 1.77%

The State Exit Poll MoE's range from 1.84% (FL) to over 4% in smaller Red states. Of course, the MOE's for the weighted national average (73,000 respondents) is smaller than 1.84%.

If you go through the links to my threads, you will see EVERY SINGLE INPUT AND CALCULATION.

[Febble]

31. 99% is appropriate for 50 states.

95% will do for now. You might like to get rid of the extra decimal places though.

But I'm convinced. The polls over-stated Kerry's vote. The question is why?

[TruthIsAll]

33. 95% is the standard; we are looking at 2 standard deviations from the mean

whereas 99% C.L. (2.58 std dev) is overkill for the task and is never used in polling (check the pre-election pollsters, they always use 95% as their basis for the required MoE.

The 99% C.L. is only applicable for calling election winners by the networks - when they were legit.

[Febble]

34. If you have fifty states

2 or three of them will be outside the 95% margin of error.

[TruthIsAll]

35. The prob = 1/40 (2.50%)for a one-tail test , and the law of large numbers

takes effect.

Assume the average Margin of error is 2.50% at a 95% CL for each state (1500 respondents). The National weighted MoE is much lower than 2.50% for the 73,000 respondents over the 50 states.

The National Exit Poll MoE is under 1% for 13,000 respondents

[Febble]

36. All that is true.

All I'm saying is that you would expect 2 or 3 states out of fifty to have a z score of greater than 1.96 or less than -1.96. How many did you get?

[TruthIsAll]

37. Here is a Z-score probability analysis

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x361617

[Febble]

38. And what factor did you use

to correct for cluster sampling?

[Name removed]

40. Deleted message

Message removed by moderator. Click here to review the message board rules.

[TruthIsAll]

42. The same one that Mitofksky used in the National Exit Poll: ZERO

Not ONE word about cluster. Read the notes.



http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=357345

[Name removed]

43. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Febble]

39. OK

forget the standard, forget tails, forget even the extra variance you need to add for cluster sampling, let's just ask the question. On your z scores, you'd expect 2 or 3 states to be more than 1.96 standard deviations from the mean.

Actually you got 16. What's more they were all on the positive side of the mean.

Josh Mitteldorf did a similar analysis in the USCV paper. Actually he and I worked on that bit together.

I agree. It is completely clear that even if any one state was within the MoE (which they all are if you use a two tailed test, which you should do, as the Kerry skew is post hoc, not a priori, and use an appropriate design factor) that Kerry's vote was overstated.

I have never queried this. It is blindingly obvious. The polls did not match the count, and the differed from the count by a margin that is well outside the margin of error.

If you simply look at the WPEs in the E-M report it is even more absurdly huge. I think I worked out that the probability of that amount of error being due to sampling error is something like 1 in 9 billion.

But the size of the discrepancy is not in itself evidence that the problem was in the count. It was not sampling error. We know that. It might have been fraud. It might have been differential non-response. But no amount of significance in the significant difference between two quantities will tell you which was wrong.

[Name removed]

3. Deleted message

Message removed by moderator. Click here to review the message board rules.

[kansasblue]

6. fine... I'll move on to 2006

But before we get there we'll make sure the lessons (and fraud) of 2000, 2002, 2004 are corrected.

[Name removed]

7. Deleted message

Message removed by moderator. Click here to review the message board rules.

[kansasblue]

4. I see

Sorry TIA, I now see you are talking directly with Febble. Great!

[autorank]

5. This line explains so much about election fraud. Kick & Recommended

Fact.
The final 613 had to have voted 100% for Bush to overcome Kerry's lead among the first 13047.

This tells us the contortions "management" can extort from pollsters and the intellectual honesty of those folks.

You work demonstrating this fact is earth shattering for our political system and is a fact that also represents a proof that there was incredible manipulation of the entire electoral process. If this can be extorted out of not-so-neutral third parties, what was "management" extorting out of election officials?

Great stuff!!!!!!!!

On edit:kick: & Recommended :kick:

[TruthIsAll]

8. No one cares to discuss this. It's not rocket science, either.

They get all wrapped up in conjectural RBR self-flagellation, running models without using REAL data, and refuse to do the simple arithmetic in analyzing the FULL National Exit poll, from 8349 respondents (good) to 11027 (good) to 13047 (good) to 13660 (bad).

Why do they avoid these simple questions arising from the 2000 voter demographic? They refuse to speak about the incongruities in the data.

Why don't they address THIS issue before moving on to anything else? They avoid discussing this at all costs. But that's not news; they have consistently avoided all the relevant issues.

Let's see if Febble will address this one.

If it cannot be refuted, there is NO point in going any further.

RBR is dead.

[Febble]

14. Do you know the mean response rates

for 2000? This year they were 53%.

That means that more than 40% of people who should have answered the questions either didn't answer (around 35%) or weren't asked (around 10%).

With non-response rates like that, you are incredibly lucky if you don't end up with a biased sample. It is no longer a random sample, and all statistics, including the statistics on which the exit polls are based, assume random sampling.

All voters are reluctant. Is it so unlikely that some will be more reluctant than others?

I work in behavioural sciences, and I worry if my non-response rates drop below 90%.

[TruthIsAll]

15. So far, your response rate in explaining the 19 facts is ZERO.

Let's discuss the FACTS stated in this post.
Not hypotheticals.

[Febble]

18. I give up.

Well, one last go.

Reweighting does not mean adding new samples. It means reweighting the data you have in the light of new information. It is, as you point out, mathematically impossible for the final sample to have been so different from the previous two. Quite apart from anything else, the MoE would go through the roof as the between sample variance would vastly outweigh the within-sample variance. So what happens is that all the data is reweighted. Not in the light of the final sample, but in the light of other information. Mostly turnout, I would guess. It's a dynamic process apparently. It's not a secret that its done. It's not like a horse race where the bets are off after the starting gun. They keep adjusting the odds as it becomes clearer and clearer which horse is going to win. Especially after 2000 where they called Florida for Gore then Bush (or was it the other way round). They want to get it right.

But I don't expect exit-poll companies like to have to do a lot of reweighting as it means their initial weights must have been "wrong".

The extent of the reweighting just tells you the extent of the problem you have to account for.

The reason could be sampling bias or fraud. Take your pick.

[TruthIsAll]

21. Your responses have me convinced.

"I give up" is a reply that I have heard often at DU - when the facts of this election stand in sharp relief and cannot be refuted by pure verbiage.

You refuse to acknowledge the FACT that the FINAL 13660 weightings are pure fiction.

Whether the argument is that the MoE is much higher due to cluster; or that the discrepancy always shows a Democratic bias, so 2004 was nothing new; or that the exit polls are not designed to predict the vote; or that Bush voters came out late; or that weather conditions favored the Dems ; or that females voted early for Kerry; or that there was a Bush bandwagon effect; or that the fundies came out in mass; or that 120 million votes are more accurate than 13,000; or that 13,000 is an insufficient sample; or that exit polls are not designed to prevent fraud (as in the Ukraine); or that Bush is a War President; or that Bush voters are reluctant to speak to pollsters - we have heard them all.


Specifically, you do not question the IMPOSSIBILITY of the FINAL 13660 NATIONAL EXIT POLL 43%/37% WEIGHTINGS of How Voted in 2000.

Of course, other demographic weights were changed as well, BECAUSE IF ONE DEMOGRAPHIC SHOWS THAT KERRY WAS A WINNER, THEY ALL WILL.

LIKE PARTY-ID, WHICH MORPHED FROM 38% DEM/ 35% REP TO 37%/37%.

IF THE WEIGHTS WERE NOT ALTERED, THEN THE PERCENTAGES HAD TO BE:
THE GENDER DEMOGRAPHIC WAS CONSTANT AT 54% FEMALE/46% MALE FROM 11027 TO 13660. BUT KERRY'S 54% FEMALE VOTE CHANGED TO 51% FOR THE FINAL 613 RESPONDENTS FROM 13047 TO 13660.

You continue to question the accuracy of the 8349, 11027 and 13047 weightings, which barely changed. Kerry won them all.

You have not said ONE word about that.
To avoid these FACTS is not very scientific.
Why am I not surprised?

[pgh_dem]

22. Before you both give up, could you answer two quick questions?

Do you (TIA and Febble) agree that the raw precinct-level data (with each precinct clearly identified), as well as data supporting the periodic re-weighting decisions, is necessary to settle this argument?

If so, do you (Febble) agree that the withholding of this information by E-M for almost 7 months (through Conyers' forum, the Arnebeck/Cobb/other Ohio lawsuits, the Jan 6 challenge, Bob Ney's *cough* hearings, the Baker/Carter Sideshow) at the very least tilts consideration toward fraud as an explanation?

In other words, E-M's reasoning behind withholding of such necessary information is so indefensible ("that information belongs to our clients...") in light of the serious questions the 'sample bias' has directed towards the integrity of the election, that the only reasonable explanation of these discrepancies will jibe far more with TIA/USCV's conclusions than the current rBr theory?

(Note: i recognize that you have clearly stated you are not a proponent of that rBr theory, just that your calculations suggest that it is not as improbable as TIA/USCV have stated).

[Febble]

25. OK

I agree that precinct level data is needed to answer these questions. I'd be happy with precinct WPEs, response rates (missed and refusals), and actual vote counts, together with some precinct characteristics, although I can see that some of these are sensitive data).

I'd also be happy if E-M would issue a report written as a scientific paper, with methods and results, and particularly if they would use my transform on the WPE before analysis.

E-M have already released, as promised, precinct level data. However it does not show the early weights, and does not identify precincts, so it is of very little use. I do understand the confidentiality issues, and don't agree that the withholding of this information is in itself evidence of fraud.

I think the biggest evidence of fraud is in Blackwell's illegal obstruction of a proper recount in Ohio. You have to ask what he was trying to hide. I also consider that voter suppression in Ohio was massive (see my paper on the USCV site here:

http://uscountvotes.org/index.php?option=com_content&task=view&id=65&Itemid=43

I think that the E-M findings that WPE was greatest in precincts where sampling protocol was most likely to be compromised is a strong finding, and supports the rBr hypothesis. However because of the problems inherent in the WPE as a variable, I'd be more convinced if the analyses were re-run using my variable.

Hope this answers your questions.

Lizzie

[pgh_dem]

27. I gotcha, but...

I'm afraid I haven't seen any plausible explanation as to why this is 'sensitive data'.

Is the suggestion that the results from very small precincts/sub-samples could positively identify a specific voter (like "there's only one chinese guy in such and such precinct 4B, and this report says 1 chinese guy voted for Kerry, therefore...")?

If so, that's a pretty far-fetched argument (not claiming it's yours...i made up the exmple), and wouldn't be material to denying legal requests (a la Arnebeck) or congressional committee requests (Conyers/Carter/et al) provided they could be obtained with confidentiality.

At any rate, I agree that Blackwell's obstruction of the recount is 'slap-in-the-face' evidence of fraud, but I feel that E-M's withholding of fully identified precinct information is no less an obstruction to resolving these concerns.

Thanks for responding, btw

[Febble]

30. Yes, that's exactly the point.

it wasn't one that had occurred to me but I deal with sensitive data myself, and I have to be very careful that nothing I publish can allow individuals to be identified.

The other reason given is that because pollsters like to use the same precincts year after year, it allows parties to game the system. There is potentially an advantage to manipulating the polls if it encourages your own side to vote and the other side not to.

I think there are ways round this, however, and maybe they will be got round.

But I don't think Arnebeck or Conyers would know what to do with the data. They'd need a statistician - and the right one! It would be much better to have some "blanched" data set available that could be analysed independently and the analyses subjected to peer review.

I don't think this is the end of the story though. I think more analyses will be done somehow, even if only by academics proposing hypotheses that they want testing and having the owners of the data test them. That won't satisfy the truly paranoid, but it would satisfy me, and would be rather good actually - you would have to have a true "a priori" hypothesis. It would not allow data dredging.

[bleever]

41. By the way, in case no one else has said so,

Welcome to DU.

Thanks for coming.

[TruthIsAll]

32. I agree.. but I also believe that the NEP time line from 8349 to 13660

along with the state exit data from 12:22am have already provided the answers. I have demonstrated this by analyzing many different permutations and combinations of the data - which lead to only one conclusion.

The precinct level detail would just be icing on the cake.
And that's why they won't let you see it.

[TruthIsAll]

44. Kick for more discussion

.

[pgh_dem]

45. Thanks to both of you TIA & Febble for replies

I don't know, TIA. My opinion of both of your analyses (as bargain-basement as my knowledge of statistics is) is that you've both generated a very detailed icing job on the cake. In essence, you've both been working at defining the shape of the unknown, and Febble seems to agree that you could be icing the same shape.

It's a bit goofy of a metaphor, but I think the precinct-level data (with enough specificity to make it useful) IS the cake, and the refusal to release this data (or even share with trustworthy sources) is just more icing giving away the shape. In other words, withholding that information through lawsuits and congressional inquiries really suggests something very damaging being hidden.

Contrary to an earlier post of mine where I stated that Blackwell's conduct was 'evidence' of fraud, I guess it would be more accurate to say that is also 'icing', revealing the shape of what is hidden.

To take a different tack (assuming the 'sensitive info' point is insurmountable), do you think that releasing the precise reasoning and mathematical model for the late-night re-weighting would give sufficient information to answer both your concerns?

If that's not clear, what I'm saying is I think you both agree that TIA proved the final 1,300 votes or so mathematically cannot produce the end results, therefore the weighting of the entire sample must have been changed to come up with those numbers. So, E-M must have had a point in which they decided their numbers were "wrong"/"out"/"mismatched", and they changed their entire weighting model, and there *should* be some record of that decision.

What I'm asking is would a complete record of that decision provide enough detail as to their raw numbers without such granularity as to compromise individual voters' identities?

Related note: I have to reject the reasoning that revealing polling locations would allow parties to "game the system" in subsequent elections. I think E-M has been proven at the very least to have screwed this job up so badly that they're going to need to change their entire methodology for subsequent elections anyway.

[Febble]

46. We don't need the weighting

We've got the WPEs.

I'm sure TIA did a good job figuring out that a reweighting occurred. I did it myself. It is no secret. The entire E-M report is devoted, in effect, to why it was necessary.

It would be good to have the whole dataset. I think it could be done if it was stripped of sensitive data, then it could be publicly available, and analyses subjected to peer review.

In lieu of that, I think it would be possible to formulate specific hypotheses and ask for specific analyses to be run. I do not think E-M are corrupt. I don't think we have any evidence for that at all.

I do think that using my index would make further analyses more informative.

Nothing in my paper supports or refutes rBr rather than fraud. It supports rBr rather than "Bush strongholds have more vote corruption", which was the inference drawn by USCV.

I do, as I have said, think it unlikely that reanalysis will undermine the strong E-M finding of an association between error and compromised sampling protocol. But that is merely an opinion, though informed of by knowledge of the nature of the confound presented by use of the WPE as the quality control variable. I could be wrong.

[pgh_dem]

47. Don't know about that (RE: E-M corruption)

I think an 'excited utterance' like Mitofsky himself encouraging skeptics to 'eat shit' constitutes reason to be concerned about the ethics of E-M's conduct with regard to the exit polls.

Not to impose my own ethical structure on the world or anything, but if I had run such an error-riddled, second-guessed poll, coming on the heels of the national polling collapse in 2002 and massive f-ups in 2000, I would have invited a peer-review process in November 2004, and not been still considering it in May 2005.

In a sane world, my international reputation for exit polling business would be on the line, and would outweigh the proprietary interests of my clients (though I would also bend over backwards to get them to consent to the peer-review process).

Of course, if it was me, I would have refunded my clients' money, too.

The fact that such information remains unavailable is, to me, evidence suggesting collusion with E-M's clients to protect either corruption or incompetence (or both).

Related: Do you think that it is possible to strip the sensitive information from the data without compromising its usefulness?

[Febble]

48. I agree that peer-review

would be preferable. I think it probably is possible to strip the sensitive information from the data without compromising its usefulness. But I do understand that there is a real issue of confidentiality at stake.

[TruthIsAll]

49. Confidentiality at stake? Or Democracy? Which is more important?

.

[intensitymedia]

50. criminal conduct of course prefers confidentiality! at all costs n/t

[pgh_dem]

51. Knowing your vote is private is important...

However, knowing your private vote is actually counted correctly is more important, as any reasonable suspicion renders the privacy counterproductive.

(Little tangent for anybody):I'm a little lost on why E-M's is trusted with this confidentiality. Is there any legal contract with federal or state governments with enforceable penalties for violation of confidentiality? Would anyone even investigate it or charge them if they did violate these laws? Is turning over raw data to their 'clients', the network media, in violation of confidentiality?

Those are just things I'm curious about. You answered the question I thought was really important, that you feel it is possible to strip sensitive info from the data, and release it for public investigation/consumption. I really don't want to put you on the spot, Febble, nor make it seem like I expect you to defend E-M or a "no-fraud" position (I'm aware of your continued activism in this matter), but I'm curious as to your response to these questions:

1) If stripping confidential data may be possible, or at least peer-review by recognized authorities, how would you characterize E-M's reticence (to be kind) as to pursuing either path?

2) If you feel there is a reasonable 'above-board' explanation for withholding (really 'hoarding') this information, is there a time-frame where it would make you suspicious of their motives? (e.g. info under lock and key until 2009, 2013, 2035?)

Again, I'm not suggesting that E-M just open their safe to any passers by, just positing that Mitofsky's personal vehement defensiveness coupled with the company's iron grip on the *useful* data... well, let's just say it doesn't avoid the "appearance of corruption"

I totally understand if this is something you just don't want to get into, and I read your position as simply feeling your noise-reducing function contributes to a greater understanding of how the poll numbers could be explained, in the absence of the actual data. Give me a thwap onna head if I'm wrong. :spank:

[Febble]

53. I have an inbuilt sympathy with the position

as I ask people to sign consent forms to participate in studies the whole time, and one of the things I assure them is that there will be no way that they can be identified from any published study. I rarely publish individual data anyway, its usually been "cooked" in some way - turned into averages and so forth.

Response rates for the exit polls are very low. I suspect they would be lower still if people thought that their answers would be publicly available.

Of course E-M is trusted with the confidentiality - they collected the data. It is their responsibility to keep it confidential.

I would read nothing into their reticence at all.

Which is not to say I wouldn't give my eye-teeth to get at the data. I think our best bet is to present hypotheses for testing.

Anyway, thanks for asking. I am not going to hang out here much longer, for fairly obvious reasons, but I really appreciate your civility.

[TruthIsAll]

61. It was not obvious on 5/2; we did not know then that you were an E-M shill

"Anyway, thanks for asking. I am not going to hang out here much longer, for fairly obvious reasons, but I really appreciate your civility."

[davidgmills]

52. A lawyer's take on confidentiality

Certainly there is no right to confidentiality in the legal sense. You tell your vote to a pollster, your right to have your vote be kept secret is waived.

That hypothetical sole Chinese female voter in precinct A whose vote might be identified certainly knew that her secret vote could be at risk when she was asked to the poll and she consented to taking it.

This is not like the attorney/client privilege, doctor/patient confidence, cpa/taxpayer confidence. Not close.

This is not on the same level as a research study of mental patients. Not anywhere close.

However, the appearance of confidentiality may enhance the percentage of response. Suckers.

[autorank]

54. PLEASE JUST STOP!

It is pure torture to see this diversion taking place. There was FRAUD in the American election, just as their was FRAUD in the legal justification for bLiar's war crimes. This isn't an academic debate about who can see what when. There are no boundaries of confidentiality when it comes to ELECTION FRAUD AND THE THEFT OF AN ELECTION. Move here, become a citizen, then you can chime in but stop diverting attention FROM THAT WHICH HAS ALREADY BEEN DEMONSTRATED WITH SUFFICIENT INTEGRITY TO WARRANT A FULL FEDERAL INVESTIGATION.

[kster]

55. Ya what "autorank" said......n/t

[passy]

60. Are you inferring that the Exit poll was wrong, because it wasn't.

The final exit poll with 13660 respondents was correct in predicting * to be the winner of the 2004 presidential election.
All the other exit poll number have entered the realm of unreality.
They simply do not exist beyond forums such as DU.
The clients have what they wanted, something that looks like it did the job.
The majority of people do not question the exit poll that shows that * won.
That I believe, is the problem we face here.

[kiwi_expat]

56. We can now match the NEP data to actual precincts.

I realize that the two most important missing pieces are NEP's pre-final weightings, and Mitofsky's weighting model. But we can now match the available NEP data to actual precincts. That should be of some use.

Thanks to skids and others on minvis' "Ohio Exit Poll Raw Data" thread, we have discovered that the NEP respondent data (from the Univ. of Mich.) is in sequence by COUNTY NAME within each state. Within a county, the NEP numbers are in sequence by CITY NAME (including the number-letter alphanumeric code following the city name).

See:
http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=
203&topic_id=329749&mesg_id=342855&page=


So far, liam_laddy and minvis have obtained the NEP precinct names for Ohio's Butler, Cuyahoga, Clermont and Hamilton counties, from the counties' BoEs. We have identified 16 Ohio NEP precincts - including 5 of the 12 Ohio NEP national-data precincts.

It would be fabulous if some high-powered researcher were able to pry the NEP precinct names from the states' SOSs. :-)
(Not likely in Ohio or Florida, unfortunately.)


Cheers and welcome to DU !
kiwi

[kiwi_expat]

57. Correct link...

The correct link to skids' NEP Ohio counties' precincts table:
http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=329749&mesg_id=342855&page=

Also, a link to blue22's Suggested Priority List for Decoding Remaining Ohio Precincts (based on raw-to-weighted %Bush gain):
http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=329749&mesg_id=342982&page=


Following is some data for the Hamilton County NEP precincts (precincts 52, 55, 57 and 63 are NEP national-data precincts)

****NOTE: liam_laddie will be examining the Cincinnati 4-M ballots, using Ohio's Sunshine laws****


...................Kerry Raw-NEP|K polling-place|Kerry loss
(NEP#52)CINCINNATI 4-M 68% 46% 22%
sample = 31
sample/poll place = 2%

(NEP#57)HARRISON C 43% 29% 14%
sample = 28
sample/poll place = 5%

(NEP#55)FAIRFAX B 50% 37% 13%
sample = 16
sample/poll place = 4%

(NEP#49)ANDERSON JJ 32% 24% 8%
sample = 34
sample/poll place = 2%

(NEP#51)CINCINNATI 22-E 88% 82% 6%
sample = 53
sample/poll place = 6%

(NEP#63)SHARONVILLE 4-C 30% 30% 0%
sample = 48
sample/poll place = 2%

(NEP#54)EVENDALE D 25% 33% -8%
sample = 52
sample/poll place = 3%


2004 HAMILTON COUNTY CANVASS
(Polling Places)
....................Bush Kerry Kerry%
CINCINNATI 4-M 312 211 40% ***
CINCINNATI 4-B 190 180 49%
CINCINNATI 4-I 207 199 49%
CINCINNATI 4-O 237 218 47%
946 808 46%

ANDERSON JJ 344 131 28% ***
ANDERSON QQ 459 132 22%
ANDERSON VV 347 106 23%
ANDERSON DDD 184 59 24%
1,334 428 24%

SHARONVILLE 4-C 321 150 32% ***
SHARONVILLE 3-B 382 198 34%
SHARONVILLE 3-C 271 117 30%
SHARONVILLE 4-A 480 188 28%
1,454 653 30%

EVENDALE D 235 89 27% ***
EVENDALE A 330 201 38%
EVENDALE B 339 178 34%
EVENDALE C 312 110 26%
EVENDALE E 116 91 44%
1,332 669 33%

CINCINNATI 22-E 65 373 85% ***
CINCINNATI 22-A 94 352 79%
159 725 82%

HARRISON C 209 108 34% ***
HARRISON E 217 69 24%
426 177 29%

FAIRFAX B 279 169 38% ***
FAIRFAX A 277 153 35%
556 322 37

[TruthIsAll]

58. I'm tired tonight. I think I'll just kick that rGr in the ass...

.

[kster]

59. kick.nt