Dr Ron Baiman: "CLEARLY A CRIME WAS COMMITTED IN OHIO"

[mod mom]

Dr Baiman posted this on another thread here but since he is new to DU, he can not post an original thread. I think it got lost in the numerous posts and think it needs to be front and center:

*********
"I would take this evidence to a trial. Clearly a crime was committed in Ohio. There is simply no other explanation for these patterns other than vote shifting. The only thing we don’t know is who did it and how. And exactly this kind of information is necessary to get serious electoral reform - that you claim to support."

RON BAIMAN

Ron Baiman is currently a Policy Research Project Development Analyst at Loyola University in Chicago, as well as an visiting assistant professor at the University of Chicago. He holds a Ph.D. in Econonomics from The New School for Social Research.





In his June 7, 2006 reply to Kennedy’s rebuttal of his earlier critique Farhad Manjoo citing Mark Blumenthal, claims that:



a) The exit poll margins of error for Iowa, Nevada, New Mexico, and Ohio were between 5% to 7%. This is preposterous. Rather than relying on Mark Blumenthal (an unreliable source for quantitative analysis), I urge Manjoo to download the National Election Pool a “Methods Statement” for the Edison Mitofsky (EM) exit polls (produced on Nov. 2 2006) at:



http://www.exit-poll.net/election-night/MethodsStatementNationalFinal.pdf



The second page of this statement sets 95% confidence intervals for these polls (for a “characteristic” held by roughly 50% of those polled, for example a Presidential candidate preference for which there is a close to even split) squarely at 4% for sample sizes of 951-2350 – the range of reported sample sizes for these states. However, as Blumenthal knows, the reported sample sizes (also in the methods statements) are about half of what they really are (see Mitofsky correspondence in Baiman June 5 Free Press AAPOR report). For these true doubled sample sizes of 2351-5250, NEP’s own estimated confidence interval falls to 3%. This clearly puts the Ohio discrepancy of about 4% outside of the margin of error - even using NEP's inflated margins of error.

My margin of error calculations (and I believe Freeman’s) find a 2% margin of error with a 30% cluster adjustment factor. As I have stated in my earlier response to Manjoo, this puts Ohio well outside the margin of sampling error with odds of less than 1,900 that Kerry’s reported result is true given the exit poll result. This is not “slight” evidence but rather highly statistically significant, especially one considered with the inexplicable pro-Bush exit poll discrepancies in the two other key battle ground states of Florida and Pennsylvania. As Freeman and I have stated, the odds that these “sampling errors” (in the same direction and of these magnitudes) would occur for these three states simultaneously in less than one in 182,000,000 (i.e virtually impossible - this number is based on doubled sample sizes). Moreover, when one looks at precinct level exit poll data , and not just aggregate state polls, the evidence in even more striking and inexplicable. A fact that Manjoo has not addressed at all.



<snip>

http://www.Baiman.blogspot.com


His posts can be found on this thread:
http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=431591&mesg_id=431591

[drm604]

1. K&R

:kick:

[babsbunny]

2. Will Keith Olbermann ever cover this?

[mod mom]

4. I think he's been told no. Remember how he all of a sudden stopped

coverage after the '04. Nothing about the GAO Report. Nothing about RFK Jr. You have to wonder.

[bleever]

43. He's quoted in the RFK article. My take is that while he's not under

any "prohibition" to cover the story, and still finds it has merit (per his remarks to RFK), he needs something very substantive that rises above the level of what we have so far. I had hoped that RFK's RS article might be such a watershed moment, and it yet may be.

My guess is that KO needs to preserve his credibility and claim to non-partisanship in the eyes of the viewing public, so his threshold for a story like this is very high, but I still have faith in him. I expect to hear e-voting stories out of him in the next few months, but this is only my guess as a regular viewer.

[Peace Patriot]

3. Thank you so much for this post! I had missed it entirely--due to

blogging for Bowen and complications at work. I am so glad Baiman has weighed in on this ignorant, lazy-minded salon.com article. As I recall, "salons" were where the rich dabbled in poetry and ideas. That's what salon is doing here--DABBLING--playing salon-ish games with the fate of our democracy. We need some tough-minded, knowledgeable advocates--and Ron Baiman is certainly that. What a hero!

For Ron:

:patriot: :applause: :applause: :hug: :grouphug: :hug: :applause: :applause: :patriot:

[autorank]

5. Ron Baiman is a gifted analyst. His work proves it. K&R

He also has a blog where the so called experts who challenge Kennedy can show up and try out their
theories. Ron also comes here on occasion and engages in dialog. His entire srticle is excellent.

[mod mom]

7. Right you are, Auto, he is highly respected. Wish TIA was here to

comment.

[OnTheOtherHand]

6. why this is wrong

First of all, Manjoo is discussing a margin of error for the difference between Kerry and Bush. So, if we otherwise accept Baiman's logic, the margin of error should be 6 points, not 3 points. By spooky coincidence, this is right in the middle of Blumenthal's "preposterous" range.

Second, Baiman attempts to refute a claim that Manjoo didn't make. Manjoo wrote that "John Kerry's lead was well within the margin of error in Iowa, Nevada, New Mexico and Ohio (the four states in which the exits showed Kerry ahead that he eventually lost)." In other words, Kerry's margin was smaller than the margin of error. Let's take Ohio. If we use Freeman's original estimates based on CNN screen shots, the exit polls gave Kerry a margin of 4.2 points (52.1% to 47.9%) in Ohio. Notice that this margin is less than the 6-point margin of error on the difference.

If we use the E/M evaluation report, we have our choice of two estimates. The Best Geo estimate based on interviews alone gives Kerry a 6.5 point lead with a standard error of 3.9 points for the difference. The nominal 95% margin of error would be about double that, i.e., plus or minus 7.8 points. The Composite estimate incorporating pre-election expectations (which is probably the estimate reflected in the CNN.com screen shots) gives Kerry a 3.4 point lead with a standard error of 2.6 points, or a nominal margin of error of plus-or-minus 5.2 points.

So, that's three different calculations with the same result: Kerry's lead was within the margin of error in Ohio. I will leave the other three states as an exercise.

Baiman makes much of the fact that the exit poll discrepancy cannot be attributed to sampling error alone. As far as I can tell, this is the fact he thinks "that Manjoo has not addressed at all." This is silly. As Manjoo pointed out in his original article, "nobody argues the errors happened by chance. Everyone in the exit poll debate agrees that there was a systematic cause for the errors in the poll. Freeman, Kennedy, et al., claim that the systematic cause was fraud, while Mitofsky and many in the polling community claim the cause was a problem with the poll."

[Joe Chi Minh]

8. "Everyone in the exit poll debate agrees that there was a

systematic cause for the errors in the poll. Freeman, Kennedy, et al., claim that the systematic cause was fraud, while Mitofsky and many in the polling community claim the cause was a problem with the poll."


Actually, a "systematic cause of error" sounds to me an extraordinary confusion of the concepts of agency, causation, commission, creation of a pattern, aleatory error, deliberate errors, false and distorting figures. A systematic occurrence of errors, however, in the sense of their forming a pattern, is an absolutely prime indicator of fraud. Fraud is proved in the courts not by a single fact, but by a marked and discernible pattern. The patterns involved here could scarcely be clearer or more statistically pronounced.

On the other hand, in the absence of a plausible explanation, the claim of "Mitofsky and many in the polling community" that the cause was a problem with the poll, is pure conjecture. Just like the conjecture that Republicans had been less ready to state who they'd voted for - when the reality proved to be the reverse. Pulling a gratuitous notion from the sky. Why would the polling of a man who'd made polling virtually an exact science all over the world, suddenly go to pieces, on his own turf? I don't believe so much as a half-plausible explanation for Mitofski's brain suddenly turning to mush has been adduced even at this late stage. You'll have to do better than that, OnTheOtherhand.

A word in your shell-like, OnTheOtherHand. * the other hand. Truth is not an equal-time resource.

[OnTheOtherHand]

9. what "pattern" are you describing?

Some "patterns" I see in the exit poll results are that red shifts were generally larger where interviewers were further from the precinct entrance; larger in more Democratic states; larger among interviewers under 35 with 1 to 3 years of college (I'm guessing most of those were actually college students). Are these absolutely prime indicators of fraud? If so, why?

Non-patterns I see in the exit poll results are that red shift is not larger on average where Bush improved most on pre-election polls, and is not larger on average where he improved most on his 2000 performance.

Thus, the "conjecture" of a problem with the poll seems better supported by patterns in the data than the conjecture that the poll evinces fraud.

"Why would the polling of a man who'd made polling virtually an exact science all over the world, suddenly go to pieces, on his own turf?"

You appear not to know what you are talking about. Mitofsky's reputation for accuracy rests on avoiding unjustified inferences from inadequate data. As has been noted many times, the raw error in the 1992 U.S. exit polls was almost as large as it was in 2004, but Mitofsky made no wrong calls. You are welcome to argue that in fact the exit polls were accurate in 1992 as well, but you should at least argue it -- not just handwave about "exact science."

People wonder why I would write about exit poll fundamentalism. Well, it's because I hear over and over again that exit polls are accurate, and that poll error theories are just a guess. When I point out that exit polls haven't always been accurate, and that there is specific evidence that tends to support poll error theories, people either get angry, or they go silent for a while... and then come back and say that exit polls are accurate, and that poll error theories are just a guess. It seems an awful lot like creation science to me.

[althecat]

10. This pattern

b) Manjoo’s claims that exit polls are not always accurate and that Mitfosky has an “explanation” for the large discrepancies (that he acknowledges were both highly significant statistically and the largest since 1988 in the January report) are both irrelevant when one considers the pattern of the exit poll discrepancy in Ohio – see graph below:



These are “within-precinct discrepancies” or ((Kerry -Bush Official Vote) minus (Kerry – Bush Exit Poll shares)) for each of the 49 precincts for which exit polls were taken in Ohio based on data released by the Election Science Institute in a report in which Mitofsky is listed as an “assisting author”. (Note that there are not 49 bars as someprecincts have the same official Kerry vote share so that the bar is an average WPD.) Negative WPD reflects large Kerry exit poll “overstatements” relative to official vote counts. Note that almost all of the discrepancies are negative (against Kerry). In fact, as Kennedy reports, 20 of the 22 statistically significant discrepancies (large enough so that they could not plausibly be the result of random sampling error) are against Kerry. But even more striking is the pattern. On the right side in “high Kerry” precincts (above 57% official vote count – right of the red vertical line) there is a more or less random pattern of Bush (positive) and Kerry (negative) discrepancies, whereas on the left side (below 57% official Kerry vote – left of red line) almost all of the discrepancies (and all of the statistically significant discrepancies) go against Kerry. Moreover the precincts on the left are densely clustered whereas they are more dispersed on the right.


This strikingly non-uniform pattern cannot be explained by either large but “unbiased” (not one-sided) exit poll discrepancies – the pattern on the right, or by a “reluctant Bush voter response to the exit Poll” (“rBr” - Mitofsky’s hypothetical explanation) as this would not even produce the pattern on the left, not to mention the un-biased discrepancy on the right. Those who have been following the 2004 exit poll debate know that rBr would produce a “U” shaped pattern of discrepancies that are larger in more competitive precincts and taper off to nearly zero in highly partisan precincts. There is no such pattern on the left of the graph. But the pattern displayed is perfectly consistent with “vote shifting” from Kerry to Bush that would “move” precincts by their official vote count to the left thus producing large negative Kerry WPD and much smaller Bush WPD, and a clustering of precincts on the left. In fact Manjoo’s (or Lindeman’s) calculations of overall exit poll response averages are simply mathematically wrong, as they don’t take into account the larger proportion of Kerry or Bush voters in the “high Kerry” or “high Bush” precincts respectively – for accurate exit poll response rate estimates, and an explanation of why Mitofsky’s data contradicts his own “rBr” hypothetical see NEDA reports.


and this pattern

[Febble]

17. OK, let me unpack this

or, if you want the whole shebang, you can read my review here:

http://www.geocities.com/lizzielid/TheGunIsSmoking_Review.pdf

Ron's plot is actually fairly misleading, as I think he himself would agree. For some unfathomable reason, it is given as a bar chart, and where two precincts shared a value on the X axis (vote share) he and his co-author decided to pool the values. In the full paper they give a rather more informative scatter plot,



from page 62 here:

http://www.freepress.org/images/departments/1996_Stolen.pdf


And if you compare the two images, you can see that there are a number of precincts above the zero line in the scatter plot that simply do not appear in the bar chart, and also a point on the zero line at the extreme Kerry end that is presumably present, but invisible, on the bar chart (ironically exemplifying that the values do indeed "taper off to nearly zero in highly partisan precincts" as Ron claims is not apparent, but actually is right there in their own plot, bang on cue in the only truly "highly partisan" precinct in the plot).

If Ron wants to draw attention to a striking pattern he needs to refer to all the data points, not just the ones that show up on his selective bar chart, and if he wants to argue a significant tendency for the WPE to be more negative at the Bush end of the plot, he needs to give the slope of the regression line, and whether it is "significant". In fact it is borderline significantly different from zero (p=.05), but unfortunately for Ron's point, not only is it only borderline, but the "expected value" of the slope under the null is slightly greater than zero, which means it is not significantly different from the expected value.

Eyeballing a pattern and calling it "striking" (especially when your plot suppresses some important variance) when it doesn't actually reach statistical significance does not require anyone else to explain it. The point about a non-significant "pattern" is that it isn't one.

Ron appears to try to get round this by arbitrarily drawing a line at 57% and pointing out that there are fewer high points to the left of this line. Well, there are, but not as few as his bar chart implies - check out the scatter plot. And even if there were, then the obvious statistical test of the claim is a chi square, and on my reckoning, not even the chi square is significant.

Look, I've nothing against making a case for corruption in Ohio. I think Ohio stinks.

But I cannot believe, as I say in my review, that the case for an investigation into what happened in Ohio is best served by dodgy plots and insignificant p values masquerading as virtually irrefutable evidence! And if the thing does come to court, I certainly hope this stuff doesn't go into the witness stand.

Ron, if you are reading this: how the heck do you defend that bar chart?

As for your own plot, althecat, that tells a better story, although it's one we already know. The exit poll had significant redshift in Ohio. What we want to know is why.

[OnTheOtherHand]

20. a quick note on response rates

Manjoo was trying to make -- more politely -- more or less the same point made here:

Again, nonsense. There are two distinct questions in that paragraph, which are being deliberately conflated:

1. In each given polling place, what percentage of people who voted were willing to participate in exit polls?
2. In each given polling place, what percentage of the people who were willing to participate in exit polls were voters for each of the major parties?

The fact that a smaller percentage of people in places that tended to vote for the democratic candidate were willing to participate in exit polls is entirely independent of whether or not in a specific polling place a larger percentage of democratic voters than republican voters were willing to participate in the exit polls. This is a deliberate attempt to mislead readers about the meanings of the results - aka, a lie.

http://scienceblogs.com/goodmath/2006/06/election_fraud_or_just_bad_mat.php

While I don't endorse that intemperate language, it underscores that the argument simply isn't compelling or convincing. One need not take my word for it, or Manjoo's, or Chu-Carroll's. One can look around and observe that political scientists have not been convinced.* I'm a political scientist, and I spend a lot of time inviting people to convince me. They rarely make good use of the opportunity.

*EDIT TO ADD: Of course, what I really want is for people to think for themselves. Ideally, looking around would help people to determine that there is a disagreement worth thinking about.

[Name removed]

24. Deleted message

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

[OnTheOtherHand]

25. if you don't want to analyze exit polls, then don't

It's really that simple. If you can't come up with the 840,000 or so stolen votes in New York -- just for a start -- then the exit polls are not working for you, except as a rhetorical device. And, unfortunately, beyond a narrow circle they aren't working well as that, either.

But you don't need to write rubbish. The issue is whether the exit polls support fraud, not whether they "disprove" it. How many times can we torch that straw man, anyway?

Good luck out there.

[Joe Chi Minh]

26. How do you respond to this?

It's part of a March 2006 article by Ernest Partridge, entitled 'An Appearance of Guilt':

"My contention might be illustrated by this parable:

Suppose that a drug-sniffing dog at an airport identifies a suspicious piece of luggage. The customs officer then locates the individual whose name is on the tag, and orders him to open it. Now suppose further that this person then proceeds to do one or more of the following:

a) He denies that the luggage is his.

b) He calls his lawyer who presents an injunction against further inspection of the luggage.

c) He claims that he is a diplomat, and thus not subject to luggage inspection.

d) He offers a bribe to the inspector if he will "forget the whole thing."

Might one not suspect that the traveler was trying to hide something?"

You can read the whole article here:

http://www.crisispapers.org/essays6p/appearance.htm


It will evidently be a revelation to you, since you don't appear to have kept at all 'au fait' with the subject.

[Febble]

29. Do you have something to contribute

to this discussion, or do you just enjoy baiting a poster who takes a more skeptical view of the evidence than you do?

How many votes do you think were stolen in 2004? Hundreds? Thousands? Tens of thousands? Hundreds of thousands? Millions?

What evidence supports your view?

Do you think that it matters whether the evidence supports a particular number?

Or is any evidence that voter was disenfranchised enough for you?

Well, I'll give you my answer. I think that any evidence that any voter was unjustly disenfranchised is enough to cause concern, and that the amount of evidence we have is enough to tell us that American democracy is seriously broken.

I also think that the amount of evidence that digital voting technology is insecure and impossible to audit, together with the apparent laxity on the part of many BoEs when it comes to ensuring safe custody of the ballots, together with the lack of transparency about the whole voting process is enough to tell us that American democracy needs urgent fixing.

But some of us, and I expect OTOH is one, and I am certainly another, feel that a case is best made with watertight arguments, and that over-reaching claims run the risk of bringing ridicule on the cause. I think there is evidence that this has already happened.

And one of the easiest claim for anyone to refute is the claim that the exit polls demonstrate that millions of votes were stolen. They don't. It would actually be remarkably difficult to postulate a scenario in which millions of votes were stolen that is even consistent with the exit poll evidence, let alone supported by it.

But when I, or other posters, attempt to make that point, then all too often someone like you comes along, pours withering scorn on the attempt, dismisses detailed explanations as academic nitpicking, failing to see the forest for the trees. And meanwhile the whole election reform movement becomes a laughing stock. This makes me angry.

When exaggerated claims made using flawed arguments from dodgy data can be cited by people like Ken Blackwell to support the assertion that the cause of election reform is the preserve of a few tinfoil hatted conspiracy theorists delusional enough to think that Kerry won the election, then we bear some responsibility for allowing flawed and exaggerated claims to be made.

If you are happy to bear that responsibility, good luck. But meanwhile I will continue to point out where the evidence cited simply does not support the case. I do not think the case depends on proving that millions of votes were digitally stolen, and I happen to think the proponderance of the evidence is that they probably weren't. But the evidence that poor Americans are systematically disenfranchised is overwhelming, and it seems to me that we owe it to those disenfranchised Americans to get our arguments right.

KCDMIII, you make me cross.

[Joe Chi Minh]

30. "When exaggerated claims made using flawed arguments

from dodgy data can be cited by people like Ken Blackwell to support the assertion that the cause of election reform is the preserve of a few tinfoil hatted conspiracy theorists delusional enough to think that Kerry won the election, then we bear some responsibility for allowing flawed and exaggerated claims to be made."

That is patent nonsense. No matter how compelling and watertight the evidence, people in bad faith will continue to remain in bad faith. Not least, of course, Mr Blackwell. The idea that you take hs opinion seriously tells me an awful lot about your level of discernment I'd rather have remained ignorant of. Nobody deserves to be that gullible.

What you and our friend Mr Ontheotherhand need to understand is that when addressing an issue of fraud such as this, the circumstantial evidence is what proves the case - a pattern biased in a naughty direction. Not single issues - much as you would like to make the polling, or rather your laughable notion of the invalidity of the polling the nub of the issue. In much the same way as the single gunman apologists like to get people to concentrate on this piece of putative, but inevitably specious, evidence and not that. Rather than get bogged down with crafty obfuscation by them, it is the most lementary common sense to consider the staistical evidence of number of grassy-knoll witnesses who suffered bizarre deaths within the space of two years - and none of whose evidence was sought or revealed, I believe. The statistics of that happening without a sinister causation are literally are of an astronomical register. GO FIGURE!

Why, why, why have you not addressed Mr Partridge's article I cited? It should be as obvious to you as to everyone with an IQ above 60 that that alone is the most compellin evidence imaginable of fraud. Why fart around with pedantic and (amazingly, but for the nature of the Neocons) endlessly controversial, putative minutiae relating to the mechanics, so-called 'hard evidence'.

This wilful reluctance of the two of you to identify the hierarchy in the spectrum of incontrovertible-controversial evidence doesn't just make me cross. It infuriates me. Read Mr Partirdge's article.

[Febble]

31. Yes, I read Partridge's article

Some of it I thought was good. But the exit poll part I found flawed. If you want to know in which way flawed, I will tell you, but only if you ask.

And yes, I take the opinion of Mr Blackwell seriously. What I don't take seriously is your pretence that you think that I am gullible. If Blackwell can point to real flaws in the case against him, then that's to his advantage, not to the advantage of Ohio voters.

And if your argument is that even a watertight argument won't convince people, so you might as well use a leaky one, well I beg to differ.

As for "identifying the hierarchy in the spectrum of incontrovertible-controversial evidence" whatever that's supposed to mean, trying to sift good arguments from bad is precisely what we've been trying to do (if a spectrum can have a hierarchy - sheesh, do something about your metaphors, man...)

And the exit poll arguments are bad arguments. Terrible arguments. The evidence that the polls were biased is, pace Mr Partridge, not circular at all, and fairly compelling. The evidence that the discrepancy was caused by fraud is mostly unsupported (except, possibly, by the finding that in urban precincts, the discrepancy was higher where older voting technology was used, suggesting a possible role of differential residual spoilage in the discrepancy), and, if anything contra-indicative of widespread, massive fraud.

But go ahead and believe that the exit polls are evidence of fraud if you want to. Just don't expect anyone to believe your assertions in the face of lucid and informed arguments to the contrary, and don't be surprised if someone like Blackwell wriggles out of valid charges of malfeasance by pointing the gaping holes in the arguments that the exit polls evince vote-switching in Ohio. They don't, and it probably didn't happen. If Blackwell stole Ohio, it wasn't in any way the exit polls could detect.

[althecat]

35. Why do you want to beleive so strongly that exit polls cannot prove fraud?

Febble you say,

"But go ahead and believe that the exit polls are evidence of fraud if you want to. Just don't expect anyone to believe your assertions in the face of lucid and informed arguments to the contrary, and don't be surprised if someone like Blackwell wriggles out of valid charges of malfeasance by pointing the gaping holes in the arguments that the exit polls evince vote-switching in Ohio. They don't, and it probably didn't happen. If Blackwell stole Ohio, it wasn't in any way the exit polls could detect."

Again... how can you say this when the exit poll data has not actually been released???

Your repeated assertions of this talking point - which rightly belongs to Blackwell not to a reasoned debate on DU - tends to suggest that you want it to be true.

Which begs the question:

Why do you want to beleive so strongly that exit polls cannot prove fraud?

FACT: Exit polls can and are used to combat electoral fraud all over the world - e.g. Ukraine and Mexico.
FACT: Exit polls are an essential tool required by anyone seeking to sophisticatedly steal an election
FACT: Exit polls have provided among the most potent evidence of vote fraud in 2004 that there is currently in cirulation
FACT: The problems in the 2000, 2002 and 2004 exit polls conducted by Mitofsky have not been adequately explained. RBR theory is a theory that has no evidence to support it and which most of the evidence that there is tends to suggest is a load of crappola....

This all means that exit polls, in the past and in the future, are right at the core of the debate over election integrity.

The fact that you, Mitofsky and a bunch of RW pundits and talking heads want desperately to convince both us and the general public that this isn't so makes me increasingly cross.

[Febble]

37. I don't "need" to believe it

I don't believe exit polls can prove or disprove fraud, but I do think they could be highly suggestive of fraud, and also suggestive of polling bias. I think both theories give rise to testable hypotheses.

You are right of course that the exit poll data has not been released. What has been released are analyses of the exit poll data in which hypotheses were tested that might well have indicated fraud had they tested positive.

And what you may not know is that I was the analyst in question, but the results are shown here:

http://inside.bard.edu/~lindeman/slides.html

with commentary by Mark Lindeman, and I also wrote it up for DU here:

http://journals.democraticunderground.com/Febble/3

I certainly don't want to convince you that exit polls are not important. I think they are. But I do consider that analyses of the exit polls, carried out, as it happens, by myself, indicate that polling bias was a very probable cause of the discrepancy and that fraud is unlikely to have been. I reached this conclusion after analysis, not before. Up until January 2005 I was strongly suspicious that the exit poll discrepancy was due to fraud.

And the reason I make these points on DU is not because they are RW talking points, but because I think that it is all too easy for the "RW pundits" to point to the polls and say "look! these Democrats and Election Reform guys are talking nonsense! There is nothing wrong with our election system" - and of course there is. It is desperately sick.

I want the movement to have good arguments. My reasons for posting are simply to try to explain to people why the exit poll evidence isn't the evidence of fraud so many think it is.

You may not trust me, and I don't see why you should, any more than you should trust a DRE that doesn't give you an auditable ballot, and the right to a fair recount.

But it happens to be the case that I am an honest UK citizen who thought that the exit polls discrepancy might have arisen because millions of votes were digitally switched, but who, on extraordinarily close inspection of the evidence, the results of some of which is now in the public domain, concluded that it was much more likely to have been due to bias in the poll, and rather unlikely, given the distribution of discrepancies relative to Bush's performance, to be due to fraud. It remains possible, but, IMO, improbable, and therefore not a useful argument to wave at the front of the urgent campaign for election reform.

Whereas there is excellent evidence of disenfranchisement all over the place, some evidence that looks like outright fraud, and copious evidence that digital voting technologies are unreliable, unauditable and vulnerable to malicious tampering.

Cheers

Lizzie

[Name removed]

45. Deleted message

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

[Febble]

46. I understand the word circular perfectly

But Partridge misrepresented the argument.

Both the theory that fraud was responsible for the discrepancy and the theory that participation bias (selection bias or non-response bias) was responsible both give rise to testable a priori hypotheses.

Testing an a priori hypothesis is not circular. A priori hypotheses were tested. Not only that, but even before a single vote was counted, the pollsters suspected bias in the poll. Hard to do that if you are deducing bias froma count. Pollsters are not as stupid as you appear to think they are, and they are not given to circular reasoning. They are, of course, scientists.

I am glad that you are without a scintilla of doubt. It must make you happy. Me, I doubt all the time. But that's because I'm a scientist, and so I regard all conclusions as provisional. I also test hypotheses. I do not simply believe what I want to believe regardless of the evidence.

But I am growing very tired of your referral to me and to others as "you people" and references to "your kind of mindset". It strikes me that you are the one with a set mind.

And if you really want to know more about me, you can google. I am not anonymous.

Elizabeth Liddle.

[Name removed]

52. Deleted message

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

[Name removed]

53. Deleted message

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

[Joe Chi Minh]

55. I am ceasing my role in this sorry saga here and now.

[WillYourVoteBCounted]

106. its

Its OnTheOtherHand not Mr Ontheotherhand

[Joe Chi Minh]

27. I'm not going to need it. Nor are you, as it's all here for your

instruction, if you have eyes to see.

DU has a comprehensive compendium of the evidence. Actually, it may not be comprehensive, but just appear so, because of the mountainous plethora of it, and its clarion appeal to the most elementary common sense.

As regards the exit polls, you're starting with the false premise that, utterly compelling proof though they provide - they are essential to prove fraud. If you read Mr partridge's article I cited, it is clear that your initial premise should be that, since they are acting as guilty as hell, it is certain - not probable - certain, that they are guilty of fraud. The alternative scenario, as we can see from Mr Partridge's parable about the drug-sniffer dog is too farcical for a childrens' comic.

In any case, it is incumbent on your people to prove there was no significant fraud - instead of doing the opposite, using every means you could, however, farcically guilty-seeming, to keep the machinery and software 100% private. Not faith-based elections, but verifiable ballots.

[OnTheOtherHand]

33. begging your pardon

I have no idea who you think my "people" are, but the idea that I am out "to prove there was no significant fraud" is unsubstantiated and untrue. Moreover, it has no discernible relationship to the thread.

I observe that Febble asked you straightforward questions in post #29, and that you made no attempt to answer them. And so, the first of those questions remains unanswered along with the rest: Do you have something to contribute to this discussion?

If you actually intend to support the proposition that the exit polls provide "utterly compelling proof," then it will not do to cite all the other proof.

[Joe Chi Minh]

56. No, it was a mistake in the first place. Except that I felt compelled

to expose the ludicrously foolish hypotheses underlying your whole position, and necessarily it equally foolish. One dolt between the two of us is enough, if we continue to argue, one of us is going to render the other one equally foolish.

We both have our own ideas as to who's who in this scenario, so I'm going to leave it at that, and refer you to my post to prebble about the chronically unscientific procedure of adopting a ludicrously inappropriate, initial hypothesis, and then presuming to dispute with those who are able to recognise the genuine experts and rely on their analyses of an individual procedure, such as polling.

[OnTheOtherHand]

58. but you don't have the privilege of misstating my views

or, rather, of course you do, but I see no reason to accede in the blunder. The hypothesis of massive fraud is infirmed by the exit poll data. It makes no difference whether one initially prefers one view or the other.

If you are going to talk science, by all means do so -- and if you are going to trash science, you are free to do that. But do choose.

[Name removed]

60. Deleted message

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

[Febble]

73. Just let me ask:

would you recognise a statistician who became president of the American Statistical Association as a "genuine expert"?

Dr. Fritz Scheuren

Fritz Scheuren, Ph.D., is a statistical consultant for the Human Rights Data Analysis Group of the AAAS Science and Human Rights Program. He is the Vice-President for Statistics at the National Opinion Research Center (NORC), and the President-Elect of the American Statistical Association (ASA).

Recently, Dr. Sheuren consulted on the methods of statistical analysis for Peru's Truth and Reconciliation Commission. He peer-reviewed the report of the analysis, and visited Peru with Dr. Ball and Jana Asher to meet with representatives of political, military, and civil society groups to explain the technical basis of the findings detailed in the report.

In recent years, Dr. Scheuren has advised on the overall direction and approach to the statistical analysis for several HRDAG projects including Guatemala and Kosovo. As a top statistician in the field, he has also provided critical peer-review of HRDAG work.

Dr. Scheuren has received numerous awards and honors, including: the ASA Founders Award (1998); Shiskin Award for contributions to U.S. Economic Statistics (1995); Finalist, Senior Executive Association Executive Excellence Award, (1992); Elected Member, the International Statistical Institute (1988); Elected Fellow, the American Association for the Advancement of Science (1984); and Elected Fellow, the American Statistical Association (1981).

Or would you prefer to choose your own?

[Joe Chi Minh]

77. Certainly not! I haven't read the rest of your letter, yet, but I can

imagine even now its purport. You really don't seem to understand at all what I have been pointing out to you and the desk sergeant.

A scientist, however pre-eminent his academic accreditations, is only as good as his integrity allows him to be. And I don't know your man from Adam. I have my fears of course.

When one of Einstein's theorems of relativity was first published, a scientist contemporary of his, who apparently is also a household name, declaimed that it was utter rubbish! Not unreminiscent of the Manchester Guardian's golf correspondent, who, as the only German-speaker, was delegated to hear Einstein personally expounding the nub of his theory. When, on his return to the Guardian, he was asked by the editor about it, he replied, "Platitudes, my boy. Nothing but platitudes"!!!!!.....!!!!!

What is more, the rigorous and pedantic reductionism of empirical science - wonderful within its own competency - is quite simply wholly inadequate, inapt for a study requiring the human dimensions in this sorry saga to be taken into account, for the framing of an appropriate initial hypothesis. Which you people steadfastly refuse to recognise, sticking to the opinion that no acknowledgement of the underlying context, the background, must be allowed to underpin the initial hypothesis concerning the constructions put on the statistics by the parties but, rather, the strictest neutrality .... as if you were all very proximately studying inanimate matter. And, incidentally, very eminent professional people are more often likely to lack integrity than the rest of the population.

If you did take it on board, you would not be content with your expert's claim that the statistics don't support the notion of Republican fraud, you would be far more exercised about the deliberately-designed unverifiability of the privately owned hardware and software, widely known by IT expert to be hackable at every level. You and our desk-bound friend seem to equate science with reason and rationality, when, as I pointed ut a long time ago, empirical science is only competent to deal with the lowest and most rudimentary aspects of the physical world. Reason, on the other hand, should counsel you both to repudiate in horror the notion that people should trust the election officials and Republican-supporting machine manufacturers, in the teeth of all the villainy that was identified even during the election process itself. A plethora of anecdotal evidence is clearly very compelling in this kind of context, where the so-called scientific method would be wholly inapplicable.

To this very day, the history of our world is very largely the history of the ministrations of desperately, if not ruthlessly ambitious psychopaths and extreme sociopaths: the Marcoses, the the Pinochets, the the Somosas, the Hitlers, the Stalins, the list is almost endless - not to even consider our own nearest and dearest Great and Good, in the West. Of course, in principle, your statistical big-shot may be a man of egregious integrity, but that does not dispense with the need for us to be less than overwhelmed merely on account of his professional accreditations.

You both really need to take on board what we have been trying to teach you.

[Febble]

78. Well, it seems I do not

understand your point, but having finally peered at your signature line, then perhaps it is becoming clearer.

But if a scientist is only as good as his integrity, how do you judge his integrity? Not by his results, clearly, because that would be circular.

It is one of the refreshing characteristics of scientific discourse that credentials take a back seat to argument, although not always as much as one would wish. However, the increasing use of blind review can only help in this regard.

For that reason, I judge all arguments on their merits, regardless of provenance. A problem clearly arises when one lacks the expertise on which to judge the argument oneself. At that point credentials matter, and for this reason, I supply the credentials of one of the authors of a paper on the Ohio exit poll data that used the kind of analysis you seem to reject.

Dr Scheuren's credentials do not guarantee that his arguments are correct, but if one is unable to evaluate two opposing arguments from the same data (in this instance, Dr Baiman's and Dr Scheuren's) then credentials may help. Actually, I find both Baiman's paper and the paper of which Scheuren is an author flawed. I find the flaws in the latter much less seroius, and this has nothing to do with Scheuren's stellar CV, nor with the nature of his conclusions, but with the a couple of the assumptions.

However, at bottom, we do take have to take much of science on trust - when papers are published, we can check out coefficients and F values, and even re-run the analyses ourselves if the data is available (which it cannot always be, in social sciences, because of confidentiality constraints), but we simply have to trust that the data itself is good. This is why falsifying data is such a grave offence in academia, and one with severe penalties.

But despite these caveats on my part, it seems that we will not agree about science, or at least the role science can play in understanding complex phenomena like human beings. I am a behavioural scientist (cognitive psychologist) and I am only too aware that much of what determines human behaviour is beyond our capacity to measure. However, I am also aware that skillful use of statistical models can give us a great deal of information about human behaviour, especially if we are rigorous in our use of a priori hypotheses, and extremely cautious in interpreting post hoc observations. If not we are at risk of discerning miraculous images in random patterns of data.

As you say, you don't know me from Adam (well, maybe Eve). Well, I was born not far from where you are sitting, and I'm not that much further away now. I hope that doesn't scare you too much.

Cheers

Lizzie

edited for sense

[Joe Chi Minh]

80. Why ON EARTH would it?

I've had our clandestine services and the CIA on my wheel, quite menacingly. Why would your urbane chatter about living near me phase me?

[Joe Chi Minh]

81. Well?

[Febble]

84. You said:

"And I don't know your man from Adam. I have my fears of course. "

Ah, misread your statement. I thought it was me you were frightened of.

As I said, I am harmless.

[Febble]

86. And apologies

it does appear that we both have problems reading the other's posts.

I simply meant to reassure you that I am a middle aged female Scottish academic, urbane, maybe, but not, we agree, scary.

[Joe Chi Minh]

88. You won't come round and break my legs, then?

We obviously don't share the same sense of humour, either. I was expecting you to come back and say something like, "No, not this time. But don't let it happen again".

[Febble]

90. No, it seems

we don't share the same sense of humour.

Just a city, and maybe the view that technology can be an engine of inequality. But I won't push my luck.

[WillYourVoteBCounted]

107. and

also it is refreshing that you don't fog up what you say with tons of rhetoric.

Thanks for talking clearly and without pompousness.

[Febble]

83. Golly, a second installment

I really do not understand your assumptions here. Why on earth would you think that any social scientist, any public opinion researcher, would "steadfastly refuse to recognise" the "human dimensions" and treat the data as though they were studying "inanimate matter"?

Do you think that they study particle physics, and then turn to social science research in their spare time?

And what makes you think that I am not "exercised about the deliberately-designed unverifiability of the privately owned hardware and software, widely known by IT expert to be hackable at every level"?

If I wasn't, why would I be posting on American discussion forums pieces like this:

http://www.dailykos.com/storyonly/2004/12/19/9056/8130

Actually, you should read that one. You will find that I myself am a not-quite-closetted Luddite. So, given that you don't, as you say, know me from Adam, perhaps you might refrain from taking these trips through what you think is my mind.

I fully agree that "a plethora of anecdotal evidence is clearly very compelling", as you would know if you had actually read any of my posts. It was precisely my point that that plethora of anecdotal evidence is much more compelling than some borderline p values on a meagre handful of datapoints.

And I'm afraid that an anonymous poster "pointing out" that the entire edifice of behavioural study is outside the competence of empirical science isn't going to shake me from my view that what I do for a living is rigorous and scientific, even though I would be the first to accept that our findings will always be subject to alternative interpretations, and our models to modification. But that is the essence of empirical science. It just happens to be even more the case in behavioural science.

[Joe Chi Minh]

87. Anonymous poster? Oh, my name's Paul Becke. It's more fun

having a nickname, isn't it?

"Do you think that they study particle physics, and then turn to social science research in their spare time?"

No, you seem to me to think the narrow focus, the degree of rigour/pedantry qua the "scientific method", is as applicable to social-science research as to the study of particle physics.

In view of the fact that Conyers, despite being a relative layman in the field, was sufficiently impressed by the polling experts who ARE convinced of fraud, I found your fears that "our" pollsters might prejudice lawsuits against the fraudsters, rather odd.

I once saw Conyers on the Larry King show with a Republican Senator. After the latter had said something - it was quite brief, he wasn't holding forth - Conyers murmured, "You have just made five errors in law..." As a loyal Neocon, Larry was clearly discomfited, but realised he had to ask him what they were. Which Conyers duly listed. I think he's familiar with the legal precept of not asking a question to which you don't already know the answer. Time will tell, of course, whose experts, ours or yours will be vindicated.

[Febble]

89. Thanks for the name

though I will still stick with my career, thanks. My nickname is from my initials: FEBL. I used it for Daily Kos when I didn't really know how these things worked, and got stuck with it.

No I don't think that "narrow focus" is applicable to social-science research. Ironically, it is because I have taken a much broader view of the data that I have come to the conclusions I have. I don't think zero-order correlations tell us much. I do think we can discover a lot when we look at the wider picture.

And Conyers' is a hero of mine. Contrary to what you might imagine, I actually sent him one of my own analyses (of machine allocation in Franklin county) back in December 2004, and was one of countless numbers who emailed various senators to try and persuade them to back the Ohio challenge. And at that time, neither Conyers, nor I, nor indeed many at E-M knew what is now known about the exit poll data. Indeed at that stage, I too thought the discrepancy was suggestive that the kinds of injustice that had clearly taken place in Ohio might extend to nationwide vote theft.

So your divisions into "ours" and "theirs" is seems inappropriate. On whose "side" would you place Walter Mebane?

http://macht.arts.cornell.edu/

Check out "papers available for downloading".

Also check out that DKos diary of mine. I think you'll find that "sides" is an inappropriate image for the range of views that are embraced by those who think that that the exit poll data do not support the argument that millions of votes were stolen in 2004.

And my chief point, on this thread, is that using bad arguments risks damaging a good case. Conyers might even agree.

[Joe Chi Minh]

92. I will check out your DKos diary, febble. I do recall however

(correct me if I'm wrong), that DKos was an "unbeliever" from day one, i.e that fraud had occurred - despite the volume, variety and enormity of the outrages reported. I've never looked at their blog since, and am unlikely to again - bar this one occasion. In a matter such as this, even to TRY to fool me just once, causes me great, nay unforgivable offence.

As regards Walter Mebane, I looked at the concluding 'Discussion' in one of his papers, and was immeasurably saddened. I suppose there must be a place for highly technical analyses by gifted academics of the polling and the fraud, generally, but what saddens me is that individual academics, naturally focus on their own specialism, where what we have here is an enormous patchwork of villainy (not measurable under laboratory conditions) and simple, final statistical probabilities so overwhelmingly clear in their import, enormity and variety that the most elementary common sense indicates massive Republican fraud.

Apparently, there were one or two cases of highly anomalous results in Kerry's favour, but common sense also indicates that the Democrats had no reason to resort to fraud. Bushco was already deeply, deeply unpopular throughout the country. And somehow, it's got worse, though who the late-comers can be must be one of life's more inscrutable mysteries. It wasn't just the exit polls that had indicated that Kerry was going to won big - which, in view of their customary bias, was pretty remarkable.

Again, the massive increase in turnout told its own story. One thing about the right is that they are - apart from the ever ingenuous and self-injuring rednecks - very worldly-wise and don't have to told to go out and vote. The more spiritual, less worldly wise are mor likely to say, as they so often do, "I'm not interested in politics. They're all the same (Rupert Murdoch said so..!)". "Ah," as someone commentd, "but politics is interested in you. So, if human nature and the past were any guide, many more people would have voted Democrat this time. I believe hundreds of thousands, if not millions of votes were not even counted, and many that were had switched from Kerry to Bush before the voters' very eyes.

And I have to say that, in the context of the whole theft of the election (two, to be precise), your apparent suggestion, earlier, that we should be very content merely to bring a strong case of voter suppression, as it suggests Democrats have strong grounds for optimism in later elections, is still something I have difficulty reconciling with the likely utterances of an intelligent person.

As for the criminal misallocation of the machines, no scholarly rationale would convince me in a fit with my leg up, that it was not yet another part of the patchwork of voter suppression and election malfeasance.

[Febble]

94. Kos

was against the case that the election was stolen. Many of the diarists were outraged, particularly that few diaries (not no diaries) on Ohio, or fraud in general, were "front-paged". One of the "front-pagers", Armando, himself a fraud-skeptic, challenged the "fraudsters" that if they could assemble a decent case, he would make sure it got to the front page. Georgia10 assembled a vast quantity of data, with the help of other posters, including posters like me and others who helped prune the bits we didn't feel stood up under scrutiny (e.g. the exit poll story).

Her compendium ended up being far more than a front page,

http://www.dailykos.com/story/2005/1/2/19512/47013

and is still downloadable here,

http://www.ajschuler.com/html/armando_s_challenge.html

and Georgia10 herself is now a DKos frontpage poster.

As for your paraphrase "that we should be very content merely to bring a strong case of voter suppression, as it suggests Democrats have strong grounds for optimism in later elections" I think it mis-states my case, particularly as regards the "as":

I don't think "content" describes my view - I simply think that case is strong.

And while I do think Democrats have grounds for optimism, I still think the possibility for digital vote theft is real, and I would positively advocate arguing the case that digital voting systems - indeed all voting systems - need to be made transparent and auditable.

All I ask - all I ask - is that the case for both these things is not sabotaged by arguments based on the exit poll story that simply do not support it. I do not think it is necessary to prove that millions of votes were digitally stolen to make the voter suppression case or the case that digital voting systems and lax BoE practices make vote theft all too easy. And not only is it not necessary, because I think the millions-of-votes story depends extremely heavily on what I consider an unjustified inference from the exit poll discrepancy, I suggest that the case is far better made without it.

That's probably as close as we are going to get to a rapprochement, so maybe we should leave it there - feel free to have the last word, as long as it's not deletable.

[Joe Chi Minh]

97. That's fine by me. I feel I've always been at a disadvantge

vis-a-vis the moderators, anyway, in this saga - doubtless due to the very nature of forums. But calling each other "The Honourable Poster" was never going to cut it, was it?

Politics is a matter of life and death to many millions of people - in the case of the US, all over the world, so passion of the part of progressives is an absolute Christian obligation. Some passionate people may be able to sublimate their emotions, but their opponents' polite and "civilised" moderation of speech, whatever their own personal merits or demerits, good faith or bad, is shared by arch villains. The Psalmist is always complaining about their honeyed words belying their concealed daggers, the war in their hearts.

"Well, if I'm right and the millions of votes weren't stolen, then it is rather good news for Democrats, because it means you have a chance of winning in November",

Those were your words, but the fact of fraud having taken place was never in question. It's been a constant since the country's earliest days. In an over-arching kind of way, all that can ever be at issue re the fraud in any US presidential election, is a) its absolute degree, and b) the degree in which it favoured one party over the other. And I find the sanguine, implicit acceptance of this in your comment above, when so many people actually there on the ground, in the polling places, were were so volubly and bitterly attesting to its extraordinary scale, leaving aside the unambiguously criminal voter suppression, well, disturbing.

If the voter suppression alone - demonic as it was - were to be the sole consideration in the matter, a chance of finally setting up an election system almost proof against significant fraud, would be missed; and in view of the increasingly despotic trend of this Government, the opportunity might not return for decades; and that perhaps subject to a revolution of some kind and much bloodshed. While gains in the civil-rights matter, the checking of voter suppression being just one such issue, would almost certainly be very short-lived. Hail and farewell, Elizabeth. I think I got that right.

[Febble]

98. Agreed

"Politics is a matter of life and death to many millions of people - in the case of the US, all over the world, so passion of the part of progressives is an absolute Christian obligation."

[althecat]

34. OTOH and Febble... I am inclined to agree strongly with your opponent here

And my reason is simply this.

Edison Mitofsky has refused to allow a proper analysis of his own data.

As I understand it he has released WPE data without identifying which precincts each data point comes from. The graph data scatter plots etc that Ron is dealing with and which you claim are not statistically significant would be able to be studied in detail across the entire country and with a great deal more likelihood of discovering whether there is PROOF OF FRAUD in the exit poll data if we actually had it.

I.E. If you want to argue that exit polls can or cannot prove fraud then lets have the argument after the data is released and we can analyse it properly. In the meantime we have a moot point in your opinion, and in the non skeptics opinions we have a series of very strong pieces of evidence in the exit poll which back up an enormous range of other pieces of evidence that have been recorded and described ad nauseum around here.

I am not a statistician but the most strong piece of evidence I have ever found in the exit polls that the results are fraudulent is that they show that Kerry received a clear majority among new voters. This occurred in an election when total turnout was enormous and new voters undoubtedly decided the result..... if Kerry had a clear majority of new voters then how could Bush have possibly won.

In these circumstances I find your attempts to brow-beat those who disagee with you by using:
a) terms of statistical art which make very little sense to us non statisticians
b) phrases like "How many times can we torch that straw man, anyway?"
..deeply intellectually dishonest.

The fact that you are so keen to do so and insist on doing so in an intellectual vacuum (i.e. considering the statistical evidence as discussed by the statisticians without considering the withholding of evidence etc.) further makes me more than a little suspicius about your motives.

[Febble]

38. As I said in my other post to you

I understand the suspicion.

I think the reasons for the non-release of the precinct level data that would make independent analysis possible are widely misunderstood; the confidentiality issue is by no means a trivial issue, which is why to date only one "blurred" dataset has been prepared, and I understand that more may be in preparation.

But I would query your statement that "Edison Mitofsky has refused to allow a proper analysis of his own data." (Edison and Mitofsky are two polling firms BTW). They analysed the data extensively themselves; they released "blurred" data on Ohio for otherwise independent analysis by ESI, and Warren Mitofsky contracted me to do a series of analyses for him. Speaking for myself, I'd have to say that I consider what I did a "proper analysis", and in fact Mitofsky did at least one of the analyses that Ron did for Ohio for the whole dataset, and presented those findings at AAPOR in Miami last year. However, if what you mean is "independent" analysis, the most independent analysis was that done by ESI, but all analyses are to an extent dependent on E-M as they collected the data. I consider myself an honest and independent thinker, but I wasn't, clearly, objectively independent.

I do agree with you, that at state level, the statistical power in the polls is very weak. 49 data points is not enough to conclude anything from either a null finding, or a finding that doesn't reach statistical significance (as Ron's fails to do). I think there are other serious problems with Ron's analyses, but the biggest problem is simply that there is not enough data for Ohio to work with. If you want to find fraud in Ohio, the exit poll data isn't it.

What the nationwide WPD data can tell us is what is generalizable on a nationwide scale, and at that scale the data has much more statistical power. And on that scale, some important findings emerge that support the polling bias theory, and some important null findings have to be taken more seriously precisely because of the statistical power to detect even small effects. One finding however, that might interest you, was a significant finding that in urban areas, there was greater discrepancy in precincts using older technology (levers, punchcards) than digital technology. It may be that this effect was due to collinearity with some other variable, but it may suggest that differential residual vote rates for older technology may have impacted on the exit poll in these precincts - in other words this is evidence for the vote miscount hypothesis, at least on a small scale (not a big effect). However, it does run counter to the idea that digital theft was the culprit. Digital machines in urban precincts had a smaller discrepancy than the analog machines.

Your point about new voters is an interesting one, and was one of the things that kept me suspicious even after the E-M January report was released. I think OTOH may have something to contribute about that. My analyses were done on precinct aggregated totals, but others have done extensive work on the questionnaire data.

Cheers

Lizzie

[OnTheOtherHand]

39. first, thank you

for offering actual arguments having to do with exit polls. Of course, your speculations about my motives are less welcome, but I respect your candor.

Since this point is so often confused, let me open with it: I am not arguing that the 2004 election was "fair" or that fraud did not occur -- still less to the view that the U.S. electoral system is fair and secure. (I do not believe that Kerry won the popular vote, but this is not a "position" that I undertake to defend.)

The exit polls do deserve serious analysis. If we treat the exit polls as authoritative, one apparently inevitable inference is that many hundreds of thousands of votes were stolen in my home state of New York (one point estimate is 840,000 votes or so). I think it is fair to ask, now, based on information already available, whether that makes any sense. I don't think so.

It seems facially paradoxical to argue both that the exit polls provide strong evidence of fraud and that we have insufficient information for proper analysis. Nevertheless, logically both statements could be true. But don't skate over the strangeness too quickly.

Between, say, Baiman and myself, both sides agree that more analytical results should be released -- so I don't want to ratchet up the disagreement on that front. Still, Baiman has presented what he deems 'virtually irrefutable evidence of vote miscount' (or so I assume, since that is the subtitle of the paper he presented at AAPOR). I really have no choice but to say that I regard that claim as ridiculous -- or else to ignore him entirely. Most political scientists are ignoring the 'exit poll debate' entirely. They are prudent. However, I think activists have a right to know what arguments hold up and what arguments don't. (I also think they have a responsibility to know this, whenever possible.)

The nationwide scatterplots at http://inside.bard.edu/~lindeman/slides.html present rather strong evidence against vote shift on the order of millions of votes, or at least against the premise that the exit polls could evince such vote shift.

One might object that Mitofsky might have rigged that scatterplot -- although, after all, we do depend on Edison/Mitofsky for all the exit poll evidence, the results we might like as well as the results we might not like. Nevertheless, if you take CNN.com screen shots or any other source of state-level "exit poll results" I have ever seen, and plot the discrepancies against the deviations from pre-election expectations or from 2000, you are likely to find either a big fat hairy nothing, or a perverse correlation -- biggest red shift where Kerry did better than expected.

There may be an exception, and I would be more than happy to have someone help me look for it. I have actually spent more time looking for exit poll evidence supporting fraud than the vast majority of people who accuse me of trying to cover it up.

I agree that Kerry received a majority of votes among new voters (although I think that the exit polls overstate the margin). I believe (obviously) that Bush received a majority of votes among repeat voters. This is not so startling to me. I don't know the basis of your confidence that "new voters undoubtedly decided the result." I surmise that you are assuming that repeat voters should have divided evenly, in what is sometimes characterized as 'a rerun of 2000.' Most political scientists would not make this assumption. How many voters could have thought about George W. Bush the same way in 2004 as they did in 2000? He could have done better, he could have done worse, but there is no real warrant for assuming no change. Incumbency perhaps could cut in either direction, but we would not assume that it was irrelevant.

If you consult post #8 (I cannot direct you to post #24, since the moderators have deleted it for cause), you may see why I consider your characterization that I attempt to brow-beat people, by using obscure terms and polemical(?) phrases, somewhat skewed. For that matter, there seems to be a fair amount of obscure terminology and polemic in the article referenced in the OP. I am here, attempting to explain myself, answering questions, responding to (or attempting to ignore) personal attacks on my motivations. If I sometimes seem annoyed, well, why would I not be? But if you want to allege my deep intellectual dishonesty, I think you should come up with better grounds than that.

I don't know what it means to say that I am considering the statistical evidence "in a vacuum." Again, by way of analogy: if someone argued that Kerry's star chart proved that he was born to be president, and if I rebutted that argument on scientific grounds, would you accuse me of ignoring the context?

I have the impression that you believe that the exit poll evidence and other evidence point in 'the same direction.' But they don't so much. I don't think anyone has presented much evidence for massive fraud in Vermont, or New York, or Delaware, apart from the exit polls.

[Name removed]

44. Deleted message

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

[AtLiberty]

11. K & R

;)

[liam_laddie]

13. OTOH...as usual you

ignore the FACT that the information accumulated over the
last nineteen months amounts to PROBABLE CAUSE to
launch an investigation of the 2004 election. You and I are
neither judge nor jury, and I for one think you intentionally
demand "proof" when in fact the data, depositions, etc., if
provided to a jury, are the basis for a case to be presented
and a verdict to be reached.
Which way it would go will have to wait until some politicians
with cojones show up in DC, or in state legislatures. You know
very well why this has not been investigated...the Reich-wing
control. Totally corrupt, and frightened.
Land Shark has laid this out clearly.

[autorank]

14. "politicians with cajones" -- we lost our best shot, Hackett

You are absolutely right, it's called PROBABLE CAUSE...which leads to an INVESTIGATION
which leads to a PUBLIC TRIAL at which the BURDEN OF PROOF is assessed...

Cheers...

[Febble]

18. I think there is a problem here

and one that I have only recently become aware of, actually thanks to Land Shark.

OTOH is not, I think (and certainly I am not) arguing that there is no cause to "launch an investigation". I don't think either of us thought that making a case for an investigation was what was what we were doing. Speaking for myself, I thought I was actually doing some investigating! I thought we all were, notably you yourself. And I think we found some horrendous things.

But certainly, if we are talking about making a case for an investigation by some other body, it is surely worth knowing what evidence is likely to stand up to scrutiny, and what is likely to be laughed out of court. You certainly don't want the second kind, and, frankly, I think Ron's evidence is the second kind. And if we want to actually bring a case, we need to know what kind of charge will actually stick.

You are welcome to disagree, but I think you are wrong to assume that those of us who try to probe arguments for validity are ignoring "the FACT that he information accumulated over the last nineteen months amounts to PROBABLE CAUSE to launch an investigation of the 2004 election."

I would be only too glad to see such an investigation; I think your democracy is seriously broken and needs urgent fixing. But I'd rather you guys didn't go into that investigation armed with evidence that will fall apart on any informed scrutiny, nor do I think that the case for an investigation is best made by claims of "virtually irrefutable" evidence of fraud that are only too easy to refute.

It's people like you, Liam_Laddie, who have collected the hard evidence, and I think we should be leading with that.

[OnTheOtherHand]

19. here is how a science blogger sees the stakes

RFK Jr's article tries to argue that the 2004 election was stolen. It does a wretched, sloppy, irresponsible job of making the argument. The shame of it is that I happen to believe, based on the information that I've seen, that the 2004 presidential election was stolen. But RFK Jr's argument is just plain bad: a classic case of how you can use bad math to support any argument you care to make. As a result, I think that the article does just about the worst thing that it could do: to utterly discredit anyone who questions the results of that disastrous election, and make it far more difficult for anyone who wanted to do a responsible job of looking into the question.

http://scienceblogs.com/goodmath/2006/06/election_fraud_or_just_bad_mat.php

Suppose someone argued that we know the election was stolen because Kerry's astrological charts show he was destined to be president. Would you cheer, or groan?

I don't think Land Shark has worked out how his preferred experts would actually play in court.

EDIT to correct formatting

[althecat]

36. Ok Febble/OTOH a change in tack - questions for you to answer?

Is there any evidence in the exit polls that you think does support the case for fraud?

What is it?

[OnTheOtherHand]

40. what "kind" of fraud?

Fraud is not a unitary monad. Different kinds of fraud or misconduct would be expected to show up in the exit polls in different ways; many kinds might not show up at all.

Trying to answer such an open-ended question does not seem likely to end well, especially given your suspicions about my motives. But if you ask a narrower question, perhaps I can be more helpful.

[althecat]

59. Vote Switching... and Ballot Box Stuffing

Note however that I do not expect the exit poll data to actually prove this fraud ocurred just that it would be helpful to:
a) help prove it
b) identify locations where suspicious activity might deserve to be looked into further.

Types of fraud to look for:

Ballot Box Stuffing...

Frankly I find it very difficult to beleive that Bush's nationwide vote increased by 11 million votes (20%) between 2004 and 2008. I also find it peculiar that most of those votes occurred on the East Coast where the votes were counted first.

See... George Bush's 8 Million New Votes Found – A STATISTICAL ANALYSIS (the 8 million was based on election night figures... excluding provisionals etc.)
http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=19390

In several of the above posts you say you find it impossible to beleive that there was widespread fraud in New York State. For me your skepticism on that count is misplaced. Where is the evidence of the extraordinary republican registration and get out the vote campaign.... most of the public evidence rather indicates that they were working very hard to suppress Democratic Vote.

In one of the above posts (in response to my observation that Kerry received a majority of new voters votes in the exit poll) Febble states that political science has no reason to consider that Bush would receive a re-run of 2000 in terms of existing voters votes. She goes further and suggests that the explanation for this apparent anomaly is that Bush received more votes from returning voters that Kerry (i.e. he received votes from voters who voted for Gore in 2000). This is plain wrong. I am not a statistician but there is loads of reasons to consider that Kerry should have received the vast majority of Gore's votes and then some.... the most obvious evidence being that come 2004 Bush's approval ratings had fallen considerably and the war was going badly.

Febble has stated that nationwide scatter plot evidence does not indicate that there were millions of votes added to the bush total. This is why a full set of data with precincts identified needs to be released. All precincts/counties will not have been effected.... almost certainly there will be anomalies evident between a set of tampered precincts/counties and ones where there was probably not tampering. Some of the WPE error will be random (noise) and some may have been caused by tampering. In order to find evidence of this you need to remove the random noise from the sample on the basis of some testing a variety of assumptions about how the theft may have occurred in various places.

& Vote Switching..

In Florida in particular there were completely anomalous results reported in the largest democratic counties Miami Dade and Broward. In these democratic strongholds Bush achieved dramatically large growth in his vote. How?

See... GWB's 1 Million New Florida Votes Found – A STATISTICAL ANALYSIS
http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=21834

This is also what is indicated to have happened in the 12(18) dodgy counties. Places where vote switching occurred - counties/precincts - would be expected to exhibit exactly the characteristics that Ron Baiman finds in the the 12 dodgy counties.... a consistent (in order of magnitude) set of anomalies which differentiates the counties in which fraud is thought to have occurred from others.

[Awsi Dooger]

63. I thought we covered this in November 2004

Nothing out of line in Dade, Broward or Palm Beach. Not that I can see. Sure, the raw vote totals went up. They did throughout the state. Florida had roughly 7,600,000 votes cast in 2004 compared to 5,900,000 in 2000.

Democrats outregistered the GOP in these counties by about 168,000 to 49,000 from 2000 to 2004. But you can't expect that difference to automatically show up in the voting margins. Kerry did 5% worse than Gore statewide and more than 3% worse nationwide. That basic change in preference will be plopped atop those counties and alter the percentages. They were exactly what I would have expected, if someone told me Bush would win by 3% nationwide: Kerry fairing several points worse than Gore, on average, in Dade, Broward and Palm Beach. Our net edge in those three counties is approximately 7000 votes higher than 2000, despite the loss of 4% in Broward and 3% in Palm Beach.

No, Kerry didn't fare 3% worse than Gore nationwide. I understand that will be the response. However, there were plenty of reasons to anticipate Bush outperforming his 2000 result in Florida. The state economy was much better than the national economy. That was evident for years, part of the reason for Jeb's big re-election number in '02, and demonstrated by the 60% excellent or good in the Florida exit polls, regarding the state economy. That same Florida exit poll had Bush at 54% approval, well above his national figure.

It's called trip handicapping, understanding how the race was run to emphasize or downplay characteristics within the numbers. Ohio 2004 had a terrible state economy. That state is still 2-4 points red leaning compared to the nation itself, with no demographic swing our way similar to Virginia or Colorado. We had a rare chance to steal Ohio in 2004 since the pathetic state economy meant it would likely fall very close to the national popular vote margin, and not 2-4 points GOP. That's exactly what happened but Kerry's 3-point lag nationwide pushed Ohio outside our grasp. On the other hand (sorry), Florida's much better than average state economy in 2004 means the +2.55 Republican partisan index of 2004 is baloney. Most likely it is still extremely close to the nation itself, evidenced by the fractional variance from the national popular vote percentages in both '96 and '00.

Statewide the Republicans had roughly 4000 more new registers than Democrats between 2000 and 2004 in Florida. Other than these Democratic strongholds, the GOP outregistered us by 123,000. That's the missing rural strength Rove prioritized post-2000.

Here are the registration figures prior to the general election, 2000 and 2004, and also the vote totals and percentages:

* Broward
2000:
Democrats: 456,789 (note: every time I've looked at that since 2000 it strikes me as sequentially bizarre)
Republicans: 266,829

2004:
Democrats: 533,976
Republicans: 283,736

Actual Vote:
2000: Gore (68%) 387,760 Bush (31%) 177,939
2004: Kerry (64%) 453,873 Bush (35%) 244,674

* Miami-Dade
2000:
Democrats: 396,518
Republicans: 338.874

2004:
Democrats:453,631
Republicans:368,334

Actual Vote:
2000: Gore (53%) 328,867 Bush (47%) 289,574
2004: Kerry (53%) 409,732 Bush (47%) 361,095

* Palm Beach
2000:
Democrats: 295,185
Republicans: 231,233

2004:
Democrats: 329,232
Republicans:233,495

Actual Vote:
2000: Gore (64%) 269,754 Bush (36%) 152,964
2004: Gore (61%) 328,687 Bush (39%) 212,688

[althecat]

64. Analysis of your data - The Figures

								
	2000				2004			
	registration		Vote		registration		Vote	
	Dem	GOP	Dem 	GOP	Dem	GOP	Dem 	GOP
Broward	456789	266829	387760	177939	533976	283736	453873	244674
Miami	396518	338874	328867	289574	453631	368334	409732	361095
P
Beach	295185	231233	269754	152964	329232	233495	328687	212688
								
								
			2000		2004			
Analysis		Dem 	GOP	Dem 	GOP	Dem 	GOP
Broward								
% of reg who voted	84.89%	66.69%			85.00%	86.23%
% reg gain 00 to 04			16.90%	6.34%		
% of vote gain 00 to 04					17.05%	37.50%
Miami								
% of reg who voted	82.94%	85.45%			90.32%	98.03%
% reg gain 00 to 04			14.40%	8.69%		
% of vote gain 00 to 04					24.59%	24.70%
Palm Beach							
% of reg who voted	91.38%	66.15%			99.83%	91.09%
% reg gain 00 to 04			11.53%	0.98%		
% of vote gain 00 to 04					21.85%	39.04%

[althecat]

66. Analysis of your data - In words

As you can see from your figures:
- The democratic party achieved 17%, 14% and 12% gains in registrations between elections in each county.
- The GOP achieved by comparison 6% 8% and 1% increases in registrations in the same period.

- The democratic party achieved a 17%, 24% and 21% gain in its actual vote between election cycles
- The GOP achieved a 37%, 24% and 39% gain in its actual vote between election cycles

On its face this suggests that the figures in Palm Beach and Broward counties in particular deserve closer scrutiny. What would convince me otherwise would be some clear reports of how exactly the GOP in Broward and Palm Beach achieved such spectacular increases in their Get out the Vote success. These are the sort of campaign organisers that the Democrats should be learning from... in both counties the number of registered GOP voters who voted increased from 66% to close to 90%.

In relation to the election exit poll debate that is underway in this thread. I would like to know how many data points the exit poll had in these three counties and what the WPEs were in each were.

As Febble will doubtless point out... the WPEs tend to scatter wildly anyway... however my assumption would be that in Broward and Palm Beach they will tend to over-estimate the Kerry Margin. And I would like to test that assumption. If you repeat the process over a bunch of other counties around the country in which you have anomalous election results then you might find a statistically significant finding....

However as it stands we cannot do this because the data remains secret.

[OnTheOtherHand]

70. just on the last part

there were only 55 exit poll precincts in the entire state of Florida. Assuming rough proportionality, there might have been 12 or 14 in Broward, Miami-Dade, and Palm Beach combined. Not much statistical power there.

Since on average there is red shift throughout the country, it isn't very useful just to see whether there is red shift in those 12 or 14 precincts.

However, nationwide, we can use 2000 returns to identify anomalous results -- and we know that those don't correlate with red shift.

These points have no bearing on whether anything bad happened in these three counties. They do explain why, analytically, I think it is a waste of time to complain about secret exit poll data in them.

[althecat]

103. Your ability to dismiss the usefullness of something you haven't seen

... astonishes me.

While the sample may not be statistically valid.... if it confirms the hypothesis then it becomes another brick in the wall.

OTOH you say:

"However, nationwide, we can use 2000 returns to identify anomalous results -- and we know that those don't correlate with red shift."

I say:

As I have attempted to explain in the past in this thread every precinct is almost certainly not be the site of vote fraud. In fact it seems probable that if vote fraud occurred it occurred in a minority of precincts. Therefore a straight national scatter-plot of WPE data is never likely to show anything obvious. And this is precisely why the data secrecy is important when it comes to analysis. I.E. If a sample of precincts which are considered to be anomalous for other reasons is selected and that sample shows a dramatic - statistically significant - red shift - then that tends to support the hypothesis that vote fraud occured in the precincts that were selected.

[OnTheOtherHand]

105. that might be because you didn't engage my argument

I'm not debating, I am reporting. Your impulse to personalize the issue trammels your ability to learn from the conversation.

The sheer paucity of exit poll precincts is one big problem; the lack of correlation between red shift and swing is pretty much a stake through the heart.

"While the sample may not be statistically valid.... if it confirms the hypothesis then it becomes another brick in the wall."

I'm not sure how to read this, if not as: "the result may be meaningless, but if we like it, we'll use it."

"As I have attempted to explain in the past in this thread every precinct is almost certainly not be the site of vote fraud."

Yes, I know that. Why would you assume that you needed to explain that the first time, never mind repeatedly?

If you believe that the exit polls are basically correct, then we need to find something like a 6% net vote shift within the exit poll precincts. Although precincts vary in size, we can get some impressionistic sense of what this entails through multiplication: e.g., a 20% shift in 30% of precincts, or a 10% shift in 60% of precincts, or a 60% shift in 10% of precincts, or a 200% shift in 3% of precincts. As you can perhaps now begin to imagine, it would not actually be very easy to miss any of these patterns in a scatterplot or the associated statistical analysis.

(EDIT to clarify first sentence)

[OnTheOtherHand]

108. oh, I thought of a more generous reading

"While the sample may not be statistically valid.... if it confirms the hypothesis then it becomes another brick in the wall."

My previous suggestion was "the result may be meaningless, but if we like it, we'll use it." That still seems like a plausible reading.

Let me try: "the data may be inadequate, but we ought to make the fullest possible use of what we have."

Yes, we should. But if that "use" is to be empirical, rather than political, we will have to acknowledge and to consider everything we find.

[OnTheOtherHand]

74. these percentage comparisons aren't very meaningful

They seem to hinge on the tacit assumption that everyone other than new registrants should have voted the same in 2004 as in 2000. It's much like your argument that Kerry must have won the popular vote because the exits indicate that new voters supported him.

An assumption that weak is grounds for setting aside an argument altogether. It may be possible to build a similar argument on better assumptions.

[Joe Chi Minh]

79. But why wouldn't the new registrants be fanatically keen on voting

Democrat, after Bush's first tenure?

Had there been ANYTHING that was not a disaster for all but the richest in your country? Even the stock-market - oil and gas related, excepted - had gone South; the national debt had gone from trillions in credit to trillions in debt; the social safety nets had been sorely vitiated; jobs are poorly paid, unemployment has, I believe, worsened. Heck, maybe the number of the poorest-paid service jobs have been boosted; disasters of historic, epoch-making dimensions, widely perceived as avoidable, such as 9/11, New Orleans, the national debt (already mentioned), Iraq and the (again widely held), world-wide growth in terrorism, the draconian, indeed, illegal impinging of civil liberties by this Administration, the roundly despotic way in which the previous election had been stolen from Gore by the Supreme court.

Look, I know you convince yourselves that urbanely-expressed demurrals and hesitant and partial acquiescences will not be noticed for the effective, unconscionable blandishments they actually represent, and the above is just one of many examples that could be cited.

[OnTheOtherHand]

91. ok, then, your arguments stink: is that better?

I can be urbane, I can be rude, people will snark either way. Whatever.

Maybe new registrants were fanatically keen on voting Democratic. Many of them, anyway. According to the unweighted Florida exit poll results, about 59% of self-reported first-time voters (not the same as new registrants) said they voted for Kerry.

The gist of your long paragraph seems to be that everyone in America should have voted against Bush. Well, not everyone did. I really don't see what else can be said about that.

[Joe Chi Minh]

99. Excellent.

[Febble]

75. It's not so much the scatter

It's the coverage. The NEP data are useless for county level analysis. In Florida, there are only about 50 precincts in the poll, to cover 67 counties. More populous counties will tend to have more precincts, as the precinct selection is stratified, but nowhere near enough for a county level analysis. As you can see, many counties will necessarily have no precincts in the poll.

It's true you can't do it because the data is secret, but it's also true, unfortunately, that you couldn't do it even if it wasn't. There is simply no sampling at county level.

But there is plenty of much better data you can analyse for anomalies, as you are doing.

[Febble]

69. I think there is some confusion here

You say:

In one of the above posts (in response to my observation that Kerry received a majority of new voters votes in the exit poll) Febble states that political science has no reason to consider that Bush would receive a re-run of 2000 in terms of existing voters votes. She goes further and suggests that the explanation for this apparent anomaly is that Bush received more votes from returning voters that Kerry (i.e. he received votes from voters who voted for Gore in 2000). This is plain wrong.

I do not recall saying this, and I cannot find it in my response to your post. I think you may be confusing my post with someone else's, and would be grateful for an actual quotation and link.


Regarding this:

Febble has stated that nationwide scatter plot evidence does not indicate that there were millions of votes added to the bush total. This is why a full set of data with precincts identified needs to be released. All precincts/counties will not have been effected.... almost certainly there will be anomalies evident between a set of tampered precincts/counties and ones where there was probably not tampering. Some of the WPE error will be random (noise) and some may have been caused by tampering. In order to find evidence of this you need to remove the random noise from the sample on the basis of some testing a variety of assumptions about how the theft may have occurred in various places.

I realise you would like to pursue particular precincts in which anomalies seem apparent, and I do understand this point. However, I would point out that the overall discrepancy is not being leveraged by a few anomalous precincts. One way we can define an "outlier" is by computing how many standard deviations from the mean it is, and whether there are more outliers than one would expect given the number in the sample. I think it is apparent from the publicly available plots of the full dataset that there are a few outliers. However, the overall discrepancy is not due to a few outliers in the redshifted direction - there are also a similar number of outliers in the blueshifted direction. Whatever is causing the overall discrepancy is not a the outliers but some slight but important factor that tends to push the whole dataset redwards. This, clearly, is what has been read as evidence of large-scale fraud; but could also arise from a slight but pervasive (but non-uniform of course) participation bias. And distinguishing between these two theories is what the analyses I have referred to have attempted to do.

Regarding the rest of this paragraph, I quite agree that some of the WPE error will be random noise, and I also agree that a "variety of assumptions about how the theft may have occurred in various places" need to be tested. Ironically, one assumption (suggested by eomer) that would be consistent with the plot is of virtually uniform fraud everywhere. There are, however, many reasons to think this is implausible, but I have nonetheless been modelling the limits of this assumption. Another is that some kind of algorithm that ensured that Bush's vote did not exceed a certain level relative to his 2000 vote, but again, there are a number of good reasons why this is hard to model in a way that is consistent with the variety of voting mechanisms used (and it might be worth noting that the majority of precinct counts on which the WPE is calculated are those collected at the precinct, where available, not county tabulator counts). But I agree that you are asking the right questions. I just want to assure you that they are ones I have been asking too.

Cheers

Lizzie

[OnTheOtherHand]

93. sorry, just getting back to this

Actually, I assumed that more of the content had been superseded by Awsi Dooger's response. Maybe it has.

I don't object in principle to using exit poll data to identify areas for further investigation, although I do point out that exit polls are a very tiny proportion of the available data.

Why do you find it peculiar that most of Bush's new votes came in the Eastern time zone (if that is what you mean)? It is hard to engage an argument that hasn't been made.

"In several of the above posts you say you find it impossible to beleive that there was widespread fraud in New York State." Well, no, I didn't. I said, inter alia, that no one had presented much evidence for massive fraud in New York apart from the exit polls, and that I didn't think it made sense to infer that on the order of 840,000 votes had been stolen there. I do not see how your point about Republican GOTV efforts has any bearing on that issue.

I entirely agree with Febble, based on a preponderance of available evidence, that Bush received more votes from returning voters than Kerry did. Your confidence that she is "plain wrong" of course does not count as evidence. (Kerry actually did receive "the vast majority of Gore's votes and then some," but this was not good enough to win.) You made no effort to present evidence that an incumbent with approval ratings in the high 40s should lose the popular vote, among returning voters or overall.

No one ever assumed that all precincts or counties were tampered with; on the contrary, it is differences in tampering that would induce a correlation between red shift and "swing." Of course the correlation will not be perfect, but it should be detectable.

In order to deem the results in Miami Dade and Broward "completely anomalous," you would need to establish a baseline from which they could deviate. The same goes for Baiman's analysis: he and others claim that the 12 dodgy counties are anomalous, but he hasn't established an empirical pattern from which they deviate. I'm not saying that you couldn't do this, only that you haven't.

[NWOhioDem]

12. Little reported knowledge about GOP fraud in NW Ohio in 2004

This is the area of Republican Tom Noe's Coingate (indicted
for stealing money from an improper $50,000,000 investment of
Ohio State BWC money) was working for Bush earning Pioneer
status ($100,000-$250,000 fundraiser).

Tom Noe pleads guilty of funneling thousands of illegal
dollars into the Bush 2004 campaign and Bush has not returned
the money to Ohio's BWC where it was suspected of being
stolen.

Ohio's Governor Bob Taft is a convicted criminal for talking
unreported gifts from Tom Noe.

Tom Noe made improper $68,000 loan to Lucas County Repo Party
(LCRP), still mostly unpaid.

The LCRP also can not account for $190,000 in donations used
in 2004 and was convicted of ethics violation.

Tom's wife was LCRP chairperson and Director of Lucas County
Board of Elections (BOE) before and during the 2004 election. 
She was forced to resign from the BOE office base on improper
activities in the Nov 2004 election by Ken Blackwell (Ohio's
SOS).

Tom was associated to Mike Drabik and is suspected of funding
his illegal and unregistered PAC called "Catholics
Against Kerry" that purchased billboards, yard signs,
bumper stickers and commercials stating "Kerry Kills
Kids"(notice the acronym)

The Lucas County Democratic Party (LCDP) and Kerry Campaign
was broken into on October 13th 2004 and only computer housing
important databases (Donors, Volunteers, Voter ID and Election
Protection Plan)was stolen.  These computers were in various
areas of the office, other computers and all printers and LCD
monitors were left behind.  Petty cash box was left untouched
on a desk.  Alarm was by passed.  Window was broken out.
ToledoGate

It is suspected that these databases were used to create
caging list to challenge voters in court and at the polls. 
Thousands of NW Ohio Democrats were subpoenaed into courts the
week before the election and hundreds had hearings before
Judge Carr threw out all challenges for lack of evidence and
being harassment.  

Thousands of Lucas County Dems were challenged at the polls.

Tom Noe's wife, BOE chair, mis-allocated voting machines and
booths from Democratic precincts at The University of Toledo
and Central Toledo into Republican suburban precincts. 

Lucas County voting machines in Nov 2004 were Diebold Optiscan
machines, processes the LC BOE used caused unreasonably caused
long lines in Democratic precincts.

Noe's Coin store (illegally bought with Ohio BWC money) was
used as the LCRP's headquarters and the Bush 2004 HQ.

Noe is suspected of illegally paying party and campaign
volunteers $10 per hour.

Noe is suspected of using the stolen BWC money to pay for Bush
campaign materials in unreported transactions.

I witnessed these volunteers throwing away a voter
registration form at a July 4th 2004 voter registration drive
in Maumee, Ohio of a person who acknowledged he was a Kerry
supporter.  This table was made to look like an non-partisan
BOE voter registration drive.  I reported it to the local
police, BOE and LCDP and there was no investigation or follow
up.

I witnessed theft of 100's of Democratic candidate and issue
signs from private property and outside the flags at the polls
by Republican party officials that went unenforced by police
who caught these people with the stolen signs.

I witnessed Republican party officials electioneering inside
the flags at the polls handing out Republican sample ballots R
BOE poll worker allowed this to continue.  This was also
witnessed by county Sheriff Deputies who did not enforce the
laws and in-turn made false claims about the placement of
Democratic signs and threatened Democratic volunteers with
arrest and force signs to be moved 500ft from polls and even
removed from private property that was 300+ feet from the
polls.

Most of these issues were covered by The Toledo Blade at
www.toledoblade.com and can be searched for verification.

NW Ohio Democrat

[autorank]

15. Welcome to DU!!! What a great composite...bookmarked!!!

Thank you for putting all this together. I've heard some of this (through the Blade) but
I've never seen it put together this well.

Thanks!!!

[mod mom]

21. Welcome to DU. I hope you have filed affidavits on the information you

provided. If you have not, please PM me (next to my user name is a little envelope icon in which you can use to send me a personal message if you like) I will refer you to attorneys who are still building the case. This information will make it to John Conyers office and the House Judiciary Dem Legal Staff who is also compiling info.

BTW...you might be interested in re-posting this info on this recent Noe thread:

http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=432521&mesg_id=432521

[Peace Patriot]

16. The exit polls are merely one indicator and set of studies in a mountain

of evidence that points overwhelmingly to a wrong outcome in 2004. This evidence begins with the deliberate creation of a NON-TRANSPARENT election system, run on "TRADE SECRET," PROPRIETARY programming code, controlled by Bushite corporations (mainly Diebold and ES&S) during the 2002-2004 period--a coup that was engineered by the biggest crooks in Congress, Tom Delay and Bob Ney (and abetted by Bilderberg 'Democrat' Christopher Dodd), with the so-called "Help America Vote Act" (HAVA). In a very short time, this $4 billion electronic voting boondoggle greatly corrupted the voting systems in almost every state. That was its intent. These legislators and their corporate partners COULD HAVE created a transparent, verifiable election system. It is not difficult to do. They did just the opposite--they fought off even the most basic transparency features, such as a paper trail, and permitted election systems to be privatized and SECRETIZED in every way, while drenching them with money, and failing to fund or require proper regulation. (For instance, they permitted secret industry "testing" of the machines!) Election officials who resisted were bullied with draconian deadlines into purchasing insecure, unreliable and extremely hackable electronic voting systems, or face penalties.

This is the CONTEXT in which Kerry won the exit polls. The exit polls are the only verification tool we had in highly non-transparent conditions, and they showed a Kerry win. The corporate news monopolies then shut down the reporting system late on election day and DOCTORED their exit poll data--in impossible ways--to make it FIT the results of Diebold's and ES&S's secret vote tabulation formulae. When they were caught, by alert bloggers and experts who took screen shots of the real exit poll numbers before they were taken down, the pollster--one pollster hired by the news monopoly consortium--then started backfilling, and dissembling, and making things up to justify what they had done.

The CONTEXT of these dastardly deeds also includes the war on Iraq. The Bush junta and the war profiteering corporate news monopolies lied about THAT, TOO! With nearly 60% of the American people against that war from the beginning, and with people flocking to the Democratic Party in 2004 to oust Bush--the Dems had a blowout success in new voter registration in 2004, nearly 60/40--there is no way they could have kept their boy, Bush, in the White House, and protected their war profits, war crimes and tax breaks WITHOUT fixing the election.

People come at this election fraud issue from the wrong direction. They presume the innocence of a government that has proven itself, time and again, to be a criminal enterprise, and they presume honesty by a corporate news establishment that has proven itself, time and again, to be a propaganda arm of the government!

You cannot and must not presume innocent by those in power when it comes to vote counting--let alone SECRET voting counting. These jokers must have been laughing all the way to Baghdad, knowing what they had cooked up here at home: TRADE SECRET VOTE COUNTING--by their bud, Wally O'Dell, at Diebold--a Bush/Cheney campaign chair and major fundraiser, who promised in writing to "deliver Ohio's electoral votes to Bush/Cheney in 2004", and by ES&S, a spinoff of Diebold (similar computer architecture) initially funded by rightwing billionaire Howard Ahmanson, who also gave one million dollars to the extremist 'christian' Chalcedon Foundation (which touts the death penalty for homosexuals, among other things).

What a coup! Can anybody doubt what all this secrecy was FOR?

And....AND...Kerry won the real exit polls (before they doctored them to suppress strong evidence of the Kerry win).

The exit poll scam does not exist in isolation. It is not some pristine, theoretical mathematical problem. The polls said one thing; they changed them to say something else--that which would protect war profits, and gas gouging, and torture, and unjust war, and tax cuts for the super-rich.

And then--perhaps their foulest deed of all--for extra insurance, they picked on the poorest and the darkest skinned among us, the black voters of Ohio, for special treatment on election day: old black grandmothers standing in the rain for ten hours to vote; black felons who had served their time and won back their right to vote, called by the Texas Rangers and told they would return to jail if they showed up at the polling place; the black, the poor and the young deprived of precincts, lied to, purged, unfairly challenged, all over Ohio--by the most corrupt Republican establishment in the country (and that's saying something); and, in Florida, black voters who had names SIMILAR to convicted felons purged by the tens of the thousands from the voting rolls, and tens of thousands of absentee ballots 'diseappeared' in the mail in Democratic areas.

The Bushites spat on Martin Luther King's grave. They destroyed our finest achievement as a people, the Civil Rights Act of 1964 and the Voting Rights Act of 1965. They dismantled them before our very eyes. I favor the theory that they did it for spite--out of sheer ugliness--but an alternative theory (and maybe both are true) is that Kerry's victory was much bigger than expected--a landslide victory, perhaps as much as by 8 million votes--and the voting machines could not handle it without exposing the rig to plain view (or because they had to be pre-programmed to certain percentages). So they had to engage in blatant, illegal vote suppression, and, of course, picked on the weakest and the poorest, as they always do.

CONTEXT, my friends. CONTEXT. Think about Katrina. Would people who would leave thousands of poor blacks to rot and die in a natural disaster be capable of such scumbag scheming as is evident in the passage of the "Help America Vote Act" by the Anthrax Congress, the FALSIFICATION of the only verification tool we had, the exit polls, and the dreadful, insulting, demeaning, Nazi treatment of black voters in the 2004 election?

Gee, I don't know. On the other hand, we could go on playing games about voting results that have been deliberately rendered non-transparent, and exit polls that were falsified to fit those non-transparent results, and never come to any conclusions at all.

[OnTheOtherHand]

22. but no

I understand your strong need to believe that the exit polls support your beliefs.

But until you can explain Kerry "winning" the New York exit poll by 30 points, I could not possibly agree. And that is just the tip of the iceberg.

No one here is presuming innocence; that is a straw man.

If you want to stop "playing games" about exit polls, then why not stop?

[glitch]

23. You see the forest, as usual. There is an agenda involved here, remarkable

that it's ignored. I guess if you ignore the agenda you can dismiss the pattern and focus solely your particular reduction.
Like many blind hands on the elephant, reductionism is not useful in every situation.

[Joe Chi Minh]

28. You got it, but he doesn't want to know. None so blind...

[Febble]

32. Five blind men and an Elephant

http://en.wikisource.org/wiki/Five_Blind_Men_and_an_Elephant

And the moral of the story is:

You can get seriously lost in a forest.

[mod mom]

41. Febble,in response to your suggest story, there are many aspects to our

suggestion of theft just like there are parts of an elephant. That is what many of us on this forum have been suggesting. I don't claim to haave expertise (or even understanding) of the exit poll controversy, yet I have read reports from those who do (like Dr Baiman) and who suggest they do point to fraud. For me, I have always concentrated on what I WITNESSED first hand, and that is the blatant civil rights abuses that were meant to disenfranchise a certain segment of the population. (If you are interested here is my suggestion how it occurred:
http://journals.democraticunderground.com/mod%20mom) I have left off exit polls and vote flipping/machine manipulation, not to my disbelief in them but due to my lack of knowledge on the subject. I also believe there is sufficient reason to question the outcome without them. They just make the case stronger.

I am not an attorney, so I have relied on the expertise of attorneys who I have worked with (Fitrakis, Arnebeck, Truitt, Peckarsky, Lehto, Bonifaz and now RFK Jr) who all believe the evidence is sufficient to prove in a FAIR cour of law that Kerry won Ohio. Shortly after the election I was part of a Liaison Committee, which included Lehto, and others involved in election reform who spoke with attorneys from the House Judiciary Dem staff (ie worked w Conyers in producing the infamous report). They also believed the evidence was sufficient. Now we have added much more evidence to support our claim. It is the expertise of these attorneys who convinced me to stay on track and work on this investigation.

So here is the issue I take with some of OTOH and your posts. You claim expertise, but there is very respected expertise to counter what you two say in questioning the stated theft. There is also legal expertise, technical expertise, procedural expertise and personal witnesses. The attorneys have looked at the entire body and have drawn the conclusion that the evidence is sufficient. Your expertise (one aspect and there is respected disagreement among experts in that area) is but one piece of the puzzle. What I don't understand is the firm committment to keep your stance despite all combined evidence. I don't understand why you don't look at the entire elephant.

[Febble]

42. Thanks for this response, mod mom,

and I actually wish more people would concentrate on what people like you witnessed first hand. I think you have an incredibly powerful case. In Franklin county alone, I think you have a civil rights case (not that I'm an attorney) without even having to prove intent. The fact is (and the evidence that it is a fact is overwhelming)that whatever allocation algorithm was used to allocate machines in Franklin it selectively disenfranchised African American voters. I think there may be a case in Cuyahoga as well, on the provisional ballot isse, but I know less about that. And that's before mentioning the lax manner of the recount, and the apparent disdain for the electoral process shown by Blackwell.

I don't know much about evidence for direct vote tampering in Ohio, but the fact that it's not evident in the exit poll evidence means very little - there are only 49 precincts in the exit poll sample, so if fraud was concentrated in one county, it would be unlikely to catch it. The exit poll hasn't much chance of catching anything other than incredibly widespread fraud anyway.

As for whether the voter suppression evidence adds up to a Ohio win for Kerry, I wouldn't know, it's not my field. Maybe it does. But whether it does or not makes not one jot of difference to the civil rights case. After all, Bush voters were also disenfranchised by those long lines. It's the disenfranchisement that matters, and of course the lack of accountability. And if you can show it was deliberate on anyone's part, then again, it doesn't matter whether it worked, or whether it didn't need to, if it happened.

OK, preamble over, to tackle your last paragraph:

I don't look at the entire elephant because I can't see the entire elephant (although I've seen enough of it to know it's an ugly brute), and in any case, my expertise is in trunks. But to leave that metaphor aside: sometimes several pieces of circumstantial evidence can be added to make a case. But not always, and I don't think the fact that there is good evidence for voter suppression in Ohio make it more likely that the nationwide exit poll discrepancy was due to fraud, except in the sense that if the voter suppression was deliberate, the perpetrators might not have been above vote switching (which is a good point). So what I have been doing is trying to find out whether the exit poll evidence suggests widespread massive fraud. And there are ways in which the exit poll data can be interrogated to at least address that question, if not answer it definitively. But what doesn't form part of the interrogation is evidence about voter suppression, except that it provides a rationale for interrogating the data in the first place. Voter suppression doesn't show up in an exit poll (although one of its close relatives - high differential residual vote rates - may do).

But the reason I may appear to have a "commitment" to my "stance" is not that I have some a priori need to defend the exit polls at all costs. Far from it. The reason I got the contract to run those analyses for Mitofsky was precisely because I'd published a paper saying that I thought the analyses reported in the E-M report needed to be redone. And that paper arose from a series of diaries on DKos devoted originally to evidence for fraud in Ohio by Georgia10 (have you read her compendium?), in which a vast team of DKos posters tried to sift the good evidence from the bad. Armando had promised to post it on the main page if he found it convincing. And he did. I ended up being associated with sifting out the bad, seeing as statistical expertise is relevant to exit polls, and it seemed pretty clear to a number of us that the arguments being made from the exit poll data were by far the weakest. But we agreed that it was nonetheless (on the basis of our knowledge back then), legitimate evidence, and it went in.

So if I do, now, appear to have a "commitment" to a "stance" it is one that I have arrived at after a considerable journey, and frankly, I would not consider it a "committed" stance at all. I 'm a scientist, after all, and while the more evidence I have for a conclusion, the more confident I am likely to be of that conclusion, scientists, by definition actually, regard all conclusions and explanations as provisional. And thus I will always regard my current "stance" on what the exit poll evidence is telling us as provisional.

OK, this is getting a bit long, but let me cut to the chase: obviously I have this extraordinary position with regard to the data, in that of the protagonists in the online debate, I have, uniquely, actually analysed it. I am therefore aware that some of my views inevitably arise from analyses which I have done, but the results of which are not in the public domain. However, the most important findings are. I taken pains to point only to findings that have been publicly presented, and, if possible, can at least be partially evaluated by objective observers, as is the case with scatterplots, one of the most transparent form the results of an analysis can take .

So, to wind up: it is my view (expressed here:

http://journals.democraticunderground.com/Febble/3)


that the exit poll evidence a) does not support nationwide vote-flipping fraud on a massive scale, and b) if anything contra-indicates it. I am aware that Ron Baiman and others have a different take on the exit poll story. I think Baiman's analysis of the exit poll data released for Ohio is really, really, seriously flawed. This is not because I am failing to consider the whole elephant. It is because in trying to discern the shape of the elephant I found a completely irrelevant object which turned out to be a turkey. And I think Ron has the turkey.

And I don't think turkeys add luster to your excellent civil rights case.

[Joe Chi Minh]

47. Still can't see beyond that trunk, can you, Febble?

"And I don't think turkeys add luster to your excellent civil rights case."

There are a lot of bad guys out there who'd be overjoyed (and not a little amused. Bad guys can have a sense of humor) to see your surreal inversion of priorities!

Civil-rights problems are a permanent 'given' under Republican administrations; it would scarcely be an exaggeration to say they are one of the salient features that defines them.

On the other hand, when the fraudulent election is exposed in all its ugliness - well that would be a tall order - the incumbent Democratic administration can be relied upon to restore the already-vitiated or lost civil rights of minorities, at least to the level obtaining during Clinton's tenure, and in fact, they would be steadily improved. AS WELL AS THE LOT OF EVERY OTHER AMERICAN CITIZEN BEING IMPROVED SIMULTANEOUSLY.

You can see the bad guys laughing with joy at your post, now, can't you?

[Febble]

48. Why would anyone laugh at my post?

Not being a bad guy, I suppose I don't see the joke.

[Joe Chi Minh]

49. Well, it would be what seems to be your implicit suggestion

that Democrats should rejoice at possibly having an excellent case of civil-rights abuses by the Republicans, and kiss the exposure of the whole multifariously fraudulent election - which would solve a HOST of other major issues - goodbye.

I mean there's often a brutish kind of common sense, even a kind of basic honesty about many even far-right wingers, sometimes absent in Democrats/liberals. Most notable by far in this regard, I think, would be Pat Buchanan. And I think they would marvel at what they would perceive as your gall, trying to pull a stunt like that; not realising you were speaking in the utmost good faith.

[Febble]

51. Well, if I'm right

and the millions of votes weren't stolen, then it is rather good news for Democrats, because it means you have a chance of winning in November. But the fact that there is an excellent case for civil-rights abuse is not terribly good news unless someone brings it to court.

I assume that yourself believe that millions of actual votes were stolen in addition, although you still haven't answered my question about what magnitude of vote-miscounting theft occurred and what evidence supports your number. Remember that the only kinds of fraud or disenfranchisement that will show up in an exit poll are kinds that affect votes cast, and not even all kinds of miscounting of votes will do that.

My position is that I don't think the exit poll evidence supports the case for millions of votes switched to Kerry. I have no views about any other form of theft. That is why I asked you how many votes you thought had been stolen and why.

But rather than answer my questions, you keep making unsupported general statements about my motivation and mindset, about which you appear to know precisely nothing.

I'm happy to debate this with you, but not very happy to simply read your erroneous views on my integrity and intelligence.

[Joe Chi Minh]

54. Well, if I'm right and the millions of votes weren't stolen, then it is

rather good news for Democrats, because it means you have a chance of winning in November."

That's even funnier! I have to say it. I'm not baiting you now.

The reason why I don't accept your singularly reductionist framing of the issues of the fraudulent election, is because I have to admit that I am singularly unimpressed by the style of writing you apologists affect, evidently intended to impress us all that you are deliberate and serious scholars. Deliberate and serious scholars can see the wood for the trees, because they adopt a plausble hypothesis.

In the teeth of the wholesale mendacity and bad faith of this Administration, which totally dwarfs what was already a sorry history of endemic election fraud in the US, the massive and incontrovertible suppression of the votes of members of minorities, the proven comments among friends of Republicans connected with the election, that they would "deliver" the Ohio election to Bush (O'Dell of Diebold, Blackwell and Jeb, in relation to Florida, and some drunken mutt saying that that he knew they'd already won, because they would be doing all the counting; in the teeth of all this and so much more, you are so misguided as to adopt as your initial hypothesis that the Republicans conducted an honest presidential election!

Ohio and The United States electorate, good, bad and indifferent, were not confounded test-pieces in a laboratory! Boasting about how rigorous your research and analysis, when your most fundamental hypothesis has remained unchanged from its pristine nescience, is, in these circumstances, a very sorry joke. Precisely because of the history of endemic election fraud in the US, in a measure, at least, unknown in Western Erope, and the maniacal level of partisan animosity - not to SPEAK of the habitual reliance of all Presidents on the financial patronage of rich individuals and businesses, invariably including major criminals, even sometimes organised crime bosses, indicted before or after such elections, then pardoned, your sanguine 'tabla rasa' hypothesis seems so unscientific in the true sense of the word (knowledge), that just responding to you all thus far has been very burdensome. However when the most basic assumptions are not shared - in this case, to name but one, a common understanding of the importance of a plausible initial hypothesis, and what would constitute it in this context, then argument is futile.

The purpose of an open mind is to close on the truth, but when such antithetical positions are held by disputants, one has to be a fool or a rogue, and the other, at least relatively sensible. The upshot is that continuing argument makes two fools, where there was only one before. There's not meeting of minds. So, if I'm the fool, I don't want to turn your mind to mush; and by the same token, if you are the fool, I'll spare my mind the indignity. Ave at que vale. Now, there's a pretentious valediction for you.

[OnTheOtherHand]

57. actually not

Febble never adopted the "initial hypothesis that the Republicans conducted an honest presidential election!" Yeesh.

If people would, from time to time, attend to what Febble actually writes, things might go better.

[Joe Chi Minh]

61. You lose. There's no getting round it. The people here are not

fools. You're not preaching to the choir here. Get over it! It's possible, if you try hard enough. Life goes on. It'll come back to bite you in the bum, when it all comes out, but you did your best. That's the thought to carry with you. "Tant Que Je Puis", my old school motto!

[Joe Chi Minh]

62. THEN WHY AREN'T YOU BOTH DEMANDING THAT THEY

PROVE THE HONESTY AND INTEGRITY OF THE ELECTION? Doesn't democracy demand at least as much?

Instead, you try to prove the impossible, that the polls were faulty and by implication that there is no evidence of fraud to be extraolated from them!!!!

[Febble]

67. I am.

And it does.

As for your last sentence, it appears that you cannot read.

[Name removed]

68. Deleted message

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

[mod mom]

50. New Baiman post on San Diego:

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=132x2671568

[RonB]

65. Why the Critics are Wrong

1) Febble's "U" shape is just a general pattern derived from partisan response rate equations that are mathematically equivalent to those used in NEDA reports to show that only highly divergent (that is statistically significantly) divergent partisan exit poll bias for high Bush versus high Kerry precincts can explain the national pattern of exit poll discrepancies shown in the Mitfosky January report. Her attempt to use essentially the same (mathematically equivalent) analysis to prove the opposite has never made any sense. If discrepancy declines for partisan districts (the "U" shape), how does this explain a massive increase to 10% in high Bush precincts?

2) OTOH's distinction between what I've called WPD and "Kerry WPD" (Kerry official vote share minus Kerry exit poll vote share) is correct but has no bearing on exit poll margin of error calculations as long as one is consistent in applying WPD margin of error to WPD, and Kerry WPD to Kerry WPD margin of error (the former is exactly twice the size of the latter). You can't mix and match. I was applying Kerry WPD's to Kerry WPD margin of error. Ohio has a large and highly statistically significant discrepancy - especially when the doubled sample size (recently acknowledged by Mitofsky) is used.

3) The Ohio precinct level data is the most accurate that is public ally available (no cluster adjustments necessary and about as close to random sampling as you can get). It shows an unbiased WPD on the right (in high Kerry precincts) and highly biased (against Kerry) WPD on the left with no "U" shape and a pattern that is strongly consistent with vote shifting. Mifosky (and Febble's and OTOH's) "reluctant Bush responder" hypothesis (rBr) doesn't come close to explaining this pattern. After applying the most explanatory rBr, 30% of Ohio's 49 exit polled precincts still have significant discrepancies that overwhelmingly against Kerry - that add up to more than double the number of votes necessary to swing the election to Bush.

4) The 12 county Ohio evidence is even more damning than the exit poll evidence. Its shows a series of coincidences for which no other explanation is possible (other than the hand of God) except vote shifting. I have challenged anyone to explain these "coincidences" in some other way. No one has. Febble's attempts to say the there is some necessary link between the C/K and B/M ratios are simply wrong. Votes for unknown and underfunded "down ballot" judgeship candidates are generally driven almost entirely by the "up ballot" candidates - especially in a race as hotly contested and contentious as Kerry/Bush in Ohio 2004! The data shows no evidence of Moyer to Connally substitution - just of Kerry to Bush vote shift.

More details on all of this are available in AAPOR paper at http://www.freepress.org and in my postings at http://www.baiman.blogspot.com

Thank you all for your support and encouragement!

Ron

[OnTheOtherHand]

71. well, no

1) I would think a glance at the WPE scatterplot would establish that the "massive increase to 10% in high Bush precincts" doesn't mean much.

2) Ron is ignoring context. We are not debating whether there was a discrepancy in Ohio; we are debating Kennedy's claim that the exit polls gave Kerry a decisive lead in Ohio. Ron may well believe that even the latter is true, but as the E/M evaluation report shows at pp. 21-22, he is wrong.

3) As we've pointed out repeatedly, Ron's description of the "pattern" in Ohio doesn't bear up to statistical examination. Possibly that is why he doesn't report a statistical test. (And Ron's assumption of constant non-response bias is simply irrelevant.)

4) It's not at all coincidental that Connally tends to do better where her opponent tends to do worse -- let's not check common sense at the door. And I have explained repeatedly, and without rebuttal, why Connally is most likely to outstrip Kerry in Republican counties. But if Ron wants to argue that the Connally result is anomalous, then he should at least present one pair of presidential and judicial races that turned out the way he would consider normal. Otherwise, again, there is really no statistical debate here at all, only Ron's subjective notions of weirdness.

EDIT: fixed link, also softened opening claim for people aren't used to estimating statistical tests from scatterplots

[Febble]

72. Hi, Ron.

1) Febble's "U" shape is just a general pattern derived from partisan response rate equations that are mathematically equivalent to those used in NEDA reports to show that only highly divergent (that is statistically significantly) divergent partisan exit poll bias for high Bush versus high Kerry precincts can explain the national pattern of exit poll discrepancies shown in the Mitfosky January report. Her attempt to use essentially the same (mathematically equivalent) analysis to prove the opposite has never made any sense. If discrepancy declines for partisan districts (the "U" shape), how does this explain a massive increase to 10% in high Bush precincts?

Answer



From Mitofsky's slide presented at AAPOR 2005, at which you were present.

The answer is on the slide, and the results of the assymmetric U shaped functions, namely family of assymmetrical U shapes, the direction of the U shape depending on the value of "alpha" aka "bias"), is apparent. As you can see, the values approach zero at the extremes; however, because the U shape for any given value of "alpha" is assymmetric, where alpha represents either pro-Kerry bias in the poll or pro-Bush bias in the count, the value remains more redshifted(in this plot redshift is shown as positive) for longer as you approach 100% Bush. Similarly, where alpha represents either pro-Bush bias in the poll or pro-Kerry bias in the count, the value remains more blueshifted (negative on this plot) for longer as you approach 100% Kerry. And as you can see there is a huge range of WPE values, and they include both extreme positive and extreme negative WPEs. Whether these are due to bias in the poll or bias in the count they clearly represent a non uniform distribution of alpha. And because of the assymmetric function of alpha with respect to vote share, the cloud of plots has maximum variance in the centre, and is slightly tilted.

The plot tells you nothing about fraud, and nothing about bias. It simply tells you that at least one of the two was present.

And to answer Ron's specific question: the reason that the mean WPE was more redshifted in the category defined as "high Bush" in the E-M report is also easy to see from the plot: four highly redshifted outliers are present in that rather small category (40 precincts IIRC), exerting high leverage on the category mean. And indeed, the overall correlation between WPE and Bush's voteshare is significant (statistics given on the slide). However, as I explained above, under the null hypothesis of randomly distributed fraud or bias, the "expected value" of that slope is in the observed direction, and the observed slope is not, as it turns out, significantly different from the expected value.

I will say, however, and Ron will no doubt agree, that a zero order correlation is fairly uninformative. If we can account for some of the variance with known factors, then we will have more statistical power to detect a real effect. And that is what was done in the analysis reported here:

http://inside.bard.edu/~lindeman/slides.html

The results of that analysis does not support the interpretation that redshift was due to fraud.
I understand that Ron disputes the validity of this analysis, and if he would like to detail his refutation on this thread, I promise to engage debate.

3) The Ohio precinct level data is the most accurate that is public ally available (no cluster adjustments necessary and about as close to random sampling as you can get). It shows an unbiased WPD on the right (in high Kerry precincts) and highly biased (against Kerry) WPD on the left with no "U" shape and a pattern that is strongly consistent with vote shifting. Mifosky (and Febble's and OTOH's) "reluctant Bush responder" hypothesis (rBr) doesn't come close to explaining this pattern. After applying the most explanatory rBr, 30% of Ohio's 49 exit polled precincts still have significant discrepancies that overwhelmingly against Kerry - that add up to more than double the number of votes necessary to swing the election to Bush.

One reason the Ohio precinct level data is highly inaccurate is that in the majority of precincts, the voters were sampled from several precincts sharing a polling place, but the result was compared only with the vote count for the selected precinct. Ohio has an unusually high number of precincts that share a polling place. This hugely increases the standard error, clearly. The sample is simply not a straight sample from the precinct of interest. The significance values you calculate in your paper are therefore, wrong for that reason alone (though there are other reasons as well, namely fairly serious problems with your sample size assumptions)

But note:

1. there are no "high Bush" precincts in the Ohio dataset (max Bush vote share 78%)


2. there is one extremely "high Kerry" precinct (Kerry 96%), and right on cue, the WPD declines to zero.


3. As explained above, the "U" shape given by any given value of "alpha" (for further details read NEDA papers or mine) is assymmetric, so that for a value of "alpha" that represents either pro Kerry bias in the poll or pro-Bush bias in the count, one would expect slightly higher values in moderate Bush precincts than in moderate Kerry precincts.


4. However for a distribution of alpha values (whether caused by bias in poll or bias in count) you would expect a tilted ovoid with maximimum variance where support for each candidate is even, but with values more negative in the moderate Bush part of the spectrum than in the moderate Kerry part of the spectrum.

These points are well illustrated in your own plot:



as well, of course, in the plot of the nationwide dataset above.

Note the data point with zero WPD in the one extreme precinct, the ovoid shape of the scatter, and the tilt, and an overall redshifted WPD. Note also that in your bar chart this data point is invisible, and that scatterplot version allows more of the variance in WPD - indeed of predicted ovoid distribution - to be observed.

The plot is thus consistent with a distribution of "bias" in the count, favoring Bush and equally with a distribution of bias in the poll. The plot does not disambiguate the two.

Moreover, if you algebraically subtract some mean value of an effect from a series of datapoints, and are left with residuals, all you demonstrate is that there is variance in your effect. Why anyone would think that non-response bias or selection bias would have a uniform value in all precincts is something I have never understood, and in any case, was ruled out by Edison-Mitofsky's own report which you yourself have analysed in detail - the degree of discrepancy varied according to a number or rather important factors. The extent to which these factors were present in any one precinct will, inter alia, affect the total discrepancy.

The 12 county Ohio evidence is even more damning than the exit poll evidence. Its shows a series of coincidences for which no other explanation is possible (other than the hand of God) except vote shifting. I have challenged anyone to explain these "coincidences" in some other way. No one has. Febble's attempts to say the there is some necessary link between the C/K and B/M ratios are simply wrong. Votes for unknown and underfunded "down ballot" judgeship candidates are generally driven almost entirely by the "up ballot" candidates - especially in a race as hotly contested and contentious as Kerry/Bush in Ohio 2004! The data shows no evidence of Moyer to Connally substitution - just of Kerry to Bush vote shift.

Well, Ron, as you know, I have demonstrated to you, actually using a simulation, that there is indeed a relationship between C/K and B/M. The lower the number of votes in the judicial race relative to the presidential race, the looser this relationship will be, but a relationship there is, simply because there are a finite number of possible judicial votes. In fact, on my calcs, the judicial vote is on average about 85% of the presidential vote, so you'd expect the relationship to betwen C/K and B/M to be fairly strong (and when I say "expect" I mean this in a mathematical sense as well as in a common English sense, as you will be aware). Where Bush is doing his best relative to Kerry, he will also tend to out run Moyer by the largest percentage, even when Moyer is doing well relative to Connally, as will tend to be the case where Bush is doing well relative to Kerry. And where Moyer is doing worse than Bush, Connally will tend to do better relative to Kerry, and will therefore be more likely actually to outrun Kerry. These are simple properties of the distributions. And to illustrate it, I would draw attention, in your own data, to four counties where Kerry not only outruns Connally but where Moyer is outrunning Bush. And these four counties comprise 467,175 presidential votes, not incomparable with the 564,935 presidential votes cast in the Baiman Twelve. And, all four of these counties are counties in which Kerry just happens to be in the lead.

If you believe in your coincidences, surely this coincidence points to pro-Kerry fraud in these counties? Except that it doesn't, it is simply the mirror of the phenomenon by which you would expect the presidential candidates to outpoll their judicial counterparts in precincts where they are doing best, and to do worse relative to their judicial counterparts where they are doing worst.

This is not to say that there is nothing suspicious about the fact that the judicial race seems to have been sufficiently non-partisan that Connally was able to outpoll Kerry where Kerry had a small share of the presidential vote. There may be. But there is nothing suspicious about the set of "coincidences" - if the judicial race is less partisan than the presidential race (and my understanding is that no partisanship was shown on the ballot) then the constellation of phenomena you observe is simply what is expected under the null hypothesis. Far from being "coincidences", they represent a simple mathematical relationship.

edited for grammar and typos

[Febble]

76. Missed a bit:

...partisan response rate equations that are mathematically equivalent to those used in NEDA reports to show that only highly divergent (that is statistically significantly) divergent partisan exit poll bias for high Bush versus high Kerry precincts can explain the national pattern of exit poll discrepancies shown in the Mitfosky January report.

Ron, as OTOH and I have already explained to you, 1) response rate and participation rate are not the same thing and 2) your partisan response rate equations generate impossible values (>1) even under conditions where the only error is sampling error within the MoE, and c) these impossible high values are more likely to be generated when the voters in question are in the minority.

Your partisanship response rate do not give you what you think they give you. For the relevant math, see Lindeman, Liddle and Brady, downloadable here.

Lizzie

[RonB]

100. The Critics are Dead Wrong and Have NO Explanations

I'm afraid OTOH and Febble are too committed to their past errors and misjudgments to acknowledge the obvious.

1) The scatter plot that they keep throwing up (for which no one has the data but them) is what's meaningless. They use "data variance" to suggest no significant correlation between precinct partisanship and WPD bias - but this is ludicrous. In Ohio data variance was 16% as I recall - nothings goings to turn up significant based on data noise - all exit polling statistics are based on "model variance".

2) Kerry did indeed have a decisive lead in Ohio based on the un-adjusted exit poll data that closely matches Mitofsky's final numbers (see NEDA March report). No amount of playing games with definitions of bias, etc., can change this fact. What Mitofsky SAYS in his ignominious Jan report is quite irrelevant to this FACT and contradicts his OWN margin of error estimates that were released BEFORE THE ELECTION as I've pointed out my responses to Manjoo. If anyone is guilty of misrepresentation its Manjoo and Blumenthal who are apparently trying to apply WPD margins of error to "Kerry WPD" rather than to double this as they should.

3) The OTOH/FEBBLE/MITOFSKY bit about "constant" bias is an old horse that I thought they had abandoned long ago. NO ONE is modeling CONSTANT bias - all calculations in NEDA reports after March are based on "AVERAGE" bias with an error term (in the bias). And the main point is that these differences in "partisan response exit poll bias" have to be highly significant (significantly larger in high Bush precincts than in high Kerry precincts)to explain the differences in mean and median WPD between this types of precincts.

4) The claim that I "don't report a statistical test" for the Ohio data is outright bunk! It just shows that OTOH has never actually read the Ohio report (I'm being charitable here!) - much of which is based on statistical tests that had to be estimated as Mitofsky AND ESI did not release any statistical tests for the precinct data (in direct violation of AAPOR ethic guidelines - as is Mitofsky's refusal to release precinct and pollster characteristics data for other states).

5) On C/K and B/M - as I've stated the notion that there should some "arithmetic" link between these ratios only makes sense if you assume SUBSTITUTION between Connally and Moyer voters - that is that voters see these candidates as actually COMPETING AGAINST EACH OTHER rather than just "down ticket" completions of a Kerry or Bush vote. Moreover the set of absolutely implausible "coincidences" does only BEGINS with comparative inequalities - it also includes rough matching of lost Kerry and excess Bush votes FOR THE TOTAL AND FOR ALL 12 COUNTIES for which C/K >1. There is absolutely no a priori reason why these numbers should match the way they do absent vote shifting.

For example, if on looks at the only 4 (for a well funded incumbent!) counties where M/B>1 one finds that:

a) in three of these four Moyer got ABOVE average votes shares ( suggesting that IN THIS CASE it is plausible that the high M/B is driven by substitution from Connally to Moyer which would also cause a lower C/K or higher K/C ratio. And this turns out to be true, 3 out of the 6 counties where K/C>1.46 are also counties where M/B>1. Note that this would be the tendency, since if Moyer got a higher than average vote share, Connally would get a lower than average share (of the total vote for supreme court judge), driving K/C up. The ratios are being driven by shifts from C to M. This pattern is not surprising and "fits" OTOH's and Febble's "substitution" argument.

b) But for the C/K pattern we have just the opposite! In 11 of the 12 C/K >1 counties Connally got a BELOW average vote share (and Moyer a corresponding ABOVE average vote share). Nine of these 12 are among the only 14 cases out of 88 where B/M>1.43, but this is not driven by shifts from Moyer to Connally, but rather by shifts from Kerry to Bush. Moyer's share is NOT below average and Connally's is NOT above average.

c) Additional "proof" of this can be seen by the fact that the "lost Kerry" and "Excess Bush" votes roughly match 12 TIMES AND IN TOTAL - see Manjoo response at www.baiman.blogspot.com whereas a similar calculation of "Lost Bush" and "Excess Kerry" votes for the four M/B>1 counties show no match at all (as one would expect as they're being driven by Connally to Moyer shifts that have little to do with Bush and Kerry vote shares).

Counties with B/M<1
COUNTY Lost Bush Vote Excess Kerry Vote
Erie -8955 5736
Mahoning -16989 8683
Stark -22095 6538
Trumbull -13349 9026
Total -61389 29982

The total "Lost Bush" vote of 61,389 is almost double the "Excess Kerry" vote and there is no rough match in orders of magnitude except maybe for Erie.

The critics are just dead wrong in their assumptions and in their claims to have, even a remotely plausible, explanation for the 12 rural county "miracle" in Ohio!

[Febble]

101. Ron, if we are going to debate this here

then starting by casting aspersions on other posters' motivations isn't a good start. We are all prone to error and misjudgement, and indeed, the more I have looked into exit poll analysis the more I have learned about the complexity not only of the data, but the way that probability distributions work, some of which has led to a revision of my earlier views and conclusions. I believe some of your own work has been revised. So let's leave accusations about others being "committed to their past errors and misjudgments" out of this.

And let's put some of this right.

1. OTOH does not have access to the data. I alone was contracted to analyse the data and it was not released to me, therefore I have no authority to release it to anyone else. You keep asserting that OTOH has the data, even though we have both told you, publicly, that he does not. Unless you can provide reasons to believe that we are lying, then you should stop making assertions that have been categorically denied.


2. If you think that nothing can be gleaned from the data, then why analyse it? If there is variance over and above sampling variance, and there is, then the way to analyse it is to model some of that variance, not to pretend it doesn't exist. And clearly, for the Ohio data, because the majority of precincts were in shared polling places, then there will be a large amount of non-sampling variance even before we start to consider bias. You simply have not addressed this issue in your own analyses.


3. The t scores that determine the confidence intervals for the projections are based on estimates of non-sampling as well as sampling error. Estimates of non-sampling error are made from the incoming data. They cannot be determined before the election.


4. I have never known what you meant by "constant" bias, Ron, but I understand from your recent paper that you algebraically subtracted a mean bias from each precinct. This will leave you with residuals, as you observe. The fact that you are left with residuals merely tells you that bias wasn't constant. It doesn't tell you that it wasn't bias.


5. What you do not state is a statistical test for your bar chart. I have stated it, however, and it isn't significantly different from the expected value. I have also tested your division non-parametrically, and that isn't significant either. You certainly do provide statistical tests for your precinct MoEs. Unfortunately they are based on inflated sample sizes, erroneously matched precincts, the assumption that the voters are sampled from a single precinct, and the assumption that a partial response rate does not affect the randomness of the sample.


6. On the Connally issue: check out Trumbull, Erie, Mahoning and Stark.

Lizzie




edited to insert a missing "not"

[OnTheOtherHand]

102. OK, Ron, just keep escalating

1) Can anyone honestly say that they know exactly what Ron's first point means? If we aren't allowed to look at the data in assessing the correlation between partisanship and WPD, what on earth are we supposed to look at?

2) Ron, let's be clear here. Are you actually asserting that Edison/Mitofsky have falsified the standard errors in the evaluation report? If so, how would you know this? Did you write the estimators for them?

Ron, the tables you cite in the methodology statements give margins of error for subgroups in the tabulation. They are not intended to supply margins of error for the estimates. It's fun to speculate about what the design effect multipliers ought to be for the estimates, but the exit pollsters actually figure the standard errors of the estimates from the data themselves. If you don't like it, go run your own exit poll. But it seems quite unreasonable to hang your hat on the accuracy of the exit polls while ignoring how the exit polls actually work.

I can't figure out why you are even arguing this point. Why are you so hell-bent on arguing that Kerry had a decisive lead in the Ohio exit poll? Are you convinced that E/M was on the verge of calling Ohio for Kerry at 7:30 PM, but then Karl Rove called and threatened to shoot Mitofsky's dog? What are you saying, man?

(Full disclosure: I have no idea whether Mitofsky has a dog.)

3) Umm, yes, there are significant differences in bias across precincts. No, Ron hasn't demonstrated that the bias, and/or the difference in bias (I can't even tell what he means any more), is significantly greater in high Bush precincts.

4) Ron, you assert the existence of a "pattern" of greater red shift in Ohio precincts with high official Bush vote proportion. At least, I think that's what you assert; the absence of a formal hypothesis makes it hard to tell. That pattern is what you haven't tested. Do please read my posts before you undertake to rebut them.

5)

...as I've stated the notion that there should some "arithmetic" link between these ratios only makes sense if you assume SUBSTITUTION between Connally and Moyer voters - that is that voters see these candidates as actually COMPETING AGAINST EACH OTHER rather than just "down ticket" completions of a Kerry or Bush vote.

The mind reels. Connally and Moyer were actually competing against each other. They weren't "just 'down ticket' completions of a Kerry or Bush vote." If Ron wants to argue that voters perceive the judicial candidates as appendages of the presidential candidates, he should at least present some evidence.

Ron, if you will kindly reread my posts on this subject, I thought I had made it rather clear that Connally would be most likely to get more votes than Kerry in Republican counties, and thus it is expected that C/K > 1 where Connally has a below average vote share. Why is this so hard to understand?

[RonB]

104. Slight Correction to Earlier Post

Two out of four M/B>1 counties have (way) above average Moyer vote share.

Basic points remain the same.

[Febble]

109. And to mine

when I pasted your post into Word to write my response, I managed to miss your final lettered points. I see you did check out the four counties I suggested.

OK. The next thing to consider is that the Republican counties are more Republican than the Democratic counties are Democratic, and that Moyer's percentage overall was higher (so the expected value of C/M will tend to be lower everywhere) and will shadow B. This makes B>M>C>K more likely than K>C>M>B. But a final check would be Bush's vote-share in each county relative to 2000.

My point being that under the null hypothesis, where B is high, B>M>C>K will tend to happen, and where K is high, K>C>M>B will tend to happen. Because, in Ohio there are more high B counties than high K, plus M is higher overall, the first will be more likely than the second. What you need to show is that what happened in rural counties goes beyond what is expected under the null. I don't see that you have done that.

edited for clarity

[Joe Chi Minh]

82. For your delectation, Prebble:

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=102x2332523

[Febble]

85. Do get my name right

It's easier than yours.

Yes, I really worry about this kind of stuff. Anything that filters science on the basis of what it finds rather than whether what it finds is true, is deeply worrying.

This is why I find the erosion of academic tenure so worrying as well.

But we can only hold on to our own integrity as best I can.

[mod mom]

95. Arnebeck to OTOH:

Perhaps the best response to OTOH comes from the RFK Jr article itself:

Indeed, the extent of the GOP's effort to rig the vote shocked even the most experienced observers of American elections. ''Ohio was as dirty an election as America has ever seen,'' Lou Harris, the father of modern political polling, told me. ''You look at the turnout and votes in individual precincts, compared to the historic patterns in those counties, and you can tell where the discrepancies are. They stand out like a sore thumb.''

[OnTheOtherHand]

96. that's not a response to OTOH at all

That's a non sequitur.

Baiman is analyzing exit poll data, badly (for reasons that have been discussed at length).

Harris is apparently analyzing election returns, although we don't know how. The statement that discrepancies "stand out like a sore thumb" doesn't do much good without an effect estimate. Does Harris think Kerry won Ohio? If so, by how many votes, and why does he think so?

As far as I know, Arnebeck has done nothing but good work, and I owe it to him not to pretend to be convinced by a Lou Harris quotation in Rolling Stone, or to pretend that anyone I know will be convinced either. Certainly there are some precinct results that make no sense at all (some of the caterpillar crawl precincts in Cuya and the high-turnout precincts in Miami come to mind). That's no answer to my questions above.

[Name removed]

110. Deleted message

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

[mod mom]

111. Voter fraud was NOT the issue-cite your numerous cases of voter fraud.

[Name removed]

112. Deleted message

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