Plausible deniability & Election eTheft

[yowzayowzayowza]

The most commonly feared election etheft methods seem to be tabulator(GEMS) hacks and Clint Curtis style coding. As a programmer of over 20 yrs I gotta say those are very unlikely scenarios. Tabulator hacks could leave traces in page files, remote access logs and telephony records. Nefarious code is risky as it persists far too long. Both methods could be discovered in a recount of the paper trail. Further, neither method utilizes the preferred right-wing modus operandi since Shrub Senior nearly got busted in Iran-Contra: plausible deniability, a tactic which legitimizes stunningly blatant illegal activity. Bush lied us into a war? 'Naw, jus bad intelligence. Oops, sorry.' Judith Miller lied us into a war? 'Naw, jus bad sources. Oops, sorry.' Etc, etc...

Plausible deniability could be injected into DRE systems with preplanned errors introduced into the software that creates the ballot definitions the DREs execute. The switch could be a redundant, missing or superfluous parameter produced within the ballot definition file by a series of seemingly innocuous functions such as an unusual edit sequence. The DREs response to this unusual parameter would be to accidentally-on-purpose behave as described by Pamela de Maigret. This pair of interlocking side-effect errors could easily be activated for a particular precinct or election (thus avoiding local elections and tests). Also, the paper trail will always match the tabulation. Response to the unlikely discovery of this error? You guessed it: 'Oops, sorry.'

IMHO, tho the tabulator & explicit code methods of etheft are certainly possible, a method relying on plausible deniability is far more conceivable. Having mentioned these scenarios a coupla times before, I reiterate in hopes of expanding understanding of the potential for etheft and that explorations of electoral results (actual & exit polls) not unduly focus on exploits bound simply by voting or tabulation system.

(...Make sense, jus more contrarian gibberish er mebbie needs more cow bell?)

[rzemanfl]

1. The only hope we have is to create a spread so wide that any

denial simply won't be plausible. I don't know where that point would be reached, 60/40 possibly, certainly a 70/30 actual vote spread couldn't plausibly be sold as 51/49 the other way.

[mirandaod]

3. I agree.

We need a 20% margin, at least. I think they'll cover their tracks by spreading the cheating around a few different ways, like they did in Ohio. Suppress registration, voter intimidation, limit number of voting machines in Democratic areas, and then, mess with the vote tabulation machines here and there. If the margin in big enough, they won't get away with it.

[rzemanfl]

21. My thought is that all the people who intend to vote Democratic

wear should something blue, preferably a hat, when they go to vote. Then when people are turned away on technicalities or stand in line for hours, there will be no question as to what their intentions were. Also, if most of the people in line are wearing blue it is going to be hard to explain how, after fixing the "humidity problem" or locking the place down because of the "terror alert," the Repuke somehow one.

[Der Blaue Engel]

4. How is it that we've come to this?

I've seen this stated a lot recently, that it's the Dems who must win by huge landslides in order to overcome the fraud. It's a sad day in America when we admit to this kind of defeat. That unless our candidates are so overwhelmingly charismatic and manage to please all of the people everywhere, we have no hope of winning (and further, I've seen many say, we "deserve" this defeat for not having won by such margins). (Not that you're saying this, just noticed it in other posts.)

The only hope we have is to expose and prove this fraud. I don't want to live in a country where one party doesn't even have to win 50% to appear to win, while the other party must always win by a 2/3 majority. We can't take this lying down.

[rzemanfl]

20. Yes, but once the Dems get subpoena power, etc. won't the fraud

then be exposed and set right?

[GuvWurld]

16. There will be no spread wide enough

The spread between the "official" story of 9/11 and what actual evidence exists is wider than any spread we should ever hope to depend upon to upend government myths and propaganda. Frankly, the same is true for what is known about the 2004 "election." There will be no better or more evidence or proof.

We have a problem drawing a line in the sand. First, we don't even think about doing it soon enough. Second, when we try to do it, we have a lousy track record of making it stick. How else can we account for 300+ anti-Patriot Act resolutions collectively failing to restore our rights? When does no mean NO, and enough, ENOUGH? If we can't say this clearly or effectively enough, we should know to expect subjugation at the hands of the brutal fascist leaders who occupy government buildings in Washington, DC.

Bottom line: plausible deniability may have been important at one time, but now the power and control has been so consolidated that plausibility itself no longer matters. Who could have imagined a levee breaking or a staged video conference with military troops and their commander-in-chief? They now blow past plausibility with impunity.

If ever there was a time when peaceful revolution was justified, how would we know? I submit we are far past it and our thinking and planning and organizing needs to revolve around this ultimate goal. Pursuit of election reform is a tactic (that I recommend and support) toward this end but cannot, in itself, account for the creation of freedom, liberty and Democracy. Read Blueprint For Peaceful Revolution.

[Blue Shark]

44. I would have submitted to you...

...the unlikelyhood of a Kerry up by 4% to a Bush up by 3% Seven point swing as just such a margin on implausibility. The MSM didn't bat an eye.

[Land Shark]

46. bingo yowzayowzayowza; E-fraud or Glitch Switch; the media will never know

Media's "gotta" give equal time to insanity, especially if it comes from Republicans. So, if there's plausible deniability the story never gets legs. Right thinking people trained not to buy it.

This is why there is "visual evidence" of vote switching -- it looks like glitches. One has to get into it fairly deep to see it is not just a glitch. Media can't get to that level of complexity, and even if they do, 290 million americans won't

[Stevepol]

2. Until somebody in on the scam comes forward it's all speculation:

what's not much in doubt in my mind is that the election was stolen. The exit polls indicate clearly, and Mitofsky himself admits it, that either the exit polls are at fault (and all over the country) or the vote counting is wrong (all over the country).

So these are your choices. There ain't no more:

( ) exit polls

( ) vote counting

Which do you think is at fault?

Or can somebody think of another option? Divine intervention? inter-dimensional mis-programming by means of a crack in time?

[Febble]

5. well, since you ask

the exit polls:

http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=397074&mesg_id=397194

But I think yowzayowzayowza is absolutely right. Plus, there is far more evidence for "glitches" of random and various sorts than there is for the kind of huge clean hack that would explain the exit polls.

But you don't have to explain the exit polls. People are so busy paying attention to the "elephant in the room" that the incriminating needles in the haystack are in danger of being overlooked.

[foo_bar]

11. why can't they both be wrong?

[yowzayowzayowza]

12. Careful!

Gettin all logical is gunna start some trouble. :evilgrin:

[Febble]

19. Actually, I think we are at the stage

of being able to exclude SOME middles.

From this:

http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=397074&mesg_id=397194

we now know that the exit poll redshift was not correlated with swing to Bush. This rules out one important scenario:

massive vote-switching using a software hack made possible since 2000

I think it rules out a few others too, but it certainly leaves several fraud options intact, including various forms of electronic sabotage, especially targetted in key places (like strongly Democratic precincts in key states).

Which is why I think you are looking in absolutely the right place - i.e. in clear sight.

[Bill Bored]

27. Wait a minute, what's the difference between

"massive vote-switching using a software hack made possible since 2000"

and

"various forms of electronic sabotage, especially targeted in key places (like strongly Democratic precincts in key states)"

Any electronic sabotage sufficient to change the outcome of an election is pretty "MASSIVE" in my book! And if it's electronic, it would not have been as possible in 2000 because the platforms on which it could be carried out were not nearly as prevalent.

[Febble]

28. Read my lips

vote-SWITCHING.

It's vote-switching that ought to be detectable in that correlation, if it was new since 2000.

Vote destruction in Dem precinct, that even-handedly destroyed both Dem and Rep votes would also push the election Bush's way. As would even-handed vote-multiplication, though I guess that would be easier to detect. As TfC pointed out, both these forms of fraud would preserve the within-precinct vote proportions and therefore not show up in the exit polls.

In effect I'm talking about electronic vote spoilage.

[yowzayowzayowza]

29. sElection 2k eTheft switch oops ...

in Gadsden County, Florida:

In November 2000, Shirley Green Knight, Hutchinson’s deputy, a soft-spoken African-American, had recently defeated him for the office of elections supervisor, though she had yet to assume the office. After the votes had been tallied, she noticed something strange: more than 2,000 ballots, out of 14,727 cast, had not been included in the registered count.

How had this happened? Because of a very technical but profoundly important detail. The central optiscan machine used in Gadsden had a sorting switch which when put in the “on” position would cause the machine to record overvotes or undervotes in a separate category for possible review. After the election, Knight says, she learned that Hutchinson had demanded that the switch be kept off. “I have no idea why he would do that,” says Knight. Seeing how many ballots never got counted, she urged him to run them through the machine again—this time with the sorting switch on—but he resisted. Hutchinson was finally overruled by the Gadsden canvassing board. They looked at the rejected ballots. Sure enough, they were overvotes—and for good reason.

Gadsden had used a variant of the caterpillar ballot,...

On edit: I was gunna include this in the OP as an example.

[Bill Bored]

31. Yes but this was a 14% undervote rate!

No scatter plots needed to find this.

And it was a central count op scan -- not precinct count, although with Diebold's junk, those ballot reject settings are turned OFF by default on the precinct count scanners.

[yowzayowzayowza]

37. Man o man.

Ya can't post anything round here w/o eliciting the Exit Poll Warz!!! Skattered here & pied there. ;)

The post was in response to Febble comment:

Vote destruction in Dem precinct, that even-handedly destroyed both Dem and Rep votes would also push the election Bush's way. ...talking about electronic vote spoilage.

I should have added a description comparing this electronic switch to the software switch described in the OP. Sorry.

[Bill Bored]

30. Let me get this straight...

...you're saying that just for the sake of avoiding detection by (smirk) an EXIT POLL, not even an EXIT POLL, but some esoteric SCATTER PLOTS produced by some geeks to try to explain an exit poll discrepancy that didn't even exist at the time the hacks were perpetrated, they would delete both Bush and Kerry votes in the same proportions as cast in Dem precincts? Good grief! That has to be the most inefficient election theft methodology imaginable!

And how exactly would they hide the excessively high undervote rate generated by deleting enough Dem AND Repub votes in these precincts to swing an election? They would stick out like a sore thumb, would they not? I don't think you'll find enough undervotes to even give Kerry the win without some vote switching in Ohio. If it's that simple, why didn't they just COUNT THE BLOODY UNDERVOTES? Unlike FL, Ohio's "hanging chad" laws are very clear; yet Kerry conceded because they knew the undervotes weren't enough to change the outcome. Give them a LITTLE CREDIT, will ya? They may be dumb, but they're not THAT DUMB!

I have to sleep on this one but it's just so inefficient and so easily detectable. It's not elegant; it's crude. I can't see it. I think you are assigning far too much importance to these scatter plots.

All they had to do was switch punch cards around in the same precincts you're talking about and they'd get at least twice as many votes in Shrub's favor than by just flushing ballots down the loo as you seem to be suggesting. And there would be far fewer undervotes and far less plumbing problems too! And if they did the same thing in 2000, the scatter plots would look fine, wouldn't they?

[Febble]

36. Well, I have to say

I don't think it's very likely either.

It's just one way I can think of (well TfC thought of it) where the fraud could be hidden from the exit poll). But as you say, it could be detected by some other method. Unless plausibly deniable.

Honestly, Bill, I can't see how any election theft method on a massive scale could be undetectable by some method that's been tried. People can disagree about what the scatterplots mean, but I CANNOT see how any massive vote-switching fraud scam wouldn't show up. And as you say, the idea that whoever planned it anticipated scatterplots is a bit far-fetched.

[Bill Bored]

38. OK, let's step back for a moment, shall we?

The assumption is that Red Shift has to be correlated with Bush doing better in '04 than he did in '00 for there to be fraud. But truthfully, all this would mean is that there was fraud detectable by the exit polls. In other words, it's a way to tell if the exit polls are indicating fraud -- not necessarily whether there was actual fraud.

ESI's headline reads, "Exit Polls Not A Smoking Gun For Fraud" or some such thing, right? And based on their somewhat convoluted data, this is a fair statement (although there were 2 precincts that may still be questionable).

Now somehow, this seems to be getting morphed into "Ain't No Fraud."

Why?

I wasn't necessarily convinced that the exit polls could prove fraud in the first place; nor do I think they can rule it out beyond reasonable doubt either. So I think I'm being consistent, albeit a bit stubborn.

But there seems to be a subtle shift in the discussion from testing whether the exit polls can prove fraud, to actually using them to prove there wasn't fraud. And I'm not sure exactly how or when this occurred, are you?

[Febble]

39. Well, sort of

and I claim some responsibility. Because I do think the ESI finding, and now this one, puts limits on the kind and/or extent of the fraud that is plausible.

I'm being very circumspect about this Bill, because I certainly don't think the exit polls rule out fraud. But I am currently persuaded they rule out massive, vote-switching fraud. I'm willing to be convinced otherwise. And I can accept it wasn't massive - targetted fraud. Or that it wasn't vote switching.

But if it only occurred in the unpolled precincts, the precinct selection should had a redshift - and it didn't, it actually had a slight blue shift. And if it was in the polled precincts, the WPE should have had a redshift - which it did - but the that redshift should correlate with swing - which it doesn't.

So yes, I'm suggesting the exit polls suggest not that there wasn't fraud, but that there wasn't massive, vote-switching fraud. If you can persuade me that there could have been massive vote-switching that could also have been invisible to to both the precinct selection finding and to that scatterplot, then I'll be happy to eat my hat on DU.

Targetted fraud I will buy. Voter suppression I will buy. Vote suppression I will buy. These things on a scale to swing the electoral college vote I might buy if the price is right. But I cannot see how massive vote-switching is consistent with what we now know about the exit polls.

[Bill Bored]

40. But what are we trying to prove?

First of all, I still don't know exactly how you define "massive." It would help to define that. If you mean there's no widespread correlation that indicates fraud, that's obvious. But has anyone even suggested that there should be? Widespread fraud is the easiest to detect by any means, so of course, fraud would have to be targeted.

So I think the whole argument that there must be some magical correlation is a strawman.

That said, there were a number of anomalous precincts in the ESI OH report which no one is talking about because you keep looking for non-existent correlations and I'd imagine the same is true of the national poll.

<http://www.votewatch.us/Members/StevenHertzberg/report.2005-05-14.4978140903/report_contents_file/>

For example, in page 4 there are 2 outliers; precincts whose poll discrepancy cannot be explained by rBr because according to ESI's own poorly presented data, it's mathematically impossible. That's 2 out of 50, or 4% of the precincts, and an UNKNOWN percentage of the vote count by the way, because the precincts are not weighted by vote count. I know there are other poll biases besides rBr but there is also the possibility of fraud in these precincts. But these 2 outliers are simply glossed over.

On page 8, where there is no correlation between (BushVote04-BushVote2K) vs. (BushVote04-BushPoll04), there are still 4 precincts where the Red Shift did in fact coincide with Bush doing better than he did in 2000. That's 4 out of 50 or 8% of the precincts, and once again, an UNKNOWN percentage of the vote count.

Now why am I focusing on these anomalies and not the ones that might seem to favor Kerry? Simple: Kerry DIDN'T WIN THE ELECTION! He owes us no explanation at this point.

Now as to the point about the precinct selection being representative of the whole, that's fine because the above indicates that it's still possible for there to have been fraud in both the polled precincts and the unpolled ones. So no conflict there either.

Unless you can prove that the anomalies stated above were not targeted fraud, I don't see the point in continuing to argue that there was no "massive" fraud. But again I don't know what your definition of massive is.

The fraud I care about is ANY fraud sufficient to change the outcome of the election.

[Febble]

41. OK

The fraud I care about is any fraud. Elections shouldn't be fraudulent. And any form of disenfranchisement is a disgrace to a democracy and an abuse of human rights.

Fraud that changed the outcome of the election is on one sense more serious, except that there is no chance of changing it. If it was not only unjust but illegal I want to see investigations and prosecutions whether or not the election was changed.

By "massive fraud" I mean fraud on a scale to win the popular vote, often indexed by the size of the exit poll disrepancy (and beyond). I do not think the exit poll discrepancy is an index of the magnitude of fraud, and I do not see any other evidence that popular-vote-winning-fraud took place. Indeed I think the exit poll evidence contra-indicates it.

I have addressed your question about the "impossible" discrepancies elsewhere - don't know if you've seen them. The point is that if the polling bias was at the level of voter selection the precincts arent' "impossible". And there is evidence that there WAS polling bias at voter selection level.

[Bill Bored]

43. I didn't say they were impossible did I?

I said they were impossible to explain with the reluctant responder hypothesis alone.

Since rBr seems to be the official story of why the polls were off, and there are 2 precincts out of 50 in one state alone where it was impossible for rBr to fully account for the discrepancy (thank you ESI), I think Mitofsky needs to come up with some other "official" explanation(s). If they were presented in the debate with Freeman, I haven't seen them yet, but it is NOT responsible of him to suggest that there is reason for confidence in the election system when we know very well that there isn't.

Mitofsky should accept that the system could have been hacked. rigged and exploited and use his position to work to improve it instead of just defending the status quo and the fallibility of his own useless exit polls.

Fritz Scheuren is another one. There were 4 precincts out of 50 in the Ohio exit poll with Red Shifts where Bush did in fact do better in 2004 than in 2000 (thanks again ESI!). That suggests the possibility of fraud in 8% of the precincts polled. Now is that 8% in addition to the 4% in which rBr could not explain the entire discrepancy, or is the 4% a subset of the 8%, or what?

Until this is explained, fraud is always a possibility, unless the integrity of the election system itself can rule it out, and believe me, it doesn't.

How many like this in the national poll then?

Sorry, but when it comes to this stuff, I'm losing my patience. These guys are either with us or against us.

[Febble]

48. OK Bill

I'm losing patience myself, but not because of you.

Look, I'm not interested in what Scheuren of Mitofsky say. I'm interested in what they show (and it would help in Scheuren's case if his plots were properly explained and labelled.)

I don't like the term "rBr" which is at least as loaded as "WPE", and I don't even like "non-response bias" because it implies that form of non-response that causes bias from refusals - people can also fail to respond to a poll by not being selected in the first place when they should have been had the selection process been properly random.

By range of possible response rates, ESI appear to mean the range of possible votes from all those who refused or missed (i.e. selected but not approached). All could have been Kerry voters, all could have been Bush voters. However, if a greater proportion of refusers/missees were Bush voters than the proportion who responded, then you'd get "rBr". And clearly the the maximum possible percentage of refusers/missees for anyone candidate is 100%.

However these are misleading limits to the possible. If more Kerry voters than Bush voters were selected for interview (because of biased sampling), then that would extend the possible range of possible response rates. And there is good evidence that at least some of the "non-response bias" which I prefer to call sampling bias, came from non-random selection, rather than from differential response rates. Mitofsky has drawn repeated attention to this.

To take your second point. Yes it is possible that in the precincts where both red shift and swing to bush were greater than average the common factor was fraud. But then you have to explain why this effect was counterbalanced by precincts in which swing to Bush was less than average and the PLD was also red shifted. And all combinations of the above. Whatever the effect was, it appears to have been counterbalanced by equal and opposite effects (Kerry fraud, anyone?) However, the ESI was not very high powered so it is possible that there was "true" shared variance between redshift and swing that was swamped by variance from other sources.

Which is why the same (essentially) analysis on the whole 1250 dataset is important. You cannot "prove" a null. A flat regression line does not show that there was no shared variance - no relationship. But it does allow you to set confidence limits on the size of that relationship. And it is very very small.

Honestly Bill, I'd love it to be bigger.

You are right. The exit polls are not an index of fraud. However this finding does put upper limits on the net effect of vote-switching fraud. I think. Eomer has opened one loophole, which you might like to check out (an algorithm that prevented Bush's vote proportion dropping below 2000 level) but I am not optimistic. I'd like to know if you think it could be done, but not convinced the predictions it makes would be supported. TfC suggested another - but you don't like that one!

And the last thing I want to do is dispel the notion that evidence of fraud needs investigating (it does), that vulnerability to fraud needs fixing (it does) or that electoral injustice needs investigating AND fixing (it certainly does).

I really don't want to fight, least of all with you.

[kiwi_expat]

42. Regarding outliers: anecdotal evidence might be useful.

"Sampling bias" can seem an abstract concept unless you have seen it in action.

Liam_laddie has reported that he interviewed a couple of Democratic Party Poll Observers from Cincinnati 4M. (Of the Ohio NEP polls we have been able to identify, Cincy4m has the biggest WPE. But it does not have the biggest WPE of all the Ohio NEP polls.)

According to the observers, the NEP interviewer at Cincy4m was totally out of his depth. He managed to conduct 31 interviews out of the 1754 voters (times 2, because they went by him twice) fanning in/out from the polling place (the polling place held 4 precincts). It was raining and he was stationed 100 feet away from the door, in a group of campaigners who were vying with each other for the attention of voters entering the polling place.

The observers said that "Republican types" just streamed by him. They didn't refuse to be interviewed. They didn't even notice him.

Maybe the guy was able to note the age, gender and ethnicity of every Nth person in the fanned mob, but I doubt it very much.

He was totally discouraged, took several long breaks and went home early.

[kiwi_expat]

45. The big WPE for Cincy4M was probably not the result of rBr, per se.

The interviewer might not have even had a high refusal rate. In fact, the worry is that he might have had a LOW refusal rate because of his selection method.

His problem was he had trouble ASKING people to participate. People didn't even notice him. It would have taken a very strong person to resist interviewing those who did. (It is a wonder that Cincy4M didn't have a bigger WPE than the -28% it had.)

As for whether the resulting WPE was *biased*, that depends upon what caused some people to notice him. He was standing in a group of mostly pro-Democrat campaigners. Perhaps Democrats/independents were more likely to give the group a look.


It is important to emphasize that the 100-foot limit (which placed the NEP interviewers with the campaigners) affected most NEP precincts in Ohio, for at least part of the day. According to the E-M Report, the last NEP precinct to lift the restriction did so at 5:00pm.

[OnTheOtherHand]

85. consensual correction

kiwi and I have conferred and agree that Cincinnati 4M was 68% Kerry in the exit poll, 40% Kerry in the count, for a WPE of approximately -56 (subject to slight rounding error) -- not -28, because WPE is calculated on the margin, not the vote share. For people marking their ESI scatterplot scorecards at home, that is the point at lower left on ESI Figure 3 (where the scale is vote share, not WPE).

[kiwi_expat]

47. I'm afraid I used the term "sampling bias" incorrectly.

No matter how non-random the sampling (and how big the WPE), I'm afraid an individual precinct's results can not be said to have "sampling bias". That is a term reserved for net sampling results. I think.

Pro-Bush-voter sampling could offset pro-Kerry-voter sampling resulting in a zero "sampling bias", if I've got that right.

[OnTheOtherHand]

49. I think it could be right either way, depending...

...on the population(s) about which one is trying to generalize.

It's perfectly reasonable to treat the voters at one precinct as a population, which the interviewer is attempting to sample without bias by adhering to a strict protocol intended to give every voter an equal probability of being included. Which is not exactly what happened in that Cincinnati precinct! Sampling bias is an appropriate term and a good conjecture given the facts you've described. (Note that we can't directly measure sampling bias, at any level. Even in the absence of fraud, interviewer error, and other "interesting" error sources, random sampling error could make the WPE larger or smaller. I know you know this.)

It's also reasonable to consider the state electorate as the population of interest, and to argue that while there may have been sampling biases at the precinct level, they may well have cancelled out so that there was no net sampling bias at the state level. (Or they may well not have.)

And one can even take this up to the national level, although the exit polls aren't designed to predict the popular vote. Sampling bias in various states might "cancel out" -- although that might be cold comfort to the candidate who learns that he got 200,000 votes too few in a swing state but 200,000 more in a safe one. The exit pollster wouldn't be very happy either.

[BeFree]

50. Really?

"....the exit polls aren't designed to predict the popular vote."

That's news to me. I thought the whole idea of exit polls was so that the companies who paid E/M $10 million for the exit polls could make a sound prediction of the popular vote - state by state.

And all this twisting about 'sampling bias' is just spin. The system is designed to keep bias to an absolute minimum. For yall to there was a great deal of sampling bias, or rbr, or whatever, is to say that E/M was full of bull. In that regard we then agree, the final exit-polls were BS.

Looking however, at the general census and comparing it to the three early sets of numbers from the exit polls, one sees the numbers match up real well, so there is congruence there. A congruence that disappears as the final BS polls came out.

Hey, when is all the raw data gonna be seen by us little folks? What are they hiding?

[Febble]

51. Do you think we might

keep this conversation civil? It is very difficult to concentrate when issues are couched in such aggressive terms.

I'll leave OTOH to address the first point, except to say that the polls are designed to predict each state, from which the electoral vote can be predicted.

Your statement about spin is simple assertion. You are right when you say the system is designed to minimize bias, given whatever cost restraints there are - but sampling bias is very difficult to exclude, especially if you allocate only one briefly trained interviewer to a crowded precinct, and ask them to stand in the rain and approach every 10th hassled voter. See Kiwi's post. You can disbelieve it if you want, but there is no intrinsic reason to disbelieve it.

Next point: specify which numbers match up "real well" to what.

Final point. You can see the raw data. What you won't see the raw data attached to anything that would allow you to identify the respondents. You won't get to see my medical records either. It's called confidentiality.

[Name removed]

52. Deleted message

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

[Febble]

53. The "early" data

isn't raw. And better versions of it are available in the E-M report, published January. Far from wishing it could have been "kept" from us, Mitofsky has been drawing repeated attention to it.

And you are right, the discrepancy was much larger in 2004 than in the previous 4 presidential elections. No-one to my knowledge is attempting to "spin" anything different. Certainly not Mitofsky. The entire E-M report was an attempt to evaluate why it was so large. You may not agree with its conclusions but no-one is saying there wasn't a discrepancy.

"Trying to make the final releases congruent" is what the exit polls are designed to do. They are designed to predict the result, not audit it. If you google Edison Mitofsky, and go to FAQ you will find out how the projections are made. This information was available well before the election. The projections include several sources of data, including vote-counts as they become available.

And yes, the electoral college is decided on the popular vote in each state. However, the term "popular vote" is usually taken to mean the nationwide vote. As in Gore won the popular vote in 2000 though he failed to win the electoral vote (even though he would have won it in a just system).

And don't accuse me of not telling the truth. It's against the rules.

It also happens to be untrue.

[BeFree]

54. Accuse?

No, I just offered advice to everyone, myself included.

Again, I ask myself: "What are they trying to prove?" The only answer I can come up with is this: I dunno. Nothing they have said has proven anything that refutes the common sense knowledge that E/M are trying to cover their asses.

As far as not 'auditing' the popular vote, show me something to that effect. If it's not in the field of auditing then it is of no use whatsoever, and then the $10 million was all a waste? A needless and frivouluos exercise? Tell that to those who spent the 10mill.

[Febble]

55. Well, I try to tell the truth as I see it

I'm sure you do too.

Not sure where you got the 10mill from,but E-M's clients were the National Election Pool i.e. the tv networks. They wanted projections of the counted results, as they always do, so that they can "call" the states before all the votes are counted. As a Brit I find that incomprehensible, but it does seem to take longer for your machines to count votes than our bank tellers.

If you want an exit poll as an audit, it will cost you more than 10mill. You'll need at least that many precincts, and at least two interviewers working in shifts. And it will still be a lousy audit, because you will still not be able to rule out non-response bias. Better to insist in transparent counting methods with paper ballots and mandatory random audits.

[BeFree]

64. I couldn't agree more, Febble

That's why we are insisting on transparent counting methods with paper ballots.

Lacking that, we have to go with what we've got. What we've got is a measly $10 mil of an exit-polling system, so we try to make the best of it.

What we got was nice raw data at first. Raw data that TIA, et al, had a go at, and came up with some interesting conclusions. After that, we got weighted data that contradicted the raw data and with that contradiction came the explanations trying to explain away those contradictions.

Ergo, this board, and these discussions. Some here are working from the raw data, others, such as yourself, are proxies for the weighted data. It is the majority's belief that the weighted data is incorrect - a lie. Unfortunately the proxies of that weighted data are painted with that broad stroke of the contradictions. There seems to be no way around this mess and some will be besmirched by their associations

We are in this mess through no making of our own, yet finding our way through the morass we've been presented is going to leave some of us with any of a various combination of paint spots. Democracy is messy at times, but I think we can all agree Democracy is worthy of the struggle.

[Febble]

67. Well, see my post below

As I say there, the data TIA has been working with is not "raw" - or not most of it. When he uses the WPE data, that is "rawer" but still not raw.

I am not a "proxy" for "the weighted data" (what a bizarre thought). I'm not very interested in the weighted data, for the same reason none of us are. What I'm interested in is the discrepancy between the raw data (not the early exit data, but the raw responses) and the count.

And I couldn't agree more that Democracy (and democracy) are worth the struggle.

Glad we seem to be getting to the same page!

[OnTheOtherHand]

58. I'm sure we appreciate the advice

which raises a question about something you said upthread. You wrote, "For yall to (say?) there was a great deal of sampling bias, or rbr, or whatever, is to say that E/M was full of bull." But, as I think you know, it's E/M that says they believe there was a great deal of sampling bias, despite their efforts to minimize it. So, umm, aren't you the one who is saying that E/M is full of bull? Is it really "common sense knowledge" that E/M is lying? and if they are lying, in what sense are they "trying to cover their asses"? and, dare I ask, can you cite any evidence for your interpretation?

(I don't entirely agree with Febble about the last five presidential elections. There were at least three exit polls in 1988, and one of them apparently -- at least in the truly "raw" returns -- showed Dukakis slightly ahead of Bush, which would be a substantially larger error than in the 2004 exit poll. TIA has reported this, but for an external source, there is
http://www.emergingdemocraticmajorityweblog.com/donkeyrising/archives/000940.php
However, as Febble mentioned, the results that were released on cnn.com as polls closed -- not "leaked" -- weren't raw returns; they were composite estimates, weighted in various ways, and also incorporating prior estimates from pre-election polls. So even if we had all the raw data from the 1988 exit polls, we still wouldn't have results comparable to the ones on cnn.com. I don't know whether that "President Dukakis" poll was actually less accurate, or the raw data are particularly misleading.)

"As far as not 'auditing' the popular vote, show me something to that effect." Well, here are some things that E/M says about their own product: "Exit Polls Provide Rich Content for News Outlets. Voter information provided for states, regions and the nation. Poll responses deliver rich details of who voted and why." (http://exit-poll.net) And, "Projections are based on models that use votes from three (3) different sources -- exit poll interviews with voters, vote returns as reported by election officials from the sample precincts, and tabulations of votes by county." (http://exit-poll.net/faq.html) In other words, E/M combines official vote totals with exit poll results to make its projections, and it uses the exit poll results to produce tables that the media subscribers can use to explain voting patterns. I don't see even a hint here that they are trying to evaluate the accuracy of the official vote returns; on the contrary, they rely on official vote returns in their projections. So, I would put the question back to you: do you have an iota of evidence that the media subscribers paid $10 million to try to audit the election results? Did some network anchor say that when I wasn't listening?

[BeFree]

62. You're welcome

"So, umm, aren't you the one who is saying that E/M is full of bull?"

Yeah, ya got that, eh? E/M is full of bull. The early exit polls (read raw) had no E/M bull applied to them. The last poll reeked of E/M. The first numbers were raw - without the E/M. So now they are covering their asses because the weight they applied to the data stinks. Their wiggling reminds me of a child being potty trained.

"..the results that were released on cnn.com as polls closed.."

See you are not one the same page as me. I am talking about the data that was released before the polls closed and the raw data that had no weighting applied, and you are talking about the data that was

"... weighted in various ways,..."

Using the term 'auditing' to describe the exit polling is a bit of a stretch. But really, why pay $10mil for something that if they wait till midnight can be had for free? I remember the previous pollsters being put out of business because they had numbers that didn't match the diebold numbers in 2002. E/M learned that lesson and figured for $10mil they'd best try to meet the diebold numbers. Unfortunately for them, raw data got leaked before they had a chance to weight it down. Unfortunate for them, but good for democracy.

Too, can you think of any other audit of the vote besides the exit-polls? If not, that means we have no audit.... any wise man would say we need some kind of audit and for years we kinda relied on the exit-polls to be that 'some kind of audit'. Now you say there was no audit whatsoever?

[Febble]

63. No, there was no audit

Look BeFree, I think you do have some slightly strange information about the exit polls, what they are for, and how they work.

The "early exit polls" were not "raw data" that then had "E/M bull applied". The inputs into the "early exit polls" was data collected from various sources, all by by Edison Mitofsky, one of the most important sources being, indeed, the "raw data" collected, by E-M at the precinct exits. This was then manipulated in various ways, including weighting by age sex and gender from the non-response demographics, and I believe pre-election polling also went into the mix. So it certainly wasn't "raw", even then. However what normally happens next is that once the vote-returns start coming in, they too are added to the mix. It's a continuous dynamic process, and because it is not ever designed as an audit, but to project the results, E-M use the "best" data at available at any given time. Once the vote-count starts, that data-stream also comes on tap.

You may not like this procedure, and indeed if the exit poll was designed as an audit it would be completely inappropriate. But the networks commission the polls so they can project the results, and so this is how it is done. And it does project the results, very accurately. It's just that, sadly, we have no basis for confidence in the results.

I agree with you that the early projections - and better still the actual response data - is worth looking at because it tells us how large the discrepancy between the raw data and the final projections were. But the idea that there was some utterly pristine data "leaked" early, then hastily covered by a blanket in embarassment later is just not they way it was. And far from having to rely on the CNN "leak", the E/M report gives us a great deal of detail about of exactly where the discrepancy lay - from this we know that it was at precinct level.

And as for paying 10mill for something that could be had for free after midnight - quite. But tv is always in a hurry for a scoop. And the entire vote-count can take a while, as we discovered in 2000.

There is no "audit" for your elections. There desperately needs to be, and exit polls, even if designed for the job, would be a very ineffective method. Random handcounts are the way to go IMO. In lieu of this, the currently designed exit polls can be reverse-engineered to a certain extent to try to shed light on whether or not there was fraud. But it's a terrible substitute.

[OnTheOtherHand]

65. your facts are still confused, as far as I can tell

although since you rarely define terms or cite sources, it is hard to know.

Even in the case of the results that were leaked before the polls closed -- say, on Drudge -- we have no reason to assume that those results were "raw." What we can say is that they weren't weighted to the official vote counts.

Evidently the 12:22am national tabulation was not weighted, or mostly not weighted, to official vote counts either. (They were weighted to other things -- they were not raw.) Obviously these are numbers from after the polls closed, and they still showed Kerry ahead. On election night, E/M's first priority is to make state calls, because the national popular vote percentages don't determine the election outcome. But those 12:22am national numbers were in line with the state numbers, which had been posted on cnn.com as the polls closed.

If there is some reason why you intend to focus on results leaked before the polls closed, as opposed to released or leaked after, maybe you could say what it is.

"Using the term 'auditing' to describe the exit polling is a bit of a stretch."

OK, we agree.

"But really, why pay $10mil for something that if they wait till midnight can be had for free?"

Probably the biggest answer is in my preceding post: "Poll responses deliver rich details of who voted and why." That information is not available for free. The exit polls in conjunction with official returns may let the networks "call" states a little sooner than they would otherwise, but that probably wouldn't justify the expense. It's mostly the ability to, say, natter about the role of "moral values" or Bush's inroads among Hispanics that the subscribers are paying for. And even if the vote totals are off by 3%, the crosstab numbers being hawked on TV (and subsequently written about in the papers) "should" be fairly accurate.

"I remember the previous pollsters being put out of business because they had numbers that didn't match the diebold numbers in 2002."

Source, please.

"E/M learned that lesson and figured for $10mil they'd best try to meet the diebold numbers."

No, the exit polls are weighted to official returns after every election. Many of the datasets are archived so one can see this for oneself. It's also part of the standard operating procedures announced in advance, as Febble pointed out -- no secrecy involved.

"Unfortunately for them, raw data got leaked before they had a chance to weight it down."

Actually, weighted data that showed Kerry ahead were released -- not leaked -- on cnn.com as the polls closed. You can say that you're not talking about that, but it is nonetheless true.

"Too, can you think of any other audit of the vote besides the exit-polls?"

No, there isn't. That's one reason why we're all here talking about election reform, if I recall correctly.

"any wise man would say we need some kind of audit..."

yes, and the wise women too...

"and for years we kinda relied on the exit-polls to be that 'some kind of audit'."

We did? Source, please.

"Now you say there was no audit whatsoever?"

Yup, pretty much. Of course there are some recount provisions, but we all know why those are inadequate.

[BeFree]

66. OTOH

I'm glad you are here. Yes, you read that right. It helps to have someone question the common knowledge that most of us are quite aware of, and your questioning does tend to reaffirm that knowledge base.

It is curious that you think the raw data that was provided before the final exit-polls were manufactured - was not raw, but weighted. I had never heard that before.

We all agree that the vote could have been stolen. Yes? It is our finding that the raw exit-polls show that it WAS. Of course, E/M doesn't want their data to be used in such a way. Too bad.

Ya know, the people who paid $10 mil must be quite pissed that they wasted their money. I guess we won't see anymore exit polling done if it isn't worth the money. I guess E/M are a bit worried they will be out of work. In that regard I can hardly blame them for spinning their inherent contradictions between the data that was leaked and their final presentation. That's business, eh?

[Febble]

68. The people who commissioned the exit poll

(I have no idea what they paid) are the TV networks. They did, in fact, get what they paid for - a projection of the counted result that enabled them to call each state, and, as OTOH, some demographic stuff to natter about while they were waiting.

What the rest of the country DIDN'T get for the networks' money was a free audit of the election.

As you said upthread, exit polls really are rather frivolous things. They can double as a smoke detectors in the absence of a decent sprinkler system, but that's about it.

[OnTheOtherHand]

69. hmm...

Since you disregard much of the post to which you are purportedly responding, it is hard to know what to think or to say. At least we have two important points of agreement: that the election results aren't audited, and that they should be.

"It is curious that you think the raw data that was provided before the final exit-polls were manufactured - was not raw, but weighted." Again, your terminology is very fuzzy. It would help if you responded to the actual text of my actual posts, and indicated whether you agreed, or even comprehended the words. Let me try once again. Your distinction between "raw" and "weighted" is incorrect. As far as we know, E/M doesn't release "raw" data until later; the datasets available via ICPSR actually incorporate "raw" information from over 70,000 questionnaires. But the estimates that E/M is preparing during election day, both before and after polls closed, incorporate all sorts of weights -- demographic, geographic, folding in absentees, taking account of pre-election polls, and ultimately incorporating official returns. So, the salient difference isn't between "raw" and "weighted," but rather between weighted to official returns or not. I can't think why it would be useful to insist that weighted results are "raw."

"It is our finding that the raw exit-polls show that it WAS" (i.e., the vote count was stolen). Well, I don't think even Freeman goes as far as "show that it WAS." You can "find" that, if you want. Most survey research experts "find" something else. Folks should know that if, say, they go to reporters saying "The exit polls show that the election was stolen," the reporters will have a hard time finding credible experts who agree. In contrast, if folks say "Paperless e-voting is inherently insecure," that view has more expert support. That is pertinent no matter what you or I think of the exit polls.

"Ya know, the people who paid $10 mil must be quite pissed that they wasted their money." This statement implies that you may not have read my previous post; maybe you can sort things out from Febble's response.

[BeFree]

70. that's pretty Raw

Here's your quote:
"Your distinction between "raw" and "weighted" is incorrect.

Raw is raw. Meaning before cooked. Meaning before slathered with sauce; before marination.

As far as we know, E/M doesn't release "raw" data until later;

That's right. Good for you. That's why the leaked data is called LEAKED. E/M didn't release the data, because, as you say, they don't release it til later. It was LEAKED. LEAKED RAW data.

...."raw" information from over 70,000 questionnaires...

Yes. Raw information.

So, the salient difference isn't between "raw" and "weighted," but rather between weighted to official returns or not.

There is alot of difference between raw and weighted. And yes, the official E/M report was weighted... it was not raw. The early LEAKED data was raw, ie. it was not weighted. Sure, it had differing sets such as gender, religion, income, but it was uncooked... it was raw. It was not weighted to official returns.

Question: Has there ever been such a controversy over exit-polls before now? Not that I can reall. Is it becuse this was the first time the raw data got out because of the internet, and got out well before the weighted, cooked data was officially dished up? Or is it that this time the data was overcooked making it taste bad?

What a waste of $10 mil. I'll bet E&M are sweating. They'll never work in this town again! LOL

[OnTheOtherHand]

71. nope, sorry, still wrong

The raw data was released (not leaked) in January, dude. January.

What was leaked on election day before the polls closed was almost certainly weighted. What was released on election day after the polls closed, but before official returns were available, was definitely weighted -- just not to official returns.

"Sure, it had differing sets such as gender, religion, income, but it was uncooked... it was raw. It was not weighted to official returns."

Well, as Senator Moynihan said, everyone is entitled to his own opinion, but not to his own facts -- or, perhaps he should have added, to his own language. The results released after the polls closed, but before official returns were available, were weighted to gender (among other things), and were not weighted to religion or income.

You may want the word "weighted" to refer only to official returns, but it doesn't.

No, I don't recall such a controversy over exit polls either. Of course, as far as most of my colleagues are concerned, there is no controversy over the exit polls. I'm fairly unusual, as far as I can tell, in attempting to take the arguments on both sides seriously.

I think E/M will do OK; as they say, they didn't call a single state wrong, and neither did any network using their data. Not that I imagine they were having a whole lot of fun that night.

[BeFree]

73. Almost certainly weighted?

Are you saying the 'leaked on the internet shortly after midnight data' was weighted? You can't be serious. It was raw. Raw means un-weighted. At least you admit it was "almost certainly." Are you also admitting you don't know?

Of course, if most of your colleagues are so certain that there is no controversy over the exit polls then you are hanging with the wrong crowd. LOL

I think E/M are going down, way down. Maybe not down and out, but controversy is bad for bizz.

Eh.. won't you tell Febble how much E/M was paid, surely you or one of your colleagues knows. I know, but coming from you it'll impress Febble.

[OnTheOtherHand]

74. the national surveys? certainly weighted

And I am dead serious. (It is only the midday data that I can't be certain about. Yes, I make every effort to admit when I don't know things, a discipline I would commend to you as well.) The national data are neither raw nor unweighted; they are weighted to quite a few things, as per my previous posts. I have no idea why you would debate this point.

If most of my colleagues are so certain that there is no controversy over the exit polls, it might be in part because certain activists insist on getting basic facts wrong, which is rarely a good way to get taken seriously.

[BeFree]

75. Ah, yes, the midday data

If that's what you want to call it then I will too. Now, if you have any facts to offer us, please, let it roll. (I mean facts that the midday data was weighted.)

Maybe, finally, we are getting somewhere. I was beginning to wonder if you and I could ever agree.

The Midday data was raw Agreed? Or almost certainly agreed?

Freeman has his basic facts wrong? And since your colleagues think so, ergo, there is no controversy? Just who DO you hang with?

[OnTheOtherHand]

76. no, the midday data were almost certainly weighted

E/M report, p. 7:

"An Age-Race-Sex adjustment is performed based upon the refusals and misses from sample voters that are observed by the interviewers at each polling location. The age, race and gender compositions in the exit poll results are adjusted to account for the differing completion rates of these demographic groups."

E/M report, p. 5:

"Early in the afternoon on November 2nd, preliminary weightings for the national exit poll overstated the proportion of women in the electorate. This problem was caused by a programming error involving the gender composition that was being used for the absentee/early voter portion of the national exit poll. This error was discovered after the first two sets of weighting; subsequent weightings were corrected. This adjustment was made before NEP members and subscribers used exit poll results on-air or in print."

So, while I can't be absolutely certain that any particular leaked numbers were weighted, we do know that E/M was weighting all day.

Freeman doesn't even use the word "raw" in his entire presentation, so have the decency to leave him out of it. You are on your own here.

We can try this in words of one syl-- short words, I mean.

The polls use weights, lots of weights, all the time! Age, race, sex, place, and more! But not vote counts: those come late. When the vote counts come, those are put in the weights too. 'K?

[BeFree]

77. Adjustments, weightings, program errors

All excuses after the raw data was leaked. Seems the weightings (as you call them) are a mask. Seems the census numbers are congruent with the raw data leaked at midday. Seems E/M are covering their asses.

Freeman was not quoted by me. It's just that he has this thing going that is part of the controversy. I see you don't want to admit there is a controversy. You or your colleagues.

E/M are going down.

[BeFree]

78. I quote Freeman here

The recent fuss over WPE and alternative analyses is about whether the E/M model proposed to account for unprecedented differences between their exit polls and the recorded vote is impossibly implausible or just highly implausible.



The bottom line remains that there is no evidence - or even theory - of differential response. There has never been any. The E/M data itself fails to substantiate the claim, and may, in fact, undermine the claim entirely. Without access to the data, it's difficult to know.



The unexplained questions about the exit poll and the election have certainly NOT been answered, either in Miami or anywhere else. In a system where campaign managers serve as election supervisors, where voting machines provide no assurance that votes are counted as cast, where a wide array of "irregularities" (a.k.a. vote suppression, vote manipulation, and mistabulation) were documented, and where counts and "recounts" are conducted in secret, the exit poll results stand out as conspicuously suspicious. The exit poll data, for all its limitations, is one of the only means to gain both national and local insights into whether, in fact, the official reported results of the presidential election were even in the ballpark.



That the relevant data has not been made available for independent analysis is inconsistent with the principals of a scientific community ? or, for that matter, a democracy.


http://www.appliedresearch.us/sf/aapor_response_to_mitofsky.htm

[OnTheOtherHand]

79. not an iota of support for your use of "raw" and "weighted"

(and, for that matter, Freeman doesn't really support his assertions here, either, but let's not quibble -- I certainly agree with him that the election system is pretty dodgy).

It's not news to anyone here that Freeman thinks the exit polls are highly suspicious. If anyone has read the thread this far, s/he will notice that you and I weren't actually arguing about whether Freeman thinks the exit polls are suspicious.

By the way, strange as this may seem, most political scientists don't read DU. Go figure.

[BeFree]

80. I don't need support

I use my own terminology, because I am free to do so. And you even agreed with me on some of it. It's all just hair-splitting, anyway.

Ya know, our little conversation here about this controversy probably won't be read by a handful of people. It's quite inconsequential. I do not expect to sway anyone who has seriously studied this matter, but I did hope to sway you. Oh well.

Too bad the hotshot political scientists don't read DU. If they partook of that aspect of democracy, and bothered to avail themselves of the common sense knowledge here, they'd all come away with a broader base from which to operate. As it is, they, by excluding some available knowledge, limit themselves. Evidence of such limitations is found in the fact that the political world they suck up too is at it's worst these days. Screw the professionals.

I have to give you credit, though. You have dipped your toe into the Underground. Evidence of your increased knowledge base may at this time be non-evident, but I am sure you will be better off in the long run.

[OnTheOtherHand]

81. how can you sway people if you insist on using your own terms?

Part of entering into dialogue is, well, entering into dialogue.

It's not "democratic" to insist that words can mean whatever you want them to mean, IMHO. And you shouldn't hope to inform anyone if you reserve the right to be unintelligible.

[BeFree]

82. Are you referring to the term 'RAW'

If so, I see what you mean. But to me the data that was leaked was raw, at least as raw as any data we have seen. So, in that term, it is quite intelligible. You even used the 'RAW' term in one of your posts to describe a data set. Your dialogue contained the term and now you want to disown the use? WTF?

[Febble]

83. Forget the terms

What are you trying to say?

That the data shouldn't have been weighted at all? That it shouldn't have been weighted to the vote returns? That the unweighted data should have been made available? That the data weighted only to demographics etc should have been made available? (it was).

That there should have been a different poll? That the networks should not have commissioned the one they did (designed to allow them to "call" each state) but rather one that would audit the election?

Look, you may not like the way the poll was done. But it wasn't, as I said in another post to you that perhaps you haven't read, "cooked" to cover fraud. The data was originally "cooked" to give as good a possible estimate of the likely outcome, based on known facts about the distribution of the "raw" data (demographics of non-respondents, geographical data about the precinct locations) and these projections were not "leaked" - they were broadcast. Later the data was "cooked" again to take into account the actual returns. It happens in the UK too. The early projections of the winning party are based on the exit polls alone, and as the results come in, the prediction is tweaked to take account of the known results. By the time most of the results are in, the predictions are pretty close to the eventual results.

The only difference between the UK and the US is that we can pretty well rely on our vote count. The fact that in the US you can't is not the fault of the pollsters.

No-one is covering up anything except the information that would make it possible to identify respondents. This is because it is quite unethical to collect data from participants in a study and then publish it in such away that the confidentiality of the participants could be violated.

The argument, really, truly, is not about fraud on the part of E-M - it is about fraud in the vote count. We can use what E-M HAVE published to analyse the discrepancy between the real RAW data and the vote count and to try to figure out whether it was due to polling bias or fraud. No-one denies the discrepancy was large, and impossible to have occurred "by chance" least of all E-M, who for perfectly valid reasons, believe it was polling bias. They may be wrong. But there is no reason to suppose they are lying.

I think you need to direct your fire at a more appropriate target.

[BeFree]

86. Is that right?

The covering up of the data, like Freeman says, could be handled in a way that protects the responders. Your view, which I've read ten times now, goes against Freeman. Fine. That's the controversy. I think they should let Freeman look at it, as it is the data that can answer some question. Anything less and I have to say they are hiding something.

What I am saying, on the other hand, is that the leaked data shows that there was something screwed up about the vote. On that we all agree. But the fact that E&M tried to spin it away like nothing was screwed up, but once caught tried to pin it on some innocent little way that say they should have been prepared for, but weren't, shows they are still spinning, and at the very least, if they are correct, means they should hang it up and get an honest job.

Release the hidden data to experts, quit the spin, and quit trying to throw off a reasoned and honorable investigation, E&M, and let the ball roll where it may. Otherwise the credibility E&M once had is trashed, and they are going down.

It's that simple.

[BeFree]

87. And furthermore...

I think it highly disingenuous to say that there was fraud and in the same breath say that the exit polls can't prove the fraud.

Maybe 'prove' is the wrong term. But yall get my drift.

[Febble]

88. Actually I don't

get your drift.

I am truly confused about what you are saying.

[BeFree]

90. Here ya go & edited

Following this is the original post. A post that may have been unfair after consideration. BeFree

The idea of using the exit-polls to discover fraud has been dis-counted here time and again. Freeman, IIRC, has stated that the polling does uncover some evidence and that if we could just look at the hidden data from E&M he thinks it would go further toward proof. So, to say, as some have, that there is nothing there that can prove fraud happened, is the disingenuous part.

&&&&&&&&&&&&&&&&&&&&&&&&&&&&

A Febble quote from upthread:
"There is no "audit" for your elections. There desperately needs to be, and exit polls, even if designed for the job, would be a very ineffective method. Random handcounts are the way to go IMO. In lieu of this, the currently designed exit polls can be reverse-engineered to a certain extent to try to shed light on whether or not there was fraud. But it's a terrible substitute.

Its what we've got - the exit polls. They can be used to provide evidence of fraud, it isn't the best, but it is the highest and best use of those numbers at this time. They are effective.

[Febble]

91. OK

And I agree. They are a lousy tool, but the best we've got.

But the reason they are a lousy tool is that they are subject to all kinds bias themselves, so it is difficult to say whether any discrepancy between the survey and the vote count is due to bias in the poll or bias in the count.

And it may even be impossible, but worth trying nonetheless.

However, the analysis that Mitofsky showed in Philadelphia strikes me as pretty conclusive evidence that it wasn't in the count.

Now listen carefully: I am not saying it proves there was no fraud. It doesn't. But I think it demonstrates that, specifically, new, vote-switching fraud was not the major reason for the discrepancy between the poll and the count. I would like it to be otherwise. But it is very hard to devise an explanation for the results of that analysis that is consistent with new, vote-switching fraud as the major explanation (read my lips - I'm not ruling out fraud in places where it happened in 2000, including vote spoilage in strongly Democratic precincts, nor other forms of fraud in 2004).

And I'm all ears for suggestions.

[OnTheOtherHand]

84. well, it depends on the context

The data that were released in January are "raw," which is to say, there is one record per respondent, so you can see what real people actually said. (The records have weights attached, but it is the analyst's choise whether or not to use them.) As far as I know, nothing that was released on Election Day was raw in this sense; it all had some kind of weighting applied.

I'm not trying to be mysterious here; we all understand that at first the results were not weighted _to the official vote counts_, and later they were -- and that made a big difference. It may seem like a nitpick to insist that even the early exit poll results were weighted, but by golly, it's true.

Note, by the way, that I am disagreeing with you on one point of fact: as far as I know, anything that came out on Election Day -- early or late, leaked or released -- was less "raw" than the data that were released in January. This point of fact isn't crucial to the exit poll debate; it is just, well, a point of fact.

While "weighted" is a technical term, "raw" is not -- but it's hard for me to imagine any survey analyst endorsing the idea of calling weighted aggregates "raw."

[Febble]

72. In answer to your last question

I expect the reason there has been such a controversy is threefold:

One is that the race was close enough that the discrepancy meant the difference between the two results. In 1992 there was also a large discrepancy, but Clinton "won" both the vote and the poll.

Second is that after 2000 everyone was alert for fraud, as they should have been.

Third is that the insecurity of the software meant that fraud on a large scale was technically possible.

So that explains the controversy. However, it is a myth that exit polls are phenomenally accurate. Ironically, they may have gained that reputation precisely because they normally are weighted to the early returns. But they are often outside their "MoE" because the MoE is only calculated on sampling error, and takes no account of non-sampling error. All polls have non-sampling error, which is why some early "cooking" is essential - to adjust for non-random sampling, e.g. polling too many women, or not enough older people. But you can only weight for visible characteristics of non-responders. So inevitably some error remains.


(edited for clarity)

[kiwi_expat]

59. Thanks! Here is my little list of possible contributions to a WPE.

As I now understand it, a particular precinct's WPE could be accounted for by a combination of factors: random sampling error plus sampling bias plus respondent bias; absentee* ballots; under/overvotes; uncounted provisional ballots; and, of course, fraud/accidental vote errors (adding, deleting, switching, mis-counting). Not to mention the fact that missed/refusals are not weighted in, and that polling place - not precinct - vote counts should be used in the comparison.


*NEP does not include absentee voters in the precinct exit poll totals.

[Febble]

15. Well, I think they both CAN be wrong,

up to a point. But there comes a point when vote switching on a big enough scale ought to show up in the WPE - which I don't think it did - unless the fraud was deliberately confined to non-exit poll precincts - in which case it should have shown up in precinct selection, which it didn't.

But that wouldn't rule out stuff that was more like sabotage than systematic theft. Especially if it was random enough that it tended to preserve the voting proportions (as TfC suggests). Machines that just look "faulty" if investigated. But if faulty machines were more likely to crop up in Dem precincts, it could still result in net loss to Kerry without showing up in the exit polls.

That's why I think that yowza's point is good - it's like BillBored's ballot rotation thingy. Just introduces noise at precinct level, but if the noisy precincts tend to be Dem precincts then Kerry will suffer more than Bush.

Not sayin' that's what happened, just that's where I'd look.

[kiwi_expat]

6. Thank you for discussing DRE fraud *vs.* tabulator(GEMS) fraud!

Many posters seem to lump them together, but they are very separate concerns.

I too am more worried about paperless DREs than I am about tabulator(GEMS) fraud. As you point out, tabulator(GEMS) fraud can always be detected by manual recounts, if there is a paper recored of the votes. Paperless DRE fraud can not be detected easily, if at all.

However, regarding DREs, I'm not sure what you mean by "also, the paper trail will always match the tabulation". Are you talking about DREs that have paper trails? Or are you saying that manual recounts can not normally count DRE votes?

Cheers!

[yowzayowzayowza]

8. Naw, naw.

DREs *with paper trails* would always match tabulation under the exploit I described. A DRE w/o at least a paper trail is suicide.

[kiwi_expat]

17. I have always wondered why a DRE paper trail couldn't be rigged also.

Sorry, I'm not following the details of the exploit you described. Would the paper trail match the votes as recorded by the DREs? If so, couldn't fraud be caught by the voter? Or if the paper trail didn't match recorded vote, couldn't that be caught in a manual recount? Which would be more plausibly denied?


I agree with your emphasis on "plausible deniability". That is one reason I think the fraud of choice in Ohio (and thus the EC) was vote suppression not tabulator fraud. (DREs did not feature in Ohio. Thus the 2004 election was not stolen by DREs. But the next one could be.)

[Febble]

18. And me.

"Plausible deniability" would also cover allocating voting machines in Franklin county on the basis of past turnout rather than "active voter". Looks fair - actually systematically selectively disenfranchises Kerry voters.

[yowzayowzayowza]

24. The term "paper trail" does...

...currently include a wide range of implementations from full VVPB to a cryptic log w/o voter verification. The described exploit would operate on the latter.

The point of the description is to emphasize the difference between explicit & covert software exploits.

Sorry, mist yer post earlier. GO ASTROS!!!!

[Bill Bored]

32. I'm rooting for the 'Stros and here's why:

First you gotta love the National League, but besides that, when the BoSox won the Series in '04, Kerry lost big time. Now, what do you suppose would happen if a team from TEXAS won the Series this year?

[yowzayowzayowza]

33. So how does...

Up by two, top of the 9th, two outs, none on, two strikes then lose by one compute? Nader mebbie? UG! I may need Ragaine after this.

[applegrove]

7. And by playing games with new technology - not stealing votes - but

playing all sorts of games in terms of access, number of machines, sorry defaults & the rest - they separate the Dems from each other.
People read into the situation according to their level of fear.

Works for them that way.

We are supposed to be apathetic & not vote.

Will not work on me.

The moment any Democrat hears an exit poll saying Dems are wining, early on election day and released illegally through the internet, each Dem should be taught to redouble their efforts to get to the polls. Bundle up the kids and walk to the poll and be prepared to wait twenty hours to vote (and think of those people in New Orleans).

All Dems need to have a little mission statement ready to "deliver" to the first snarky looking matron bully who tries to push them into "not voting" with "rule-making" and other intimidation.

[yowzayowzayowza]

10. I know what ya mean.

In almost a quarter centry of voting in Tx, I've never cast a ballot for federal office werth showing up for, but I'll be gawd-damned if anyonez gunne keep me from voting.

[applegrove]

13. Get your "little speech" ready for the bully matron "rule-maker" you

tries to dis your ID.

[eomer]

9. I agree with your basic premise but may quibble with some details.

I'm also a programmer of over 20 years and agree with your point about plausible deniability. It would be very easy to write code that looks innocent but does the job. It wouldn't have a handy comment that says:
// Fraudulently switch votes to pre-selected winner

Just like the chaotic mess of an election system we have in this country is perfect for hiding fraud, a mess of a software program would be an easy place to hide a vote-switching algorithm through some combination of complicated functions that each look innocent but work together to switch votes. This kind of code, if caught, could be explained away as an innocent bug even if it were in fact an intentional hack.

Where I don't agree with you is your conclusion that this points to DREs as the culprit and exonerates tabulators. Tabulators could have a plausibly deniable error in them just as easily as DREs. Either hack is as prone to detection by examination of code as the other. I take your point about the paper trail making it trickier for them to steal paper votes than e-votes but it looks like that is what they did. Maybe they sized up the situation, realized they owned all the important governmental functions in some key places (like Ohio) and also knew there weren't enough DREs being used to let them steal the election by stealing votes in just DREs.

There also was some measure of old fashioned low tech fraud and also fraud that crossed the low tech / high tech boundary, like stuffing paper op scan ballots and physically altering op scan ballots with markers and stickers.

When you get down to it, every aspect of our election system is effed up. Even paper ballots, hand counted won't solve the problem if the ballots are hauled around in some dude's van for hours or sit around in boxes in the hallway without effective controls to avoid ballot stuffing, ballot defacing, ballot theft, ...

And what good are any control mechanisms in the system if the judiciary branch at the state and/or federal level refuses to address blatant violations in a timely way?

So, innocent looking code that steals votes with plausible deniability? Agreed. That being the only way they stole votes and only in DREs? Disagree.

[yowzayowzayowza]

14. The difference between...

front & back-end exploit is paper trail support. I can't imagine even Repubz are gunna support audit trailless systems fer very long. If I wuz puttin my azz on the line, I'd code for the front end.

I totally agree, however, the methods for electoral fraud are manifold. "The price of freedom is eternal vigilance." - Thomas Jefferson

[autorank]

34. Tabulators...Diebold Plausible Deniability

Doesn't Diebold have a lock on the tabulator market? People would believe their errors without blinking, although it's pretty obvious that the errors all seem to fall on the side of the Republicans.

I think that this is an excellent thread.

Now let me remind everybody of this innovative idea:

http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=197869

Nice huh...

[yowzayowzayowza]

35. Innovative?

No. These "ratios" only "matched" in 18 of 756 and 22 of 1419 precincts. Looks as lame now as it did then. The term that comes to mind is "spreadsheet jockey."

Do you have support for this claim:

(Diebold tabulator) errors all seem to fall on the side of the Republicans.

[BeFree]

22. Contrarian gibberish.

Also, the paper trail will always match the tabulation."

What paper 'trail'? In the last election, any 'trail' there was, wasn't worth the paper that carried it.

neither method utilizes the preferred right-wing modus operandi since Shrub Senior nearly got busted in Iran-Contra: plausible deniability,

Just plain gibberish, having nothing to do with e-voting. They got away with it because they pardoned everyone.

I reiterate in hopes of expanding understanding of the potential for etheft and that explorations of electoral results (actual & exit polls) not unduly focus on exploits bound simply by voting or tabulation system.

I rest my case. Verdict? Gibberish

[yowzayowzayowza]

23. No Way!

Iran-Contra had nothing to do with e-voting? I must need a new :tinfoilhat:

Verdict? Comprehension Issues.

[Ellipsis]

25. Cow Bell.

We can never get enough press on this point ... until we get some press that is.

Read your post the other day ...did'nt get a chance to say thanks.

So ...thanks for you presence talent and effort.

[Bill Bored]

26. Agree completely.

But first you have to understand that most people here don't know what ballot definitions are or how they're programmed. I don't know what they think, but the way it's done is on GEMS or a similar system BEFORE THE ELECTION. And it's done for every machine in the jurisdiction from one source, sometimes even for an entire state such as Georgia.

One thing you may be incorrect about is that Clint Curtis style coding at the DRE level would not be detected by a recount of the paper trail. Actually it would -- but only if the paper trail were voter-verified. If it's not, then of course it's useless except as a way to canvass the precincts and compare precinct totals to central tabulator totals. And that's why switching totals in the central tabulators is so lame.

So I agree based on the evidence in EIRS and elsewhere and the actual features of the software, that ballot definition misconfiguration, possibly on a grand scale, or at least in a targeted way to favor Republicans over Democrats probably accounts for most of the observable glitches and some unobservable ones too.

Given the relatively small number of votes needed to swing a close election and the relatively small number of people setting up ballot definition files, it's pretty easy to see how the whole thing could have been stolen, esp. in states or counties where no voter-verified paper audit records or ballots are available or used for independent audits or recounts. And there are plenty of those!

[Time for change]

56. Thank you yowza -- it seems that this contains some important clues, but

I don't understand it.

Can you tell us how this kind of vote theft would manifest itself in a different way than the kinds of vote theft that would not allow plausable deniability -- or is my question not clear enough?

[Time for change]

57. Oh, wait a minute -- maybe I get it from reading some other posts

Are you saying that the fraud was done on individual DREs rather than on central tabulators? If so, that's kind of hard to swallow for me based on grounds not related to your post.

Wouldn't that require a tremendously greater amount of coordination and planning. I mean, since in order to make a huge difference, this would have had to have been done on so many different machines?

And why would this allow the paper count to match the machine count? If someone votes for Kerry but the machine is programmed to register Bush, wouldn't the paper still say Kerry? And if not, then what the hell is the point of a paper trail?

[yowzayowzayowza]

60. Two points:

The ballot definition files the DREs operate from are centrally created and race set unique.

As a result, some sort of unusual edit procedure in the ballot definition software could create a "glitch switch" (good one Land Shark!!!) in a single definition file that could in turn trigger abhorrent DRE behavior throughout a very specific region.

[kiwi_expat]

61. yowza(x3) explains his use of the term "paper trail" in post #24

"24. The term 'paper trail' does...
...currently include a wide range of implementations from full VVPB to a cryptic log w/o voter verification. The described exploit would operate on the latter." -yowzayowzayowza


So, if someone votes for Kerry but the machine is programmed to register Bush, the paper would say "Bush", in the described exploit.

[yowzayowzayowza]

89. FYI: Even tho I'm obviously not privy to ...

... the internals of the systems involved, a "glitch switch" could **without a doubt** produce ALL the types of errors & distribution ratios described in your EIRS vote switching thread