Direct Material Proof of Massive Election Fraud in Ohio in 2004

[RonB]

Friends,

See: www.baiman.blogspot.com for a link to this power point presentation.

It includes new evidence just found in Ohio.

I haven't yet been able to scan in all of the supporting documentation
(mostly the absentee audit reports and certified vote reports - but have all of
this available as backup and will scan this in as soon as I can.

We need funds now to continue gathering data and in order to launch a campaign
to save the ballots that will be destroyed in many counties on Sept. 3,
2006. See pictures from Darke County at the end of the spreadsheet.

Best,

Ron

[NVMojo]

1. Rank and KICK!!!

[iconoclastic cat]

4. What? How do you figure? nt

[autorank]

101. Ron Baiman is a national treasure, but that's just my opinion and I

happen to be 100% correct. This is just terrific. I'm glad you posted it here and you are able to see the response. It's nice to seem some new users, e.g., L.Coyote, chiming in with some major input.

I'm not a mathematician, I just play one on TV but you don't need a PhD in the obvious to see that
Ohio was taken off the deep end of democracy and held under water until it aligned itself with the
wishes of the "masters and commanders" (in their own minds) who are so greedy for power and so incapable
of handling it.

You do this country a great service. Thanks you!!!

K&R

(and NVMOJO :hi:)

[iconoclastic cat]

2. The link to the powerpoint is dead. nt

[alfredo]

3. Thank you. This might be locked, then moved to GD or

some other forum.

No big thing, but do get acquainted with posting rules in Late Breaking News. I made that mistake when I was a new poster here.


Anyway welcome to DU.

[uppityperson]

5. k&r for clean elections, without them we are screwed.

even with them we may be screwed, but without them there is no hope.

[Kip Humphrey]

6. Ron, having trouble with the Powerpoint link:"bcvrtf.yahoo.com not found"

[iconoclastic cat]

7. And yet everyone here is kicking and recommending.

I'd like some content before I rec.

[drm604]

16. Same here.

He could also post a text explanation of this proof that can be read without Power Point. That would be helpful.

[Land Shark]

60. I had no problem downloading it K&R n/t

[bleever]

8. Ron, I trust that you are in touch with

RFK Jr.'s team re: evidence that may be material in their own qui tam filings, and the need for financing to preserve it.

Not sure if this is indeed directly relevant to their actions, but worth exploring.

:thumbsup:

[IndyOp]

9. I can't get to the Powerpoint... (n/t)

[bear425]

11. Ron comments on his blog post to download the link

and then load it in powerpoint. I, for one, don't have powerpoint on my machine. :(

[L. Coyote]

13. PowerPoint viewer is a free MS download. LINK only n/t

http://www.microsoft.com/downloads/details.aspx?FamilyID=428d5727-43ab-4f24-90b7-a94784af71a4&displaylang=en

[bear425]

24. Hey, thank you very much. That was nice of you. n/t

[Nictuku]

10. Caution: New Powerpoint Virus

Microsoft Warns Against PowerPoint Virus

July 23, 2006 9:10 a.m. EST

Nicole King - All Headline News Staff Writer

Seattle, WA (AHN) - Microsoft Corp. is warning customers about a new virus that's affecting its PowerPoint presentation software. The virus allows hackers into computer systems.

The virus affects PowerPoint 2000, 2002 and 2003. The company has reported a limited number of attacks.

When a user launches a PowerPoint attachment to an e-mail or opens a file that a hacker has provided, the virus is launched. Hackers can also lure users to a Web page that contains the virus.

The virus then installs a keystroks logging system that captures everything that is typed on the computer. It leaves the computer open to having a hacker install other programs.

Microsoct officials say they're developing a security update to guard against the virus and it should be available on August 8 or sooner.

[Donnachaidh]

12. if it's stored on yahoo there may be traffic problems

Yahoo always comes to a screeching halt if it's pounded too much.

The powerpoint should be uploaded to rapidshare, or some other download point. I wouldn't depend on yahoo.

[Patsy Stone]

14. K&R

Thanks! Best of luck, Ron!

[drm604]

15. Ron, your link appears to be bad.

Either that or Yahoo is being overloaded. Maybe you should make it available elsewhere. You could upload it to http://upload.mp3mx.net/ for example, and post that link.

[RonB]

17. I changed the link.

It should work now.

Sorry!

[Melissa G]

18. Glad to see you back, Ron! Thanks for posting! K&R! n/t

[go west young man]

22. Works for me!

Thanks for your hard work Ron. Kick, kick, kick!

[Febble]

43. Why don't you

download PDF995 (free pdf creation software) and make a pdf copy. Almost everyone has a pdf reader on their computer. PDF995 installs itself as a printer, and you just "print" your powerpoint to PDF995, and it will save it as a pdf.

But before you do that, for goodness sake redact the personal details in the poll book photos. You are exposing those people to identity theft.

[L. Coyote]

19. Ohio Supreme Court Vote Anomaly

Are these counties with punch card ballots?

If so, what is the effect of counting the ballots with the wrong template (this has to be a precinct by precinct analysis)? Would that have an opposite effect on two races?

Here is a result set from Cuyahoga County, from the locations with 2 precincts and 2 ballot orders, therefor easily segregated into probabilies of precinct cross-voting:

			 		% b	 % p	 % K
K-B P=0		all			0.39	0.48	69.17
K-B P=0		K-p = 1.0		0.20	1.60	70.50
K-B P=0		K-b = 1.0		1.02	0.25	69.85
K-B P=1		all			0.25	0.15	70.26

Note how when there is a 1.0 probability that Kerry cross-votes are recorded as Badnarik votes (b), the Badnarik percentage rises from 0.20 to 1.02

Note how when there is a 1.0 probability that Kerry cross-votes are recorded as Peroutka votes (p), the Badnarik percentage rises from 0.25 to 1.60

Note also that when sorting only to Kerry to Bush probability = 1.0 (K-B P=1), the %b and %p are low, and when K-B P=0, their percentages rise. These differences reflect the amount of cross-voting, and reveal the actual low level of third-party candidate support.

Is this what is happening in these rural counties? The same sorts can be done with down-ticket races as comparatives. In situations with multiple ballot orders at locations, cross-precinct voting (and malfeasance switching ballots between precincts), can have opposite effects on different races, D-R or R-D.

STOP the ballot destruction!

Data from: http://jqjacobs.net/politics/xls/cuyahoga_2.xls

[rumpel]

20. Ron, thank you for this great analysis

and welcome to DU

Please keep us posted on your additions.

Question do you see any correlating formula used or do you think the results are random "tweaking" ?

[farmbo]

21. Nails Warren Mitofsky as the bumbling, self righteous a$$hole he really is.

Based on a string of emails in the presentation, Mitofsky consistently refuses to honor requests to release anonymous raw polling data citing "respondent confidentiality." Then he calls the author "unethical" for even requesting it, and tells the author to "F&@k Off" in the end.

In a state where there was a rule of law, the author's Miami County poll book investigation would lead to a full blown criminal investigation. It shows irrefutable evidence, precinct by precinct, that more votes were tallied than could be justified by the poll book signatures.

Alas, this is Ohio...the New Law West of the Pecos.

:grr:

[Febble]

23. And your problem is?

Mitofsky insists on maintaining the ethical principles regarding respondent confidentiality laid down by his own professional organisation (AAPOR), Ron repeatedly asks him to violate them, and you wonder why Mitofsky tells Ron to F&@k off?

Ron appears to have so little regard for confidentiality that he posts confidential emails and unredacted signatures, addresses and social security numbers in a public link.

Frankly, if I were Mitofsky, or any of the people whose signatures appear in that powerpoint, I'd tell Ron to F&@k off too.

As for his analysis:

I have a review of some of the exit poll stuff here:

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

Shorter version: all Ron has done is demonstrate that the exit poll discrepancy cannot be due to chance, which no-one disputes. He has not demonstrated that it must be due to vote miscounts. None of his analyses even start to disambiguate the effect of miscounts from the effects of bias in the poll.

Regarding his Connally argument: Ron seems to be still unable to understand that to the extent that the judicial race was not collinear with the presidential race (and as partisanship was not on the ballot, it is not surprising that it isn't), the better Bush does, the more likely it will be that Connally will outrun Kerry; similarly, the better Kerry does, the more likely it will be that Moyer will outrun Bush. Both these things are evident in the data, and are of course not a "coincidence" but a direct result of the fact that the two races are not collinear.

He may be right about Miami County, I don't know. But the fact that the rest of his arguments don't hold water does't fill me with hope.

[organik]

25. You've got to be kidding, right?

At this point in time, arguing that the exit poll discrepancy is not an indicator of fraud, in Ohio especially - is just silly.

The 2004 Election was stolen.

Kerry won, Gore won - get over it!

[Febble]

26. It's certainly not silly

Whatever you think happened in Ohio, and I think a lot of rotten stuff happened in Ohio, it is not "silly" to consider that the exit polls discrepancy is not an indicator of it.

See the report by these guys:

http://www.electionscience.org/Members/stevenhertzberg/report.2005-07-19.7420722886/view?searchterm=None

And read my own review of Baiman's paper:

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

Feel free to come up with a counter argument, but the 49 NEP precincts in Ohio offer no support for the fraud hypothesis over the hypothesis of bias in the poll. There is not enough statistical power to rule out fraud as a contributor, but that fact that Bush did not perform any better in precincts where the exit poll shift was "red" to those in which his shift was "blue" suggests it was not a major contributor.



Plus, much of what we know went wrong in Ohio was stuff that wouldn't show up in the exit poll anyway.

[Joe Chi Minh]

27. Your loyalty to you erstwhile patron does you great credit, Prebs.

We may have had our differences in the past, but nobody can take that away from you. It's beyond admirable; it's poignant.

[Febble]

28. Oh jeez

it's you again.

[farmbo]

29. It's silly that anonymous polling data can't be released owing to...

..."respondent confidentiality." Even if there were names on the raw polling data-- which has never been the case when I have been exit polled-- those names could be easily be redacted. (I suspect there are no names since it is an anonymous survey, but I will defer to you on that aspect of the email exchange).

As for the accusations by Mitofsky of the author being "unethical" for requesting them, if the emails are accurate, that's just the height of arrogance.

After all, Mitofsky is, at best, the bumbling Edsel Ford of exit polls, or, at worst, a co-conspirator in a vile and unAmerican election heist: your choice.

Either way, if the emails are true, this turkey could use a little humility.

[Febble]

34. The names are not on the responses

And the responses are (or were until recently) available for free download to anyone.

But because those responses are in the public domain, and because they include an extraordinary amount of personal detail, it is important that no precinct identifiers are released, because in some cases respondents could be identified from those details. That is why ESI commissioned a "blurred" dataset in which the vote-shares were, in effect, rounded, making it much more difficult to identify individual precincts, and yet preserving the statistical properties of the dataset.

And yes, to demand, repeatedly, data that would compromise the confidentiality of the respondents, despite having been told of the ethical constraints regarding the release of that data, is, IMO, itself unethical. I don't wonder that Mitofsky lost his temper. The fact that Ron has now publicly released data that could put poll workers at risk of identity theft would seem to confirm Mitofsky's verdict. He seems to have no clue as to what kind of duty of care is owed by researchers to their participants.

It is unethical to divulge data that compromises the confidentiality of participants in any research project.

[salvorhardin]

44. Given how everyone here feels about Choicepoint (rightfully so!)

Would you feel that the author of a medical study involving very detailed polling data in very small towns was being unethical in not releasing the raw data along with geographic identifiers?

[L. Coyote]

30. RE: wouldn't show up in the exit poll anyway ...

".... much of what we know went wrong in Ohio was stuff that wouldn't show up in the exit poll anyway."

Remember, you have to ADD all that to the exit poll discrepancy!

And keep in mind, what does show in the exit polls was enough, in its own right, to reverse the outcome.

[Febble]

32. Well, you can certainly add it.

But my point is that I consider that Ron's inference that the exit poll discrepancy indicates fraud in Ohio is deeply flawed. I consider the ESI conclusion that the discrepancy was unlikely to be due to fraud is a valid one, for the reasons they give, although I also have some criticisms of that study.

There are only two things that could have produced the exit poll discrepancy in Ohio, and elsewhere: miscounted votes, or biased poll. No-one disputest that the discrepancy could have been due to sampling error.

But Ron claims that his analysis disambiguates these two alternatives. I do not believe it does, for reasons I give in my review. And the finding (reported by ESI) that there was no tendency for Bush to do better in redshifted precinct in Ohio (and some were actually blue-shifted) suggests that it is more likely that bias was responsible for the discrepancy.

On the other hand I am very confident that more Kerry votes than Bush votes were lost in Franklin County due to inequitable machine allocation.

[L. Coyote]

37. RE: The qualifier "for which there is best evidence in Ohio"

Liddle writes, "I will start by saying that what follows
is in no way an argument that there was no vote corruption or electoral injustice, in Ohio.
I believe that there was."

So far, so good.

" ....However, the kinds of corruption of the democratic process for
which there is best evidence in Ohio are of a kind that would not be manifest as a
discrepancy between the exit poll and the counted vote."

The qualifier "for which there is best evidence in Ohio" helps here, but what of the obvious evidence of cross-voting?

Were the exit polls conducted at punch card precincts?

At punch card precincts, were there 2 or more ballot orders at the location?

If there were 2 or more ballot orders at the precinct's location, how would cross-voting impact that precinct?

Have these considerations entered the debate?

All the players in this debate, Liddle included, have overlooked this aspect, as best I can tell.

[Febble]

41. Nope

not overlooked, and yes, exit polls were conducted at punchcard precincts in Ohio.

Punchcards and lever machine precincts, nationally, had higher discrepancies than DREs or optical scanners, when urban areas were compared. But there isn't enough statistical power to find small effects at state level. So if punchcard problems affected the exit poll in Ohio, the effect size was too small to be statistically detectable with the small amount of statistical power available.

But that's not saying much. As I said, 49 precincts can't tell you much except that the discrepancy wasn't due to chance, and that it wasn't correlated with advantage to Bush. And as it wasn't correlated with advantage to Bush nationally either, and WAS correlated with factors likely to make selection bias more likely, it seems at least reasonable to conclude that the exit poll data doesn't support the fraud narrative.

Which is not to say that Ohio wasn't stolen. Or that Kerry wouldn't have won it on a level playing field.

BTW I hugely appreciate the fact that you have actually read my review. I am perfectly ready to debate it.

[L. Coyote]

48. Thanks, this helps

I have no quibbles with the review article.

I do have concerns that the cross-vote (or precinct-switch malfeasance) impact has been ignored in that debate. There is plenty of evidence of cross-voting where the analysis has been performed (Cuyahoga). My question is, has this been done for Ohio??? I have not seen it.

Using the stats from Cuyahga precincts with zero probability of cross-voting to third party candidates, their actual support levels are determinable. That is a useful baseline to compare with other counties.

[Febble]

42. I should also add:

that the sentence that begins the paragraph following the last sentence of mine that you cite reads:

A paper that claims that “Precinct-level Exit Poll Data Show Virtually Irrefutable Evidence of Vote Miscount” is therefore of potentially great interest.

My point being that while the evidence for voter suppression in Ohio is copious, the evidence for the kind of stuff that would show up in exit polls is in shorter supply. Which is why, had Ron been right, his paper would have been of "great interest". But, sadly, he's pretty consistently wrong from start to finish.

[L. Coyote]

33. a 17% shift in their margin

ESI Brief - Analysis of the 2004 Ohio Exit Polls and Election Results
State: Ohio
Author(s): Susan Kyle, Douglas A. Samuelson, Fritz Scheuren, and Nicole Vicinanza with Scott Dingman and Warren Mitofsky Date: 06-06-2005

"Summary: .... To some observers, the discrepancy between the polls and the results suggested that there was a problem at the core of the electoral mechanism: the counting and tabulation of votes AFTER they were cast. The discrepancy gave rise to widespread accusations that votes were shifted from John Kerry to George W. Bush..." (CAPS added for emphasis.)

"...counting and tabulation of votes AFTER they were cast..." may have been a problem, but so was casting votes in the wrong precinct, cross-voting.

Here is a sample precinct study from Cuyahoga County:

(from: http://jqjacobs.net/politics/ohio.html)

".... Bush and Kerry votes could be switched by cross-precinct voters ... notice that K and B are in the same column, collocated one above the other. Kerry voters from precinct 2908 will punch the third position if they vote on 2909 machines, but the third position in their precinct is counted as a Bush vote.

"... Kerry voters in 2908 will punch the third position in 2909, and those votes will be switched to Bush votes, the same position in 2908. Bush voters from precinct 2908 voting in precinct 2909 will vote the second position, Badnarik (b) when tallied in their own precinct, while Bush voter from precinct 2909 will count as Kerry votes in 2908.

.... apparently 14 Bush voters in precinct 2908 cross-voted in precinct 2909, punching the second candidate in the order, but their votes counted as Badnarik votes in their own precinct tally. Bush just lost 14 votes, 14 of the 15 votes now in Badnarik's tally.

Unlike the Badnarik, Peroutka, or disqualified votes, Bush-Kerry and Kerry-Bush switched-votes are not as readily apparent. To determine that amount, assumptions are needed. Assuming an equal proportion of Kerry voters, a proportion equal to candidate support in the precinct, cross-voted in 2909, 28 of them would punch the third position in 2908, mistakenly cross-voting for Bush when their punch cards are tallied in their own precinct. In this scenario—assuming random cross-voting in the two precincts—Badnarik gains 14 Bush votes while Kerry loses 28 switched votes to Bush (Bush nets +14 votes). The spread between Bush and Kerry has changed by 42 votes in Bush's favor, a 17% shift in their margin...."

That's just the one precinct, but it does illustrate what actually happened. Cross-voting does impact exit polls, and the problem is not a POST-voting problem.

[Febble]

35. Yes indeed

there is some evidence of the kind of problem that would show up in exit polls, if it were widespread. Another example is uncounted provisional ballots. Interviewers do not distinguish between voters who cast a provisional ballot from those who do not.

But if those problems were concentrated in one or two counties (like Cuyahoga), then it is unlikely to have impacted on the exit polls - there were only 49 precincts sampled from the entire state, although Cuyahoga, being large, is likely to have been represented.

My point is not that there were no problems (or corruption) in Ohio, but that the exit poll data does not shed any light on it. In other word, Ron's "virtually irrefutable evidence" from the exit polls is nothing of the kind. I do not believe that the case for fraud in Ohio is best made by flawed statistical analysis, and I do not think that the 49 exit poll precincts in Ohio tell us anything very useful at all about fraud in Ohio. What little they do tell us goes in the other direction.

[L. Coyote]

38. RE: "Ron's "virtually irrefutable evidence" from the exit polls is nothing

"... Ron's "virtually irrefutable evidence" from the exit polls is nothing of the kind. ..."

Granted. I see nothing original there. His ideas could benefit from an actual exposition in writing. We ae left with "trying to figure out what he said."

I disagree with, "...I do not think that the 49 exit poll precincts in Ohio tell us anything very useful at all about fraud in Ohio."

Are those 49 precincts identified? If so, then analysis with other factors considered might tell us a lot.

[Febble]

40. Well, he's done it in writing

that's what my review was of.

It's here:

http://uscountvotes.org/ucvAnalysis/OH/Ohio-Exit-Polls-2004.pdf

No, the 49 precincts are not officially identified (which is rather the point of the confidentiality issue), although at least one was identified by DUers, and is discussed by Mark Lindeman on page 10 of his paper here:

http://inside.bard.edu/~lindeman/beyond-epf.pdf

However, in their evaluation exercise, Edison-Mitofsky gathered a large amount of information about each precinct, and while there is very little statistical power at state level, it did enable some inferences to be drawn at a nationwide level. Unfortunately, their report does not give statistical details, and in any case I think they did it wrong. Which was why (to cut a long story short) I ended up being contracted to analyse it again.

But, as I said, that analysis can tell us very little at state level, as there are so few precincts per state. What is interesting, however, from the state level data, is that redshift was greatest in the bluest states, which seems to be a general tendency, at least in the five years given in the E-M report. And when you plot a regression line through the states in 2004, Ohio is not far from the regression line. So there is no particular reason to suppose it bucked a nationwide trend. Which may mean that the nationwide trends can tell us something about Ohio.

But what the nationwide data show is that there is, as in Ohio, no tendency for redshift to be correlated with advantage to Bush. And there is a strong tendency for redshift to be associated with factors likely to make any underlying differential response tendency manifest (i.e. greater opportunities for selection bias to creep in).

There's other stuff too.

[Awsi Dooger]

36. In regard to exit polls, I've noticed plenty of silly, the other way

Let's take that national exit poll, the one embraced by fraud theorists asserting Kerry won nationally 51-48. I haven't mentioned this on DU in probably a year, and perhaps never in this forum, but time for a recap.

One category conveniently overlooked is "When Decided." If you analyze those numbers based on how the race unfolded and the pre-election poll consensus, the adjusted 1:25 PM Monday results make much more real world sense than 12:22 AM, or either of the two previous calls.

All three non-adjusted "When Decided" numbers are remarkably similar, Kerry with 50% among the 79% of voters who decided more than 30 days before election day. Then Kerry takes 60 or 61% among the 10% of voters deciding "Last Month." So if those numbers are true, and this is a representative sample, then how did Kerry ever trail in the pre-election polls? Certainly in the final weeks he should have led, but that's hardly the case in the late poll consensus: http://www.realclearpolitics.com/bush_vs_kerry.html

Let me guess the response; the pre-election polls were wrong, tilted to Bush. Funny how that wasn't the argument post-2002, when pre-election polls were gospel around here, and the older the better. I still see posters insisting Cleland was up double digits over Chambliss in Georgia, even though the late polls had it tied with all the momentum toward Chambliss.

Anyway, that 12:22 AM NEP, the one preferred as the ultimate truth around here, gives Kerry 53% of the late deciders, those who made up their mind on election day or within the previous three days. His number along "Last Week" is even worse, 48%. Sorry, but there wasn't a pre-election model on this planet that gave Kerry 51% of the popular vote if he managed only a slight majority of the late undecideds. TIA was championing 67% in his election model. I warned him several times that was wildly optimistic, since the challenger has been defined and hardly charismatic. In my model I struggled between 55 and 62%. Only the highest numbers in that range pushed Kerry slightly over the top. The wild cards were party ID, white women and late undecideds, and all indications are we lost all three, compared to the numbers we needed.

On the other hand, the weighted NEP describes the race exactly as I remember it. Kerry has 46% of those deciding beyond 30 days. Then a 54% majority among "Last Month." That probably reflects Kerry's debate superiority. But over the final week, with the release of the Bin Laden tape and 9/11 providing a fear-based ally for the incumbent, the vital late undecideds don't break sharply our way and Kerry comes up short.

[Melissa G]

55. Hi Awsi, you seem to merit a TIA refuting...

PM me if you need directions to view the dissection..

[Awsi Dooger]

76. I seriously doubt TIA looked at the "When Decided" category prior to 2004

Melissa, that's my beef with TIA and Steven Freeman, among others. Instant authority, even though they had no background in the subject. TIA certainly has done extremely impressive mathematical work on DU and elsewhere for years, but I debated with him often enough to know he placed extreme emphasis on pre-election polls before 2004, virtually no mention of exit polls. Of course, that's partially attributable to the exit poll collapse of 2002. I'm sure he would have managed dozens of threads if the 2002 exit polling had been released.

But this is what I'm getting at, not all of this is pure math. It helps to know the applicable trends and evaluate them sensibly and with historical background. I'm posted previously I think you could hand Steven Freeman an exit poll allowing John Kerry 60% in Utah and he would merely type it in his database and scream foul if the actual vote tally didn't match.

I've studied that "When Decided" category since '96. Hundreds of views, including presidential, governor and senate. Not once have I seen a Democrat with 50+% in the "Over 30 Days" category who hadn't lead in the pre-election polls. Let me repeat that, not a single time. In fact, I can't remember a Democrat with 49% who didn't lead the early polling, nor a Democrat who was tied with the GOP opponent even if both were slightly below 50%.

That category is dominated by the Republican. I should have emphasized that in my post last night. Sometimes I leave out a point when I've mentioned it on so many forums it seems cliche. But that's what I was referring to when I asked how Kerry didn't lead the pre-election polls if he truly earned 50% beyond 30 days, and 61% in the "Last Month" category. If Democrats had to rely on the election result matching that "Over 30 Days" margin, we would be a 30% party in congress. I don't know if Republicans are truly less flexible, or merely like to assert they knew all along, but just check out the state by state exit polls for 2004 or earlier and you'll see what I mean. It's hardly uncommon for a Democrat to trail that category by 8 or 10%, even if the race was tight throughout.

Of course, it's not always a perfect comparison or easy to evaluate because the pre-election polls have much lower percentage of undecideds than the exit polls reveal. The pollsters include leaners and I think it's human nature to voice an opinion and not whimper out with a "sorry, no idea." Men, in particular, seem reluctant to say they're undecided. But in 2004 you had an extraordinary percentage of knew-all-alongs in the exit polls, mid to high 70s everywhere including 79% in the presidential race. That's maybe 10% higher than typical.

Admittedly, I've looked at that "When Decided" number primarily for betting purposes, plus understanding how states vote in relation to how they were polled. Sometimes the exit poll number in that "When Decided" category hints the pre-election polling was flawed all along. The most dramatic example is the '04 Alaska senate race between Tony Knowles and Lisa Murkowski. Conventional wisdom is Knowles led for an entire year, normally a few points, then senator Ted Stevens changed the outcome with a late fear flurry. But the beyond 30 days number was overwhelming in favor of Murkowski, something like +15, then Knowles grabbed a vast majority of the final month and late deciders. I've read many pollsters say Alaska is very difficult to poll and that's further confirmation. I'm convinced Knowles never led at all, or not in the final months. Determining where the polls are likely to err is my hobby, not screaming fraud if they do err. I wagered on Lisa Murkowski in that race, since my PAN or Partisan Adjustment Number for Alaska indicated she was probably leading.

I know, Diebold in Alaska. It's always Diebold. But the same thing happened in 2002 when Fran Ulmer narrowly trailed Frank Murkowski in the polls, then he blasted her by 15%.

I realize this has rambled. A TIA response is more than welcome if he addresses the historical issue, and finds another example(s) of a Democrat at 50+% in that beyond 30 days "When Decided" slot who did not lead the pre-election polls. Frankly, if someone had told me in early 2004 that Kerry would get 50% in that category, I would have projected him to lead the pre-election polling by 2-4%. Again, this is poll based, not result based. Certainly possible for a Democrat to lead that category, then lose via late deciders, and the reverse is true.

So my dilemma regarding 2004 is do I trust the relationship I've noticed since '96, that a Democrat with 50+% long term strength naturally led the pre-election polling, or since that didn't happen, not even close, do I reject that 50% as a flawed representation and conclude the 46% adjusted figure is more likely accurate? I choose the latter, with the caveat that virtually all my 10 years of sample have been with "weighted" exit poll figures. Maybe they've been fooling me all along.

[Melissa G]

98. I was going to give you a long answer until I read your down

thread post..
So short answer...My experience of Steve Freeman does not lead me to agree with your assessment of what he would do with that data..
and
I'll pm you with the TIA response link if you don't have it..

Edit to say that one of my main problems with exit polls is that they can institutionalize fraud.

[Melissa G]

99. You have been refuted.. again...Check your in box.n/t

[Awsi Dooger]

100. Thank you for the link in your PM

I sent a lengthy reply. My comment in the other forum that influenced that thread was overboard, at least in terms of 2004.

[Melissa G]

107. Hi Awsi ,TIA likes your math, if not your conclusions...

I have all kinds of problems with your assumptions but not enough attachment to the conversation to enumerate all of them..
Check your in box..
Best,
Melissa

[Awsi Dooger]

109. Admittedly, that "When Decided" has been less varied in '00 and '04

In the presidential race. Tons of 48s and 49s all over the place in the 30+ days category. Emblematic of the polarized era. So I guess it's not as unlikely/impossible Kerry could have had 50% in that category and not led the polls. But a very typical statewide breakdown is a Democrat who led the polls by 3-5 points ending up with 48 to 50% in that number. When a Democrat trails the pre-election statewide polls by a few points, he normally receives 43-46% in 30+ days. But in the '00 NEP I believe Gore was higher than that, maybe 47 or 48, even though he trailed the late national polls also.

When I first started looking at it, the difference between 30+ days and the late deciders was incredible. And it can still be that way in isolated gov and senate races, normally the Republican significantly ahead in the early deciders then the Democrat surging with 56+% in the final month and days. It frequently doesn't jive with the polling or conventional wisdom of the race, that's what puzzles me.

[autorank]

102. Has that ever made a difference...

Hello MelissaG.

[OnTheOtherHand]

103. nope, serious discussion here is very, very difficult n/t

[BeFree]

58. Yeah, right

Out of 16m new voters, bushco got 10m and Kerry got just 6m?

No way, Jose.

Papers across the country that in 2000 were bushed, canceled their subscriptions and bushco still got 10m new?

And of the early votes that I've seen, Kerry won by a large margin and that from around 30% of the votes, but the other 70% on election day bushco carried?

You want to talk early numbers? Lets talk about those early numbers.

[Awsi Dooger]

77. I thought you were ignoring me

Not sure which I prefer :)

Bush had plenty of advantages as an incumbent. That wasn't always recognized on DU.

Here's a very good link I saved on election day 2004, from Mark Mellman detailing the uphill climb we faced. I remember being numb when I read it, recognizing the truth of what he was indicating. I recommend the entire article but I'll paste some key paragraphs:http://www.thehill.com/mellman/110204.aspx

"First, we simply do not defeat an incumbent president in wartime. After wars surely, but never in their midst. Republicans have been spinning this fact for months, and they are correct.

Democrats have spoken often and powerfully about the nation’s economic problems. But by historical standards, they are not that bad. The “misery index” is 7.8 today but was 20.5 when Jimmy Carter was defeated. Economic models of elections show Bush winning 52-58 percent of the vote.

One could simply suggest that the models are off, but there is more to it than that.

These models essentially confirm that the level of economic pain we are now feeling is not commensurate with voting an incumbent president out of office.

Unemployment and inflation are lower than they have been when incumbents have been defeated. Growth is higher than it has been when presidents have been tossed out of office."

<snip>

"Only 39 percent give the economy a positive rating, a problem for the incumbent.

Yet in 1992, only about 10 percent were positive about the economy.

Taking all that and more into account, an expert forecasting model suggests that Bush will get 51.6 percent of the two-party vote."

[BeFree]

82. Made up? No. Mistake? yep

See:
http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=439645

"Yet the official 2004 tallies showed Bush with 62 million votes, about 11.6 million more than he got in 2000. Meanwhile Kerry showed only eight million more votes than Gore received in 2000. To have achieved his remarkable 2004 tally, Bush would needed to have kept all his 50.4 million from 2000, plus a majority of the new voters, plus a large share of the very liberal Nader defectors.

Nothing in the campaign and in the opinion polls suggest such a mass crossover. The numbers simply do not add up."

What about those votes that came in from the early voting numbers? Are you willfully ignoring those, too?

[OnTheOtherHand]

83. presumably you meant this to reply to me

OK, I stand corrected: Michael Parenti just made it up.

But I think Awsi did point out on another thread why Parenti's argument is wrong: it assumes (without mentioning the assumption) that all the Gore voters would vote for Kerry. Arguments that depend on unstated and incorrect assumptions are very, very bad arguments. Parenti's argument took a big hit there.

Cite the early voting numbers, and then we can discuss them.

[BeFree]

84. There goes that integrity thing, again

Those are the numbers, not made up....well, yes they were, by diebold...anyway, what is there but assumptions? The pre-polling? Assumptions. Most of your arguments? Assumptions.

The closest we can get to real concrete numbers are the raw exit polls and the early voting numbers. Both of which show Kerry winning.

The rest of this crap is assuming the e-voting correctly counted the votes, and that assumption is the most ridiculous.

[OnTheOtherHand]

85. nope, it's a fact thing

I actually don't give a rat's ass about Parenti's "integrity." I want to know the chances that I can learn anything from him. And in my experience, it is hard to learn from people who either don't reveal, or don't notice, the assumptions underlying their arguments. So, his credibility took a hit.

"Those are the numbers, not made up...."

Did you actually read the post you appear to be replying to? The point is, Parenti claims that Bush's 62 million votes would have to come by "(keeping) all his 50.4 million from 2000, plus a majority of the new voters, plus a large share of the very liberal Nader defectors." And that is flatly, blatantly wrong, because it simply ignores the Gore voters.

You again refer to "early voting numbers," but don't say what you are talking about. Looks like you've got nothing much, beyond your charming faith in exit polls.

"The rest of this crap is assuming the e-voting correctly counted the votes...."

No, it isn't. You keep repeating it despite many corrections. Hmm.

[BeFree]

86. It is your credibility in question

You haven't produced one bit of evidence of anything, and everything others have produced that goes against your assumptions is met with nothing but negative responses.

You have NO credibility. Except for a negative credibility.

And you show negative intellectual integrity, too boot!

[OnTheOtherHand]

87. again, if you don't have an argument, then you don't have an argument

I've presented plenty of evidence that exit polls are inaccurate.

It's true that when I disagree with people, I disagree with them. Maybe you could present an example of someone producing something that "goes against your assumptions" that you met with a positive response, and then we could compare. But actually, I was more interested in talking about the election. Funny how you always get around to changing the subject. Hmmmm.

[BeFree]

88. Your intellectual integrity is the subject

Or rather the lack thereof.

All you are is exit polls. No real HAVA stuff, nothing much about the machines, no support for CA-50, I could go on.

It seems you are here only to protect the exit polls.

That's it, a one-shot pony. Well, son, you shot your load.

[OnTheOtherHand]

89. that's just dumb

In what universe am I "protect(ing) the exit polls" by arguing that they aren't accurate? *

Actually, I have been trying to protect the movement's credibility: I think it's stupid to make stupid arguments. Your mileage may vary.

I've got nothing to say about CA-50, so I'm not saying anything. Imagine that.

* EDIT TO ADD: On the off chance that any newbies are reading the thread, my often-stated opinion is that the exit polls aren't guaranteed to be accurate within sampling error. Like other measurements, they need to be used judiciously.

[Name removed]

90. Deleted message

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

[OnTheOtherHand]

91. huh? n/t

[OnTheOtherHand]

81. and then there is the part you just made up

"Out of 16m new voters, bushco got 10m and Kerry got just 6m?"

Uh, no. Probably Kerry won the new voters, but lost anyway. Didn't we just go over this?

If you want to be taken seriously, why do you make stuff up?

[RonB]

59. Election records are PUBLIC records

By definition public records are available to the public for anyone to peruse and publicize at will. These judges have nothing to worry about if they did nothing wrong. Their names and addresses are also in the telephone book!

Plus Mitofsky "confidentiality ruse" is clearly just an excuse to avoid public responsibility. It would be almost impossible to identify individual respondents in most precincts and he could blur the one's where such a remote possibility exists.

[Febble]

61. They have everything to worry about

Sure, anyone can break into my house and steal my purse. But I try not to keep it in full view of an open window.

You just the left signatures, social security numbers, names, and addresses of several innocent people in full view of an open window, without their permission. You've never heard of "identity theft"?

And your assertion (in quotes - who are you quoting - yourself?) that Mitofsky is using a "'confidentiaility ruse'" as "an excuse to avoid public responsibility" has absolutely no basis in evidence. Read the guidelines of his own professional organisation. In addition, Mitofsky consulted the ethics committee of the American Statistical Association, and they agreed with his position.

Identity theft and data-mining, as we know only too well, is a serious issue, one you do not seem to have begun to grasp, as is all too apparent from your own careless public posting of other people's personal data. If you want access to confidential data, then acquiring rigorous habits of data protection yourself would be an excellent start. Actually, realising that personal data actually requires some level of protection would be an excellent start. One reason the data you want should not be publicly released is because there would be nothing to stop some idiot publishing it in a powerpoint on the internet.

If you want access to the data, then join AAPOR (if you have not already done so) and argue it out with the ethics committee. You may learn something.

[RonB]

78. I removed that slide from the power point

OK Lizzie, I get your point. My error. The one offensive slide has been removed. The evidence is still very much there.

Regarding Mitofsky, if you think he's withholding all of the national precinct level official results for exit poll data (except for Ohio which somehow doesn't compromise confidentiality) because of his great fastidious concern for "respondent confidentiality", I've got a bridge in Brooklyn to sell you!

He also has led the ASA and other august bodies to believe that he has released all "the raw data" when he's actually only released about half of it with weights that are adjusted to match the official data.

Great sense of public responsibility - anyway all this is water under the bridge, we're well beyond exit poll arguments at this point!

[Febble]

79. Thanks Ron, that's decent

But let me try to explain (again): The Ohio data don't compromise confidentiality "somehow" - they don't compromise confidentiality because the data were blurred.

The data were blurred because of a "fastidious concern for 'respondent confidentiality'". (Respondent confidentiality is not something that should be put in scare quotes. What is scary is when it is not respected.) And of course it isn't some weird coincidence that the blurred Ohio data were released, those data were specifically prepared for release according to the procedure developed by Fritz Scheuren*, one of the authors of the ESI paper, because Ohio was targetted for research into whether the exit poll discrepancy there was evidence of fraud. I understand further blurred datasets may have been commissioned. But it's a complicated and labour intensive procedure (in order to preserve the statistical properties of the data) and involves actual data collection (statewide precinct totals). You don't just press a button and put the data through a blender.

AS for your third paragraph: no-one has been misled (except, it seems, you and other non-polling experts like me - until I bothered to find out), and all the data that is actually available is released. Nobody is going to enter the data from the remainder of the questionnaires - they were never intended for use. They are not needed for any purpose other than tallying the presidential vote responses on election day. The pollsters have an efficient method for selecting a usable random sub-sample of the questionnaires for use in the crosstabs and that is all they need. It gives them enormous statistical power as it is. What you actually want is the presidential tallies, not the raw questionnaires. It's the percentages from the presidential tally totals that were used by ESI.

Actually, I'm wondering whether you've actually LOOKED at the raw questionnaires- weights are provided, but it's a simple matter to turn the weights off in SPSS, and if you aren't using SPSS, the weights are just another column you are perfectly entitled to ignore. Many do.


*Link to the 1986 blurring paper here:

http://www.amstat.org/sections/srms/proceedings/papers/1986_070.pdf

[BeFree]

31. Kick

Is this indeed the bullet from the smoking gun? I hope so. Thanks, good work, Ron.

[redacted]

39. BIG KICK

[roseBudd]

45. In OH the precinct on the punchcard is identified with a printed or...

stamped precinct number. In order for the tabulator to know what precinct and what hole corresponds to a particular candidate that precincts ballots must be preceded with a pink header card.

Suspicous minds (mine) want to know why a punch card system would introduce the possibility of human error or malfeasance since the incorrect header card would go unnoticed. Indiana punch card ballots have the precinct code punched in the ballots so there is no opportunity for a BoE employee to "mistabulate" a precinct.

Edited to add I have tried unsuccessfully to determine if OH punch cards used to have precinct codes punched under Sherrod Brown. Brunner's people did not get back to me.

[L. Coyote]

46. Suspicous minds want to know

RE: "Suspicous minds (mine) want to know why a punch card system would introduce the possibility of human error or malfeasance since the incorrect header card would go unnoticed."

Add my mind to the list. I want to know a lot more too, especially why a location could have more than one ballot order.

We need a seminar/gathering of those who have focused on these Ohio issues. A gathering of minds might advance the ongoing inquiries. The Internet is great, but a weekend with everyone in one location .....

[Febble]

47. Excellent idea

but it could be a virtual location, couldn't it? Why not set up a date and do it here:

http://www.electionauditinstitute.org/

[L. Coyote]

49. But do they have hot tubs?

There is no substitute for a weekend of minds gathered in a remote location, severed from everyday concernes and focused on one thing. Of course, a virtual component could be possible too.

[farmbo]

50. Febble...your take on the Miami county poll book discrepancy?

If you ask me, that was most damning evidence of fraud in the presentation. Looks like somebody fed some pre-punched ballots through the tabulator. Oldest trick in the book, but since the recount precincts were cherry picked, no one found out about it until now.

And I'm now convinced there's something to the Connelly anomaly, since it only appears in the SW Ohio counties where Bush received his late night victory margin. Too convenient.

I swear we spend too much time contemplating our navels on exit polling and we're ignoring the first hand evidence of fraud staring us in the face.

[Febble]

51. Well, I am waiting to see

what Ron has come up with here. I don't have a take yet. I confess that I'd really like to see something that would put someone in jail (preferably Blackwell). And I have absolutely no faith in the recount.


But the Connally thing doesn't strike me as anomalous, for reasons I gave. It's not coincidence that she should run ahead of Kerry (but behind Moyer) in places where Bush was doing best - it's exactly what you'd expect if the judicial race was less partisan than the presidential race, and my understanding is that partisanship wasn't even on the ballot.

If you are going to cite something as an anomaly, statistically, you have to figure out what would be expected under the null hypothesis.

Mark Lindeman has a plot here that illustrates the relationship:

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

which I just doctored:



What you can see is that while there is a strong broad positive relationship between votes for Kerry and votes for Connally, there is a wide spread as well, across the vote-share range, i.e. while Connally generally does well where Kerry is doing well, there is plenty of variability in the way her voteshare compares to Kerry's. And what you can also see is that the judicial race is less polarized- the range of Connally's vote share is considerably less than the range of Kerry's.

I've drawn a diagonal line - any datapoint that falls on the diagonal is where Kerry and Connally got equal voteshares. Datapoints above the line are where Kerry is doing better than Connally, and datapoints below the line are where Kerry is doing worse than Connally. And you can see that because the judicial race is less polarized than the presidential race, that Kerry runs ahead of Connally where his own support is highest, and tends to run behind her where his own support is lowest.

What is sort of interesting is that the plot is fatter at the bottom than the top. It looks as though the judicial race was more partisan at the Kerry end of the presidential range than at the Bush end (but bear in mind that of course there are more extreme Kerry precincts than extreme Bush precincts). In other words it does look as though Connally's vote is dragged upwards more strongly by a high Kerry vote, than dragged downwards by a high Bush vote. So you might want to argue that this assymmetry is what is suspicious. But it doesn't strike me as inherently implausible that the race should have been more partisan in highly Democratic precincts (where Connally's partisanship was more likely to be known) - and precincts are certainly more partisan at the Dem end of the range.

But I couldn't agree with you more about exit polls, certainly at state level. At national level they were certainly potentially interesting, which was why I chased after them. But the data strongly suggest that the sample was biased, and actually contra-indicate vote-theft on a multi-million vote scale.

So it's the first hand evidence that is important, and, frankly, it doesn't matter whether it was result turning or not. Theft is theft. As I say, I'd like to see someone prosecuted.

[farmbo]

53. These stats. don't reflect drop off rates from top to bottom of the ticket

You can drive yourself nuts comparing the relative vote between Kerry and C. Ellen Connally. As your chart suggests, it is not "inherently implausible" that voters in the SW counties--but not elsewhere-- voted for her and against Moyer, the four term GOP incumbent. After all, in Ohio the Supreme Court races are nominally Non-partisan.

But what I don't buy is the stark deviation from normal drop off rates in the suspect counties for Kerry voters.

Let's face it. Voters went to the polls in 2004 to vote for the presidential candidate of their choice. Many voted for the entire tickets, but there was/is always, always, always a significant drop off in the sheer number of votes from the "top" of the ticket to the "bottom."

That's where they screwed up. Those drop off rates were, I suspect, largely the same in all the other counties outside the suspect 12, but were actually reversed in the suspect counties, where --we are asked to believe-- those conscientious exurban voters cast more votes for the Bottom of the ticket than the top.

I don't buy it.

[Febble]

62. Well, if it's "stark"

then you (or someone) needs to demonstrate that. You may be right. But to make a convincing case that something is anomalous - you need to demonstrate there is a discontinuity from the pattern expected given the broad patterns in the data. It looks all of a piece to me. But I am prepared to be persuaded otherwise.

And remember that the plot is not of numbers, but of proportions. It doesn't show you the "drop off" - it shows you the extent to which Connally's share of the Connally+Moyer vote was/was not correlated with Kerry's share of the Kerry+Bush vote. And what it also tells you is that regardless of "drop-off", the judicial race was less polarized. In other words, even amongst those who voted in the judicial race, opinion as to who was the best candidate appears to have been more mixed, geographically.

In other words, the plot tells us that the judicial race was:

• highly, but not completely, correlated with the presidential race
  
• less polarized than the presidential race.

These are independent facts about the data. Given these facts, it is not a coincidence, it is merely follows logically, that in high Kerry territory, Kerry will tend to run ahead of Connally, and Moyer ahead of Bush, and that in high Bush territory, Bush will run ahead of Moyer, and Conally ahead of Kerry. In terms of actual votes, rather than vote share, it is slightly more complicated, because votes for the judicial race did indeed "drop off" (but were still high). Whether Moyer runs ahead of Bush in vote terms in high Kerry territory, or Connally ahead of Kerry in high Bush territory, will depend on the proportion of voters who cast votes for the judicial race.

This is what I mean by establishing what you would expect under the null hypothesis. Once you have done so, you may be able to demonstrate that the discontinuity in suspect counties is greater than you'd expect from the distribution you'd see under the null. But it's certainly not what Ron has done.

[farmbo]

65. Sorry...I have a day job and I'm not a statistician. Part of the indignity

... of being an Ohioan is that Nobody in power shows the slightest interest is demonstrating, thru transparency, that the results in these Southwest counties were legitimate. Everywhere we look there's stonewalling, late night terror lockdowns, computer/ tabutator tampering and oughtright lies.

It seems to be up to we few (We Happy Few?) election fraud aficionados to demonstrate that their numbers were cooked.

This is bass-ackwards. (Alas, an argument for another day.)

Still, I believe the Connally anomoly is good for one thing: pinpointing the "scene of the crime." These counties had the means, motive and opportunity to turn the election. And many here believe they did.

[Febble]

66. Sorry about the stats-speak

and yes, it is bass-ackwards.

Nobody should feel they have to use inferential statistics to find out whether their vote was counted.

And I'd really like to see a civil rights case brought regarding the provision of voting machines in Ohio. I don't think you even have to prove intent to disenfranchise, just a systematic effect - but I'm not a lawyer!

Cheers

Lizzie

[roseBudd]

70. Exactly SW OH should have been least likely to vote Connally, it's

not like she is the offspring of a Taft, Dewine, Deter or Luken. Nmae recognition is important and even though Supreme Court Justice isn't a race you expect to see a yard sign for, Warren County DUer lizzieforkerry told me Moyer signs were all over Warren in Bush/Cheney sign yards.

I belive the Connaly votes were Moyer votes that were lost to ballot crawl when Kerry votes were tabulated as Bush votes. I think the ballots were culled, separated and a portion fed through behind the wrong header card.

Or Kerry votes were flipped to Bush in the tabulation but Connaly votes weren't flipped to Moyer.

No matter how red Warren is portrayed lizzieforkerry said no way Kerry did no better than Gore as she worked both campaigns for months. BTW there was no Gore campaign in Warren. He pulled out. She also said P&G has brought a lot of socially liberal East Coasters to Warren in the last few years and that Kerry signs were everywhere. Warren County is no Mormonbot Utah.

[Febble]

73. Well, stats are no substitute

for on-the-spot observation. And much as I dislike your DREs, I don't like your punchcards any better.

Good luck.

[L. Coyote]

92. RE: "Kerry votes were tabulated as Bush votes"

That should be easy to demonstrate.

Rosebud writes" "I believe the Connaly votes were Moyer votes that were lost to ballot crawl when Kerry votes were tabulated as Bush votes. I think the ballots were culled, separated and a portion fed through behind the wrong header card. Or Kerry votes were flipped to Bush in the tabulation but Connaly votes weren't flipped to Moyer...."

Ballot crawl, either due to 'cross-precinct voting' or 'cross-precinct tabulation malfeasance' should be very easy to detect, as it was in Cuyahoga County. Because there are five candidate positions for President, there are five ballot orders. Down-ticket races do not all have the same number of candidate positions. An analysis of these orders and their relationships is required. Sorting precincts according to which cross-votes are possible and which not (probability 1 or 0) in different races, precincts, and locations will reveal if what you allege is so.

Has anyone prepared spreadsheets of the official results for the punch card counties with the ballot orders, a la Joe Knapp in Cuyahoga?

There are two main divisions in statistics:

Descriptive Statistics: describing what the data shows when all the data is available.
Inferential Statistics: making judgments of probabilities using a sample of the data.

Exit polls are inferential, hence the margin of error.
Vote tabulation studies are of the entire population (data set) and not necessarily inferential. This makes for very convincing argumentation, w/o margins of error (or non-responders).

[OnTheOtherHand]

93. a master file of ballot orders would help

I don't know how Knapp initially compiled his data, but I do know that the Cuyahoga BoE posted sample ballots on its website, so I could spot-check and verify that (1) Knapp seemed to have it all right and (2) it followed a deterministic pattern based on precinct name. If we assume that (2) holds for other counties, it would be pretty simple to extend the analysis -- but much less simple to test the assumption.

If caterpillar crawl appreciably affected the results in Warren, it must have been very well targeted: Badnarik and Peroutka's combined vote share was well below the already small state average (0.36% in Warren, 0.47% statewide, subject to slight rounding error). Miami comes in even lower, at 0.31%. I do realize that ideally from the standpoint of an election-stealer, most of the votes would fall from Kerry to Bush, not to third-party candidates. Of course, Mebane and Herron were looking for signs of vote-switching in some precincts, and generally found not much.

[L. Coyote]

94. deterministic pattern based on precinct name

RE: "...followed a deterministic pattern based on precinct name."

This is written into law.

Precinct order overlay with candidates in alphabetic sequence.

001 = bBKdp
002 = BK...

etc.

Where d = Nader.

So, the law was not strictly followed as Nader was not a candidate!! Blackwell went to court to avert changing how everything was set-up after Nader was disqualified.

[OnTheOtherHand]

95. two concerns

One is whether we can be absolutely confident about how "alphabetic sequence" of precinct name is applied. For instance, I have two different statewide datasets; in one, every precinct in Warren has a geographical name; in the other, every precinct is numbered. I can see that the sequences aren't the same.

The other is that when I started to look at precinct sequences for the Senate race in Cuya, they didn't seem to follow the same deterministic sequence as the presidential race. Mostly they alternated, but there were breaks in sequence. Finally I gave up because I didn't have time to check all the PDFs.

We can try it, but how will we know whether we are wrong?

Yeah, the Nader/DQ business is pretty smelly.

[L. Coyote]

96. the statute specifies precinct number

Of course, such things can be manipulated, by lumping 2 precincts, arbitrarily combining precincts at location w/o regard to number order, changing boundaries, etc. etc.

RE: "...The other is that when I started to look at precinct sequences for the Senate race in Cuya, they didn't seem to follow the same deterministic sequence as the presidential race. Mostly they alternated, but there were breaks in sequence...."

This is a good thing to use for probability sorting. In locations where we know a Kerry vote would be switched to Bush by cross-precinct voting, what happens to the Senate race in the same precinct?? This is the kind of probability sorting that may unravel the anomalous voting patterns in the SW counties. It could also be used in Cuyahoga for the K-B vote-switch precincts if the down-ticket races produce an opposite switch direction in some down-ticket races and a same direction in others.

There is so much that can be done, and still needs to be done. Don't wait on the Dems, go after them or Just Do It....

[OnTheOtherHand]

104. yes, I know the statute

but since it is up to BoEs to number the precincts, it is an open question whether any statewide list authoritatively contains the sequences necessary to reconstruct the ballot orders. There has been some work on trying to pin this down.

I see no conceivable way that caterpillar crawl could account for the "anomalous voting patterns in the SW counties." Still no one has made a serious effort to demonstrate an anomaly, viz, a deviation from a normal baseline. That isn't to say that caterpillar crawl didn't affect the SW -- although as I've noted, the SW counties don't seem to stand out in Badnarik and Peroutka proportions, which raises some questions. It might make more sense to start with counties that do have high Badnarik and/or Peroutka (and/or undervote) proportions.

[L. Coyote]

105. caterpillar crawl could account for the "anomalous voting patterns

Cross-precinct voting will only be reflected in the 3rd-party votes if it is a random problem, rather than malfeasance.

We only see if we look!! Have you looked at the situation closely enough to know how many precincts shared locations, if they had different ballot orders, etc.?

In a malfeasance scenario, the vote-switching would be accomplished by counting with the wrong template, switching Kerry votes to Bush votes. Doing this will not move any votes into 3rd-party slots for President. At the same time, some down-ticket races could be reversed.

Kerry-to-Bush vote switching does not show, except in the final tally. It does not leave a trail in that particular race, except in statistical analysis the inference of shift is apparent in the percentages after probability sorting.

Random or directed cross-precinct voting vs. ballot tally malfeasance are very different beasts.

To those who are concerned about these SW counties, I must say LOOK FIRST! Do an analysis. Noone has commented on whether or not there are multiple ballot orders at locations. The most fundamental inquiry has not even been started in this regard, as best I can determine.

[OnTheOtherHand]

106. who are you lecturing, and why?

As I stated, no one has demonstrated anomalous voting patterns in SW Ohio, so it makes no sense to say that caterpillar crawl could account for them.

Mebane and Herron did an entire analysis looking inter alia for Kerry-to-Bush vote switching. You've studied it, I assume?

[L. Coyote]

97. RE: "Finally I gave up because I didn't have time to check all the PDFs."

At some point we need more organization in this effort. There's lots of duplication of effort, venting and ranting, but no leadership keeping a big picture organized, assigning tasks, etc. Nothing gets accomplished on a grand scale as a result, and little synthesis results.

This is why we have political parties. But they need leadership that understands these matters. That is evidently lacking.

[roseBudd]

54. To dismiss Connaly you have to believe SW OH voters are more likely to

vote for a name they have never heard. We never heard her name, never saw any ads, no yard signs, no bumper stickers. You needed a D cheat sheet to remember who she was. I tried to memorize her name but I drew a blank and skipped SC race.

[Febble]

63. Well, no.

The reverse is true. Think about it. You skipped the race. It looks as though others guessed.

on edit:

Sorry, that's not very clear. What I mean is that the less well known the candidates, the more you'd expect the pattern you actually see. If there'd been lot of Dem posters for Connally, Bush supporters SW OH would have known who she was, and the race would have been more polarized. But there wasn't, so it wasn't.

Think of it this way: If you are a Republican, and you know Connally is a Dem, then you are unlikely to vote for her. If you don't know Connally is a Dem, you might. It seems that they didn't know, so some of them did.

But the data seem to demonstrate that more people voted along party lines than didn't, but a subtantial proportion didn't, throughout the state. This could have been because the race wasn't highly publicised, and so a lot of people guessed, or wanted a change, or something not to do with partisanship. But whatever the reason, if stuff other than partisanship influenced the judicial race (and it clearly did), then what you'd see is exactly what you do see - Connally doing better than Kerry in Bush territory, and Moyer doing better than Bush in Kerry territory.

[roseBudd]

71. If you were a cop you would say "I smell a rat in SW OH", Connaly +

Warren County lockdown lies about FBI/Homeland Security which was preplanned and shared with only Rs + Clermont County stickers on ballots. It is all way too suspicous.

I was at the Warren BoE with Andy. The R in charge did not want us there, not withstanding Andy's charm. She was a bAtch.

Edited to add since I almost forgot:

Warren held the polls open lkong enough and managed to tabulate with no outside independent observation slow enough to post results after Cuyahoga the biggest most populous and most D county in OH. If you are going to cheat you need to come in after Cuyahoga.

Just like Hackett/Schmidt in '05 even though only part of Hamilton is in the 2nd, Clermont needed to delay until after Hamilton reported. Which is exactly what they did by claiming moist ballots in 91 precincts.

BTW Warren BoE has an R Susan and a D Susan. The preplanned terror alert memo was accidentally sent to the D Susan. Someone needs to interview her.

[Febble]

80. I smell rats too

Good sniffing. But I'm trained to find statistical rats, and I'm not sniffing those. Doesn't mean they aren't hiding somewhere (or using deodorant....)

[OnTheOtherHand]

74. the poll book thing needs more analysis

Baiman sees more votes than voters in some precincts, more voters than votes in others, and apparently regards it all as evidence of fraud favoring Bush. He could be right in any or all cases, but the reasoning and evidence are far from obvious.

For instance, he reports massive "phantom votes" in Concord South West and "disenfranchised voters" in Concord South East. But Kerry ran about 10 points better than Hagan 2002 (more like 20 on the margin) in Concord South East and 9 points better in Concord South West -- both comparable to Kerry's statewide improvement over Hagan. So it certainly doesn't appear that lots of Kerry votes were gobbled in one precinct and lots of Bush votes were stuffed in the other.

Something was wrong in Concord South West -- and presumably in Concord South, at the same polling place -- both of which reported bizarrely high turnouts. I think these both rank among the top ten precincts in reported turnout (out of over 11,000). It would be excellent to resolve that mystery.

Something strange seems to have happened in South East as well; facially, one possibility is that the absentee votes weren't included in the precinct totals (which would probably cost Bush net votes).

[Bill Bored]

56. Glad to see you guys talking.

Just a couple of questions before I get back to work:

Where's the link for this dropoff study? Is that one of Ron Baiman's? It seems like there were about 20 posts on this thread on navigation of Ron's links alone! I must have missed it on first reading. Sounds interesting.

One other question for you Febble:

I understand the theory that because Bush didn't do better than he did in 2000 in the NEP precincts with the Red shifts in 2004, this means the exit polls don't indicate fraud. But we don't really know whether he would have done better or worse in these precincts in 2004 (after 4 years of the worst presidency in history) than in 2000, do we?

So how do we know that he wouldn't have done even worse than he did the NEP precincts were it not for vote shifting? Just because he didn't do better in these precincts than he did in 2000 doesn't mean the red shifts were not due to vote shifting does it?

I haven't thought about this for a while (thankfully!) so forgive me if I'm missing or forgetting something. I'm just trying to poke a whole in the premise of this theory, which always relies on how Bush did in 2000 but not how he should have done in '04. I guess for us it just seems impossible that a guy who was relatively unknown in 2000, and actually lost to Gore, could have possibly done any better in 2004 after the country got to know the man and his "work" -- and that INCLUDES Republicans.

[Febble]

64. Well, it's not a question

of whether he did better or worse than 2000 as such. Even if he'd bombed in 2004 in the true vote, and even if fraud had failed to rescue him from his bomb, you'd still expect a better performance (or a less bad performance) from him in precincts where there was a big redshift than in precincts where the shift was blue - if the redness or blueness of the shift an indicator of fraud.

In other words it's the slope of the regression line that tells the story (or would, if there was one), not the intercept. UNLESS, as eomer has pointed out, all VARIANCE in fraud is due to polling bias, and only the INTERCEPT is due to fraud (i.e. fraud was uniform). But Ron is making the opposite argument, and saying it is irrefutable. :shrug:

I'm hoping OTOH will come in on the Connally story, as he has the data.

[eomer]

67. Unless...

the fraud causing the 2004 redshift also occurred in 2000. Or maybe it occurred in even greater amount in 2000 in some precincts, causing a negative correlation.

I had a discussion with OTOH while you were out a few day ago where I was suggesting that you could compare the change in redshift with swing rather than just redshift with swing. That way you would be comparing the increase in apples to the increase in oranges rather than apples to the increase in oranges. I even had a catchy term that was going to make me famous - "redshift swing" - but, unfortunately, OTOH said that the precincts polled in 2000 are not the same ones polled in 2004 so you can't do the comparison that I suggested.

I do think you've made an assumption that fraud in 2004 is new fraud and I think that assumption is a flaw in the logic. If the fraud is not new fraud then you wouldn't expect correlation with swing. So you haven't really ruled out fraud as a cause of the discrepancy -- you've only ruled out new fraud as a cause of the discrepancy.

BTW, a noteworthy event just occurred -- Bill Bored participated in an exit poll discussion (other than to tell us to cut it out). I just looked out my window to make sure it is not snowing in Miami in July.

[Febble]

68. Yes, and the sun is shining in a British summer

the endtimes are nigh....

Yes, I caught that discussion. I had in fact checked it out, but only at state level, as only state mean WPEs are available for 2000. It doesn't make any difference to the nationwide finding - the correlation is still zero.

But you do have another good point (though I will lay claim to it myself - I've been saying it for ages) that the swing shift correlation would only pick up NEW fraud/miscounts. So it wouldn't, for example, pick up differential residual vote rates associated with, say punchcard technology, where that has been going on for years, although interestingly, analysis of urban precincts by machine may indeed have picked it up - in urban precincts, (particularly black urban precincts) the discrepancy was higher on levers and punchcards than on DREs or optical scanners.

But if we are trying to explain the increase in precinct level discrepancy between 2000 and 2004 in Ohio, then that doesn't look as though it's correlated with benefit to Bush. But the statistical power is low (although the increase is large).

[eomer]

69. Regarding the lack of correlation at the state level,

I'm not sure what that finding tells us.

The thing about doing it at the precinct level is that the increase in WPE is a direct measure of what happened in a particular count. New fraud will directly impact WPE in a single precinct.

Doing it at the state level, though, is not a direct measure because you're comparing the results of one sampling of precincts with the results of a different sampling of precincts so you're really just looking at averages anyway. Subtracting one from another at this level does not isolate the change since the change in one place has already been aggregated with change, maybe of the opposite sign, of another place. There may have been new fraud in some precincts and a discontinuation of fraud (negative new fraud) in others.

Meanwhile, the swing at the state level will have a component that is true swing in the voters' intentions, which may have ups and downs that are orthogonal to the fraud component.

There is just too much going on here for me to be convinced so far. Feel free to elaborate if there is some reason I should be.

[Febble]

72. Well I compromised

I baselined precinct level discrepancies in 2004 to mean state WPEs in 2000. Not brilliant, but the best I could do, and it might have sucked up enough variance to allow a latent positive correlation to peek through. But it didn't.

But I think the thing is that if there was some stuff going one way and some stuff going the other way, at state level, it could have cancelled out, just through luck. As, possibly, in Ohio. The thing that bothers me is the complete lack of any hint of that correlation at nationwide level, because the statistical power is so great. If there really is exit-poll visible fraud in there, then it has to be systematically designed to fool the plot (or more be uniform than seems to me plausible).

Let me think some more about your post.

[Melissa G]

75. Did lightening just strike twice in the same place?BB on exit polls?

[roseBudd]

52. When I initially stared researching punch cards in Dec. '04 I assumed that

the precinct code was punched because I corresponded with a preelection tabulation testing citizen volunteer from Indiana. Contiguous states, punch card ballots, and who wouldn't assume that the existing technololgy, a punch means something wouldn't be used. It doesn't make sense to not use the punch unless you want to be able to feed the ballots behind the wrong header card.

Incidentally Fritakis found Warren County pink header card with just a Bush punch.

[AtLiberty]

57. ???

Have you hugged a statistician today? ;)

[Awsi Dooger]

108. Does anyone know exactly when that ballot rotation started?

I've seen references to 30 years ago but not a specific election cycle. I found one link to a Plain Dealer article on the subject, but it was a very old article and the link was dead.

The reason I ask is we're focusing on Ohio 2004, but in '76 the state was also very pivotal and much closer in the official vote. Carter beat Ford in Ohio by something like 10,000 votes among 4+ million cast. Ohio and Mississippi were the two states to determine the election, both called in the wee hours. Ford needed both of them for an unlikely win, and to take the electoral college while losing the popular vote, but Carter held on narrowly in each state, similar margin in Mississippi.

I was just curious if the ballot rotation was introduced before or after that election. Which side pushed for it, and the rationale.