Chaos in Cuyahoga? 49,000 Votes Disappear into the Ether ...then found?

[Iceburg]

Please provide some feedback comrades...

Chaos in Cuyahoga? 49,000 Votes Disappear into the Ether

Intro

In Cuyahoga, the ballot order was rotated and evenly distributed within reason amongst the 1436 precincts. We know that ballot order in and of itself is not a significant factor in predicting the outcome of a contest if the ballot order is rotated and evenly distributed. Further, the analysis of number of precincts versus % votes awarded returns somewhat mixed results (See Appendix A - % of Votes Cast versus Graph of Number of Precincts) . However, when we examine voting locations (and their precincts) from a "uniqueness of the ballot order" perspective, there appears to be a significant relationship between the two factors. The Cuyahoga data indicates that the greater the chaos (i.e. the greater the number of ballot arrangements within a voting location), the more "one candidate" appears to benefit.

To illustrate the ballot groupings consider:

Location A has 1 precinct and thus only 1 unique ballot order combination
Location B has 3 precincts but only 1 unique ballot order combinations
Location C has 5 precincts and 3 unique ballot orders combinations
Location D has 10 precincts and 4 unique ballot orders combinations
Location E has 8 precincts and 4 unique ballot orders combinations
Location F has 7 precincts and 4 unique ballot orders combinations

Therefore location A & B belong to the same class of ballot arrangements (1); Location C belongs to class 3, and Location D, E, and F belong to class 4.

Method

Using the Copperas & Cuyahoga BOE data (excluding the absentee vote), I constructed a cross tabulation of the votes cast by the number of "unique ballot arrangements within voting location". Since there were only 5 candidate slots there is a minimum of 1 and a maximum of 5 ballot groupings in any one vote location whether there are 2 or 25 precincts voting in that same location.

Assumptions made:

1. The field (in the Copperas data file:Ref#1) Street Address is unique to one and only one location (583 distinct addresses identified)
2. Ballot order (in the Copperas data file:Ref#1) is correct.

Findings and Observations

From Graph 1, we see that in voting locations with one or more precincts but only 1 unique ballot order across precincts, Kerry picks up > 75% of the votes cast for president.



In the 2nd group of Table 1 (2 distinct ballot arrangements) Kerry's returns drop to 71%. (Note: This group includes the now infamous Cleveland 4N and 4F precincts at vote location "BENEDICTINE HIGH SCHOOL at 2900 MLK JR DR " where Peroutka picks up 215 votes and Badnarik picks up 164 votes.)

Further, as the number of ballot arrangements increases within location (i.e. the chaos), Kerry’s returns diminish significantly (the drop is > 21% in this scenario with 5 distict ballot arrangements within vote location)

A Closer look at Homogeneous Ballot Order in Voting Locations

Table 2 presents the detail data for all precincts that have homogeneous ballot order within vote location. From voting location to voting location, the ballot order may vary but within this subset the ballot order remains constant within the specific voting locations.



The first group of votes cast (in 136 precincts), is the simplest case -- vote locations that have only one precinct.

The 2nd group in the table presents the votes cast for vote locations with 2 precincts, all with the same ballot order within vote location.

In the last group in the table, 2 voting locations have been identified that have 3 precincts each, but within each vote location the order of the ballots remains the same.

Thus, Table 2 represents the scenarios where the opportunity to "shuffle the ballots" i.e. mix up the ballots either intentionally or by accident is minimized. (Ballot stuffing and machine manipulation is still a possibility in all cases).

What if?

If the homogenous data from the 182 precincts in Ballot Class 1 above represents a more accurate picture of how citizens intended to vote in Cuyahoga, then we using the percentage votes cast from that group we can recalculate that Kerry would have picked up another 49,000 votes in Cuyahoga alone had ballot mix up not been a factor.

Conclusions

Number of unique ballot arrangements within voting locations appears to be a significant predictor of vote distribution. Further study is required to determine whether other influencing factors may be involved.

References

1. Cuyahoga Precinct Results: http://copperas.com/cuyahoga/cuyahogaofficial.zip
2a. Ballot Order: http://boe.cuyahogacounty.us/boe/ballots/
2b. Precinct Results: http://boe.cuyahogacounty.us/BOE/results/history/2004/EL52.txt
2c. Vote Locations: http://boe.cuyahogacounty.us/boe/PDF/votinglocations.pdf

Appendix A

[DoYouEverWonder]

1. Was Cuyahoga County the only place

that allowed multiple preciencts to vote in the same place?

One request,could you make your charts a little smaller? The thread is falling off the screen.

BTW: Great work. I love data.

[Iceburg]

2. Multiple precincts in vote locations are common in most counties

Thanks for the tip on chart size. I will try to edit before it is too late.

[DoYouEverWonder]

3. I would assume that they only do this

in the city centers? Not out in the suburbs?

[skids]

4. No, suburbs too.

Although it doesn't use punchcards, I looked this up for Pickaway,
and there were several precincts with shared locations. Not as many
as the city, but still, it occurs quite often in the burbs.

[DoYouEverWonder]

5. A couple of differences that might be worth considering

Is is true that only the suburbs had the newer electronic voting equipment? If used correctly it should be easier to handle multiple precinct voting.

Second, was this sharing of precincts normal for OH or was it a result of budget cuts and such in the last 5 - 10 years? I heard somewhere that was why the OH legislature kept the punchcard systems, because they couldn't affort the newer systems? Did they also cut back on polling places?

[jmknapp]

6. 'Burbs have more pcts per polling place on average

For example:







The pattern continues with 6-10.

[Iceburg]

7. jm -- your maps give rise to an idea on how to eliminate

any geographic or party bias with respect to examining whether ballot order uniqueness influences vote outcome...

If we did the analysis within City/Ward (as opposed to an entire county)in those cities and wards where there was enough data to merit a comparison.

As a first step, do you think you could plot the "ballot uniqeness" classes as you have done above with "number of precincts" classes?

[AirAmFan]

36. Joe--Do you have precinct tallies for "Disqualified candidate" for President?

I'm taking a bottom-up approach to get estimates of miscounted votes on another thread. Unexpectedly high votes for minor candidates and, most of all, "Disqualified", are tell-tale signs that voters may have used a "wrong" punch machine. But the EL52.TXT file on the BOE site just has a zeroed-out column for "Disqualified".

Do you have it up, or do you know where I could find it? Or have you never seen these counts?

Thanks.

[berniew1]

50. When are precinct exit poll numbers going to become available??

[L. Coyote]

8. Cross-precinct voting switched votes to wrong candidate

http://www.truthout.org/docs_04/121204U.shtml

Cleveland Paper Cites Voter Problems, Votes Assigned to Wrong Candidates, AP, 12/11/04,

"approximately 1,000 voters in the two precincts cast ballots just steps away at machines meant for the other precinct."

This incident was detected.

How much went unnoticed? Can this be manipulated to switch votes purposefully?

More links

[idealista]

9. amazing and brings many questions to mind

This is fascinating, and could be damning.

Unfortunately, if the ballots were mixed up in these polling places there is maybe no way other than statistics to determine voters' intent. If 49,000 votes were taken from Bush and given back to Kerry, it is most of what would be necessary to tip the election.

I had no idea that this mixing of ballot order could amount to such a huge windfall for the Republicans! Amazing, and horrible.

BIG QUESTION: Assuming that these multipe-ballot locations do not conincide with a different demographic (i.e. locations with more real Bush voters), WHY ARE THERE NOT MORE 3RD PARTY VOTES GENERATED BY BALLOT CONFUSION?

Is it possible that the order of presidential candidates was carefully planned so that Kerry votes would slip to Bush, not to any other candidate on the ticket? Is it possible to do that within the law requiring random rotation? It seems like there would have to have been either careful ballot planning or else the handing of wrong ballots to voters was only done in specific cases, not just randomly, and involved collusion of poll workers?

Thanks for this amazing piece of analysis.

[MarkusQ]

13. Do we have detailed information on the ballot order in each precinct?

Or can we get it?

Rather than "reasoning backwards" we could

1. Determine what "interesting properties" the actual ballot orderings had
   
2. Determine the odds of that happening "by chance"
   

I'm not sure how systematic errors could have been reliably introduced, but looking at what was done might give us the Ah ha! moment. Putting it another way, motive and opportunity are well established...let's dig into the means here.

--MarkusQ

[AirAmFan]

19. GREAT question. Here's the beginning of an answer, using ballot order

information from a directory of 1458 sample ballots on the Cuyahoga BOE website.

I'm calling Ohio's ballot a "caterpillar" because, at a given polling place, when you compare voting machines with each other, Kerry's name (and the names of the other candidates) may climb from one line on the ballot to another, like a caterpillar climbing from twig to twig on a tree branch. Using the wrong machine will result in a vote for a candidate other than the one the voter intends.

Ostensibly to comply with an obscure state law requiring "ballot rotation", to avoid giving any single candidate the advantage of being at the top of the ballot everywhere, Cuyahoga county does something so INCREDIBLY STUPID you'll find it hard to believe. The punchcards used by different precincts at a voting location are indistinguishable from one another, except for a precinct stamp that is supposed to be applied to the back of each card. But EACH PRECINCT'S PUNCH MACHINES ARE SET UP DIFFERENTLY! If a voter uses the wrong voting machine, her vote WILL BE MISCOUNTED!

Last Friday, the Cleveland Plain Dealer ran an article on this issue, at
http://www.cleveland.com/news/plaindealer/index.ssf?/base/cuyahoga/1102674912293811.xml .

In compliance with the "ballot rotation" law, it would have been just as easy to standardize the equipment to be used at each location, and to "rotate" the ballot by POLLING PLACE. But, instead, "ballot rotation" becomes a recipe for miscounted votes. Strange patterns of precinct results for minor candidates provide telltale signs that some voters have used the "wrong" machines to punch out their candidates' chads.

In principle, exactly the same "caterpillar ballot" problem may apply where optical scan machines are used rather than punchcards. Ohio has 88 counties. If other counties have implemented "ballot rotation" in the same manner as Cuyahoga, and co-locate substantial numbers of precincts on Election Day, then the "caterpillar ballot" may be responsible for miscounting tens of thousands or even hundreds of thousands of Ohio votes.

In addition to "innocent" errors caused by overcrowding, confusion, and lines of different lengths for different voting machines, the "caterpillar ballot" setup invites deliberate abuse. For example, piles of pre-stamped ballots could be switched from one table to another, wrong stamps could be applied AFTER voters have submitted their completed punchcards, voting machines could be switched to incorrect booths, etc.

--------------------------------------------------------------------------------

DETAILS:

According to yesterday's Washington Post, 70 percent of OH's voting machines are simple cardpunches. This figure includes all of the machines used in Cuyahoga County (Cleveland). According to the official list of Cuyahoga County polling places at http://boe.cuyahogacounty.us/BOE/PDF/votinglocations.pdf , only 136 of 583 polling places housed a single precinct. 1322 of 1458 precincts (more than 90 percent) voted together at 447 clustered polling places, where from 2 to 10 precincts were co-located.

For example, Precincts numbers 1806 ("Cleveland 04 F") and 1814 ("Cleveland 04 N") voted together at Benedictine High School on MLK Jr Drive in zipcode 44104. As of 11/30/04, official election results at http://boe.cuyahogacounty.us/boe/results/history/2004/EL52.TXT for this location were

Precinct B_ W_ K__ P__
Cleve4F 0 20 299 215
Cleve4N 164 12 334 10

where "B" is short for Badnarik, "W" means Bush, "K" stands for Kerry, and "P" stands for Peroutka.

As you'd expect in a polling place located on a street named after Dr. King, Kerry won both precincts handily. But, though Badnarik got no votes in 4F, Peroutka got 215 to Kerry's 299. And though Peroutka got only 10 votes in 4N, Badnarik got 164 to Kerry's 334. This is a very strange pattern. What in the world could be going on?

The answer lies in the specific ballot orders for these precincts. See their sample ballots at

http://boe.cuyahogacounty.us/BOE/ballots/pdf/CLEVE04F.pdf ---> KDPBW
http://boe.cuyahogacounty.us/BOE/ballots/pdf/CLEVE04N.pdf ---> BWKDP

"D" stands for "Disqualified", and is in alphabetical order between Kerry and Peroutka because it was Nader who was removed from the ballot.

The overwhelming majority of voters at this location chose John Kerry. To do so in precinct 4F, they had to punch a hole next to Kerry's name in the Presidential column on line 1. If they received a punchcard with "Cleveland 4 F" stamped on the back, but erroneously used Precinct 4N's equipment, they would have punched a hole next to Kerry's name on line 3. When the ballot was routed to the counting machine for precinct 4F, the vote would have registered for Peroutka. Not Kerry, not Badnarik, not Disqualified, and not Bush. This explanation is consistent with the observed pattern of results for 4F: Peroutka was credited with 215 votes, while Badnarik got none and Bush got 20. The "caterpillar" explanation of this vote would be that the votes for Bush and for Badnarik in 4F are accurate, but that virtually all of Peroutka's 215 really were miscounted KERRY votes.

Similarly, In 4N, you had to punch a hole next to Kerry's name on line 3. If you received a punchcard with "Cleveland 4 N" stamped on the back, but erroneously used Precinct 4F's equipment, you would have punched a hole next to Kerry's name on line 1. When your ballot was routed to the counting machine for precinct 4N, your vote would have registered for Badnarik. Not Kerry, not Peroutka, not Disqualified, and not Bush. This explanation is consistent with the observed pattern of results for 4F. Badnarik was credited with 164 votes, while Peroutka got 10 and Bush got 12. The "caterpillar" explanation of this vote would be that the votes for Bush and for Peroutka in 4N are accurate, but that virtually all of Badnarik's 164 really were miscounted KERRY votes.

This polling place had a configuration of ballot orders that is very simple to analyze, for three reasons.

(1) The underlying intent of virtually all voters who used this polling place is quite clear--there was near unanimous support for Kerry.

(2) The caterpillar explanation does not entail any Kerry votes being miscounted for Bush. And

(3) For each precinct there is only one type of "wrong" punch machine that could have been used in error.

Some other configurations are more complex, because one or more of the simplifying factors at Benedictine High are missing. But the principle of using strange patterns of votes for minor candidates to find miscounted votes is the same.

See also more details in a DU '2004 Election' thread at http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x147718

[MarkusQ]

21. Ah ha!!!! (?)

So, as long as this only happens in Kerry's strongholds, it could drastically reduce his support without any fancy math needed.

Were there multi-ballot polling places in Bush strongholds?

If not, why not?

If so, what happened in them?

--MarkusQ

[Iceburg]

46. You have crafted an excellent description of the transfer

[AirAmFan]

49. Thank you. What do you make of my 'back of the envelope' calculation

in post #47 on this thread?

Does it make sense?

Does it raise any questions in your mind?

[Iceburg]

51. I haven't got to it yet -- but I most ceratinly intend to.

I will get back to you later in the day. I greatly appreciate your comments and clarity of thought.

[Iceburg]

45. Yes we have the ballot order for each precinct from Cuyahoga's

BOE web site. Poster "jmknapp" published the data two days ago in excel. Without it, I could not have performed the analysis.

What I really would like to get my hands on are some official ballots and templates from Cuyahoga.

[L. Coyote]

23. Just pick up a voting machine and move it to the next table (precinct) ...

given you know the candidate order, just swap two machines with a specific order. The rest is undetectable history in "both" precincts.

So the analysis should be by locations, not precincts. How do the locations where multiple precincts with several machines per precinct compare with not in this category.

Two precincts with two machines each and you can neutralize the outcome, 2 precincts with 3 machines each and the shift will be less detactable.

Do any locations fit these parameters?

[imaginary girl]

43. I say this with all respect and sincerity to you, but ...

"Is it possible that the order of presidential candidates was carefully planned so that Kerry votes would slip to Bush, not to any other candidate on the ticket? Is it possible to do that within the law requiring random rotation? It seems like there would have to have been either careful ballot planning or else the handing of wrong ballots to voters was only done in specific cases, not just randomly, and involved collusion of poll workers?"

... Is it possible that the precincts selected for the 3% recount were not selected randomly even though it is required by law? ...

I think we all know the answer to the second one ...

[Iceburg]

52. The short answer to your question ....

"Is it possible that the order of presidential candidates was carefully planned so that Kerry votes would slip to Bush, not to any other candidate on the ticket? ...YES!!!

A carefully planned operation to switch ballots (some not all -- that would be to obvious) that minimizes risk of being caught would have to occur in the voting locations that had
a) multiple precincts
b) 2 or more unique ballot order arrangements
c) Kerry's relative position on ballot X equal to Bush's relation position on ballot Y

A,B, and C are necessary conditions for shuffling votes without getting into machine tabulation tampering and involving many staff. Although if machines are swapped out of a precinct (due to planned failures)that of course opens the door to another method of shuffling the vote and it is highly likely they used both methods. The bottom line is that without programmatic tampering of the machine tabulators one only needs to have the ballot read by the wrong machine, or the wrong ballot fed into the right machine. Swapping machines is just a special case of the ballot shuffle -- no re-programming is required.

Now back to the necessary conditions:
To minimize the risk of failure we require multiple precincts collocated in a voting location where at least 2 distinct ballot orders and they are ordered in such a manner that Kerry's position on ballot is interchangeable with W's position on another ballot.

As you know, there are 5 and 5 only ballot order arrangements in Cuyahoga:
1)BKxyz
2)zBKxy
3)yzBKx
4)xyzBK
5)KxyzB

Where B=Bush K=Kerry x=disqualified y=Peroutka z=Badnarik

Note that the relative candidate position to another candidate remains constant for all ballots in Cuyahoga County. Kerry's position is always adjacent to Bush although on the final rotation it appears that Kerry and Bush are separated, mathematically speaking they are adjacent because the series repeats itself.

Thus in voting areas with 2 distinct ballot orders, these 5 ballot order pairs qualify as easy switches:
..1.....2.....3.....4.....5
BKxyz BKxyz zBKxy yzBKx xyzBK
KxyzB zBKxy yzBKx xyzBK KxyzB

So if I give the Kerry voter
a) the wrong template, he punches for Bush.
b) the wrong ballot, he punches for Bush.
c) the right template & right ballot but direct him to the wrong tabulator, he in essence voted Bush
etc.
If halfway through the day, the tabulator for precinct X breaks down and is replaced by one that was programmed to read ballot order Y then again the Kerry votes are shifted to Bush.

In the above pairings we would see no change in votes cast versus ballot order arrangements to the third part votes.(That smells) Which is exactly what my graph illustrates.

So what we need to know in what locations were machines swapped out due to "malfunction".

Were any staff rotated from one voting area to another? I do believe that there needs to be at least one person involved in the skulduggery.

If ballot switching did occur, it is unlikely that the perpetrators made any effort to cover their tracks on the other races/contests such as the Ban on Gay Marriage issue. I have found some data that suggests the % votes cast do not jive with the bush% increases. More on that later.

[L. Coyote]

65. Ballot switching

You write;

"If ballot switching did occur, it is unlikely that the perpetrators made any effort to cover their tracks on the other races..."

This should be detectable given what happens in parallel for US Representative races, other races. These should also have a parallel shift. Since those races have different numbers of candidates/options, herein may be the proof sought.

Of course, with careful preplanning of ballot design in combination with machine placement, switching might benefit other candidates of the same party. How does the Kerry to Bush switch affect other votes in individual races/questions?

Delineating where this is detectable, where not might be difficult.

[Iceburg]

73. There may or may not be a parallel shift ...

I have not seen an actual ballot or template. Nor have I read the actual procedures for ballot preparation and subsequent counting. If the procedures only requires that the presidential candidates be rotated within a prescribed area on the ballot and all other parts of the ballot remain constant, then we may not see a shift in their respective votes ... that said, then the lack of shift in the remaining contests/races could potentially become an "shuffle" indicator. However, being a lowly rural Canadian, I am not up on the inside politics of Ohio enough to know for example that a substantial number of democrats would support the republican Senate hopeful as opposed to the democrat senate candidate. However, votes cast for the constitutional amendment to ban gay marriage may turn out to be a beacon. Greater than 11% of the precincts cast more votes as "NO" to the proposed amendment then voted for Kerry. That seems odd to me, but there could perhaps be a reasonable explanation.

So to answer your question "How does the Kerry to Bush switch affect other votes in individual races/questions?
I don't know at this point -- one would need a copy of the ballot card, template and the associated procedures before making any definitive statements.

[mulethree]

79. All the races rotate

So the order of senators rotate as well as the order of president.

But theres not a lot of correlation. None of the senator, state judges and such had a real big margin that can really be used.

I was trying to use the state board of education for this, #5 and #11. In both cases the races went roughly 60:30:10 so a precinct where the third candidate did much better that 10% can be used to raise a flag, but not very precisely since a lot of people, especially rebubs, don't bother to vote for their state board of ed.

[Iceburg]

94. Question: Is there a specific area on the card designated

for each contest and is it the same area allocation on all cards for each contest? In which case the first 5 slots would be reserved for the presidential contest, the next 3 slots reserved for the senate contest and so on. Then, within the designated contest areas the candidate names are rotated?

Or do the contest areas change their absolute positions, while maintaining relative positions with respect to each contest area?


Any clarification on this matter is greatly appreciated.

[AirAmFan]

98. How to vote on a punchcard machine, in OH2004 and FL2000 elections

The Columbus Dispatch ran this graphic the day before Election Day; see http://www.dispatch.com/election/election-president.php?story=dispatch/2004/11/01/20041101-B4-00.html

It mainly repeats the instructions on the first page of the twenty-some page template BOE staff were supposed to slip into the metal "pages" of every machine before they opened the polls. Individual PDF images of these template pages for each of 1458 precincts still are up at the BOE website.

For example, the templates for the Benedictine High precincts analyzed in post #19 in this thread are at the following URLs:

http://boe.cuyahogacounty.us/BOE/ballots/pdf/CLEVE04F.pdf
http://boe.cuyahogacounty.us/BOE/ballots/pdf/CLEVE04N.pdf

If you would download one or both of these files and print out at least the first two pages, you'd see what voters at that notorious polling place were SUPPOSED to see on Election Day. Then take another look at post #19 and the mechanics of Cuyahoga voting should become much clearer.

Four years ago, the Palm Beach Post was THE source for explaining the subtleties of punch-card voting. I wonder whether their online archive goes back that far. If it is, it's well worth exploring. What Cuyahoga used this year is very similar to what Palm Beach used four years ago. Except they had a 'butterfly' problem, while Cuyahoga had 'CATERPILLAR CRAWL'.

[Iceburg]

100. Greatly appreciated but I still am in search of an answer to this

particular question:

We know that the candidate names are rotated in each contest. But what I don't know is whether or not they are rotated through the entire length of the ballot or whether they are rotated within an area a specified area designated for that contest whereby the contest areas do not change just the virtual order of the candidates within the fixed areas.

The answer to this question is VERY important because if the names are rotated within a fixed designated area on the ballot for each contest, then taking a ballot marked for Kerry and reading it on a 2nd machine where Bush's position = Kerry's position would result in a extra vote for Bush and but MAY not affect the outcome of any other contests.

To illustrate, assume that
1) the 5 slots on the card represent the presidential contest
2) there have only 2 contests -- for president and for senate

Scenario 1
For President....For Senate
punch mark position *.................*
precinct A KxyzB.............dr
precinct B BKxyz.............rd

Bush gets 1 extra vote and the republican candidate for senator gets 1 extra vote

Scenario 2
For President....For Senate
punch mark position *.................*
precinct A KxyzB.............dr
precinct B BKxyz.............dr

Here, Bush gets 1 extra vote and the democrat candidate for senator gets 1 extra vote. It is possible with a lot of grunt work to identify every precinct that has a ballotorder such that the other contest remain relatively unaffected.

Another related point... the relative order of yes/no on the gay marriage ban issue does not change from precinct to precinct. And curiously, when the chaos is at its maximum state of disorder (i.e. 5 unique ballot orders within a voting area), the percentage vote for the "NO" group swells to a whopping 5 % (48% vs 43%), while Kerry's return in that same ballot group avg 53%!!!!! This fact runs contrary to all the data. Did this group of bush supporters suddenly become compassionate conservatives? ...I THINK NOT!!!!

[mulethree]

77. Would still be detectable if used in meanigful quantities

"Thus in voting areas with 2 distinct ballot orders, these 5 ballot order pairs qualify as easy switches:
..1.....2.....3.....4.....5
BKxyz BKxyz zBKxy yzBKx xyzBK
KxyzB zBKxy yzBKx xyzBK KxyzB "

Every one of these would result in 3rd party inflation. In #1 Kerry votes go to X, in 2-5 Bush votes go to Z.

#1 makes no sense for fraudsters, if you do it in a Kerry precinct then you get a large noticeable X vote. If you do it in a really strong Bush precinct then Kerry gains.

#2-5 You would use this only in places where the Bush vote is going to be extremely low such that it would not stand out when they show up as Z. These would show up as precincts with an abnormally low Kerry vote in relation to the polling place as a whole. Assuming an actual 5% Bush vote, you could transpose 20% of ballots before the Z numbers would start stinking and this would result in Kerry numbers being 19% below that of other precincts at the polling place. But there were polling places where Bush got really low numbers (in stand-alone precincts), these would be the places to look?

Cleveland 2,4,5,6,7,8,9,13
East Cleveland 2,3,4
WARRENSVILLE HEIGHTS 2,3
Highland Hills 0


Note that a) and b) are irretrievable, while c)(wrong tabulator) and (tabulator switched partway) could be retrieved.

In case c) you would have to hand sort them so the recount is done with the right scanner setup.

In case d) you could just re-count with a proper scanner setup.

[Iceburg]

44. You hit nail on the head idealista with your question ...

"WHY ARE THERE NOT MORE 3RD PARTY VOTES GENERATED BY BALLOT CONFUSION?"

The graph of 3rd party votes is flat ... ie this indicates that there is no correlation between the number of distinct ballot orders and votes cast by the 3rd party candidates. Further, if there was no strong correlation between Kerry's % of votes and number of distinct ballot his line of data points would be relatively flat too.

The reason ... high vote counts for these candidates would be to easy to detect (recall the early identification of irregularities in precincts Cleveland 4N and 4F at voting address 2900 MLK JR DR.

"Is it possible that the order of presidential candidates was carefully planned so that Kerry votes would slip to Bush, not to any other candidate on the ticket? "
The short answer to your question is -- yes!!! To prove or disprove this theory I will analyze the subset of voting locations with multiple precincts where the ballot orders arrangements is > 1 and Kerry's position on one ballot is the same as Bush ballot position on at least one of the other ballots in that same location.

[L. Coyote]

85. The Non-Votes (undervoting) seem to be in strongest Kerry Precincts.

Sorting Cuyahoga County precincts by standard deviation of percentage of non-votes, it is clear that the skewed precincts are in 85%-Kerry-voting precincts.



This is based on the raw data, certified (w/o the absentee votes).

This is not statistically normal.

Why is a separate question.

[donkeyotay]

10. Thanks for this work, Iceburg

Some other people are looking into some of the companies entrusted with giving a fair count. They, along with Blackwell, appear to be Republican in nature. If one tries out the hypothesis that our country's shift to the right correlates to republican control of computerized elections, how would one produce evidence?

I think the approach you've taken is the way it will eventually be proven that our elections are being unduly influenced by the people who control the technology. They might not leave a smoking memo, but they will leave statistical evidence that will form a pattern clear enough to convince people.

I think it's good work. THANKS!
:thumbsup:

[Straight Shooter]

11. Kick for iceburg's great work

:kick:

[smartvoter]

12. Big KICK

[Critical Thinker]

14. Intriguing analysis - very insightful.

Great work; I think this certainly merits further examination.

Nominated for homepage.

Edit: BTW, I think you could find a better title, more to the point - maybe something like "Cuyahoga confusion: did ballot shuffling lead to cross voting?" or similar. Just my 2 cents.

[texpatriot2004]

15. Good Job Iceburg, well done, great work...Thank you. n/t

[smartvoter]

16. You need to get this to Arnebeck. nt

[Dolphyn]

17. Hmm ... I wonder ...

what the results would be if we did the same kind of analysis in a county that is overwhelmingly Republican. If there's no such correlation in a Republican county, it might point to a deliberate attempt to "scramble the ballots" in Democratic counties.

[smartvoter]

18. All: This ties to the Caterpillar Ballot thread:

http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=147718&mesg_id=147718

[Emit]

20. Kick n/t

[Quakerfriend]

22. Very well done, Iceberg

Thank you for cont'g to hammer away at the data!!

[Iceburg]

53. I agree we needd to get our hands on a Republican county

and perform the same analysis on whether or not there is a correlation between % votes cast versus number of unique ballot arrangements. In theory (in a perfect world) there should not be one.

The lines in the graph would be more or less flat (as was the case for the 3rd party votes cast)

[MarkusQ]

124. Joe come through...

http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=177045#177188

[RaulVB]

24. Very nice job!

Send it to the authorities.

[MarkusQ]

25. Can you look at this in a Republican stronghold for comparison? (n/t)

[Iceburg]

54. I would love nothing more that to have my hands on

data from a Republican county using punch cards, ballot order rotation, with multiple precincts collocated at the same voting location.

If you have some or know where I can get it let me know.

[MarkusQ]

86. Where did you get the data you are using here? (n/t)

[Iceburg]

92. The data comes from jmknapps post ...

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x106743#117088

There is a link to an excel spreadsheet and links to Cuyahoga BOE results, ballots, locations etc.

All of these links plus a few more are in my original post above under the section "References"

[minvis]

135. Butler County Information

I have the information for Butler County with their precinct co-location sites, but I am still waiting from their BOE on the ballot order in those precincts.

Precinct by Precinct Vote Counts
http://www.butlercountyelections.org/txt/1104canvas.txt

Polling Place Locations
http://www.butlercountyelections.org/index.cfm?page=pollingPlaces

I have also run the numbers on Bush and Kerry Precincts won by Turnout percentage. I will post those later.

[Dcitizen]

26. I'M SCARED for what I see this statistics!

Higher unique ballots, lower votes for Kerry
Lesser Locations per precinct, lower votes for Kerry
Absolutely Non Sense.

Locations/Precinct
158/182 86.81
225/501 44.91
140/464 30.17
52/240 2.8
8/49 16.32

[mulethree]

30. Why isn't the 3rd party vote affected more?

1. The votes seem to be moving from Kerry to Bush as the 'confusion' increases. But if it is happening randomly - then the third parties should be picking up more votes, Kerry votes should be spread across all 4 (5?) of the other candidates.

2. Also, what factors affect the number of precincts in a polling place? Is it possible that pro-kerry areas tended to have fewer precincts at a poll and Bush districts tended to have more?


re: 1. To avoid the third party vote escalating, you could have it happening non-randomly. Precincts where a direct kerry-bush mapping occurred would be encouraged to mingle, while precincts where a mingling would inflate 3rd party votes, would be properly discouraged from mingling. But this would show spikes within the polling place, where some precincts had 'clean' kerry/vs bush preference and other - the 'transposable' precincts showed much less preference.

re:2. Simplest assumption to fit this, is that urban areas have fewer precincts per polling place, and rural or suburban areas have more precincts per polling place. This fits with 'suburbanites are used to driving everywhere' while urbanites insist on being able to walk.


I don't mean to put your work down, and do mean to check out these theories myself. I just hope to discourage the tendency we get here to go 'WOW! send 50 copies to kerry, conyers, arnback and so on before vetting and peer review can weed out problems.

[MarkusQ]

31. There was also an "empty" slot to suck up votes. (n/t)

[jmknapp]

56. Also, some precincts are so heavily Kerry

...that any Bush votes switched to Badnarik/Peroutka would not be noticeable. For example, if 100 random votes were switched in a 95% Kerry (inner city) precinct, 95 Kerry votes might go to Bush, 5 Bush votes to Badnarik--no red flag.

[L. Coyote]

76. Pattern in % undervoting = correlation to Kerry precincts

Statistical Summary for Cuyahoga County:
Percent Kerry Sorted by Standard Deviations in Percentage of Non-Votes.

stat.......... < 1.5 .......... > 1.5 ........ County wide

mean ......... 66.45 .......... 87.56 .......... 67.79
median ....... 64.24 .......... 93.96 .......... 65.40
stdev ........ 17.81 .......... 13.24 .......... 18.29

precincts ... 1366 ............. 92 ............. 1458

If the undervote was caused by fraudulent machine and precinct swapping, it was definitely targeted at Kerry precincts.

In the precincts with greater than 1.5 standard deviations from the county mean percentage of undervotes, the Kerry vote is over 21 % higher. What would this be without undervoting? And check the median differences!

[AirAmFan]

83. This post brings back memories of Florida 2000

There are other explanations for this correlation

(1) I remember some news reports four years ago about how the voting equipment in worst shape in Palm Beach County ended up in the poorest neighborhoods. Maybe it's because bureaucrats know they're less likely to get in trouble if they ignore complaints from poor people rather than from somebody who might be tight with someone "important". One wizened Riviera Beach resident was eloquent about how it was IMPOSSIBLE to punch out chads with the bent-tip styluses provided by the precinct on Election Day. People were using ball-point pens, toothpicks, hairpins, anything to try to remove those chads. Your vote was sure to wind up "spoiled" if you followed the directions, rather than take the ballot out of the holder to work on it. And many elderly lacked the eyesight and hand-eye coordination to do this.

(2) At the Congressional Quarterly stack in the library recently, I ran into someone who was looking for the same material I wanted. We chatted as we shared, and I learned that, not only was he a Republican politician's staffer, he had run as an unsuccessful Republican candidate for some local office. The subject of spoiled ballots somehow came up and he had a stock catchphrase he kept repeating, "TOO STUPID TO VOTE". Our conversation soon ended in awkward politeness.

[mulethree]

34. Done

Post #32 shows that you cannot transpose bush and kerry votes directly, and shuffle has to leave evidence in the form of third party votes.

Post #6 shows that the 'confused' polling places are mainly suburban, hence the increase in Bush votes is expected.

[Dcitizen]

37. Votes can spread to other parties only .2%

(1) 3rd parties can only get average .2% (1828, 1667/648,886 < .02%)
but the difference here is (75+71+66)/3 - (57+53)/2 = 15.6%+

(2) Further researches a/o geo experts.

[Iceburg]

57. Good questions ... See post #43

The scullduggery would have been invoked on a special subset of ballot order pairs. I will be conducting that analysis this weekend.

http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=166913&mesg_id=171612&page=

[thedutch]

27. HOLY CRAP!!

GREAT work Iceberg! what a horrible symmetry.

do this for every county in the state. I command it :)

[Iceburg]

58. Gosh ...If I only had the data ...I would accept your command

I am a lowly Canadian in rural Ontario -- we don't get out much. Thankfully posters such as jmknapp have made it possible for me to participate by posting some of the BOE data

[Critical Thinker]

28. kick! n/t

[truehawk]

29. Another Conlcusion: mutiple ballot orders made votes easy to shift !!

[mulethree]

32. No

A(Kerry)    	E	D	C	B
B(disq)  	A	E	D	C
C(Peroutka)	B	A	E	D
D(Badnarik)	C	B	A	E
E(Bush)     	D	C	B	A


There is no combination of ballots where you could directly
swap bush votes for Kerry votes.  
Take the first and second for example.  If you swap them, then
an "A" Kerry vote becomes an "E" Bush
vote, but the "E" Kerry votes become "D"
Badnarik votes. 

So any scramblage would show up as third parties getting more
votes. 

[bruised]

35. 3rd party votes

Pity. If you're right then this could not give Kerry as many votes as stated before. But there was an adormally high number of 3rd party votes in Cleveland. This could be the explanation

[MarkusQ]

40. It could if they went in the "Disqualified" slot. (n/t)

[truehawk]

38. Remember those 93,000 "spoiled ballots with no vote for President?

How many of those were Kerry votes that wound in the Disqualified slot?

[AirAmFan]

47. This is a CRUCIAL, insightful question, of statewide and national importance

The 93,000 "spoiled" figure is a statewide total. What was it for Cuyahoga alone? Do you know? The key point is that the proportion of "spoiled" that is "disqualified" gets multiplied by up to 5 by the "caterpillar ballot" phenomenon. If there were 20,000 statewide votes for "disqualified", Bush's lead could shrink by on the order of 100,000 due to "caterpillar" miscounts alone, and there still would be 73,000 BLANK "spoiled" ballots from which to mine pregnant and hanging chads.

IMO the most likely sources of "caterpillar ballot" miscounting are voting queues of unequal length within polling places that clustered more than one precinct. People in long lines at one precinct may have taken their punchcards to shorter lines for other precincts, where they would punch their candidate's name in the "wrong" space.

Since polling places in pro-Kerry areas tended to be overcrowded much more often than polling places in pro-Bush areas, the vast majority of miscounted votes would have been intended for Kerry, though some intended for Bush doubtless wound up being miscounted too.

Miscounted Kerry votes had to go SOMEWHERE, and there were only four possibilities: Disqualified, Badnarik, Peroutka, and Bush.

Miscounted Bush votes had to go SOMEWHERE, and there were only four possibilities: Disqualified, Badnarik, Peroutka, and Kerry.

Given a voter's intent and a miscount due to overcrowding, if all these possibilities were equally likely, the number of punches for "disqulified" at a given location is CRUCIAL. It can be used to derive estimates of the number of miscounted votes for Badnarik, Peroutka, Kerry, and Bush at those overcrowded locations. See messages #30 and after at http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x147718

In Cuyahoga alone, Badnarik got 1886 votes, including 164 at the Benedictine High precinct, and Peroutka got 1751 votes, including 225 at the Benedictine High precinct. If virtually all minor-candidate votes in Cuyahoga were miscounts, a good estimate of "Disqualified" for Cuyahoga would be about 1500, half the total excess vote for Badnarik and Peroutka. If virtually all the overcrowding leading to miscounts were in Kerry areas, an estimate of the number of Kerry votes given in error to Bush also would be about 1500. But such votes transferred from one major candidate to another count double in reducing Bush's "lead".

So a back-of-the-envelope estimate of the "caterpillar" effect on Kerry in Cuyahoga is 7500 votes, including 1500 votes for "disqualified". But Cuyahoga is only one of 88 counties. The statewide effect of caterpillar miscounts on Bush's "lead" easily could get into six figures!

[mulethree]

66. How many disqualified were intentional?

Nader got almost 17,000 votes in Cuyahoga in 2000. If people who insisted on voting for him, voted the 'disqualified' column intentionally.... I don't think you can automatically assume all 'disqualified' are errors or caterpillar droppings. Somewhere around here jmknapp had a county that was reporting the disqualified instead of zeroing them out, that might be a gauge as to how many Nader2000 voters voted Disqualified2004

Cuyahoga had
Ballots cast = 687,260
Pres. Votes = 673,729
Diff = 13,531

Would include disqualified, under votes, over votes and write-ins


The Badnarik votes in Cuyahoga, 1886 makes 0.28%
The Peroutka votes in Cuyahoga, 1751 makes 0.26%

We know several clearly shuffled precincts added artificially to these.

They got 0.31% and 0.21% in Franklin county with primitive electronic machines. And 0.26% and 0.21% statewide.

The Franklin ratios should? be unaffected by caterpillar, while the statewide ratios include the cuyahoga and probably some other caterpillar votes.

So I don't think you can automatically consider the majority of their Cuyahoga votes to be caterpillar droppings. I think you should aim at 500 instead of 1500. Making total caterpillar vote 2,500 instead of 7,500


Hmm, I don't know that Franklin was immune from caterpillar. If they rotated ballot orders, then the paper ballot label face sheets could be swapped between two machines that were programmed for different orders, and produce a caterpillar effect.

So I have stripped off multi-precinct polls and looked at the Badnarik and Peroutka ratios for {Cuyahoga where only one precinct at polling place} = 0.22% and 0.20% Multiplied by entire Cuyahoga vote = 1,500 and 1,352 respectively - so 400 fits even better than 500
and we have 400 * 5 = 2,000 confused ballots.

[AirAmFan]

82. Another valuable, very responsive post. Thank you! Have you uncov-

ered outlier tallies and/or a geographic concentration of minor-candidate votes, or use of "wrong" templates even at single-precinct polling places?

Your frequencies for Badnarik and Peroutka at 136 single-precinct polling places seem rather high. I was in the process of trying to remove outlier clusters from the Badnarik and Peroutka tallies, and then recalculating mean and standard deviation for Badnarik, Peroutka, and both minor candidates together. I expected to get at least down into the "teens", Your post prompts me to examine the single-precinct locations for outliers, too.

And thanks for anwering my question about spoiled ballots within Cuyahoga. Subsequently, I looked on the SOS website and found the following, which deviates slightly from your answer. Is my arithmetic faulty, or are you using prior-to-final tallies? We're not that far apart.

The total number of votes cast for president in Cuyahoga was 673,766, according to http://serform2.sos.state.oh.us/sos/results/2004/gen/pres.htm
But the number of votes cast overall in Cuyahoga was 687,260, according to http://serform2.sos.state.oh.us/sos/results/2004/gen/voterTurn.htm
This leaves a difference of 13,494.

[mulethree]

88. ..

The data I've been using came from here : dated 11/30
http://boe.cuyahogacounty.us/BOE/results/history/2004/EL52.txt
from
http://boe.cuyahogacounty.us/BOE/results/currentresults1.htm

It was supposed to be the final numbers Cuyahoga submitted, so I don't know If I have messed up some data or if it changed before the state certification. File is a bit painful to work with so I made spreadsheets a while back.

The Badnarik+Peroutka in the single-precincts is 0% for 28% of precincts. Then a pretty regular distribution between 0% and 1.12% for 66% of the precincts. The last 6% the chart knees up pretty sharp, but they're between 1.22 and 1.75% and we're talking about 4-8 votes per precinct for both candidates combined. One precinct Cleveland 13H had 4.71% but that was 85 votes overall with 4 for Peroutka. Peroutka_max=6 and Badnarik_max=5 I don't see any geographical clustering of the precincts with high B+P% scores, 7 different towns for the 10 precincts.

If the average for B+P% = 0.42% = 0.0042 = 2 votes in a 480 vote precinct. I suspect you could end up higher than the teens just from people who hit the wrong hole by mistake. The 1.75% upper limit of the single precincts is 4X the average and perhaps a better place to start looking, there are 36 precincts that exceed 1.75%

[AirAmFan]

90. Thanks for the minor candidate distributions. See post #49 in the other

thread, which was inspired by your very helpful critique using single-precinct polling places as a counterfactual.

There seems to be geographic clustering both for votes for minor candidates AND for spoiled ballots. Dividing the county up into 30 or more areas roughly similar in demographics might allow using the AREA's single-precinct data as a counterfactual for the AREA's multi-clusters, both for deciding how much of the minor candidates' tallies to treat as "caterpillar error", and for deciding how much of "Disqualified" to consider "caterpillar error".

It certainly seems reasonable not to have the same counterfactuals for all multi-precinct clusters.

Does this sound "kludgy" to you? I've convinced myself it's a considerably more efficient strategy, even though single-precinct polling places are rather rare in the data.

[AirAmFan]

99. The 'NOT-ALL-THE-WAY-DOWN' hypothesis for single-precinct polling

places.

In addition to the "caterpillar" problem, another pervasive problem may have led to miscounts, in single-precinct voting places as well as multiple-precinct polling places.

EMPIRICAL QUESTION: At the 136 single-precinct locations, what was the distribution of nonvotes by ballot order? The "not-all-the-way-down" theory would predict that, in pro-Kerry areas, the pattern with DISQUALIFIED at the top and KERRY at the bottom would have a significantly lower proportion of spoiled ballots than the four other patterns.

I remember from news reports about the Florida fiasco four years ago that one common problem with punchcard machines comes from not pushing the ballot card all the way to the bottom of the holder. See the graphic in post #98.

If the card is not ALL THE WAY DOWN, the voter may punch a hole for a candidate listed BELOW the candidate she wanted. ABOVE is impossible, unless the bottom of the card slot has been removed, or the voter has "scrunched" and damaged the card.

What would have happened if some voters unintentionally punched holes for the candidate listed IMMEDIATELY BELOW KERRY?

That Ken Blackwell is a sly devil. In all ballot orders except the one that has Kerry at the bottom, the voter would have punched DISQUALIFIED, because DISQUALIFIED (Nader) "just happened" to be the next name after Kerry this year in alphabetical order. Since, as the letter from BOE (post #67) points out, the counting machines were programmed not to even tally DISQUALIFIED, votes that disappeared (or WERE 'disappeared') via this route would have left no trace!

Tinfoil time: "Trickster" volunteers might even have shoved small pieces of cardboard down to the bottom of card slots to produce this problem intentionally, without the knowledge of precinct staff or other volunteers.

[mulethree]

106. how much tolerance do those machines have?

If it occurred naturally, with say - loose chads falling into the wrong place and accumulating so the card won't seat properly, then it would occur gradually over time, and perhaps be corrected one or more times during the day as a sharp-eyed voter realized the problem and cleaned, or insisted on a poll worker cleaning the mess out.

I wonder how much tolerance the card readers have for this? How far off the mark could the punch be before it fails to read or reads as the next candidate down the list?

If it was intentional I'd still imagine some sharp-eyed person might discover it and it would hopefully be reported and show up on some voting problem reports databases as blatant tampering.

But I guess I don't get your theorem. If Disqualified were at the bottom, then it would catch votes intended for the candidate above. If Kerry were at the bottom then shifted votes intended for Kerry would be below all the Pres candidates and either not read, or reading as a Senator vote? Producing under-votes for president rather than a vote that appears to be for another candidate?

[AirAmFan]

111. Sounds as though you DO understand the mechanism I'm suggesting

for miscounting votes, but not my proposed simple test.

Define a binary variable (call it KERRYLAST) that is 1 for ballot order 'DPBWK' and 0 for all other ballot orders. The question I want to ask the data is whether, in precincts where the vast majority intended to vote for Kerry, there is a significant difference in the percentage vote for 'DISQUALIFIED' by KERRYLAST.

Yes, votes intended for ANY candidate could wind up being mispunched for DISQUALIFIED. But where the vast majority intended to vote for Kerry, there would be a path to DISQUALIFIED where KERRYLAST=0 that does not exist where KERRYLAST=1.

There's an implicit assumption in my theory that, while ANY candidate listed below Kerry could get some of his votes if the ballot was not all the way down, the candidate IMMEDIATELY below him is much more likely to get his vote than the candidate two, three, or four positions down. We might be able to test for some of those farther-away misvotes too, but I'm assuming the adjacent ballot position below would get mmost of the action.

In strongly pro-Bush areas, the same 'not-all-the-way-down' mechanism could lead to miscasting votes for Kerry instead of Bush. But at least there would be an unexpectedly higher Kerry vote left behind by such mishaps. The disturbing thing about not-all-the-way-down in Kerry areas is that DISQUALIFIED was not even tallied. It is in a black hole, commingled with dimpled, pregnant, and hanging chads, smooth intact punchcards, and all other nonvotes.

We don't have precinct tallies of DISQUALIFIED yet, though Joes letter from Michael Vu leaves hope we may get such data eventually. In the meantime, total nonvotes can serve as a substitute.

The basic idea is that many fewer vot

[AirAmFan]

118. This topic 'not all the way down' now has its own thread

I googled and posted a news report on this very problem in Florida four years ago. Someone looked for one-off punches all across the ballot for people who probabliy intended to vote straight Democratic or Republican tickets.

New thread on "not all the way down" in Florida 2000 and Ohio 04

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x176358

[L. Coyote]

84. Non-Vote comparison - Cuyahoga County and Statewide

5,574,476 less.. 5,481,756.. = 92,720 = 1.663% Non-Votes State

602,048 ..less.. 591347..... = 10,701 = 1.777% Non-Votes County

Pre-certifid State stats.

[L. Coyote]

87. Non-Votes and Candidates % Statewide Sorted by Voting Machine

Statewide, punch card non-votes (undervoting) is near 2 percent. If E-Voting is a true baseline ....

[AirAmFan]

48. I disagree. "ballot rotation" is not a simple "swap". See post #19 above and

and posts numbers 30 and following at http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x147718

[mulethree]

69. I think we do agree

I said there's no simple swap, all rotation-related errors will produce 3rd party votes as artifacts.

From the posts you refer to, we seem to agree on that.

Note : you can avoid the smiley faces and align things easier if you check the "Check here if you want to format your message in plain text. Use for code snippets" Up at the top of the compose

[Iceburg]

74. I beg to differ, there is a subset of ballot order pairs that

make shifting very easy.

Thus in voting areas with 2 distinct ballot orders, these 5 ballot order pairs qualify as easy shift:

Precinct1 KxyzB
Precinct2 BKxyz

If I take ballots that have been punched for kerry in Precinct 1 and have them read on Precinct 2's tabulator, then the 3rd party voted is unaffected. However, I can not take a ballot from precinct 2 marked for Kerry and read them on precinct 2's tabulator without adding to the third party vote.

The entire set of pairs that can be shifted without altering 3rd party votes are:

. . . . ........1 ... . . 2 .....3 ..... 4 .....5
Precinct1 KxyzB BKxyz zBKxy yzBKx xyzBK
Precinct2 BKxyz zBKxy yzBKx xyzBK KxyzB

The term "swap" is a bit misleading as the transaction must be oneway between pairs or the 3rd party vote is affected.

[mulethree]

78. So you're assuming only Kerry ballots as opposed to random

You're assuming that whatever means of swapping it will only include Kerry ballots.

Not random - voter used the wrong machine swapping; but someone picked out a bunch of Kerry ballots and moved them to the wrong pile sort of swapping.

I don't see how to do it other than going thru ballots once they've already been punched, seperating out Kerry ballots and then moving them to another precinct. Then you need to replace them and such with ballots from another precinct of nearly equal number. Creating a 2-way or circular swap that moves kerry votes to bush in all affected precincts.

Pretty conspicuous though fingering through ballots and picking out particular ones. And it would be reversible by re-sorting the ballots back to their own precinct and then recounting.

[Iceburg]

93. I think the intent was to only shift the Kerry ballots

That said, I have to admit I am not at all familiar with the punch card voting process.

1. are the punch cards identical before being distributed to the precincts

2. when, how and by whom are the punch cards given their precinct identifier?

3. are the templates fixed in the precinct or are the templates handed to a voter by a poll worker?

4. do voters routinely check the precinct identifier before voting?

5. after voting do voters deposit their cards in a box or feed it into a tabulator?

6. is there a tabulator for each precinct or can a single tabulator read multiple precincts with variant ballot orders?

7. do voters sign a log book when they check in or after they cast their vote?

8. what procedures are invoked to secure the ballots, machines, poll books during and after the vote?

9. why did the Cuyahoga not publish the votes for "Candidate Disqualified" ...surely some votes must have been recorded for Nader even if it was by accident?

Until I know the answers to these questions and others it would be very presumptuous on my part to speculate on "how the shift occurred"

If you can clarify any of the above or point me to where the documentation exists I would be grateful indeed.

[AirAmFan]

80. EXCELLENT! Agreement is good. And thanks for the posting tip

I guess I misunderstood what you were saying. Thanks for reading my long post here, and for delving into my algebra-fest in the other thread.

Does what I;m donig on the other thread make general sense to you? I realize delving into minute details is tedious and a time-investment.

Your posting tip will save me from potentially disastrous experimentation in future threads. I had gone to VIEW -- SOURCE, found the HTML for a fixed-font post, and gotten the idea to insert (inside HTML code brackets) PRE CLASS="DCPLAIN" at the beginning of my post, and /PRE at the end. Your way is much simpler, and more likely to actually work!

[Iceburg]

59. Not another ... that is exactly my premise ...

[bruised]

33. great work - the exit poll Cuyahoga percentages could prove the theory!

Well done iceburg that was a brilliant analysis. You may have hit on something big. Your statistics provide a theory on how votes could be switched in punch card precincts. Not only is the correlation suspicious but the fact that Kerry's share dropped from 75 average
to only 55% in precincts "with 5 distinct ballot arrangements within the vote location -lets call them 5* locations)" is even more suspicious.
The reason I say that is first that the Ohio exit poll found that Kerry got an average of 65% in Cuyahoga. If as seems likely a good number of these 5* locations were in low populated low income areas - then also cnopsider the following those with income under $15K voted 73% for Kerry those under $30K 66 % for Kerry. Finally Afro-Americans voted 84% in favor of Kerry. (from the data on www.exitpollz.org/ohio_pres1.html -the 7.32pm update)

A tough job to put that all together but it could help to build a picture. This work can be certainly linked to those (who know Cuyahoga) who are looking at precincts with a surprisingly low Kerry vote -which I imagine is already being done by the activists in Ohio.
Great post- hope it gets to the right places.

[Scout1071]

39. Kickin' it up

[senegal1]

41. Very interesting! Kick.

[Iceburg]

60. Thanks Bruised. I will be further refining the analysis over the weekend

to hone in on the ballot order pairing that make the swap easier and less detectable. See post 52 (my answer to post 43)
http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=166913#171612

[L. Coyote]

42. Is your spreadsheet available? Kicking too!

[Iceburg]

61. I can email it to you or you can pick up jmknapp's data

(a link to an excel spreadheet)http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x106743#117088

I took Joe's excel spreadsheet and filled in a few of the ballot orders where they were missing. I then dumped all of the data into MSAccess and performed most of the analyis from there, since it is much easier to work with in database format. Keep in mind that it takes dozens of twists in the rubiks cube to synthesize the data into is final presentation format.

[L. Coyote]

63. I have the spreadsheet. Thanks

and do not use Access. I'm on a Mac. Thanks.

[lonestarnot]

55. CONSTITUTIONAL CRISIS!

CONYERS prepared to challenge Jan 6. We must be there! Make your reservations!

[Iceburg]

62. Followup: I thank all of you for the feedback and request some assistance

Your remarks and questions have been immensely helpful.

The analysis will continue over the weekend. My reply to post 49 pretty much explains where I will be going with the next level of inquiry.

http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=166913&mesg_id=171612&page=

Areas in which you can assist:

1) Identify which precincts had machine malfunctions and either had to be swapped out or repaired on site.

2) Find out whether any polling staff worked in more than 1 location on voting day (the names of the precincts are required)

3) Find out whether the Cuyahoga machine repairs & swaps were performed by one or more persons.

4) Identify voting locations that are right next door to one another

5) Find data for other counties using punch cards (preferably known republican areas) that have
.......a) rotate the ballot order in the same manner as Cuyahoga
.......b) multiple precincts co-located in voting areas
.......c) published the ballot order for every precinct

[L. Coyote]

64. Ohio statistical summary sorted by red / blue counties. Let's compare ...

with precincts at the individual county level.

This chart might provide some approximate baselines for comparison.



For those who want to crunch more numbers.

The Ohio spreadsheet.

The Cuyahoga County spreadsheet.

[L. Coyote]

89. Adding to this Stats Summary: the Statewide Non-Vote percentages.

Standard deviations:

All Counties:	1.02
Kerry Counties	0.68
Bush Counties	1.08

Means:

All Counties:	1.99
Kerry Counties	1.94
Bush Counties	2.03

Hey, the pre and /pre tags work!

[jmknapp]

67. Vote shifts to "Disqualified"

I sent the following email to Michael Vu, director of the Cuyahoga County Board of Elections:

I notice in the canvass results that your software automatically zeroes out all entries under "Disqualified," that is, the Ralph Nader position. There is no count in that column for any of Cuyahoga County's 1436 precincts. The documented ballot mixups (http://www.cleveland.com/news/plaindealer/index.ssf?/base/cuyahoga/1102674912293811.xml) would have resulted in votes in the Disqualified column in many cases. Is it possible to modify your software to correctly report the vote totals in the Disqualified column? In additional to abnormally high Badnarik/Peroutka totals, any Disqualified count at all would be an excellent way to identify precincts where the ballot mixups occurred.

Sincerely,
Joe Knapp
Columbus
Cc: Diane Solov, Plain Dealer
-------------------------

I got this in reply today:

Mr. Knapp:

Thank you for your email and interest in the Plain Dealer article. As you know, the tabulation system did not pick up any votes for the row listed Disqualified Candidate. Although it is possible to modify software to pickup the DQ Candidate row, we are currently conducting a recount of the presidential race and will not consider any change in programming to analyze that portion of the election.

We are currently analyzing the entire election and will be breaking all components of the election process in order to implement preventive measures, so this will not occur in future election.

Again, thank you for your interest if I can be of further assistance, please do not hesitate to contact me.

Michael
-------------------------

So it would appear we might get to the bottom of this eventually.

[Critical Thinker]

70. I got a chuckle out of his reply

"...(we) will be breaking all components of the election process in order to implement preventive measures..."

Say what?

I think I understand what he meant, but his sentence sure was poorly worded!

[jmknapp]

71. I know! Obviously a typo...

No doubt meant to say they are breaking their analysis up into a study of each component.

[L. Coyote]

75. Maybe, maybe not

Or the wording could be carefully chosen so as not to indicate anything with certainty.

And, they're already broken!

[L. Coyote]

72. Analysis of non-votes (undervoting) should reveal suspect precincts.

County summary:
Percentage of non-votes (undervoting) by precinct.

mean 1.83
median 1.48
stdev 1.43

The scatterplot of the Lat/Lon of these precincts shows a concentration. The mean location is 41.4855° N., 81.6360° W.

Precincts with greater than TWO standard deviation in percent of undervotes:

SORTED IN DESCENDING ORDER (12.2 stdev !! to 2.0 stdev):

CLEVELAND 6M
CLEVELAND 8I
CLEVELAND 19A
BEACHWOOD J
CLEVELAND 13H
EUCLID 3-C
CLEVELAND 5F
STRONGSVILLE 1-J
CLEVELAND 6J
CLEVELAND 15I
CLEVELAND 1D
CLEVELAND 6D
EUCLID 4-D
CLEVELAND 13X
NORTH ROYALTON 3-A
CLEVELAND 11R
CLEVELAND 9M
ABSENTEE VOTERS - 11-11-21
CLEVELAND 14F
CLEVELAND 7A
CLEVELAND 9R
CLEVELAND 13A
CLEVELAND 7U
CLEVELAND 12L
NORTH ROYALTON 5-B
EAST CLEVELAND 2-D
CLEVELAND 17N
CLEVELAND 4O
CLEVELAND 6U
CLEVELAND 2K
CLEVELAND 14J
EUCLID 3-A
ABSENTEE VOTERS - 11-12-25
CLEVELAND 8L
CLEVELAND 4F
CLEVELAND 5A
EAST CLEVELAND 4-G
CLEVELAND 7F
CLEVELAND 5K
CLEVELAND 5U
EAST CLEVELAND 2-F
GARFIELD HTS 1-B
EAST CLEVELAND 4-D
ABSENTEE VOTERS - 11-10-21
CLEVELAND 3K
CLEVELAND 16A

[L. Coyote]

91. Non-Votes equals number of Voters minus candidate votes

I realize I have been mixing terms. My stat "non-votes" is based on the total of voters minus the total number of votes for the four candidates. So non-votes includes hanging chads, the none-of-the-above crowd, the hypothetical too-stupid-to-vote crowd in some neighborhoods (yeah right), the why-doesn't-this-damn-stylus-fit-in-this-hole problem, and, of course, "caterpillar crawl."

[AirAmFan]

81. Thanks for posting this. On my "caterpillar ballot" thread, data on Dis-

qualified would be crucial, though mulethree has an excellent point in in post #66. See post #47 here for a boil-down of what's in the "caterpillar" thread, at http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x147718

The formula I've derived for a "bottom-up" estimate of miscounts at a clustered precinct, and the candidates they came from, is in posts numbers 30 through 46, especially 44, 45, and 46. What do you think?

[jmknapp]

95. great thread

I hadn't looked at it until just now (so many thrwds, so little time!). I'll read it through and post maybe tomorrow. Looks like you've done a lot of revealing mathematical dissection.

[L. Coyote]

96. Anyone reading this thread should also read ....

To get the full story on "Caterpillar Crawl" in the land of "Cuyahoga Chaos" follow the link below. I wish I had been following it earlier in this discussion.

Great clarity. The other thread:

Ohio's 'CATERPILLAR BALLOT': As bad as Florida's Butterfly?

http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=147718&mesg_id=147718

[goclark]

97. Kick for more awareness nt

[AirAmFan]

103. Thanks. I'd really appreciate your feedback. Do you know anything

about the HISTORY of Cleveland's implementation of the ballot shuffle? What they're doing seems UNBELIEVABLY STUPID! If they wanted to avoid giving any candidate without name-recognition the advantage of superior ballot position, they just as easily could have shuffled BY POLLING PLACE, rather than BY PRECINCT, avoiding "caterpillar crawl" completely. The current implementation just invites mishap and deliberate mischief.

Has this been going on for a long time? Or did it come in with HAVA, with Ken Blackwell, or with Gov. Taft?

[StephanieMarie]

68. kick n/t

[dlaliberte]

101. Correlate with precinct size

Excellent study.

Can we do a breakdown by precinct size? The reason I ask is that in the recount process, they selected precincts that had only 550 votes or more. See my previous post of a report from a recount coordinator: http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=175361&mesg_id=175434

Thus, I would expect to see very little variation in those larger precincts, and lots of smaller variations in the smaller precincts.
Did the polling places with more precincts also have smaller precincts?

[Iceburg]

102. Interesting report ... in particular, the statement about the poll books

"Anomalies were found. Almost all of the witnesses that I spoke with felt that the ballots were not in random order, that they had been previously sorted. There would be long spurts of votes for only one candidate and then long spurts for another, which seemed statistically improbable to most.

From what they were able to get through, witnesses found that signature counts were very much different from the official recorded number of ballots."

If bunches of "Kerry" ballots were moved at specific periods say during the day (as opposed to randomly), then what the recounters observed is what we would expect ...long spurts of votes for only one candidate, and signature count mismatches.

Now about your request for a breakdown by precinct size... what is it you have in mind --- % votes for candidates versus number of voters (i.e. votes cast)??

[EMunster]

104. link to...

http://bellaciao.org/en/article.php3?id_article=4815

[dlaliberte]

107. Size vs anomalies

The key thing to find is anomalies. If they chose precincts randomly among the set of precincts that had 550 or more votes cast, then I would expect there would be less anomalies in those precincts. But perhaps the precincts were selectively chosen even among that set of precincts, in which case we may not find any correlation based on size alone.

I was thinking that if the larger precincts (550 or more votes) were in polling places with few or no other precincts, these would have less chance of anomalies caused by swapping ballot books. So correlate the size of the precincts with the number of other precincts in the same polling place. Also correlate the size of the precincts with the difference in the number of expected votes for each candidate, if that can be determined. It seems that only the exit polls are a good measure of the expected votes for each candidate.

However, the anomalies found by a recount as compared to the official count would indicate either the tallies were tampered with or the ballots themselves were changed, but not both (because if the ballots and tallies are made consistent, and the tally is consistent with official tally, then a recount will not detect any problem).

And swapping of the ballot books among precincts at the same polling place will not be detectable by a recount, especially since the disqualified column is being ignored in the recount. Because the candidate that voters thought they were voting for was not what was recorded in the ballots and tallied, only exit polls will show a difference.

Statistical evidence, as in your study, shows that there must be some problem, but how can we get hard evidence? Looking at the number of votes for the disqualified candidate helps. If votes for disqualified translate to undervotes, that helps. What else can we look for?

[L. Coyote]

109. General statistics for Cuyahoga .. part of the answer is

# Voters, % Cast, and % undervote

mean	700.4	60.0	1.8
median	695.0	62.2	1.5
stdev	193.8	9.5	1.4

The average precinct is about 700 registered voters with 60% turnout.

PS. Pasting out of Excel between pre and /pre tags works!

[L. Coyote]

110. U write "if we compare the signature books with the tallies, we will find

In that post you state:

"My opinion: the election officials know that if we compare the signature books with the tallies, we will find substantial differences."

This is certainly an important aspect to detecting the rate of cross precinct ballot punching. All the stats in this thread have an underlying assumption that the number of votes and the number of voters recorded as voting in the precinct are in agreement.

Even if they are, the ballot crawl factor can be significant. If not, the margin of erroneous voting amplifies.

Another thread that ties in. The hive mind is getting too complex to keep up.

[L. Coyote]

105. Comparison of two 'Caterpillar Crawl" Precincts .. and the loser is ..

I wish I had Access to speed up filtering. This follows filtering for high percents of non-votes. Then filtering to multiple precincts at one location. These two cases illustrate the winners and losers, and the relative number of votes caterpillar crawl alters:



It is obvious who the winners and losers are.

In the first precinct Kerry loses 93 or less to Bush's loss of 9 or less. In the second case Kerry losses up to 24 to "disqualified" while Scrubya loses up to 4 to Badnarik.

No wonder the Chaos was focused on specific precincts. "It's the 10 to 1 ratio, stupid," could have been the Ohio Republican poll challengers mantra!

At some point a statwide "Caterpillar Crawl Margin of Error" needs to be calculated. Is it enough to throw out the election??

[L. Coyote]

113. In 80% of precincts, ballot crawl shifts Kerry vote to "d" .. AirAmFan ??

AirAmFan (Sat Dec-18-04 11:21 AM) writes in another thread:

"For every ballot rotation position except the one where DISQUALIFIED comes first and KERRY comes last, votes intended for Kerry well could have been mispunched for Disqualified. This is because it just so happens that Disqualified was the ballot postion immediately below Kerry in 80 percent of precincts!"

How was this arrived at? 80% of precincts at locations with multiple precincts?

[AirAmFan]

114. I'm just assuming the BOE's precinct ballot shuffle was unbiased overall

They allocated one of five ballot orders to each of 1,458 preciincts:

Badnarik, Bush, Kerry, Disqualified, Peroutka (BWKDP, or B for short);
Bush, Kerry, Disqualified, Peroutka, Badnarik (WKDPB, or W for short);
Kerry, Disqualified, Peroutka, Badnarik, Bush (KDPBW, or K for short);
Disqualified, Peroutka, Badnarik, Bush, Kerry (DPBWK, or D for short); and
Peroutka, Badnarik, Bush, Kerry, Disqualified (PBWKD, or P for short).

Notice that, in 80 percent of ballot orders (all but DPBWK), D is immediately below K.

Your post #105 shows that you have a database with ballot order in it. You could run a crosstab, to check how close the proportion of DPBWK frequencies comes to the theoretical 20 percent.

I can't run such statistics so easily myself yet. I'm still at the 'theory' stage, preparing an ASCII database to be read into SAS when I'm satisfied I know how best to structure a database to test theories.

[L. Coyote]

117. "You have a database with ballot order in it..."

An Excel file, based on Joe Knapp's original file linked from this thread. I just uploaded the smaller version, with only the certified results, sorted by ballot orders. I'm about to work on it. Here is the spreadsheet before I entangle it.

After reflection, I thought you meant that all ballots not pushed in adequately would vote down one candidate.

I have not used SAS yet. Systat will import other file formats, such as Access and Excel (at least one of these), plus others like tab delimited files. You should be able to make the steps easy by converting file types. Is anyone on this thread using Systat?

Access has powerful filtering mechanisms. They are easy to learn with the Help feature. Just create a "Query" to multiple parameters.

Excel has many statistical tests built right in. Just go Insert > Function > then in the dialog menu Statistical. The functions are described in the dialog box when you click them, and in greater depth in Help. Click the Help icon in the dialog box to go directly to the function you want.

[AirAmFan]

119. To minimize confusion, I've created a new thread for 'not all the way down'

Since virtually all the posts on this thread and the "caterpillar" thread deal with confusion at multi-precinct polling places, I decided the "all the way down" problem should be kept separate.

Considering the two problems together could become very complex very fast. For example a voter might make TWO mistakes. She might take a ballot stamped for one precinct to a machine set up for another precinct. And then, IN ADDITION, she might not push the ballot all the way down!

But let's say each problem had a 5 percent chance of happening. The chance that both problems would affect the same voter would be only 5 percent of 5 percent, or 0.25 percent. So treating the problems separately would not much affect the ability to do valid empirical work on each, and would avoid a great deal of confusion on the bigger threads.

New thread on "not all the way down" in Florida 2000 and Ohio 04:

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x176358

[AirAmFan]

116. Clarification: I'm assuming there could have been 2 separate sources of

mispunches, "caterpillar crawl" at multiple-precinct polling places, and "not all the way down" at ANY location, including single-precinct polling places.

To quantify and separate the effects of the two problems, single-precinct locations could be looked at JUST for "not all the way down". Then any problems at multi-clusters beyond the level seen at comparable single-precinct clusters could be attributed to caterpillar crawl.

I don't know at this point which problem may have been worse: 'not all the way down' or 'caterpillar crawl'.

[mgr]

130. Which direction are you looking?

Nice analysis, however, these caterpillar ballots need to be read in both directions up and down.

For the Euclid precincts: I count 103 miscounts for Kerry and 49 for Bush. For the Cleveland precincts which are more complex, votes swap between Kerry and Bush column and with 20 disqualified for Kerry, bush loses 1 at Badanarik.

If what I am seeing is correct, the vote swaps are even greater! Need to get affidavits as to voter intent in at least two precincts to clinch it. Should not be too hard in a 90% democratic precinct?

Mike

[Chili]

108. Iceburg, I can answer some of these questions:

1. are the punch cards identical before being distributed to the precincts

From what I understand, yes, the punchcards should be identical - in other words, a Kerry vote should always be in position "6" on the ballot itself. But the rotations on the ballot pages are different in the precincts (5 possible rotations, always the same order in the rotation, as I think somebody said above)

3. are the templates fixed in the precinct or are the templates handed to a voter by a poll worker?

Not sure what you mean by template. You mean the ballots into the machines? The ballot "books" are fixed at the voting machine, attached to the booth. In my precinct, poll workers were sliding the ballots in for the voters; in other precincts, voters were putting them in themselves, which allowed room for upside-down ballots, backwards ballots, etc. There's no room for error if the poll worker puts the ballot into the machine for the voter, there are tabs that have to match and fix the ballot into place at the top.

4. do voters routinely check the precinct identifier before voting?

Speaking for myself? I had no idea what a precinct identifier was before the election.

5. after voting do voters deposit their cards in a box or feed it into a tabulator?

Into a metal box. In some places, voters put it in the boxes themselves. In my precinct, you handed it to a poll worker, who then put it into the box.

6. is there a tabulator for each precinct or can a single tabulator read multiple precincts with variant ballot orders?

Don't know this one, but I doubt very seriously there was a tabulator for each precinct. But please don't take my word for this.

7. do voters sign a log book when they check in or after they cast their vote?

Yes, it is required. In fact, my poll book had my signature from the last election, from the last time I voted (2000). I had a little moment of anxiety about it, too, because my signature has changed since then, I've shortened it.

8. what procedures are invoked to secure the ballots, machines, poll books during and after the vote?

From what I understand, the ballots and books are locked in metal boxes identifying the precinct, then taken back to the Bd. of E. for the count.

Anything question I didn't list, I didn't want to guess an answer.

I hope this helps.

Also, you asked in another post about malfunctioning machines in Cuyahoga County? Here's a list of them:

http://shadowbox.i8.com/Suppression/ohio/ohiomachines.htm#table4

[AirAmFan]

112. Chili: Thans for this. Do you know anything about the HISTORY

of Cleveland's implementation of the ballot shuffle? What the BOE does with ballot ordering seems UNBELIEVABLY STUPID! If they wanted to avoid giving any candidate without name-recognition the advantage of superior ballot position, they just as easily could have shuffled BY POLLING PLACE, rather than BY PRECINCT, avoiding possible "caterpillar crawl" completely by standardizing machines within the same polling place. The current implementation just invites mishap and deliberate mischief.

Has this been going on for a long time? Or did it come in with HAVA, with Ken Blackwell, or with Gov. Taft?

[Chili]

115. I wish I knew the answer to that...

...but I'll try to look it up somewhere. Amazingly, I paid very little attention to any of this prior to November 3RD. I knew nothing of rotations. Even before the election, I didn't believe that we'd get screwed in Ohio. I was very aware of Blackwell's attempts, but the actual vote... I naively trusted in not only the power of a huge turnout, but in a belief that MOST of the officials in Ohio, even Republicans, were more honest than they're appearing to be. *whacks self in the head*

I'll never do that again.

If I find anything, I'll post it!

[AirAmFan]

120. Thanks for looking for more info. This 'ballot order' confusion

really takes the cake.

[Iceburg]

121. Chili ... I greatly appreciate your post

I thank you in advance , if I may impose on your time one more time ... I am seeking further clarification on the terminology
and the process. Explain it to me as if I'm a child ... I will not be offended -- it is VERY important that I get these details right.

Please comment on my assertions:
1) If the ballots are identical, then the precincts to which the ballot belongs is not punched on the card"

2) The "templates" I refer to in my post may be the equivalent of what you call "ballot pages"

3) The contest areas and the candidates areas on the ballot are constant in so far as how a central tabulator will treat them. That is to say, regardless of precinct or "ballot page" order, the nth position punched on the ballot will always be treated as a vote for Kerry.

Once again Chili, thankyou!!!!!

[Chili]

129. Iceburg, I lost track of this thread - sorry!

I'll try to answer these to the best of my knowledge, but if anyone knows better, please chime in:

1. There MAY be holes punched at the bottom identifying the precinct, but I wouldn't swear to that, I just remember reading that somewhere. However, there definitely is a stamp on the back of the ballot with the precinct name in clear blue ink (the ballots are manila in color). It will read "Cleveland 5-T" for instance, right in the middle of the ballot on the back.

2. Then, yes, we were talking about the same things... the devices that the ballots are actually slid into and fit into place by the tab holes in the top stub (each ballot has a 3-inch or so stub at the top, which is removed)

3. I believe that's right, that's what I meant about Kerry always being in the 6th position.

You're very welcome, sorry I didn't see the thread again sooner...

[Iceburg]

137. Interesting ... a) if there is only a precinct ink stamp on the back and

no punches for precinct then putting it in the right box (initially) is up to the voter or the poll worker, since there is nothing readable for the tablulator in terms of precinct identfier. In which case, only a hand recount could resolve the issue.

b) If there are holes punched to identify precinct AND there is a stamp on the back, then it would be possible to determine whether any on the cards were stamped incorrectly. Again this would involve a hand recount by stamp, to which one would compare a machine recount.


Question Chili: Do you know whether or not there is a serial number on the ballot? If so how is it allocated? Are precincts alloted a certain range of serial numbers for future auditing? (On our paper ballots in Canada, the ballot has a serial number stamped on it twice -- one on a tear of stub. The poll worker takes the serialized stub and the voter places the ballot in the ballot box.)

[Chili]

149. okay, now that it's over, I'll admit it...

...I was "OhioWitness" the other day, posting about the Cuyahoga County recount. Since I didn't have a lot of nice things to say about the election officials ("random???"), I thought it best not to say anything until I was sure I wouldn't be going back, so I posted under another nic.

Anyway... I don't think there is any serial number on the ballots. There may have been some kind of code that I didn't detect, or that they didn't tell us about. But I didn't see anything obvious - however, there HAS to be something that ties that ballot back to the person who voted with it. There has to be. And if there isn't, there SHOULD be. In fact, we should be required to sign the ballot ITSELF as well.

But that's for the reform...

Anyway: right, about the voter and the pollworker. I wouldn't have known what the correct box should be for my ballot. And where I voted, I couldn't tell you if the booths were sufficiently separated according to precinct. Could I have voted at the wrong precinct? Maybe so. Where I voted, the poll worker walked me to my booth and put the ballot into the template. When I was finished, the poll worker put it into the box (it was on the same table where I signed the pollbook, so there SHOULD be no margin for error there... in MY polling place. But polling places did different things, there is no standard... and if there was, it wasn't adhered to).

The part about the stamp on the back of the ballot: no way for US to tell is the ballot was stamped incorrectly. But then, if each ballot is the same, it should've matter beyond making sure the ballot stamp on the back matches the template its slid into. Does that make sense? During the recount, if a ballot from another precinct found it's way into the stack, we would've seen it on the back, the stamp wouldn't have matched. But you're right - only a hand recount would catch that.

I hope this helped...

(PS: I didn't get a chance to see any of the polling books, our precinct was so large we were the last to finish counting, and the pollbooks had been looked at already - but not all of them)

[AirAmFan]

138. This is a really IMPORTANT detail: WHERE was the voter supposed to put

the ballot once she'd finished removing the chads?

If there were a "lock box" next to each punch machine ('Deposit Ballots Here'), and ballots for each separate precinct were stored separately, then at some point staff very likely would have discovered the "caterpillar" ballots. Some ballots would have been seen to have the "wrong" stamp, unlike all the other ballots from lock-boxes for that particular precinct. In this event, depending on how ballots were stored after counting, IT STILL MIGHT BE POSSIBLE FOR HAND RECOUNTS TO IDENTIFY ALL MISCOUNTS AND ROUTE THEM TO THE PROPER CANDIDATE!

But I suspect that in any case all ballots from Cuyahoga now have been commingled and SORTED SO THAT RECOVERY OF MISCOUNTS IS IMPOSSIBLE.

I also surmise that ballots were not originally sorted by the precinct where they were punched but rather commingled in one pile per polling place from the start, or gathered by the precinct where they were issued, so that it never was possible for the miscounts to be reversed. In the future, if BOE refuses to rotate ballots by polling place rather than by precinct, then separate 'lock-boxes' at each punch machine, and storage of ballots by lock-box number until time for hand recounts has expired, would be a second-best solution to the "caterpillar" problem.

One remaining detail of interest: Didn't each voter get an envelope along with the ballot? Was this envelope stamped with name of the precinct where it was issued?

[Chili]

151. yes, we got an envelope...

...and that's where I THINK I saw the number that might've matched the one next to my name - on the envelope. I'm afraid I wasn't as aware of what I should be looking for, I wish I had been more observant.

And in my precinct, one box served maybe 5-6 voting booths, and we stood there as the poll worker put the ballot into the box.

[Dancing_Dave]

122. Reminds me of a funny Al Camus piece against corruption

In French Algeria in the '30's. One of his lines describing an election there was "Parts of Algeria exist in a state of amazing chaos."

[L. Coyote]

126. kick n/t

[L. Coyote]

123. Number of ballot orders at each location could be manipulated to skew...

the results. I've started looking at locations for number of precincts, number of ballot orders, and which ballot orders are placed together. Here is an example location:

				b	Bush	non	Kerry	p

6501	PARMA HTS A	BKdpb	1	279	10	379	2
6511	PARMA HTS K	BKdpb	1	157	7	209	2
6522	PARMA HTS V	KdpbB	3	168	3	205	0
6512	PARMA HTS L	KdpbB	1	170	5	219	1
6502	PARMA HTS B	KdpbB	0	138	6	189	0
6518	PARMA HTS R	dpbBK	1	196	2	162	0
6503	PARMA HTS C	dpbBK	0	192	3	218	1

5 of 7 can change votes from Kerry to Bush, the last two cannot. Even though there are 7 precincts, only 3 ballot orders are used instead of 5. This can arbitarily delimit which votes can be reassigned to which candidate.

[Straight Shooter]

125. Kick for the Sunday a.m. insomniacs

:dem:

[Iceburg]

127. I'll never get any sleep ...

[L. Coyote]

128. kick n/t

[mgr]

131. Graph 2, number of polling places?

Nice study.

Graph 2 might be more meaningful if the number of polling places with the various numbers of precinct ballot was included. It might also be interesting to map the locations as well versus party registration.

Mike

[jmknapp]

134. Maps

These don't show Bush/Kerry leanings, but you can see the distribution of precincts for 1, 2, 3, 4 and 5-order polling places:

[L. Coyote]

139. Some Cuyahoga ballot order statistics

All numbers are average percentage.
First column is number of ballot orders.
Last row is all precincts.

       Kerry  Non-vote	 Cast	   Badnarik	Peroutka

5	53.03	1.10	 61.34	   0.19		0.09

4	57.28	1.25	 63.60	   0.46		0.22

3	65.84	1.74	 60.97	   0.27		0.23
	
2	70.02	2.04	 58.48	   0.35		0.38

1	74.31	2.22	 56.61	   0.21		0.21

All	66.49	1.80	 60.01	   0.32		0.27

It seems that a precinct-by-precinct analysis will reveal the true number of cross-precinct votes lost to "disqualified."

[Iceburg]

164. LC, I have all the ballot orders plus a new findings ....

Subject: LC, I have all the ballot orders plus a new findings ....



I provide a bit more detail in post 158 on the sources and sinks
in thread http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=166913&mesg_id=184692&page=

Now that I have all the souces and sinks identified in my data set, I am comiling and comparing my results. Even though I have a hunch one candidate made his greatest % gains in the Class 5 Ballot Order arrangments , I'm sticking with Class 2 for now is that is where the largest number of votes are (and the easiest to analyse ;-)

Don't cha wish ya had a nickel for every hour you've spent on this?
Democracy is neither easy nor cheap.

[L. Coyote]

190. More indications of where to analyze.

I like this type of graph for this pattern.

You write: "I'm sticking with Class 2 for now is that is where the largest number of votes are (and the easiest to analyse ;-)"

I've figured out that it is not the place to look, at least not in Cuyahoga. Check out the stats in my 189-190 posts. The place to look is in the 3 and 4 groups. That's where the points are being shaved.

The skew of Badnarik, peroutka, and non-votes is a great indicator. And the percent tables tell all too.

We need to look at precinct voter and registration numbers in relation to ballot sequences. We know that crawl 1 & 4 is where the highway robbery takes place. What are the relative proportions of the potential cross-overs wherever this combo occurs? Is that what skews the data?

Since one precinct is voting Bush by mistake while the other is voting Kerry by mistake, which machine would you put in the larger precinct?? And with one precinct larger than the other, is there a tendency to cross-over??

[L. Coyote]

140. Distribution of Ballot Orders in Cuyahoga-Sorted by Ballot Number

The numbers of each ballot type allocated to locations.
Sorted by the number of ballot types per location.

	5	4	3	2	1	Total

bBKdp	9	47	85	115	31	287

BKdpb	10	46	99	102	31	288

KdpbB	11	49	90	101	35	286

dpbBK	10	47	94	87	49	287

pbBKd	9	49	97	98	31	284

They are not evenly distributed!
Does the last column meet the "letter of the law" criteria?
Or is this fraudulent?

Note the 49 ballot types ending in Kerry at locations with only one ballot type, the highest number, compared to 87, the lowest number in the 2 types column.

[AirAmFan]

142. The last column is the one that really counts, and that looks random

It looks as if the BOE performed "unrestricted" randomization, or at least were smart enough to make sure your last column was 280-something for every ballot order.

You haven't really shown why we'd expect the other columns to be as even as the last column. "Number of ballot orders" is entangled with cluster size in a non-random way, by construction. For example, it's impossible for single-precinct polling places to have more than one ballot order, so they dominate your next-to last column. This column borrows a few observations from clusters of other sizes. Most likely, most of the single-order clusters in addition to the single precincts are of size two. Among the size two clusters, BB, KK, etc are more likely than are BBB, KKK, etc among size three clusters. These in turn are more likely than BBBB, KKKK, etc in clusters of size four, and so on.

If you'd post the same table, only substituting CLUSTER SIZE for NUMBER OF BALLOT POSITIONS, we'd see whether the distribution of ballot orders within each group of polling places of the same size was even. If it is, then your columns cannot be even, because they move observations in one even column of the cluster-size subgroup table to another, in a systematic fashion. Think of the table you've posted here as a revised version of the table I'm suggesting that you run. If you start with even columns, and then move observations from one column to another in a non-random way, you can't expect to wind up with even columns, except for the last one.

Let me be very specific. Define a variable for distribution of cluster size, taking value "5" for clusters of sizes 5-10, "4" for clusters of size 4, "3" for 3, "2" for 2, and "1" for 1. Then substitute this column for number of ballot orders in your software and see what you get. If the columns in that table are even, then you wouldn't expect these columns to be even.

[jmknapp]

143. The ballot orders are treated equally

And there should be no association with cluster size or any other factor. I would think that the distribution in Coyote's table should be even. It's hard to argue I think that a particular order should be favored in any particular grouping.

The main question would be is it just random fluctuation or is it statistically significant. Taking coyote's table, there are 177 single-order cases, so each of the five orders should average 177/5 = 37.4

The standard deviation would be sqrt(177*0.2*0.8) = 5.3

So the observed value of 49 would be (49-37.4)/5.3 = 2.2 standard deviations above the mean. That's significant. Whether it's significant in terms of influence on the results is another question.

[AirAmFan]

144. Before I answer this post, let's return to a related issue: Did you see my

reply -- at http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=147718&mesg_id=178809&page= -- to your post about the need for Cuyahoga neighborhood demographic data? Has anybody made any progress on this, or come up with a plan?

in linking neighborhood demographics to precinct clusters, what IDs would be used for precinct clusters? I've used the cardinal order of the cluster in the BOE file at http://boe.cuyahogacounty.us/BOE/PDF/votinglocations.pdf . Thus 001=BAY_VILLAGE_1-A 583=WOODMERE_A. There was one precinct in EL52.txt that matched no location in the other file. Thus I set 000=CHAGRIN_TWP_A.

[jmknapp]

146. Yes, but the idea of a manual process for 1436 precincts

...is somewhat daunting. I did find something pretty good though: a web site that lets one enter a geographic coordinate (lat, lon) and a radius in miles and it outputs the estimated black % in that radius. The site is http://mcdc2.missouri.edu/websas/caps.html

I automated a little script that hit that web site with the lat,lon for each of the 1436 precincts, for a radius of 0.5 miles around the center of each precinct. This isn't perfect, since precincts can have some pretty odd shapes, but it's probably as good as any other option.

An updated spreadsheet with the estimated black % in each precinct is at http://copperas.com/cuyahoga/cuyahogaofficial.zip

Hope you can use!

Here's one chart of interst I made from that data:



Pretty interesting how that works out. I guess it goes along with the observation that there tend to be more precincts per polling place out in the suburbs?

[AirAmFan]

147. Thanks! Very helpful. Do you have the latitude and longitude of each precinct up

in that file, or in another file? Those variables for each precint or polling place would permit linking to the CANDO database directly, and get not only current demographics but demographics from the 1990 census to identify declining and yuppified neighborhoods. And programming the link in Access or even in Excel would not be difficult.

I suggested downloading all 94 canned profiles only as a timesaver on going through the CANDO documentation and finding variablenames for what they consider their best neighborhood measures.

Thanks for the quick reply. I've got to work on something else now, but I'll get back to you by midnight tonight on the statistical dependence of ballot order and number of ballot orders.

[jmknapp]

148. Yes, lat/lon data are in the file linked above n/t

[AirAmFan]

157. Joe--How would you suggest finding the census block group closest

to your coordinates? I have all the linking info, but I want to double-check the definition of "CLOSEST".

Least sum of squared deviation between latitutes and longitudes, ie

MIN{(LAT(Census) - LAT(JMXLS))**2 + (LONG(Census) - LONG(JMXLS))**2} ?

Is a degree of latitude the same as a degree of longitude? Or am I missing a coefficient.

Here's how I'd link your coordinates to one of CANDO's neighborhoods:

http://povertycenter.cwru.edu/doc/ngctdir/mlist00.txt links census tract.block-group to one of CANDO's 94 Cuyahoga neighborhoods

And http://www.census.gov/geo/www/cenpop/blkgrp/bg_cenpop.html
links latitudes and longitudes of centers of population to census tract.block-group (TTTTTT.B):

"Centers of Population by Block Group

This page contains links to Census 2000 Block Group centers of population. The files are in comma delimited ASCII format, with one file for each state. The record layout is as follows:

ST: The 2-digit State/State equivalent entity FIPS Code.
COU: The 3-digit County/County equivalent FIPS Code.

TRACT: The 6-digit Census 2000 Census Tract number
BG: The 1-digit Census 2000 Block Group number

POP: The Census 2000 Block Group population.

LAT: The latitude coordinate of the center of population for the Block Group

LONG: The longitude coordinate of the center of population for the Block Group"

[L. Coyote]

156. Ballot Order, Machine Distribution, and How to Fix an Election for Dummies

Well, maybe for statisticians, but that's not as catchy! It may be difficult to follow even for not-so-dummies. Bear with me.

As posted before, remember the percent Kerry sorted by standard deviations in Percentage of Non-Votes is:

stat	< 1.5	> 1.5	County wide

mean	66.45	87.56	67.79

This made me suspicious.

There is no combination of ballots where you can directly swap Bush votes for Kerry votes across both precincts in a two ballot order location. However, if Kerry and Bush are in the same position on two orders at one location, cross-precinct voting will switch their votes. So, basically you want cross-precinct voting to occur where your opponent has the highest support. If 90 percent are voting Kerry, for every 9 votes switched from Kerry to Bush, 1 vote switches from Bush to Kerry. (Anyway, that's when both precincts cross over equally. If you can get just one precinct to cross over, say by shorting just that precinct of equipment for example, the equations change dramatically.)

With only two ballot types at a location, if Bush and Kerry swap votes in one position, from their other positions Kerry and Bush votes go to Badnarik and Peroutka.

b B d K	p --- b	B d K p	
d K p b B --- K p b B d 

These are "crawl 2" and "crawl 3" scenarios. In the first instance the lower ballot order has crawled 2 positions, the next 3. In 90 percent Kerry precincts, for every extra Badnarik vote, Kerry lost 9 and Bush gained 9.

In precincts with a 50/50 support split for Kerry and Bush, they lose equally and the third party candidates gain equally. No problem. Disqualified has no real role in the two ballot scenario that swaps Bush and Kerry votes.

Disqualified can get candidate votes from "crawl 1" and "crawl 4" scenarios as follows:

b B d K	p --- b	B d K p
B d K p	b --- p b B d K	

In these two crawl scenarios, Kerry losses votes to disqualified and Peroutka, Bush losses to disqualified and Badnarik. No Kerry votes go to Bush, no Bush votes go to Kerry. This should be detectable as equal skews in disqualified and Peroutka.

So, which "crawl" would you place in the 90 percent Kerry precincts if you had the opportunity to allocate voting machines and you wanted to skew the vote? Placing as many "crawl 2' and "crawl 3" machines as possible the two ballot type locations with highest Kerry support would shift votes to Bush. The other two deprive Kerry of votes, but do not switch them to Bush. Maybe this is for Dummies after all!!

Given this understanding (which I did not clearly communicate), I wanted to know more about the distribution of ballot orders. I found this to be the county wide pattern:

#	Ballots	Count
10	5	1
9	4	1
7	5	1
7	3	1
6	5	2
6	4	3
6	3	1
5	5	4
5	4	22
5	3	6
4	4	28
4	3	26
4	2	6
3	3	106
3	2	43
3	1	2
2	2	176
2	1	18
1	1	134

Here we see that there are 6 locations with 5 precinct and 3 ballot orders, 22 locations with 5 precincts and 4 ballot orders. I wondered why locations with 5 precincts did not have all 5 ballot orders, and why all locations with 4 precincts did not have 4 ballot orders, etc!! I wondered how they were combined--which "crawls" with which!

The post above shows the distribution numbers. That finding prompted more statistical questions (easier than checking every location).

So I asked, "What is the correlation between Bush and Kerry percentages?" Statewide in Ohio it is -0.9938, a near perfect inverse relationship, as expected without Perot. For E-voting statewide it is a startling -0.9994, as if the voting was pre-programmed. Here are the Pearson correlations for Cuyahoga precincts with high Kerry voting:

Kerry vote	> 90%	80-90%	80-60%	

Bush		-0.63	-0.76	-0.87

In precincts with higher than 90% Kerry support, the correlation drops to -0.63

Here is the precinct count and average number of non-votes:

		> 90	80-90	80-60	<60

Count 		106	59	183	155

Non-Vote	12.11	10.03	7.69	5.91

The higher the Kerry percentage, the more non-votes there are for those precincts.

So, questions remain. Next step is to have Excel count the ballot types in these categories. Meanwhile, this post.

[L. Coyote]

160. The ballot sequence is b . B . K . d . p

This sequence in the previous post is wrong:

b B d K	p 

It should be alphabetic with Nader after Kerry, not B d K... Oophs, so this much changes:

Crawl 2 & 3 = Bush to d and p, and Kerry to b and p.

Bush and Kerry switch votes in "crawl 1" and "crawl 4" ----with the opposite precinct's Bush votes going to Badnarik (crawl 1) and Kerry votes going to disqualified (crawl 4).

So, when Kerry votes go to Bush in one direction, Kerry votes go to disqualified in the other direction.

That's exactly where the stats keep pointing. That's why the non-votes correlate with Kerry better. Here is that stat in another form.

The mean number of non-votes per precinct sorted by ballot numbers is:

Ballots		5	4	3	2
	
Non-Vote	4.47	5.50	7.14	8.33

In crawl 1 & 4 locations, Bush votes go to Badnarik when and Kerry votes go to Bush. This is where the multiplier lies hidden. In a 90% Kerry location, for every one extra Badnarik vote, nine Kerry votes were swapped to Bush votes. We only detect 1/10 of the cross-precinct voting in the 3rd party vote. This is what to look for. This is the crawl you would want to have in high Kerry precincts to skew the cross-precinct voting in your favor!!! If I finally have it right. I'm sleepy.

This is definitely not for Dummies. Either dat or I is one! And with three ballot types at a location, the complexity begins!!

Now here's another zinger I just noticed. Pairing a large precinct with a small precinct. I have to rest before contemplating the implication of that trend.

[Iceburg]

161. I have analyzed a subse of 2-2-176 data set ... interesting relationship

with 3rd party votes ...it is described in post 153
http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=166913&mesg_id=184692&page=

[L. Coyote]

165. In the two ballot subset there are four combinations

even though there are five candidates that must be evenly allocated to the 1st position. The sequence of the candidates is constant. So I reduced the possible combinations to just four, the number that the second ballot shifts in relation to the first is the key. I'm applying the term "crawl" to this.

You are describing a "4 crawl" voting location. Irrespective of which candidate is in front of the line, this is a "4 crawl"

b	B	K	d	p
B	K	d	p	b

The lower row has crawled four space to the right. It doesn't matter which one is top or bottom either. So I've adopted the ascending sort convention, with b above B, and I'll call the top row the 'upper" precinct. The bottom row precinct is the "lower." (B is taken by Bush.) So we have U and L, upper and lower.

So a "4 crawl" location can have b & B in the first position, or d & p, or p & b, or B & K, or K & d. The cross-voting results are the same. Am I right??

If so, this reduces the discussion down to the four "crawls" in consideration. Five crawls and you have only one ballot type.

In the case of your crawl in the referenced post and as above, that is "crawl 4," if Kerry votes go to Bush from lower to upper, Bush voters in lower will cross vote to Badnarik. Strategically, you'd place these machines in 90% Kerry precincts. That way, for every 9 votes Kerry loses to Bush, Badnarik gains one that Bush loses (given equal cross-voting and equal precinct sizes). So the Bush-Kerry margin shifts 17 votes. At the same time the upper voters using the lower machine will cast 9 votes for disqualified instead of Kerry along with the one for Bush. Kerry loses 26, Badnarik gains 1. Is this ratio correct?? Am I seeing this correctly???

You would definitely want this combination in a 90% Kerry precinct where there are the fewest possible Bush votes to move to Kerry. Likewise with "crawl 1." Crawl one is the same as crawl four if you switch upper and lower and sort columns to ascending order!! The same can be done with crawls 2 and 3.

Then are some locations with three precincts and two ballot types....

[Iceburg]

166. LC ...I'm not following - perhaps I need a little shut eye but ...

What I was describing via my diagram was the 5 possible pairings that I would be searching for in the 2x2 set (2 ballot Order pairs, 2 precincts) ..which essentially boiled down to is the K positiion in ballot order X = the B position in ballot order Y of the same location. I am assuming they didn't want to get caught so those mapped pairs fullfil that requirement. That was the plan, but in reality ballots shifted in both directions. Further, I am assuming these shift occured for the most part within the voting location and not at some central counting office. If there wre no respect for those shiftable pairs then we would see much more 3rd party votes (unless the deed was done using other cards from other locations)

You state "In the case of your crawl in the referenced post and as above, that is "crawl 4," if Kerry votes go to Bush from lower to upper, Bush voters in lower will cross vote to Badnarik."
My response: Yes, they would and that would NOT be part of the plan ...those are blunders by the planning team or accidents...the "cigarette butts left at the crime scene as it were." Poll workers have a good eye for knowing who is Dem and who is not -- it then becomes easy to give the right card to the right person ...you might get it wrong every now and then but if you are agood pollworker working for the dark side its possible. Therefore, if the ballots were stamped with the precinct number before handing the ballot to the voter only through a manual recount along with verifying the signatures in the poll book will we know with some degree of certainty that our theories were correct. Ther are many other ways to achieve the same end.

Perhaps your crawl will make more sense to be in the morning when I have fresh eyes and all neurons firing.

PS..you do outstanding work!!!

[L. Coyote]

176. Here is a subset of the 2 ballot type locations

I just uploaded this Excel file to the server after doing the sort. It's still a bit of a mess, and may be missin a precinct or 2! The color coding is by ballot sequence. That makes the combination patterns readily apparent. One worksheet has the 2 ballot location precincts, the other a sort to only the "Crawl 1 and 4" combination, the one with Bush and Kerry votes swapping. I have not yet done stats on it, though it includes stdev for disqualified, a good indicator of cross-voting.

Back to your referenced post where you wrote:

"Note that Kerry's returns go down disproportionately to the increase in 3rd party votes ... A DROP OF OVER 30 PERECENTAGE POINTS!!!" You've got the picture!

In a 2:1 Kerry location, a 4% cross-over can shift the Kerry-Bush outcome 15%.

I'm still not at the answer to the question, "Which ballot patterns are in Kerry precincts?" Your graph seems to indicate what we can expect to find. We already know that with 5 positions, things are not equal given particular combinations at a locality!!

And this summary shows where the greatest number of non-votes are located, in Kerry precincts. Put 2 and 2 together....



I'm trying to write this up with more clarity than a threaded discussion. I'm uploading now, so you can check out the first example and the selection of precincts I'm using. I'll upload as I progress.

[Iceburg]

177. Example & Selection --Right on LC!!!!

Can you check your link to the excel file -- I can't access it.

And I need clarification on:
"In a 2:1 Kerry location, a 4% cross-over can shift the Kerry-Bush outcome 15%. "

[L. Coyote]

180. The Excel should be uploaded now. I uploaded the wrong file. n/t

[AirAmFan]

155. Sorry, LC. Joe's right in #143, and I was confused in #142

The ballot orders are treated symmetrically when counting the number of different ballot orders per cluster.

In the comparison I was making, between distribution of ballot orders by cluster size and distribution of ballot orders by number of different orders, the latter should shift locations from column to column equally from all rows.

In your table, it looks as though the columns are roughly even except for two pairs of adjacent cells.

In the first row, column 3 is 85 and column 4 is 115. If these statistics were instead 100 and 100, the columns would be closer to balancing.

In the fourth row, columns 4 and 5 are 87 and 49. If instead these cells were 100 and 36, the columns most likely would reject an F-test for the hypothesis that the columns are different.

Sorry for my confusion--I was tired and rushed when I wrote #142.

[L. Coyote]

141. Number of Votes for Candidates Sorted by # Ballot Types at Locations

The actual vote numbers give a good idea of the relative weight of the categories: 'the number of ballot types at a location.'

	Kerry	Badnarik  Bush	    Non-Vote	Peroutla

5	10,163	   38	   8,682        219	 19

4	58,382	  256	  43,684      1,308	 230

3	125,631	  505	  63,090      3,329	 422

2	145,473	  719	  58,922      4,200	 798

1	 55,263	  165	  18,108      1,633	 132

All	394,912	 1,683	 192,486     10,689	1,601 

The totals are off because I lack the ballot order at several precincts.

[mgr]

153. Interesting pattern

Excellent use of GIS.

Am I seeing a pattern that voting locations with larger numbers of precincts mostly occur outside of Cleveland? This would roughly correlate to the suburbs having more Bush support than the city, and might explain the lighter Kerry support in part. Still needs party identification.

Have you compared this to the number of voting machines available at these voting locations? I wonder if the precincts with high concentrations of African Americans were discriminated against by having fewer voting machines, and or relocated voting locations, or more likely visited by Republican challengers. Your data manipulation is so close to showing whether there is Fraud due to lack of equal access by African Americans.

One could justify the smaller number of machines due to the voting location being just one precinct, while the suburbs need more because there are more precincts at the location. I doubt such an argument would stand in court.

Mike

[mikelewis]

132. Question - Why does the Precinct totals differ between the

Cuyahoga County information and the Sec. of State's totals.

The Cuyuhoga County BOE states that it has a precinct count of 1458 and the SOS says there is 1436 precincts. Why the disparity?

[AirAmFan]

133. Cuyahoga reported its absentee results in 22-line blocks

of its HUGE precinct level file for each contest. But they are not physical precincts. There are 1436 physical precincts.

If you want to download the actual file, you can right-click a URL link at http://boe.cuyahogacounty.us/boe/results/canvas.htm .

But be prepared for megabytes of data!

See also my notes on the layout of this file starting at post #21 in the parallel thread at http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x147718

[mikelewis]

136. Question - Why does the total number of people registered change

In the canvass report there are 1005,807 registered voters.


Why is there 1007187 registered voters listed on thier summary and the SOS website.

Furthermore, were you aware of the total number of people registered in Cuyahoga County as a percentage of the voting age population was 93.5%. In many precincts, the number of people registered exceeded 100%.

Also, I'd like to thank you for your previous answer to the precinct issue.

Sincerely,
Michael Lewis

[minvis]

145. Butler County Info

In response to previous posts that asked what grouped precincts looked like in Bush counties, I've looked at the info on Butler County. This county went nearly 2-1 for Bush. I have written to their Board of Elections asking for the ballot order in each of the precincts. They also had many locations that had multiple precincts in them, much like Cuyahoga County. There are some interesting bits of information here.

Voter Suppression was evident here as well. Of the 289 precincts, Kerry only won 34, Bush 255. Of the 34 that Kerry won, only 5 had turnouts at or above the average turnout for Butler County (70.99%). The two that were 80% or more (81.01 & 81.90) turnout that Kerry won were very close (357 to 415 & 273 to 287). The other 29 Kerry precincts were as follows: 1 under 40% (Oxford 4HL - Miami of Ohio University), 9 under 50%, 16 under 60% (which included many of Kerry's widest margins), and 3 under 70% (of those 3, none were above 65%).

Contrast these numbers to the 255 precincts that Bush won. 1 was under 40% (37.17% - 171 Bush to 164 Kerry), 11 under 50% (none of which Bush won by more than 80 votes), 18 under 60% (again most of them fairly close), 45 under 70%, 126 under 80% (118 of which are above the 70.99% average), 53 under 90% and remarkably 1 with over 90% turnout. Additionally, the 4 largest precincts (Liberty Township 4DI <1,834 registered voters>, Oxford City 4HK <1,522>, Fairfield Township 4AL <1,459> and Liberty Township 4DO <1,451>) had the following turnouts: 86.15%, 40.74%, 82.39%, 90.21%. The 40.74% turnout was again in Oxford at Miami of Ohio University. This precinct had a shared location with 3 other precincts. Only one other location in Butler County had 4 precincts in the same place (West Chester Township). All 4 of those precincts in West Chester were solidly Bush (over 2-1 margins. The 4 precincts in Oxford that were grouped together, 2 of them were won by Bush, 2 by Kerry, none of them by wide margins.

I'm still waiting for the ballot order breakdown from the elections board, but there were no large third-party votes in particular precincts (no more than 6 votes for either Badnarik or Peroutka). However, the two times that either one of them got 6 votes were in precincts that Kerry won. I'd be very interested to check this out though once I get the ballot breakdown.

I'll post once I've got the ballot order info from the Butler BOE. If someone can tell me how to attach the Excel files, I can post the info for everyone.

[jmknapp]

150. What about TRENTON 5EA?

I have 30 votes for Badnarik there.

Joe

[minvis]

152. Trenton 5EA

My mistake. You're right Trenton 5EA had 30 votes for Badnarik and it did share the same location with Madison Township 4EJ. Also Fairfield 1BC had 8 Peroutka votes and it shared with Fairfield 1BA. I also am waiting on information from Clermont County, another 2-1 Bush county for further confirmation of my findings.

[Iceburg]

158. That's excellent news Minvis. Plus New FINDINGS!!!!!

I am working on a new sub-theory of the ballot order uniqueness -- "sinks and sources" within ballot oder pairs. Assuming there was a planed shift of the ballots (the lack of randomness in itself leads me to believe this was no accident)...
I have mapped all the ballot pairs that within a voting location that can be moved from one ballot box to another without altering the third party presidential votes.



I have analyzed the first pair KxyzB + BKxyz in all voting areas that have only 2 precincts ... the evidence is mounting.

Believing that the votes cast for 3rd parties are like cigarettes butts left at a crime scene, the more butts found the longer the perpetrator stayed at the scene or, in our case -- the more 3rd party votes found the more shift took place. However, merely adding the "found" 3rd party votes back into Kerry's column will not put a dint in Kerry's total because these "butts" represent only a small portion of the switch activity.

I'm getting ahead of the actual data ... I'll let this graph of the dataset (all precincts within voting areas that have only 2 precincts: one with ballot order KxyzB, and one with ballot order BKxyz. -- there are 25 voting areas with a combined 50 precincts that meet this criteria)



Note that Kerrys returns go down disproportionately to the increase in 3rd party votes ... A DROP OF OVER 30 PERECENTAGE POINTS!!!

[minvis]

192. More Butler County Info

Here's some more interesting tidbits from the info from Butler County.

As elsewhere in Ohio, voter registration was up in Butler County as well. I went back to 1996, 2000 and even the primary of 2004 each of those elections the number of registered voters in Butler County went from 187,186 in 1996 to 216,275 in 2000 to 211,492 for the primary in March of this year. By the time of the general election in November, that number had increased over 21,000 from 2000 and nearly 27,000 from the March primary to 238,022. This, in and of itself, is not proof of fraud since interest was extremely high in Ohio and both the Republicans and Democrats were pushing increased voter registration. What I tried to find out, however, is were the increase came from. The top five precinct groups in numbers of registered voters compared to 2000 were: Liberty Township (+5,693), West Chester Township (+3,949), Fairfield Township (+3,340), Hamilton City (+1,536) and Monroe City (+1,467). That accounts for about 15,000 of the nearly 22,000 voter increase in the county.

Looking at these five precinct groups, I then looked at differences in number of ballots cast between 2000 and 2004 and the number of votes for Bush between 2000 and 2004. For example, Liberty Township had an increase of 5,083 ballots cast between 2000 and 2004. Of those 5,083 ballots 4,333 were cast for Bush. That's 85.2%, by the way. The others were similarly extremely large percentages for Bush. West Chester Township had 4,406 more votes cast than in 2000, 80.4% of which went to Bush. Fairfield Township, 3,383, 76.8% to Bush. Hamilton City, 3,214 more votes cast than in 2000, 91.8% of which went to Bush. I could go on and on. Even granting that this is a very pro-Bush county, no Republican presidential candidate pulled more than 65% of the vote in this county since 1992.

Now, one could argue that there will be fluctuations in percentages as you cherry pick precincts, but remember these are precinct groups, whole townships or cities. These groups contained at least half a dozen precincts and in the case of West Chester Township, it contained 45 precincts. With these large groupings, the percentages for Bush and Kerry should closely mirror the county wide percentages.

Anybody have any logical explanation for this or not?

[L. Coyote]

193. I would like to see these numbers as percentages of the vote.

They seem to assume that the number of people who voted for Gore would vote for Kerry.

What percentage of vote did Bush get in 2000, what percent in 2004. What is the z-score of these in relation to the county and in relation to the 1996 to 2000 differences.

It is a false premise to say a percent of the difference in number of voters went to a candidate. Those votes were not kept separate and counted apart from the others. I get the point though.

BTW, the Cuyahoga analysis continues, beginning with a distillation of this thread, at:

http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=196518&mesg_id=196518

[jmknapp]

194. It really is a fascinating phenomenon at least

...that the high-growth exurbs went so overwhelmingly for Bush. Realize that it's not just registration drives, but population shifts behind these localized increases in registered voters. IMO, if there is fraud to be found, it's in them thar exurban hills. If no fraud, then we are left to contemplate how either through self-selection or the local water supply (Kool-Aid?) that settlers in these areas become Bush-bots.

The same thing happened around Columbus. In Franklin County, the exurb of Dublin had over 100% increase in voters in one precinct, and went very heavy for Bush, and also strongly in favor of the gay marriage ban. As Republicans, these voters differ from the old-money GOP in suburbs like Upper Arlington and Worthington, who tended to vote against the gay marriage ban. Fascinating dichotomy I think. When Issue 1 is plotted against Kerry support, this split becomes graphic:



For Butler, a similar pattern is seen:

[IndyPriest]

154. Doesn't this lead us to likely suspects?

Or at least narrow the focus of possible suspects? It seems to me that the average poll workers would probably not be the ones who are switching ballots to the wrong machines for counting. Likely being "average, honest citizens," they are probably dropping punched cards into correct loading boxes. (It's POSSIBLE that poll workers were showing people to the wrong precinct voting booths, and so dropping resulting punched cards in wrong boxes. But they'd have to do that in a pattern to get the results we're looking at.) So it seems more likely that the perps are at the other end, the counting end, loading cards from, say, "precinct 1" into machines meant for counting, say, "precinct 2."

But if those counting areas/rooms are like those I've seen in Michigan, they're still fairly crowded with workers unloading boxes of cards, lining the stacks up by precinct number, handing them off to the machine loaders, etc. Everyone in a room like that seems to know which precincts are supposed to be counted by which machines. So even if someone tried to load a stack into the wrong machine, I think it's likely they would have been caught. Maybe not, but it would have been VERY risky behavior.

So, I'm led back to what seems to me a more likely scenario: somehow the final tally tabulating software was fixed. How that accounts for the PATTERN we're seeing - the vote being off in co-location precincts with multiple order ballots - I don't know. But it seems that fixing the final tabulation software is far less likely to be caught.

Am I missing something obvious?

[L. Coyote]

162. Am I missing something obvious?

Elections are decided by that one or two percent margin, or one hundred or two hundred votes. So factors like ballot spoilage of 2% in one neighborhood and 0.5 percent in another neighborhood can be very significant, Spoiling the enemy wins more elections than pretty faces!

When considering cross-precinct voting, a certain amount of it is going to happen no mattter how closely things are controlled or watched when you have two or more precincts in one room. To prevent spoilage, you need all the ballot orders the same in each location no matter how many precincts are in the room.

If you want to increase spoilage from cross-precinct voting, start by having different ballots in a location where the enemy vote is highest. This automatically weighs in the underdogs favor. Next, don't train the poll workers. You don't want them to understand these complexities, at least not in enemy territory.

Next, with five candidates, figure out which combinations of 2, 3, or four ballot orders in one location favor your candidate. Lump them together where they have the best effect for you. Next, create as much confusion as possible at that location. Send in the vote challengers in force, or some such strategy to cause more cross-precinct voting. Long lines work well!

Next, in the precinct where cross-voters will deny your candidate votes, take away a machine and put it in the other precinct, where those tired of waiting will be tempted to cross-vote for your candidate. Create a one-way cross-precinct street by booth demand.

Next, evaluate the sizes of the precincts. Put the machine that moves votes to your candidate in the smallest precinct. The other precinct has more potential cros-overs.

Also, make sure none of this happens in the areas where your candidate has the most support.

Next thing you know, you've shaved enough points to make the difference. Especially when combined with all the other things, like who gets Op-Scan, who gets punch cards. Cross-precinct voting is just one of many things in the trickster's bag. The spoilage rate difference between technologies shaves more than a point!!

[Iceburg]

163. And to make sure those ballot shifts go as planned

just make sure you have one or two insiders that rotate from polling area to polling area while the 99% of their colleausges remain unkowledgeable about their peers activities. Toss in few new machines at various location to replace one that doesn't need repairs. Presto we have a winner!

[L. Coyote]

169. RE: Cross-Precinct Voting. Is the game fixed by voting machine types?

Even without sleezebag, cigar-smoking, back-room hanky-panksters....

Consider where cross-precinct voting can occur. With punch card voting, but not with E-Touch. Check out the Ohio non-vote total and the breakdown by machines. Where cross-precinct voting cannot occur, the baseline quantity of non-votes is 3/4 of a percent.



If you cross precinct vote in an E-Touch location, your votes go to whomever you cast them for, but the poll books and precinct tallies do not concur on number of voters (they can cancel each other out rather than reflect actual numbers of cross-voters).

What's happening with Op-Scan voting? Are the candidates always in the same ballot bubble? By the look of the stats, they can cross-vote.

Nevada, with all E-Voting, has a statewide non-vote for president percentage of 0.6811, while Florida has 0.3993 percent. In 2000, attention centered on spoilage patterns in Florida, so they fixed it with higher tech ways to skew elections. Now here's New Mexico, a real mystery.



If every county performed like the 10 Op-Scan couties on the right, the number of Nuevo Mejicanos whose votes would also have been counted is 16,091 (revised number. And the skewed couties are Kerry country, so it also alters the vote margin significantly, down 1235 votes to a 4800 vote margin.

We need to get to that number with some statistical precision in Cuyahoga and in Ohio. I think that is the ultimate goal of this analysis. However, it will certainly reveal any operatives in the Tricky Dickster tradition.

[Iceburg]

170. Very interesting analysis LC

One point to keep in mind is that there exists a very transparent paper trail with the opscan ballots whether they rotate the ballot order or not. The ballot papers are typically serialized and ranges of serialized ballots are then distributed to precincts. Thus regardless of how you voted, if your serial number is out of sync with the precinct's serial number range that vote must be tossed into its rightful pile. I have planned and delivered elections using the opscan equipment but I have never even set eyes on punch card machinery.

Back to the punch cards: With the punched cards we know that--

1) there was no precinct identifier stamped on the absentee vote ...that was to be filled in later by an honest poll worker. (ref. Chili)

2) we still don't know with certainty that every punch card in the live polls had a precinct identifier stamped or punched on it

3) If there was a stamp we do not know what percentage of people actually paid attention to it

4) if the precinct identifier on the card was both punched and stamped, we do not know whether the punches actually mapped to the stamp. How many people look at punch marks and know their significance?

5) with punch cards (in Cuyahoga) we have a very fuzzy audit trail to follow -- considering all the unknowns above PLUS the fact that the poll signature books are being denied access, and the votes cast for Nader have not been published.

6) At this point, we do not know if the swapped out machines (due to alleged failure) and the new machines associated with a precinct factor into our current findings.

Only a hand recount of the punch cards combined with comparison to poll signature books is likely to(but not with absolute certainty) tell the truth. What we have now, is mounting circumstantial evidence ...not absolute proof. However, that circumstantial evidence can help the recount team pinpoint voting areas that are most likely to reveal the truth, whereby a hand recount can be conducted.

Further, I agree with one of your earlier posts that the strategy was to diversify-diversify-diversify.

[L. Coyote]

175. Many strategies to shave points adds up.

You write, "Further, I agree with one of your earlier posts that the strategy was to diversify-diversify-diversify."

I see a mosaic of strategies, each shaving points. Voting machines, suppression, multiple ballots, keeping Nader on the ballot, fewer machines, fewer locations, resorted locations, ... It's a long list.

Think of the ballot order difference if Nader was dropped and only 4 positions rotated. Didn't someone fight to keep the fifth position? Who's idea was that?

It's the cumulative effect stupid, eh?

[minvis]

159. Code or Poll Workers

I think you're correct, Indy. The idea that poll workers in each of those shared polling locations would knowingly put ballots into the wrong precinct box seems a bit farfetched. I know, at least here in Chicago, that each of the poll workers rotate throughout the day to each voting area (i.e. checking names in the registration book, handing out ballots, collecting and depositing completed ballots into the ballot box, etc.). Maybe it's different in Ohio, but either way, it would involve too many people to make it feasible and/or safe to steal votes in that way.

It's possible that once I look into the ballot order at each of these shared locations, a pattern will emerge of not only votes being stolen but possibly how.

[Chili]

167. Iceburg...

I admit it, I'm a visual person, so when it comes to numbers and stats, I'm out of the loop!

But I just remembered this: I took a photocopy of my Mom's absentee ballot before we mailed it out, and of course we still have the ballot pages... in all the chaos, I had forgotten where I had put it, but I just found it among some old bills... I'm unsure of the legality of displaying even a photocopy of an actual ballot that was a part of an election, or I would post a scan of it (not to mention that it's my Mom's ballot, though no secret as to who she voted for, LOL)...

But, here's what I see: there is NO serial number or identifying number of any kind on the bottom of the ballot. There is NO punched-out pattern anywhere that would seem to indicate precinct. There IS, at the bottom of the ballot, the following:

TO BE FILLED IN BY ELECTION BOARD ONLY:

Precinct No._______ Write-in No. _______

The actual rotation in the ballot booklet:

Candidate Disqualified
Peroutka ===> 10
Badnarik ===> 2
Bush ===> 4
Kerry ===> 6

(Candidate Disqualified - Nader - is always the 8 position)

I hope that helps, though you probably already had the exact rotation somewhere in the thread, and the ballot info too. Also, naturally the "Absentee Voter Booklet" would be different from what's placed with the template, since it's in book form; on the template, you only see one page at a time.

[Iceburg]

168. Goof Info Chili -- "they leave out the precinct so it can be filled in lat

in later by a nice poll worker. I find it stunning that the precinct was not premarked. Neverthless, the ballot your mother filled out had a dpbBK ballot book order rotation and if it was deposited into a precinct with a KdpbB ballot box for subsequent reading, bush would gain an additional vote.

I do have all of the ballot orders for each precinct, however I do not have the ballot orders for the absentee ballots. Any idea where I could get them? And I certainly don't dont have a copy of the ballot for the Ohio election although I do have some authentic ballots from other parts of the world -- my favourite being the 1994 South African election.


This ballot design is a marvel in that it takes into consideration that 80% of the population was illiterate, further there are 11 "official" languages in SA. To overcome these challenges, the ballot includes simple instructions in 11 the languages at the top of the ballot ("Mark your ballot next to the party you choose"), followed by a picture of the leader of the party, plus the party's sympbol, ancronym, and fullname in english. The party names are in apha order starting with the PAC (by random draw who gets first slot). The ballot order rotation was then constant for all voting locations.

Each SA ballot had a serial number in the top left hand corner -- ranges of serial numbers were distributed to voting locations. Thus the serial number could be used for audit purpose later.
In SA less than 1% of the ballots cast were spoiled while greater that 87% of the voter eligible population turned out to vote. That is AMAZING!!!

More about the SA Ballot here ...
http://images.google.ca/imgres?imgurl=&imgrefurl=http://www.folkart.com/henderson/&h=208&w=120&sz=8&tbnid=9KDBh5rMEycJ:&tbnh=98&tbnw=57&start=7&prev=/images%3Fq%3Dsouth%2Bafrican%2Bballot%2B1994%26hl%3Den%26lr%3D%26sa%3DG

Now about your mother's ballot:
a) were there any markings on the back?
b) were there any pre-punched holes whatsoever on the ballot?

If you are comfortable with this idea, I would love to get a copy of photocopy you have. I can send you my email via pm if you so choose?

ps. believe it or not i am a very visual person too!

[Iceburg]

171. "Good Info Chili - "they leave out the precinct so it can be filled in

later" is what I meant to say ... my fingers have a severe case of keyboarditis this morning.

[jmknapp]

172. Simulation

Ran a simulation of random vote-switching to see what happens in Cuyahoga County

* pick one of the 591,348 votes cast at polling places
* pick another precinct at the same polling place to switch it to
* if the order is the same, ignore
* if the order is different, tally result

When done a large number of times in the simulation, it turns out that Kerry loses 0.55 vote on the average for each vote switched. Bush loses 0.17 vote for each vote switched.

Overall, for each vote switched, on the average:

Kerry -0.55
Bush -0.17
Badnarik +0.24
Peroutka +0.22
Disqualified +0.26

Since Badnarik got 1,886 votes countywide, if all were due to random switches, an estimate of the total switched would be 1886/.24 = 7,858 ballots.

That's consistent with the Peroutka total since the random switch model would project 7858*.22 = 1728 votes for Peroutka. He got 1751.

If that's the case, then Kerry would have lost 7858*0.55 = 4,322 votes due to random ballot switches. Bush would have lost 7858*0.17 = 1,336.

Net, Kerry's margin might have thus been shrunk by 4322-1336 = 2,986 votes due to random ballot switches.

This would be an upper bound because of the assumption that all Badnarik/Peroutka votes are due to switches.

[Iceburg]

173. JM can you run the simulation with this criteria ...

* pick one of the 591,348 votes cast at polling places
* pick another precinct at the same polling place to switch it to
* if the position of Kerry(K) is different from the position of Bush in the other ballot, ignore it
* if the position of Kerry is equal to the position of Bush in the other ballot, tally result

[jmknapp]

174. Hi Iceberg

Not sure what you're getting at, but this is what I get:

K -0.66
B +0.33
b +0.33
p 0.00
d 0.00

[L. Coyote]

179. In Two Ballot Precincts, there are only Two Patterns of Switching

So the next question is how do they compare. Can you run the scenario for the "crawl 1 and 4" pattern versus the "crawl 2 & 3" pattern.

The "crawl 1 and 4" pattern swaps Kerry and Bush votes both ways, and benefits Badnarik one way and disqualified the other way.

The "crawl 2 and 3" pattern has Kerry and Bush votes going to the other 3 positions.

The big payoff for Bush is the "crawl 1 and 4" combination in 95% Kerry precincts.

It's this simple in two ballot locations. In three and four ballot locations it is far more complicated, with more combinations.

Some of the two ballot locations have three or four precincts too.

[L. Coyote]

178. Non-Random Ballot Order Combinations Produce Even More Shift ...

as does the amount of support in a precinct.

This is truly awesome!! The implications are incredible.

NOW, regarding your Kerry margin difference numbers:

The amount Kerry loses on a county wide basis reflects his level of support in the county, about 2:1. In a 2:1 Bush county, a different effect would prevail.

With 5 ballot positions, not all combinations are equal. Consider this typical precinct, about 2/3 Kerry voters, with a type "crawl 1 and 4" ballot combination. (One order moves "one" or "four" positions in relation to the other, producing the same cross-over pattern.)

Y. M. C. A. 3881 Pearl Road, Cleveland -- 634 votes recorded.

#                b       B       K       d      p       % Kerry

2908	pbBKd	15	80	216	13	7	65.26

2909	bBKdp	 4	85	188	24	2	62.05

If 14 Bush voters in precinct 2909 cross-voted in precinct 2908 and punched the second candidate in the order, their votes count as Badnarik in their own precinct tally. Bush just lost 14 votes. If an equal proportion of 2909 Kerry voters cross-voted, 28 of them (the 2:1 Kerry/Bush ratio in the precinct and county) would punch the third position in 2908, mistakenly voting for Bush when their punch cards are tallied in their own precinct. Badnarik gains 14 votes, Kerry loses 28 votes, and Bush nets 14 votes. (Cross-voting the opposite way in this precinct produces disqualified, non-votes.) The spread between Bush and Kerry has changed by 42 votes in Bush's favor, a 15% shift in the result.

Only 6.6 percent of the voters cross-voted in this scenario. Kerry's 15% loss of margin is triple Badnarick's gain.

That's in a 2:1 precinct, the county average. This is why I've been ranting about how so many of the highest standard deviation non-vote precincts are 90% Kerry precincts. In 90% Kerry precincts, the ratio jumps from 2:1 to 9:1!!!! That blows the 15% shift right out the window.

Your simulation is great, and it really makes the point crystal clear. It may even prove that something was quite felonious in the kitchen that served up the portions.

We know that the pattern of machine distribution is not even. And the different ballot combinations do not produce equal switching either. Did their placement in key precincts skew the pattern?



I assume you have not factored in county or precinct Kerry vote percentages yet. ????

Do your assumptions include equal cross-voting in every precinct, or do the weight for number of voters in each precinct?

[jmknapp]

181. Assumptions

Since I selected "ballots" at random from the 500,000-odd universe, the switched ballots would automatically reflect the Kerry percent.
And the precincts that had larger number of votes would thereby suffer more cross-voting.

It gets a bit trickier picking a precinct "at random" to switch to, within the polling place. Should that be weighted by votes cast in each precinct? Or give all precincts in the polling place an equal chance of being selected?

In the simulation the selection of the cross-voting precinct was weighted by the precinct size. For example, say a ballot from precinct A in a polling place was selected, comprised of precincts A (500 votes), B (500 votes) and C (250 votes). The simulation would then pick B or C at random as the cross-voting precinct, weighted by the votes cast, so B in this case would have twice as much chance of being selected.

Turns out I had a little bug of sorts in the original simulation, fixing which changes the results slightly:

K -0.56
B -0.15
b +0.24
p +0.22
d +0.26

The original method underweighted the two-precinct polling places.

If the cross-voting precinct is chosen not weighted by precinct size, then the result is the same, so I guess it averages out.

[L. Coyote]

182. How many votes would Badnarik have gotten in Cuyahoga?

Certainly, less than his state-wide percentage. The state-wide percentage incorporates cross-voting, and Cuyahoga is different than the rest of Ohio, esp. given 66% Kerry support. Perhaps one-ballot precincts areour best clue.

This Cuyahoga statistical summary might help resolve the problem.

					
County percent: Cast   Kerry	Bush	Badn.   Per.   Non-vote

count		1435	1435	1435	1435	1435	1435
mean		60.0	66.6	31.0	0.320	0.274	1.804
median		62.2	64.4	33.3	0.182	0.000	1.464
skew		-1.082	0.113	-0.117	22.43	18.79	3.231
.05 normdist	0.000	0.000	0.044	0.437	0.436	0.109
.1 trmean	60.6	66.7	30.9	0.189	0.143	1.671
.5 trmean	61.7	65.5	32.2	0.136	0.084	1.527
avedev		7.36	14.91	15.56	0.345	0.327	0.979
stdevp		9.33	17.45	18.19	1.70	1.38	1.42
stdev		9.34	17.46	18.19	1.70	1.38	1.42
.05 pconfid.	0.48	0.90	0.94	0.088	0.072	0.074
min		7.85	27.21	0.00	0.00	0.00	0.00
max		70.0	65.7	71.5	0.59	0.33	1.78
range		62.1	38.5	71.5	0.591	0.328	1.777
confidence	0.483	0.903	0.941	0.088	0.072	0.074
quart 1		50.72	61.33	4.66	0.000	0.000	1.163
quart 2		57.85	75.16	21.13	0.178	0.000	1.920
quart 3		66.73	81.16	45.55	0.354	0.261	2.366

One Ballot subset

count		177	177	177	177	177	177
mean		56.6	74.3	23.1	0.212	0.211	2.216
median		57.9	75.2	21.1	0.178	0.000	1.920
skew		-0.562	-0.487	0.487	1.291	11.210	1.583
.05 normdist	0.000	0.000	0.103	0.256	0.410	0.082
.1 trmean	57.0	75.1	22.2	0.187	0.135	2.102
.5 trmean	57.8	76.4	20.8	0.143	0.078	1.989
avedev		7.80	15.10	15.74	0.194	0.243	1.184
stdevp		9.77	17.51	18.17	0.247	0.710	1.556
stdev		9.80	17.56	18.22	0.248	0.712	1.561
.05 pconfid.	1.44	2.58	2.68	0.036	0.105	0.229
min		28.83	27.21	0.96	0.000	0.000	0.000
max		86.77	96.23	71.48	1.068	9.091	11.00
range		57.94	69.02	70.51	1.068	9.091	11.00
confidence	1.440	2.579	2.676	0.036	0.105	0.229
quart 1		50.72	61.33	4.661	0.0000	0.0000	1.163
quart 2		57.85	75.16	21.13	0.178	0.0000	1.920
quart 3		63.80	91.90	37.10	0.347	0.253	3.030

And now a summary of actual numbers.

One Ballot	Kerry	b	Bush	non	p				
mean		312.2	0.93	102.3	9.226	0.746
median		298.0	1.00	76.0	8.000	0.000
skew		0.212	1.241	1.598	1.394	1.918
.05 normdist	0.003	0.202	0.147	0.086	0.264
.1 trmean	311.5	0.839	94.2	8.708	0.621
.5 trmean	304.1	0.640	81.6	8.157	0.371
avedev		92.4	0.811	78.8	5.089	0.843
stdevp		114.7	1.055	97.6	6.706	1.104
stdev		115.0	1.058	97.9	6.725	1.107
.05 pconfid.	16.90	0.155	14.4	0.988	0.163
min		21	0	4	0	0
max		620	5.000	662	41.000	6.000
range		599	5.000	658	41.000	6.000
confidence	16.9	0.155	14.4	0.988	0.163
quart 1		232	0.000	18.0	5.000	0.000
quart 2		298	1.000	76.0	8.000	0.000
quart 3		401	2.000	169	13.000	1.000

Countywide	
				
mean		275.6	1.174	134.2	7.457	1.116
median		264.0	1.000	130.0	6.000	0.000
skew		0.313	26.370	0.743	3.335	25.037
.05 normdist	0.004	0.410	0.081	0.107	0.437
.1 trmean	274.3	0.807	128.6	6.921	0.609
.5 trmean	267.1	0.579	127.4	6.300	0.395
avedev		81.18	1.184	78.07	4.154	1.296
stdevp		103.19	4.950	96.06	5.964	6.752
stdev		103.23	4.952	96.09	5.966	6.754
.05 pconfid.	5.34	0.256	4.970	0.309	0.349
min		1	0	0	0	0
max		395543	1684	192518	10701	1602
range		395542	1684	192518	10701	1602
confidence	5.339	0.256	4.970	0.309	0.349
quart 1		232.00	0.000	18.000	5.000	0.000
quart 2		298.00	1.000	76.000	8.000	0.000
quart 3		342.00	1.000	198.500	10.000	1.000

[L. Coyote]

183. A rough guess...

Using the one ballot group, the 0.5 trimmed mean is 0.640 votes. Given the skew, this is my best choice. Assuming honest ballots for these 171 precincts, this multiplier yields 920 actual Badnarik votes in the precinct voting.

This is 0.153% compared to the inflated statwide 0.257%. It might be a little low.

That means about 765 extra precinct votes for Babnarik were recorded.

This does not seem like a lot, but it is multiplied for effect on the Kerry/Bush margin.

Using 0.150% Badnarik votes for all punch card votes in the state would then produce over 4100 cross-votes for Badnarik. Albeit Cuyahoga may be a skewed sample to apply to the state.

Badnarik's State Op-Scan percentage of 0.176 would mean a 3100 excess of Badnarik punch card votes statewide. So thes two multipliers produce a 1000 ballot range, very rough estimating. Their mean times an adjusted multiplier (excess Badnarik/Joe's number roughly adjusted for Ohio) yields about 10,000 cross-precinct votes in Ohio. That's a really really rough estimate. Any refinements??

What's 10,000 cross-precinct votes in light of the fact that the Washington state governor's race is being decided by whose car didn't start!!

Joe, given your simulation, how would you translate this into votes altered statewide? Kerry vs. Bush.

[jmknapp]

186. Ballpark

Well, in Cuyahoga I figured 0.4 votes loss of Kerry margin for each vote switched (randomly). So if 10,000 votes were switched statewide that would be about 4,000 votes of margin. Since Kerry is short by 118,000, not about to change anything.

The only hope would be that non-random mixups occurred. In fact, the Cuyahoga data show that the whiter, more Republican precincts were less prone to mixups. That is odd because the Republican precincts tended to have more opportunity for confusion, with more precincts at such polling places on the average.

This could get down to pollworker training. The urban polling places might have had relatively untrained pollworkers.

[L. Coyote]

185. Correction: Applause for the reader who notices these fine points!!

People actually read these numbers, AND detect errors!

The following lines in the above post change:

County		% Cast	% Kerry	% Bush	% b	% p	% non-vote
						
min		7.85	27.21	0.00	0.00	0.00	0.00
max		86.8	100.0	71.5	50.00	38.19	19.27
						
quart 1		54.61	52.69	15.96	0.000	0.000	0.925
quart 2		62.22	64.41	33.27	0.182	0.000	1.464


County		Kerry	b	Bush	non	p

quart 1		208	0	50	4	0
quart 2		264	1	130	6	0
quart 3		342	1	199	10	1

Copy, paste, and fill is only as great as the user!!.

[L. Coyote]

188. Not all things being random impacts simulation models/complexities ...

I can think of several off hand, some of which we have discussed in this thread starting in the first post:

My post #76 and 85 illustrate the correlation between high percent Kerry precincts and non-votes. Other factors may contribute to this, of course. This is not in the simulation, right?

In post #139 we see that the average percent of Kerry vote shifts with the number of machines at the precincts. We still have not found the mechanism/reason for this.

In post #140 we have the numbers of each ballot type allocated to locations. This illustrates the uneveness of the ballot type distributions. Your random ballot sampling should have incorporated any effect from this in the results, right?

In your post #146 we see the correlation to race demographic. This is a mystery still, unless there is a difference in the simulation results if you separate out the location classes by the number of ballots. Cab you run this easily. Comparing 1, 2, 3, 4, 5 ballot order groups. Do they produce distinct results?

In post #160 we see that the number of non-votes also changes by ballot order numbers per location. Fewer ballots orders (5, 4, 3, 2) means higher non-votes. Does this ave anything to do with your observation in #146? It means a correlation between number of non-votes and Kerry support, already demonstrated with the stdev sort, the graph reposted in #176 & 178.

In post #141 we saw that the number of ballots in the ballot order groups is also uneven. When the votes are converted to percentages another image emerges. First with the columns converted:

	Votes		Kerry	 b	Bush	Non      p
						
5	3.180		2.57	2.26	4.51	2.05	1.19
						
4	17.271		14.78	15.21	22.69	12.24	14.37
						
3	32.090		31.81	30.01	32.78	31.14	26.36
						
2	34.939		36.84	42.72	30.61	39.29	49.84
						
1	12.522		13.99	9.80	9.41	15.28	8.24
						
All	100		100	100	100	100	100
						

Now compare that to the rows converted;

							
	Total		Kerry	 b	Bush   Non-Vote	  p	
							
5	19,121		53.15	0.20	45.41	1.15	0.10	100
							
4	103,860		56.21	0.25	42.06	1.26	0.22	100
							
3	192,977		65.10	0.26	32.69	1.73	0.22	100
							
2	210,112		69.24	0.34	28.04	2.00	0.38	100
							
1	75,301		73.39	0.22	24.05	2.17	0.18	100
							
	601,371		65.67	0.28	32.01	1.78	0.27    100

We are still assuming that ballots are marked by precinct and cannot be switched after they are marked. It would be easy to move ballots to the precinct where the punch matches your candidate if that were so, and swap back an equal number with the same effect.

So, my scenario questions.

Can you play with the simulation to create a level playing field to compare with the actual random cast vote scenario?

And how about one based on the number of registered voters per precinct, or equal turnout by precincts?

[L. Coyote]

184. Blackwell has some words about the reasons for non-votes

http://www.knoxstudio.com/shns/story.cfm?pk=MISCOUNT-OHIO-12-22-04&cat=AN

SNIP

"Given human nature, when you're talking about 5.8 million people casting a vote, it wouldn't be too far-fetched to think that you have a small percentage of people who would say, 'A pox on both of your houses,' " Blackwell said.

"I just hear, as I crisscross the state talking with voters, some people don't think they have a clear choice and they think it's (between) tweedledee and tweedledum. Sometimes they just take a pass and focus on those issues and candidates that they know and that they see have a clear difference."

END SNIP

Brilliant!

[jmknapp]

187. Since the non-votes were more prevalent in black precincts

...this theory requires one to believe that the demographic most solidly for Kerry, 95% and more, would be the most likely not to see a difference between the candidates. Blackwell's explanation is inane.

[L. Coyote]

189. Statistical Summaries for the "Number of Ballot Orders" Subsets

		% Cast	% K	% B     % b	% p    % non    stdev
1 order								
count		177						
mean		56.61	74.31	23.05	0.212	0.211	2.22	0.29
median		57.85	75.16	21.13	0.178	0.000	1.92	0.08
skew		-0.56	-0.49	0.49	1.291	11.210	1.58	1.58
.5 trmean	57.80	76.44	20.75	0.143	0.078	1.99	0.13
stdevp		9.77	17.51	18.17	0.247	0.710	1.56	1.09
min		28.83	27.21	0.96	0.000	0.000	0.000	-1.27
max		86.77	96.23	71.48	1.068	9.091	11.00	6.46
range		57.94	69.02	70.51	1.068	9.091	11.00	7.73
								
2 orders								
count		503						
mean		58.59	70.16	27.07	0.350	0.380	2.04	0.17
median		60.07	70.03	26.95	0.166	0.000	1.67	-0.09
skew		-0.83	-0.12	0.14	15.64	13.68	3.63	3.63
.5 trmean	59.76	70.49	26.69	0.133	0.085	1.75	-0.04
stdevp		9.73	17.46	18.18	1.577	2.145	1.63	1.15
min		14.59	31.37	0	0	0	0	-1.27
max		82.32	97.73	66.51	30.827	38.188	19.27	12.27
range		67.74	66.36	66.51	30.827	38.188	19.27	13.54
								
3 orders								
count		466						
mean		60.97	65.84	31.93	0.271	0.230	1.74	-0.05
median		63.28	64.36	33.57	0.189	0.000	1.46	-0.24
skew		33.38	27.05	27.19	37.18	37.27	37.34	3.21
.5 trmean	62.87	64.89	33.04	0.131	0.090	1.51	-0.20
stdevp		9.15	16.97	17.64	0.901	0.730	1.30	0.91
min		7.85	27.41	0	0	0	0	-1.27
max		77.55	100	70.18	16.294	9.113	12.50	7.51
range		69.70	72.59	70.18	16.294	9.113	12.50	8.78
								
4 orders								
count		238						
mean		63.60	57.28	40.79	0.462	0.219	1.25	-0.39
median		65.03	56.81	41.05	0.196	0.000	1.11	-0.48
skew		-1.59	0.55	-0.55	15.126	8.697	2.12	2.12
.5 trmean	65.04	56.29	41.85	0.164	0.094	1.13	-0.47
stdevp		7.17	13.51	14.08	3.239	0.591	0.83	0.58
min		28.57	30.05	0	0	0	0	-1.27
max		74.82	100	69.51	50.000	7.252	6.49	3.29
range		46.25	69.95	69.51	50.000	7.252	6.49	4.56
								
5 orders								
count		49						
mean		61.34	53.03	45.59	0.189	0.090	1.103	-0.491
median		63.47	52.46	46.55	0.000	0.000	1.014	-0.554
skew		-1.56	0.28	-0.27	2.007	1.753	0.481	0.481
.5 trmean	63.03	52.69	45.95	0.095	0.008	1.067	-0.517
stdevp		6.74	8.87	9.12	0.284	0.165	0.551	0.387
min		41.63	29.85	16.78	0.000	0.000	0.000	-1.266
max		69.69	80.82	70.15	1.208	0.648	2.653	0.597
range		28.06	50.97	53.37	1.208	0.648	2.653	1.863

The last column is the "standard deviations of the non-votes" column. So it's standard deviation of the population (stdevp) stat is the stdevpof the subset for the individual precinct's stdev from the entire population mean. It shows the skew very well.

[L. Coyote]

191. Uncounted votes in Cuyahoga County by Richard Hayes Phillips

Just out, the following analysis.

http://www.freepress.org/departments/display/19/2004/1033