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

[AirAmFan]

Here on the eve of the scheduled meeting of Ohio's Presidential Electors, I'm amazed that a very curious feature of voting in Ohio has gotten so little publicity.

Maybe that's because it has no catchy name. You've heard the expression, "Give it a name!" In this post, I call it "the Caterpillar Ballot", because, at the same polling place, a valid vote for John Kerry (or any other candidate) might have to go in ANY OF FIVE DIFFERENT LINES ON THE BALLOT. Kerry's name climbs from ballot line to ballot line, just like a caterpillar moving from twig to twig on a tree. The Cleveland Plain Dealer explained it all in an article (EXCERPTED BELOW) that remains obscure. Here's my explanation, which may or may not be clearer to you:

Let's say Precinct 33 has Kerry on line 3, Precinct 34 has Kerry on line 4, Precinct 35 has Kerry on line 5, Precinct 36 has Kerry on line 6, and Precinct 37 has kerry on line 37. Friday's Cleveland Plain Dealer says, "Voters from multiple precincts typically share a polling place."

Let's say that Precincts 33, 34, 35, 36, and 37 are assigned to vote together in one huge school gymnasium. Everything will be fine as long as voters from each precinct use only the punch machines set up for that precinct's candidate order.

But if the pollling place is crowded and chaotic (as hundreds of polling places in pro Kerry areas (but not so much pro-Bush areas) were on Election Day, when it rained hard all day all over Ohio), a Precinct 33 voter might take her ballot to a Precinct 34 machine. She'd punch a hole where it says "Kerry", on Line 4. But a stamp on the back of her ballot would route it to a counting machine where a Kerry vote must be on Line 3. If she took it to another precinct's machine, she'd cast a Kerry vote on another line, but still not on line 3 where it would have to be, to be counted for Kerry.

Do you see the problem? Kerry's name moves like a caterpillar from one ballot line to another, depending on which machine is used in a room full of machines set up for different precincts.

Just like the "Butterfly Ballot" that made Palm Beach County famous in the 2000 Florida fiasco, the Ohio Caterpillar Ballot has great potential for mischief. And it is the most straghtforward explanation I've heard for the incredibly large vote totals minor candidates got in Ohio.

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

From http://www.cleveland.com/news/plaindealer/index.ssf?/base/cuyahoga/1102674912293811.xml

"Friday, December 10, 2004

"The stage for the mix-ups was set by a state law that requires candidates' names be rotated on ballots so that each candidate gets a turn at the top position. The rotation is done in the name of fair play, a nod to conventional wisdom that undecided voters tend to choose the name at the top.

In Cuyahoga County, where punch-card voting machines are used, the names are rotated on the pages in voting books that guide voters to the proper position on the punch cards. THERE WERE FIVE VERSIONS OF THE PAGE FOR THE PRESIDENTIAL CANDIDATES.

The first version lists the candidates' names alphabetically... In each subsequent version, the candidate at the top of the list moves to the bottom, bumping the rest up one notch. VOTERS FROM MULTIPLE PRECINCTS TYPICALLY SHARE A POLLING PLACE. Candidates' names in voting books are ROTATED BY PRECINCT, so there are DIFFERENT VERSIONS AT THE SAME POLLING PLACE.

Voters are supposed to use polling booths, and the voting books in them, that are specific to their precinct, not just any booth in the polling place. The problem comes when a punch-card ballot for one precinct is inserted in the voting device for another precinct. Because of the name rotation, a voter unknowingly punches a hole for the wrong candidate.

THE PUNCH CARDS THAT VOTERS SLIDE INTO THE DEVICE ARE THE SAME, BUT THEIR BACKS ARE STAMPED WITH THE PRECINCT SO THEY WILL BE COUNTED PROPERLY."

[vpigrad]

1. Any pictures?

Anyone have any pictures of these ballots? After all, freepers need to be shown pictures in order for them to understand.

[AirAmFan]

3. Sample ballots for Cuyahoga county still seem to be up at URL

http://boe.cuyahogacounty.us/BOE/ballots/ballots.htm . I found this by googling "Cuyahoga 'sample ballots'". Doing VIEW--SOURCE, I see that all sample ballots for that county are in subdirectory http://boe.cuyahogacounty.us/BOE/ballots/pdf . For example, the ballot for precinct 12H is file http://boe.cuyahogacounty.us/BOE/ballots/pdf/cleve12h.pdf .

It would be a lot of work to go through each file in that directory and see which precincts have which ballot order. So far, I haven't found a list of polling places that show which preceincts were co-located.

[mulethree]

7. voting locations

http://boe.cuyahogacounty.us/boe/PDF/votinglocations.pdf

Other info from jfern
http://www.dailykos.com/story/2004/12/2/44647/2043

A spreadsheet with the precincts grouped by polling place
http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=172x4800#4843


Still haven't found a ballot-order source other than the PDF files.

[AirAmFan]

11. THANK YOU! I haven't downloaded the spreadheet yet, but the other

two URLS are EXCELLENT, and I saw daytes as early as December 2nd on them.

Your first URL shows 583 polling places in Cuyahoga County, but enough precincts co-located at those places to take up 52 pages! Co-location was the RULE, not the EXCEPTION in Cuyahoga.

Has anyone done this same analysis for other urban counties that used punch machines?

It's early for an overall assessment, but it seems as though this one problem, "caterpillar/shuffle" votes for the wrong candidate, could account for HUNDREDS OF THOUSANDS of lost votes for Kerry.

In principle, Bush could suffer from the same problem, if hundreds of thousands of people who wanted to vote for Bush had been herded into overcrowded places to wait in line for hours and hours in the chaos of pouring rain on Election Day. But the relative wealth of pro-Bush areas assured that such overcrowding had to be rare. For example, during Wednesday's Conyers hearing, a very articulate Kenyon student pointed out that, while the Kenyon campus had only two machines for 1300 registered voters, surrounding pro-Bush counties had enough voting machines to cap registered voters per machine at 100.

[mulethree]

16. yes, it dilutes the vote, so the 'favorite' at that poll looses most

If rural locations don't have co-locations, then it affects urban locations disproportionately :(

Assuming rural = more repub, and urban = more Dem then the dems generally loose out.

But it tends to show up as high tallies for third-party candidates which, this year at least, stand out amoung otherwise low third party votes.

Affects the whole ballot too not just the presidential race, so some local tax measures and ordinances probably have wrong results.

[berniew1]

64. Just one of many things done in minority precincts to reduce Kerry vote

But though it cost Kerry many thousands of votes and disinfranchised lots of minority voters; it wasn't as big as the systematic vote machine fraud: www.flcv.com/fraudpat.html

[tinfoil_beret]

2. Do these ballots have any print on them?

If they use blank ballots for these punchcards, it doesn't give the voters any more opportunity to verify their ballots than paperless DRE machines.

If the ballots have no print showing the names of the candidates, and different machines at a polling place merely mark the voters' choices in different patterns, then this leaves absolutely no means to determine the intentions of the voters in a full audit of the results, ie. in hand recounts of the ballots of respective precincts cast at the same polling place.

If stamps on the ballots differentiate ballots for different precincts, why wouldn't the ballots have print showing the names of the candidates varying with each precinct? (I presume that a poll worker physically hands each voter a ballot for the correct precinct, either pre-stamped or stamped by a poll worker.)

If print on the ballots varies with each precinct then the voters do have the opportunity to verify their ballots, and this argument makes little sense.

If the ballots do have print showing the arrangement of the candidates, and voters using the wrong machines do not verify their ballots, then the blame sits with those voters.

[Dolphyn]

4. I've heard this called the "ballot shuffle"

and it has been used to explain why some precincts showed extremely high vote counts for some minor candidates. For example:
http://www.dailykos.com/story/2004/12/2/44647/2043

Someone on DU posted an "interesting" graph a few days ago, showing third-party votes as a function of the number of precincts sharing a polling location. Just don't ask me to find the link, LOL.

[mulethree]

8. can look for author jmknapp

He has lots of other great 'visual aids' that illustrate
these 'whole lotta numbers' problems well.

[AirAmFan]

9. THANK YOU! I KNEW other people had to have been thinking about this

also. Does the statistician working for Arnebeck and Jackson know about this issue? Is he doing anything about it?

A thorough analysis of the 'caterpillar' or 'shuffle' would start with a list of all polling places where more than one precinct voted. Then a square table could be constructed for each such polling place, with the precinct ID's running down the rows "R" and across the columns "C"

The cells of the table would show which candidate would be awarded a vote when a voter punched 'KERRY' on a precinct "C" machine using a card stamped for precinct "R". The diagonal entries all would be "KERRY", because the card-stamp and machine precinct IDs would match on the diagonal. But off-diagonal elements would include Bush, the Greeen candidate, the Libertarian, and anyone else on the ballot for President in any of those precincts.

The pattern of wrong votes would not be randome, but would be a determinate function of the ballot-orders assigned to each of the precincts co-located at a particular polling place.

[AirAmFan]

5. From what the Cleveland Plain Dealer said Friday (URL in my lead-in),

I gather that the punch cards are identical for all precincts, but that each punch card is supposed to have a precinct number stamped on the back. The only printing of candidates names is done on guide pages that are affixed to each machine. It is these guide pages that are online in the "pdf" subdirectory I list in post #3. I haven't found mages of the punch cards themselves yet.

One way mischief could be created with the connivance of the Secretary of State would be not to have any precinct number at all stamped on a certain percentage of the punch cards distributed at polling places with multiple precincts and multiple ballot-orders. Then arbitrary precinct numbers could be stamped on the cards AFTER the voters have returned them. Properly-stamped cards would also have to made available for experienced voters who would discover the "mistake", but novice voters well might succumb to the no-stamp gambit.

If the Secretary of State's office wanted to make this kind of error impossible, it could do so very easily: Just rotate candidates' names BY POLLING LOCATION, NOT BY PRECINCT, so that all co-located precincts would have the same ballot-order. Then a voter would have to take a punch card form one polling place and punch/cast it in another polling place.to vote for the wrong candidate. Even then, there would be only an 80 percent chance of error, if the second polling place were chosen at random.

I find it difficult to believe that this complicated, error-prone setup arose innocently, with no connivance from the Secretary of State's office.

[tinfoil_beret]

10. I noticed the wording in the Cleveland Plain Dealer.

I couldn't determine differently from the Cuyahoga PDF files.

With different patterns on identical ballots for different precincts at the same polling location, this polling method has the same faults as the paperless electronic polling machines. It leaves absolutely no means to identify systematic errors or fraud.

High ratios of third-party votes in individual precincts (as documented elsewhere) suggest mistabulated votes and demonstrate the fundamental flaw of this polling method.

They might as well have randomly punched these ballots.

[Higans]

6. Brilliant

How many ways must we expose the shell game before some one calls it for what it is?

[AirAmFan]

13. Why isn't this story about the "caterpillar/shuffle" ballot in Ohio as big

as the 2000 story on the Butterfly Ballot?

Does the statistician working for Arnebeck and Jackson know about this data issue?

How long has this been going on in Ohio? Was the "rotation" introduced since the "Help America Vote Act" of 2002? Who is responsible for the decision to rotate by PRECINCT, and not by POLLING LOCATION, the latter being a trivially simple way to prevent this very foreseeable problem?

Do Krugman, Blumner, Conason, Scheer, Ivins, Corn, Hightower, Palast, and the other progressive MSM columnists linked at http://www.buzzflash.com know about this? This could be the story of the year!!

[OutLiarJD]

41. Indeed. Well said.

you put me in mind of Bob Dylan, Higans.

except that all the media plays nowadays is "The Sound of Silence"...

[Chili]

12. I remember a thread about this...

...around 2 days before the election. I noticed that my Mom's absentee ballot - which matched ballots at the same precinct on election day, you'd think - looked different from the examples on the Ohio BoE site. Kerry's line #, in other words, was jumping around from precinct to precinct. Someone here assured us that this was normal and not to worry. It may be normal, but I can't see how you can avoid also thinking that it's stupid. For heaven's sake, how difficult is it to have every ballot look like every other ballot, at least in the same POLLING LOCATION? Or at least in the same COUNTY? Or, God forbid, in the same STATE?

WHO in the hell plans these things! Honestly, not trying to be flippant, but it defies common sense. I'm sure it wasn't designed to invite fraud, but if it's so easily screwed up unintentionally, anyone intending to create chaos could do so by merely switching ballot stacks from one table to another.

[jamboi]

14. Great expose! n/t

[AirAmFan]

17. Thanks! Be sure to NOMINATE this thread for homepage, if you want

more people to see it today! Also, you might send the URL and a descriptive paragraph to the progressive columnists listed at the end of the http://www.buzzflash.com homepage, such as Joe Conason, Paul Krugman, Ellen Blumner, Robert Scheer, etc.

[Cronus Protagonist]

15. How about a "ROTATING BALLOT"?

Seems a little more direct and understandable, no?




http://brainbuttons.com/home.asp?stashid=13
Buttons for brainy people - educate your local freepers today!

[AirAmFan]

18. "Give it a name!"

[AirAmFan]

20. I was trying to be witty and referential by calling it a "caterpillar", since

a caterpillar is an inchoate butterfly !

[Cronus Protagonist]

22. I realize that, but you want to reach the masses, don't you?

Why confuse the issue with pretty metaphors when one can be more direct?

[skids]

56. Caterpillar is great for the masses.

A "rolling ballot" sounds like a bland technical term.

What the masses want, is something trendy. I think it's a great name.

[AirAmFan]

19. LET's BEGIN analyzing the 'caterpillar ballot' problem at the first polling place

listed for Cuyahoga County.

This is just the start of "caterpillar/shuffle/rotation" ballot problem analysis for one of 583 polling places in one of 88 counties. So a thorough study of the problem would repeat our sample analysis thousands of times for all the other polling places in Cuyahogoa and all the other counties where it could have happened.

Look at the list of precincts by polling place in the URL mulethree gave us in post #7, at http://boe.cuyahogacounty.us/BOE/PDF/votinglocations.pdf

The first of 583 polling places on the list is Bay Presbyterian Church, in Bay Village OH 44140.

FIVE different precincts voted together at this church. The punch machines at this location thus were set up with five different ballot orders.

Look at the second pages of the sample ballots for these precincts at the following URLs:

http://boe.cuyahogacounty.us/BOE/ballots/pdf/BAYV01A.pdf ---> BWKDP http://boe.cuyahogacounty.us/BOE/ballots/pdf/BAYV01B.pdf ---> WKDPB http://boe.cuyahogacounty.us/BOE/ballots/pdf/BAYV01C.pdf ---> KDPBW http://boe.cuyahogacounty.us/BOE/ballots/pdf/BAYV01D.pdf ---> DPBWK http://boe.cuyahogacounty.us/BOE/ballots/pdf/BAYV01E.pdf ---> PBWKD

Here "BWKDP" indicates Badnarik("B") on line one, Bush("W") on line 2, Kerry("K") on line 3, "Disqualified("D") on line 4, and Peroutka("P") on line 5.

Thus for "Bay Village 01 A", the candidates were listed in alphabetical order (Since the "Disqualified" candidate was Nader, he's listed between Kerry and Peroutka.).

For "Bay village 01 B", the order is shifted up by one compared to precinct 01 A, with Badnarik going to the end of the presidential ballot.

01 C is shifted two places, 01 D three places, and 01 E four places.

Someone with a ballot stamped for 01 A had to punch Kerry on line 3 to vote Democratic. Had she gone to the punch machine for 01 B by mistake, she would have punched Kerry on Line 2, and her vote would have been counted for BUSH. Trying to vote for Kerry using precinct 01 C's machine would have registered a vote for Badnarik. 01 D would have registered for Peroutka, and 01 E for "Disqualified".

The next step in the analysis of this one polling place would be to look at the vote tally to see whether "Disqualified" and minor candidates got many more votes than would have been expected given the demographics and political leanings of Bay Village and the history of its vote for minor candidates in past elections.

[AirAmFan]

21. Continuing, final precinct vote tallies are in a HUGE text file at URL

http://boe.cuyahogacounty.us/boe/results/history/2004/EL52.TXT .

downloadable from http://boe.cuyahogacounty.us/boe/results/canvas.htm .

A note at the top of the file indicates that 1458 of 1458 precincts have reported and are included in this file as of today (12/13/04).

[AirAmFan]

23. But the presidential results are confined to a relatively small portion of that

file, fewer than 2 percent of its lines.

About a quarter of the way through the file, the names of the candidates for President appear. A line for each of the 1458 precincts follows, with tallies for Badnarik(B), Bush(W), Kerry(K), Disqualified(D), and Peroutka(P). Interspersed are ward and city subtotals.

For a very informative example of the apparent miscounting of Kerry votes due to the "caterpillar ballot", take a close look at Precincts numbers 1806 ("Cleveland 04 F") and 1814 ("Cleveland 04 N"). They voted together at Benedictine High School on MLK Jr Drive in zipcode 44104.

Even though some votes must have been cast for "Disqualified", and those tallies would have been quite valuable for the kind of analysis we're doing, the November 30th version of the file zeroes out that column completely. The reported results for this polling place are thus

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

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?

As you'll see in my next post, a very clear explanation derives from a look at the sample "caterpillar" ballots for these two precincts.

[AirAmFan]

24. Take a look at the ballot orders for the presidential candidates in these 2

precincts, on the appropriate sample ballots at the county BOE website:

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

In 4F, you had to punch a hole next to Kerry's name on line 1. If you received a punchcard with "Cleveland 4 F" stamped on the back, but erroneously used Precinct 4N's equipment, you would have punched a hole next to Kerry's name on line 3. When your ballot was routed to the counting machine for precinct 4F, your 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, bu 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, bu 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.

In all, there are 447 multi-precinct clusters involving a total of 1322 precincts, just in Cuyahoga County alone. Since punch cards were used to count 70 percent of Ohio's ballots this year, and optical scan equipment is subject to very similar "caterpillar ballot problems", the statewide problem could be HUGE, accounting for tens or hundreds of thousands of miscounted votes.

[AirAmFan]

25. Similar results nearby, and a precise ESTIMATE of the miscount

A few blocks down King Drive from Benedictine High School is Mt Haven Baptist Church, where Pecincts "Cleveland 03 B" and "Cleveland 03 I" voted together. Their results were

Precinct B_ W_ K__ P__
Cleve3B 44 18 471 1
Cleve3I 1 13 467 72

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

Ballot orders are

http://boe.cuyahogacounty.us/BOE/ballots/pdf/CLEVE03B.pdf ---> KDPBW
http://boe.cuyahogacounty.us/BOE/ballots/pdf/CLEVE03I.pdf ---> PBWKD

As at the high school, the overwhelming majority of voters at this location chose John Kerry. To do so in precinct 3B, 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 3 B" stamped on the back, but erroneously used Precinct 3I's equipment, they would have punched a hole next to Kerry's name on line 4. When the ballot was routed to the counting machine for precinct 3B, the 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 3B: Badnarik was credited with 44 votes, while Peroutka got one and Bush got 18. The "caterpillar" explanation of this vote would be that the votes for Bush and for Peroutka in 3B are accurate, but that virtually all of Badnarik's 44 really were miscounted KERRY votes.

Similarly, In 3I, you had to punch a hole next to Kerry's name on line 4. If you received a punchcard with "Cleveland 3I" stamped on the back, but erroneously used Precinct 3B'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 3I, your 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 3I. Peroutka was credited with 72 votes, while Badnarik got 1 and Bush got 13. The "caterpillar" explanation of this vote would be that the votes for Bush and for Badnarik in 3I are accurate, but that virtually all of Peroutka's 72 really were miscounted KERRY votes.

Since the two precincts here are of roughly equal size, it seems reasonable to apply the accurately measured minor candidates' totals to BOTH precincts. Replacing the "44" with "1" makes the 3B estimated Kerry miscount 43. Replacing the "72" with "1" makes the 3I estimated Kerry miscount 71. Thus a reasonable estimate of the number of votes Kerry lost here because of the caterpillar ballot is 114.

[smartvoter]

26. HEY!!!! Here is a thread analyzing its effect. It's DAMNING as hell!

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x166913

[AirAmFan]

27. THANK YOU! We need to unify the analyses in these two threads.

THIS thread essentially answers the question raised in post #13 of the other thread.

[AirAmFan]

28. Proposed division of labor: Continue "bottom-up" approach here, and

continue "top-down" approach at the other thread.

To avoid wasteful duplication of the same work, the left hand should know what the right hand is doing. While the other thread seems to take a top-down approach, I'm definitely taking a bottom-up approach.

Among the 583 polling places in Cuyahoga county, I counted 136 single-precinct locations. That leaves 477 clusters of from two to ten precincts, accounting for 1322 precincts.

My strategy is to provide a general framwork for looking at clusters, and then use it to solve the case of clusters of size two, clusters of size three, etc. I'll pick out examples of each cluster size to illustrate the principles involved.

[Freddie Stubbs]

29. Do we have any evidence that Bush's name was on the same line

regardless of precinct? That would definitley show that there was bias.

[AirAmFan]

31. No, Bush's name moved from line to line just like Kerry's

There were only five different ballot orders:

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).

Page 2 of each sample ballot shows which of the 5 ballot orders each precinct had. All 1458 sample ballots still are up at the BOE website.

I realize how confusing all this is. Try clicking on the first link in post #24 to look at a real sample ballot. What I'm trying to say in this post then may become clearer.

[Freddie Stubbs]

32. So it wasn't a purposeful attempt to hurt Kerry?

:shrug:

[AirAmFan]

35. It's hard to say what the Republicans intended here. I find it hard to believe

such a stupid procedure was put in place innocently. I don't even know how long things have been this way in Cleveland. Was this "ballot rotation" put in place as part of so-called "election reform" since Bush was selected for the White House? Sonce Taft became Governor? Since Blackwell became SOS? Maybe somebody who reads this and isn't into math and statistics could research this aspect of the Caterpillar Ballot, while we gearheads crunch the numbers.

For the most part, I think the miscounting because of the "caterpillar ballot" arose from overcrowding of pro-Kerry polling places, and there's evidence the overcrowding WAS deliberate. The Washington Post story on the DU homepage yesterday said that, in Columbus, "27 of the 30 wards with the most machines per registered voter showed majorities for Bush. At the other end of the spectrum, six of the seven wards with the fewest machines delivered large margins for Kerry. Voters in most Democratic wards experienced five-hour waits".

If there was a similar imbalance in overcrowding in Cuyahoga, the most likely way for miscounting to have arisen would have been lines of uneven length for different precincts at the same location. If you had to get back to work, the line you were in was 30 people deep, and there was a punch machine with no line across the room, wouldn't you go over there to punch out your chad?

On the other hand, tricksters could have switched piles of ballot cards from one table to another, when the county election staff and volunteers weren't looking. Or officials could have done something to screw things up deliberately. The problem with the Caterpillar Ballot is that it offers so many possibilities for mishap and for mischief.

[AirAmFan]

30. General framework for 'Caterpillar ballot' clusters: Part One

How many ways could a Kerry voter erroneously mark a ballot for another candidate using a wrong machine?

This question came to me while looking at the 171st of 583 Cuyahoga clusters.

It involves THREE precincts co-located at Cory United Methodist Church in Cleveland 44108.

Results were as follows:

Precinct B_ W_ K__ P__
Cleve8G 51 20 232 1
Cleve8H 0 45 343 2
Cleve8I 0 8 251 28

Ballot orders were as follows:

http://boe.cuyahogacounty.us/BOE/ballots/pdf/CLEVE08G.pdf ---> WKDPB
http://boe.cuyahogacounty.us/BOE/ballots/pdf/CLEVE08H.pdf ---> KDPBW
http://boe.cuyahogacounty.us/BOE/ballots/pdf/CLEVE08I.pdf ---> DPBWK

Ways Kerry voters could have gone wrong were as follows

8G ---> 8H:W, 8I:B
8H ---> 8G:D, 8I:B
8I ---> 8G:P, 8H:D

I've used a shorthand notation here. The first line says, consider a voter with a punchcard stamped for 8G. There are two wrong machines, 8H and 8I. Using 8H would result in a vote for Kerry on line 1, which would register for Bush. Using 8I would result in a vote for Kerry on line 5, which would register for Badnarik.

In all, there are SIX wrong possibilities for registering a vote whenever the cluster size is three. Whenever the configuration is WKD = (WKDPB,KDPBW,DPBWK),

the values of the following variables are as follows

wmbo_KB = 2 (there are two wrong-machine Kerry ballot orders matching Badnarik)
wmbo_KW = 1 (there is one wrong-machine Kerry ballot order matching Bush)
wmbo_KK = 0 (there is no wrong-machine Kerry ballot order matching Kerry)
wmbo_KD = 2 (there are two wrong-machine Kerry ballot orders matching Disqualified)
wmbo_KP = 1 (there is one wrong-machine Kerry ballot order matching Peroutka)

If all these wrong-machine ballot orders are equally likely, note that getting Badnarik by mistake is twice as likely as getting Peroutka by mistake. Now take a look at the results: In fact Badnarik DID get roughly twice as many votes as Peroutka, 51 compared to 31! The Cuyahoga BOE has censored the vote for "Disqualified", but the theory being developed here would estimate that the vote for disqualified was closer to the 51 than to the 31. Also, the number of votes switched from Kerry to Bush in error would be close to 31.

[AirAmFan]

33. Whoops! By mistake, I've discovered the DU codes for two happy faces

I meant:

8G ---> 8H: W, 8I: B
8H ---> 8G: D, 8I: B
8I ---> 8G: P, 8H: D

[AtLiberty]

34. Why is voting in America...

...rocket science?

[AirAmFan]

38. Because incumbents often "practice to deceive", as the old saying goes

Have you ever seen Alexander Keyssar's unique history of US voting, called something like, "The Right to Vote: The Contested History...". He shows how ingenious those in power have been when it comes to whittling back voting rights won by successive groups of Americans grudgingly admitted to involvement in public decisionmaking.

"Felon disfranchisement" laws (like those still enforced vigorously in Florida) were passed in the South shortly after the Civil War explicitly to take away voting for conviction of those crimes thought to be largely the province of African-Americans. Voter registration was initiated explicitly to curb African American and immigrant voting. The closure of the registration rolls a certain period of time before an election (now 30 days except in a few states) was meant to take away voting rights from the poor (a southern official said, "the Negro is a nomadic tribe").

These barriers ot fairness in voting have been so ingenious and subtle that its at the very least tedious even to uncover their effect.

[jmknapp]

57. Well, this is analysis of voting results

And messed-up voting results at that. So it gets complicated.

But the act of voting isn't as complicated.

[AnIndependentTexan]

36. Tod A. Rapp sold buttefly ballots

http://www.democraticunderground.com/demopedia/index.php/Triad_Governmental_Systems,_Inc.

Overview
Founded by Tod A. Rapp. Based in Xenia, Ohio. A family-owned company managed by his sons. Rapp is a financial contributor to the Republican Party and the campaign of George W. Bush. He is is also associated with (http://www.opensecrets.org/indivs/search.asp?NumOfThou=0&txtName=rapp&txtState=OH&txtZip=&txtEmploy=&txtCand=&txt2004=Y&txt2002=Y&txt2000=Y&Order=N) the companies Rapp Systems and Rapp Systems Corporation.

Voting Systems
Triad sells punch card voting systems and wrote the computer program that tallies the punch-card ballots in the centralized counting systems used in 41 counties in Ohio. Ohio granted them certification in 1990. (http://www.google.ca/search?q=cache:_k3L_IiymRAJ:www.votingindustry.com/Ohio_Corner/ohio%2520certified%2520voting%2520machines.pdf+Triad+Governmental&hl=en) Their Y2K readiness disclosure for the VIS Voter Information System Release 6R is here. (http://www.rappsystems.com/year2000/VISY2KReadinessDisclosure.html) and for the Electab here. (http://www.rappsystems.com/year2000/ELECtabYear2000.html)

Triad appears (http://fox.wikis.com/wc.dll?Wiki~WhoUsesFoxPro) to use Fox Pro as a development tool.

In 1993 Triad was awarded (http://64.233.179.104/search?q=cache:hqtRZ3Uc3OYJ:www.co.gilchrist.fl.us/minutes/html/1993-05-03_bocc.html+%22Triad+Governmental+Systems,+Inc%22+bush&hl=en) a contract to upgrade the software systems in the elections office in Gilchrist County, Florida.

Rapp Family
The family (http://64.233.161.104/search?q=cache:t_d8cDtMjFIJ:www.edgingtonfuneralhomes.com/FuneralNoticeDisplay.asp%3FName%3DMary%2520Dalena%2520Guthrie+Brett+Rapp+Xenia&hl=en) includes daughter (or niece?) Cheryl Belluci, Dwayne Rapp, (married to Michelle) and Brett Rapp (married to Chris).

Cheryl Bellucci (http://www.foxite.com/archives/0000010046.htm) is or was the Vice President of Rapp Systems Corporation. She is or was a Visual Fox Pro programmer. She made an interesting comment (http://www.taxidermy.net/forums/IndustryArticles/9912B81EEB.html) on an internet forum about counting ballots in an election in Alaska in 1999.

Connections
The web page http://www.rapps.org/ is registered to Tod Rapp: Domain ID:D1370314-LROR Domain Name:RAPPS.ORG Created On:04-Jun-1998 04:00:00 UTC Last Updated On:15-Jan-2004 21:14:04 UTC Expiration Date:03-Jun-2005 04:00:00 UTC Sponsoring Registrar:Go Daddy Software, Inc. (R91-LROR) Status:OK Registrant ID:GODA-05003276 Registrant Name:Tod Rapp Registrant Organization:Rapp Systems Corporation

Tod Rapp is the president of the Psephos corporation, whose "Elections General Store" (http://www.psephoscorp.com/) sells the "Palm Beach County Florida "Butterfly" Ballot Pages, Ballot Cards & Page Holder Kit" to the public, apparently as souvenirs. Said Rapp: (http://www.sptimes.com/News/011101/news_pf/Floridian/A_fad_called_chad.shtml) "We have a bunch of used vote recorders in stock, and this looked like a way to get some of them sold." The domain psephoscorp.com is registered to Tod Rapp.

Tod Rapp appears in the acknowlegement section (http://www.votefraud.org/saltman_roy_1988_report.htm) of the Roy Saltman report on "Accuracy, Integrity, and Security in Computerized Vote-Tallying," NBS Special Publication 500-158.

[AirAmFan]

37. Do you know how long "ballot rotation" has been going on in Ohio?

Do you know how long Cleveland has been rotating ballots WITHIN polling places, rather than jus ACROSS polling places? I find what they're doing INCREDIBLY STUPID at best, or ingeniously devious at worst.

[jmknapp]

58. I don't know but

to rule that ballots within precincts must be of the same order (due to the possibility of mix-ups) would be distasteful to the election boards I think, because the ballot may differ between precincts in other ways (different local issues, etc.). So they have to pretend that somehow these switches can be prevented in the first place.

[AirAmFan]

66. I trust your intuition on this, but it seems to me that such a reaction

to learning about "caterpillar crawl" would be a prime example of bureaucrats hunkering down behind desks and file cabinets. What a dysfunctional official culture!

[AirAmFan]

39. General framework, Part Two: Ways Bush voters could go wrong

whenever the cluster size is three and the configuration of ballot orders is WKD = (WKDPB,KDPBW,DPBWK) (just as in the location discussed in post #30), there are six ways people who intended to vote for BUSH could have gone wrong, had their polling places been overcrowded or chaotic, or had a trickster or an official or volunteer switched piles of stamped cards or voting machines:

8G ---> 8H: B, 8I: P
8H ---> 8G: K, 8I: B
8I ---> 8G: D, 8H: K

I've used the same shorthand notation here. The first line says, consider a BUSH voter with a punchcard stamped for 8G. There are two wrong machines, 8H and 8I. Using 8H's punch would result in a vote for Bush on line 4, which would register for Badnarik. Using 8I would result in a vote for Bush on line 3, which would register for Peroutka.

Thus the values of the following variables are as follows

wmbo_WB = 2 (there are two wrong-machine Bush ballot orders matching Badnarik)
wmbo_WW = 0 (there is no wrong-machine Bush ballot order matching Bush)
wmbo_WK = 2 (there is no wrong-machine Bush ballot order matching Kerry)
wmbo_WD = 1 (there are two wrong-machine Bush ballot orders matching Disqualified)
wmbo_WP = 1 (there is one wrong-machine Bush ballot order matching Peroutka)

Notice that the values of the 10 wmbo_?? variables, both the 5 for Kerry voters and the 5 for Bush voters, all flow from the configuration of ballot orders at a polling place.

For clusters of two precincts there are in principle 25 = 5 x 5 possible configurations. In practice, all the two-precinct clusters I've seen so far have had two different ballot orders, so that the 5 configurations BB,WW,KK,DD, and PP have not arisen in practice. A tedious exercise for anyone attempting a complete analysis would be to derive the values of the ten variables for each of the 25 possible two-precinct configurations.

For clusters of three precincts there are in principle 125 = 5 x 5 x 5 possible configurations. In practice, all the three-precinct clusters I've seen so far have had three different ballot orders, so that only 60 = 5 x 4 x 3 configurations might have arisen in practice this year. A tedious exercise for anyone attempting a complete analysis would be to derive the values of the ten variables for each of the 125 possible three-precinct configurations.

Similar analyses could be done for 625 possible four-precinct, 3125 possible five-precinct configurations, and so on for six-, seven-, eight-, nine-, and ten-precinct configurations.

But, since Cuyahoga only had 477 clusters, it likely would be more efficient to calculate the 10 variables only for those clusters of size four or more that actually appear in the data, unless someone writes a piece of code to do it.

[AirAmFan]

40. Part Three: Using "Disqualified" to estimate the total number of miscounts

Let c represent the number of precincts in a cluster. Then there are c*(c-1) ways to use a wrong machine. For example, when c=2, there are two ways to use a wrong machine: A voter with a punchcard from precinct 1 could use the punch machine for precinct 2, and vice versa.

When c=3, there are 3*2=6 wrong possibilities, as shown in the example (discussed above) of Cleveland 8G, 8H, and 8I.

Assume that any presidential votes recorded for "Disqualified" (numbering n(D) in all for a cluster) are completely unintentional, and that all unintentional votes are recorded by voting on a "wrong" machine.

If all voters intended to vote for Kerry, an estimate of the total number of ballots punched on "wrong" machines would be

n(D)*c*(c-1)/wmbp_KD , where wmbo_KD is the number of wrong-machine Kerry ballot positions matching "Disqualified". This total error would be allocated to erroneous votes for the other candidate according to the values of the wmbp_K? variables, and then an estimate of the revised totals could be obtained.

On the other hand, if all voters intended to vote for Bush, an estimate of the total number of ballots punched on "wrong" machines would be

n(D)*c*(c-1)/wmbp_WD , where wmbo_KD is the number of wrong-machine Bush ballot positions matching "Disqualified".

Take m(K) = n(K)/(n(K)+n(W)) = the cluster's proportion of major-party votes for Kerry as a measure of the probability a random voter intends to vote for Kerry. Then the expected value of the total number of erroneous votes is

n(D)c*(c-1)*<((m(K)/wmbp_KD)+(((1-m(K))/wmbp_WD)>.

For the 3-precinct cluster example discussed in previous posts, this works out to be

n(D)*6<((753/826)/2)+(73/826)/1)> = n(D)*6<0.456+0.004> = 2.76*n(D).

When the number of votes for "disqualified" is not observed (as in the results file on the BOE website), further progress requires additional assumptions. The most fruitful strategy would be to specify some maximum number of votes for minor candidates possible considered accurate rather than as reflecting miscounts.

[AirAmFan]

42. Empirical issue: Just how high is "too high" for minor candidates?

Overall, Badnarik got 1886 votes, 0.28 percent of the Cuyahoga total, and Peroutka got 1751, 0.26 percent of the total. Within the Benedictine High cluster alone, Badnarik got 164 votes, 15.6 percent of the cluster total, and Peroutka got 225, or 21.3 percent.

Clearly the Benedictine percentages for minor candidates were "WAY too high". But precisely how many of those Badnarik and Peroutka votes at Benedictine should be considered erroneous?

This is a question about "statistical outliers". One of the most common ways of dealing with outliers is to calculate the mean and standard deviation and then consider any observation 3 or more standard deviations above the mean as an outlier, above any "reasonable" proportion that might by some stretch of the imagination be considered genuine. We can do these calculations later, but for now let's say that the excess vote for Badnarik in any cluster is (1-rB)*n(B), and the excess vote for Peroutka in any cluster is (1-rP)*n(P), where rB is a maximum reasonable proportion for Badnarik and rP is a maximum reasonable proportion for Peroutka. For the moment, let's arbitrarily set rB = 4*0.0028 = 0.0112, and arbitrarily set rP = 4*0.0026 = 0.0104. We can always change these values later, and analyze the sensitivity of miscount estimates to assumptions about these parameters.

Then for Benedictine High, the number of excess votes for Badnarik is tentatively (1-0.0112)*164 = 162.13, and the number of excess votes for Peroutka is tentatively (1-0.0104)*225 = 222.66.

[L. Coyote]

47. Just how high is "too high" for Disqualified (read Non-Votes)?

I'm just reading this thread. I've been on the parallel thread at:

http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=166913

I just sorted Cuyahoga precincts by standard deviation from the mean percentage of non-votes (undervotes for president), then analyzed the two groups for Kerry vote.



There is a correlation of non-votes and pro-Kerry precincts.

There are 3,637 3rd party votes in Cuyahoga and 13,531 non-votes.

For absentee ballots: Percent of non-votes = 3.51
For the precincts: Percent of non-votes = 2.03

Statewide, the percent of non-votes = 1.66

[AirAmFan]

48. Thanx4 the insight &the graphic. Nonvotes= votes for Disqualified +under-

undervotes + overvotes + write-ins, in Cuyahoga as in the other 67 punchcard counties, according to mulethree in post #66 on Iceburg's parallel thread.

Overvotes have more than one punch in the Presidential column, while undervotes include pregnant and hanging chads as well as smooth intact ballots with no discernable indentation for President.

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. Most of these are quite likely to be undervotes rather than overvotes or votes for DISQUALIFIED.

Each component of nonvotes has an associated set of theories of their origin.

As I pointed out in post #83 on the other thread, my favorite theory of undervotes is inferior equipment shunted off on poorer neighborhoods, for the same reason public schools get less experienced teachers there, sanitation workers leave more trash on the streets there, police more readily use their weapons there, etc.

We have at this point no disaggregated data on the components of non-votes, though Joe Knapp's correspondence with Michael Vu (post #67 on Iceburg's thread) left him somewhat optimistic that partially disaggregated non-vote data may become available eventually.

Correlations of nonvotes with Kerry votes shed light on the undervote component, but I'm more interested in the DISQUALIFIED component.

To test the bottom-up "caterpillar ballot" theory developed in this thread, I'd be most interested in the correlations between the DISQUALIFIED portion of the non-vote and three other variables: BADNARIK, PEROUTKA, and BADNARIK+PEROUTKA, both in the data on 1436 precincts and in the data on more than 400 polling places (clusters of precincts), by cluster size.

In addition, I've developed a theory of the inverse of the ratio between DISQUALIFIED and each of the other three variables I've mentioned (Invert because there are so many precincts and clusters with zero votes for minor candidates).

While we're waiting for more disaggregated data, the total nonvote in each precinct and cluster is an interesting substitute. Please keep up the good work!

[L. Coyote]

50. Non-Votes are an indicator ...

I'm using Votes Cast minus votes for candidates. I realize that the reasons for non-votes numbers are complex, and not all can be considered as "caterpiller crawl" to disqualified. (Or pseudo-Nader voting. What was the exit poll number for Nader? Like zip!)

They do point to which precincts to examine in detail, and should be analyzed on a precinct by precinct basis. Question 1: In which locations does a Kerry vote register for Bush while a Bush vote registers for disqualified? Start with the precinct list I placed in another thread....

The precinct group with over 1.5 standard deviations certainly points to this 'phenomena' happening at a high rate in precincts with the highest Kerry support.

You write: " I'd be most interested in the correlations between the DISQUALIFIED portion of the non-vote and three other variables: BADNARIK, PEROUTKA, and BADNARIK+PEROUTKA..."

Bearing in mind the low numbers and zero votes, there are certainly very strong and interesting inverse correlations of non-votes with percentages for Badnarik and Peroutka in the >1.5 precincts. And with turnout a different picture emerges. In all cases, the differences between the >1.5 and <1.5 precincts are noteworthy.



I guess this is my MAIN POINT. If you were to do this sort of thing on purpose to fraudulently influence the election outcome, you would target the precincts where Kerry has the highest proportion of support. This way, the ratio of "Kerry votes that tally as Bush votes" to "disqualified" or the 3rd party candidates is as high as possible. In other words, as undetectable as possible. If the precinct was a 90 % Kerry precinct to begin with, the yield ratio is minus 9 Kerry, plus 9 Bush, minus 1 Bush, plus 1 disqualified.

Note below the mean and median percentage of Kerry vote (and their difference) for the high stdev non-vote precincts. They are in the 90 percent range, with an even higher median.



Do the math given 13,531 non-votes in just this one county. And start with the Statewide E-Vote percent non-votes as a baseline, 0.760 percent. The mean percent in the >1.5 group is 5.573 percent.

To get all this data in one message, here is the statewide stats image again (pre-certified).



The Republican challengers had an agenda coupled with this understanding. As such, it seems to constitues actual fraud to me. Given the court rulings, I'd call it "legalized election fraud." Cuyahoga had 3.73 percent provisionals, compared to the all county median of 2.28 percent. This certainly contributed to the "Cuyahoga Chaos."

You write: "Invert because there are so many precincts and clusters with zero votes for minor candidates."

I dropped a precinct showing no votes. I had to substitute 0.000000000001 for zero values to make percentage divisions work. This solved the zero quantities dilemna.

Back to the numbers. And maybe, pulling some of this into a Web page.

[AirAmFan]

53. Re q #1: What stat software pkg are you using? CODING TASK

It sounds as though you want to do substantial analysisi of configurations of ballot orders at different precincts. Thus you need a way to code up the eight theoretical variables I've defined for each clustered polling place configuration:

wmbp_KB (number of wrong-machine Kerry ballot positions matching Badnarik)
wmbp_KW
wmbp_KD
wmbp_KP

wmbp_WB (number of wrong-machine Bush ballot positions matching Badnarik)
wmbp_WK
wmbp_WD
wmbp_WP

I've just begun to think about how to do this. In SAS, there is an INDEX function, which gives the numeric position at which a given substring begins in a character variable. For example, consider the infamous Benedictine High ballot orders. For precinct "Cleveland 4 F", we could define a variable called RIGHT that would give the ballot order that will be used to count ballots stamped "Cleveland 4 F" on the back:

RIGHT = 'KDPBW'

If a voter went to the other precinct's machine by mistake, the ballot order would have been

WRONG = 'BWKDP'

I believe the SAS to discover that the right ballot position for Kerry is the FIRST would be

KPOSR = INDEX(RIGHT,'K')

Similarly we would define

BPOSR = INDEX(RIGHT,'B')
WPOSR = INDEX(RIGHT,'W')
DPOSR = INDEX(RIGHT,'D')
PPOSR = INDEX(RIGHT,'P')

Now, I haven't tried this out yet, but I believe SAS notation for the candidate's initial corresponding to Kerry's ballot position on the WRONG machine could be

KMATCHW = SUBSTR(WRONG,KPOSR(RIGHT,'K'),1)

Cluster size at Benedictine is only two, so arrays of RIGHT and WRONG would have to be defined to handle clusters of sizes 3 to 10.

The wmbp_?? variables could be coded as counts in arrays of ?MATCHW variables. This task is not yet in my critical path.

Are you using SAS? If not, does your package have functions with the same capablility as INDEX() and SUBSTR() in SAS? Or could somebody code up functions that provide this capacity n your package?

Sometime within the next few days I'll get back to this, if nobody has posted code for this task in the meantime.

[AirAmFan]

65. Which Disqual votes were meant for Nader, as a protest? The most

promising, but also the most tedious, approach to this question might use information from other races. See http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=176358&mesg_id=176358

[AirAmFan]

49. Another approach to 'reasonableness': Use nearby single-precinct

polling places to try to isolate the effect of clustering precincts, without introducing geographical biases. This idea was inspired by mulethree's post #66 on the other thread.

One question I have is whether zipcode would be a good variable for dividing Cuyahoga county into clusters of polling places with the same demographics and propensities to vote for minor-party candidates, once outliers like Benedictine High are removed from the calculation of minor-candidate proportions. Or would it be necessary to read pollling place addresses into a GIS address module to get a finer level of geographical detail, such as census tract or census block group? Joe, do you know the answer to this question? You're definitely the Cuyahoga GIS maven on this board.

[mulethree]

51. Map source

http://boe.cuyahogacounty.us/boe/pollmaps/countywide.htm

I'd convert it into a jpg or similar image file.

Then you get your demographics info. You can look up "census Tiger map" for a system at the census bureau that makes fairly crude maps.

These maps are a little better resolution and come up pretty decent in regards to colors and scale of values.
http://216.55.182.132/FairData/SocioEconMapper/map.asp?command=scope&annot=%7C&vis=2%7C1%7C1%7C2%7C2%7C2%7C1%7C1%7C&map=0&width=0&table=&scope=-81671591%7C41479497%7C11.377624%7C10.322608&zoomrect=&panshift=&ml0=2&ml3=2&ml4=2&ml5=2


The tiger maps give you control over colors so you could - for instance make a green map of income and a blue map of race and a red map of education and overlay them.

So now you need to take your graphics program and find a way to make layers and handle partial transparency and you can lay the wards or precinct map on top of a demographic map.

I wonder why our splendid government can't have zip code = census block = precinct or some such correlation?

Cuyahoga is great in having precinct map, precinct votes, poll locations and ballot orders for this. It looks like 0.3% of the county got mangled by this. But it's probably corrupted some issue votes.

[L. Coyote]

52. Map of Precincts with >1.75 StDev in Non-Voting Percentage

Non-vote precincts above 2.5 percent show a concentration centered at:

	Average Locations	
StDev	Lat	Lon
>1.75	41.491	81.627
>1.5	41.493	81.630

and the worst offender is:

OPEN DOOR BAPTIST CHURCH (41.49° N., 81.62° W.) with 19.72% Non-Votes of 462 voters. So I used Mapquest.com to view where this centerpoint of the concentration is located. (Dutch Alley? I don't know Cleveland.) Here is the Mapquest map online.

Below is the scatterplot of the concentration of >1.75 stdev non-voting precincts. One-tenth degree N-S equals 6.9 miles, or 11.12 km.



Two suburban precincts were cropped to reduce file size. All but four of the >1.75 stdev precincts are within about a 10 mile radius of Open Door Baptist Church.

I thought it might be useful to illustrate exactly the Bush vote in the precincts with the highest non-vote percentages. Here are the suspects in a graph with precinct names, % Bush, and % non-votes.



Any comments from readers who know Cleveland?

[AirAmFan]

54. GREAT! Thank you! Have you seen my post on the NOT-ALL-THE-WAY-DOWN

Hyposthesis? It's #99 on Iceburg's thread, and it has what I think may be a crucial insight about miscounts in single-precinct polling places as well as multi-clusters.

[jmknapp]

60. Another way of looking at the non-vote distribution

[L. Coyote]

63. The patterns are a match. Nice map Joe. n/t

[AirAmFan]

55. Another SUPERB GIS resource that just happens to focus on the Cleveland

Cleveland area is 'CANDO' at Case Western Resuerve U, URL http://povertycenter.cwru.edu/cando.htm .

A few years back, this was a state-of-the-art webpage, way ahead of its time, where you could specify the neighborhood demographic variables you wanted. Clicking SUBMIT would trigger a custom SAS run to produce input for the maps you wanted. I haven't used it in a long while, but I'd surmise that it still has this capability and has been upgraded. I don't know whether the data are limited to Cleveland proper or include all of Cuyahoga county, so that's the first thing to check out.

[jmknapp]

59. There is a lot of desire for pct-level demographic info

Which I unfortunately don't have. Several people are looking hard for this right now! I had two calls yesterday from people working in Cuyahoga County on these issues.

There is so much discrimination. I have heard that over half the people whose provisional ballots were rejected were at the right polling place, but wrong table (precinct)! Democratic challengers were not allowed to help them or direct them. Furthermore, around 20% *were* at the right precinct but their ballot was rejected for wrong precinct anyway. Analysis of these is still going on, but I have heard that most of the cases concentrate in Cleveland (51% black).

So to some extent the municipality can be used for a coarse approximatation to demographics.

[AirAmFan]

61. Joe--I'm pretty sure you'd find every demographic you'd want at URL

http://povertycenter.cwru.edu/cando.htm , for each of 36 "SPAs" in Cleveland proper and for each of 58 suburban municipalities in the rest of Cuyahoga. 'CANDO' has been available at Case Western Reserve U for many years, and is a national leader in gathering neighborhood data and making it accessible, so their "profile" data have been cleaned by day-to-day use of hundreds of people over more than 15 years.

With I first became aware of your expertise in GIS, I suspected you might be affiliated with CANDO yourself.

[AirAmFan]

62. Fastest way to get demographics into a spreadsheet may be this link

used 94 times, one for each of 58 Cuyahoga municipalities and each of 36 Cleveland neighborhoods: http://povertycenter.cwru.edu/doc/profile/profile.html :

"NEIGHBORHOOD PROFILE The Neighborhood Profiles are intended to provide an overview of the demographic, social and economic characteristics in a neighborhood or municipality within Cuyahoga County. The profiles are available for each of the 36 neighborhoods in the City of Cleveland and the 58 municipalities within Cuyahoga County. There are 8 data tables and 41 indicators contained within each profile. ...

The definitions of the indicators and data sources for each of the indicators in the Profile can be view and printed by clicking here. It suggested the user read these definitions in order to understand how these indicators are defined and calculated...In order to view or print a neighborhood profile, choose the appropriate neighborhood or municipality below. Only one neighborhood or municipality can be selected at a time...."

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

There would also be the issue of linking each of the 583 polling place addresses at http://boe.cuyahogacounty.us/BOE/PDF/votinglocations.pdf with one of the 94 neighborhoods. If the CANDO neghborhoods are census tracts or block groups, there are free online pages where street addresses can be mapped a handful at a session, or someone may have ArcView or one of the other heavy-duty GIS address modules. Or somebody who knows Cleveland could take a stab at doing the mapping by eye from the maps on the CANDO site.

[AirAmFan]

43. Using 'excess' votes for minor candidates to estimate total miscount: I

This post answers one very practical question, albeit in a form that may seem very theoretical. But it's just high-school algebra:

In a given cluster, what is the ratio we'd expect to find between votes for "Disqualified" and "excess" votes for minor candidates?

The answer is that it depends on the configuration of ballot orders at the cluster, which as a matter of mathematics generate numbers of wrong-machine ballot positions matching each of the major candidates.

Using the notation in posts #30 and 42 above, if the voter intended Kerry, the ratio is

(n(D) / {(1-rB)n(B) + (1-rP(n(P)}) = wmbp_KD / (wmbp_KB + wmbp_KP).

To make this expression seem less theoretical, take another look at the last paragraph in post #30, where we applied specific values of the wmbp_K? parameters to Cuyahoga voting cluster 171.

If the voter intended Bush, the ratio is

(n(D) / {(1-rB)n(B) + (1-rP(n(P)}) = wmbp_WD / (wmbp_WB + wmbp_WP)

Taking the expected value of the two cases gives us

(n(D) / {(1-rB)n(B) + (1-rP(n(P)})

= m(K){wmbp_KD / (wmbp_KB + wmbp_KP)} + (1-m(K)){wmbp_WD / (wmbp_WB + wmbp_WP)}

[AirAmFan]

44. Estimating the total miscount for a cluster: CONCLUSION!!!

From post #40, the total number of erroneous votes is

c*(c-1)*{((m(K)/wmbp_KD)+(((1-m(K))/wmbp_WD)}*n(D)

Plugging in the estimate of the number of votes for disqualified (n(D)) from post #43

gives us the ugly but very useful expression

{c*(c-1)}*{(1-rB)n(B) + (1-rP)n(P)}*{((m(K)/wmbp_KD)+(((1-m(K))/wmbp_WD)}*

{m(K){wmbp_KD / (wmbp_KB + wmbp_KP)} + (1-m(K)){wmbp_WD / (wmbp_WB + wmbp_WP)}}

Let's go through it in detail, illustrating the concepts with values from cluster 171:

c = Number of precincts in the cluster (3 for cluster 171--see post#30)

rB = a "reasonable" proportion of the cluster's vote that might legitimately go to Badnarik (we've assumed 0.0112)

n(B) = total cluster votes for Badnarik

rP = a "reasonable" proportion of the cluster's vote that might legitimately go to Peroutka (we've assumed 0.0104)

n(P) = total cluster votes for Peroutka

m(K) = the proportion of the cluster's major-candidate vote that was recorded for Kerry (826/899 = 0.919)

The rest of the expression involves parameters derived from the cluster's particular configuration of ballot orders:

If the ballot was intended as a vote for Kerry
wmbp_KB = number of Wrong-Machine Kerry Ballot Positions matching Badnarik (2 for cluster 171)
wmbp_KW = number of Wrong-Machine Kerry Ballot Positions matching Bush (1)
wmbp_KD = number of Wrong-Machine Kerry Ballot Positions matching Disqualified (2)
wmbp_KP = number of Wrong-Machine Kerry Ballot Positions matching Peroutka (1)

If the ballot was intended as a vote for Bush
wmbp_WB = number of Wrong-Machine Bush Ballot Positions matching Badnarik (2)
wmbp_WK = number of Wrong-Machine Bush Ballot Positions matching Kerry (2)
wmbp_WD = number of Wrong-Machine Bush Ballot Positions matching Disqualified (1)
wmbp_WP = number of Wrong-Machine Bush Ballot Positions matching Peroutka (1)

Now all that remains to be done is to allocate the total miscount to the four candidates!

[AirAmFan]

45. Corrected total miscount formula, applied to cluster 171

Taking some redundant punctuation out of the total miscount formula as stated in post #44, it becomes

{c*(c-1)} * {(1-rB)n(B)+(1-rP)n(P)} * {(m(K)/wmbp_KD)+((1-m(K))/wmbp_WD)} *

{m(K){wmbp_KD/(wmbp_KB+wmbp_KP)} + (1-m(K)){wmbp_WD/(wmbp_WB + wmbp_WP)}}

Thus the total miscount in cluster 171 is

{3*2} * {(0.9888*51)+(0.9896*31)} * {(0.919/2)+(0.081/1)} *
{0.919{2/(1+1)} + (0.081){1/(2+1)}} = 6*81.107*0.541*0.946 = 249.057

(Note that wmbp_KW = 2, not 1, and wmbp_KB = 1, not 2 as incorrectly asserted in post #30. This is because the second wrong-machine Kerry ballot position for precinct 8H matches W, not B as incorrectly stated in post #33.)

[AirAmFan]

46. FINALLY, the estimated total miscount is divided up and allocated to the 5

candidates (including the Disqualified candidate).

First, it is split into two shares, using the proportions of major-candidate votes Kerry and Bush were awarded at the cluster. We have denoted these proportions as

m(K) = n(K) / (n(K) + n(W)) and (1-m(K)) = n(W) / (n(K) + n(W)) .

Then each major candidate's share is divided into c(c-1) pieces (where c is the number of clusters), and the pieces are allocated according to the values of the "number of wrong-machine ballot positions" variables wmbp_K? and wmbp_W? .

Continuing with the example of cluster 171, the estimate of the total miscount is 249.057. 91.9 percent of this total, or 228.883, goes to Kerry ballot miscounts. The other 8.1 percent, or 20.174, goes to Bush miscounts.

Because there are six wrong-machine ballot positions in any 3-precinct cluster, divide the Kerry miscounts into 6 pieces, each of size 38.147. Divide the Bush miscounts into 6 pieces of size 3.362 apiece.

Since wmbp_KD = 2 (see the end of post #44), 76.294 extra votes for Kerry are pulled from the pile of spoiled ballots, and in particular that portion of the pile attributable to votes for the Disqualified candidate. 38.147 votes are transferred to Kerry from Badnarik, and the same number from Peroutka. The other 76.294 come from Bush's vote, since wmbp_KW = 2 for the configuration of ballot orders at cluster 171.

Similarly, Bush gets 3.362 votes from each of Disqualified and Peroutka, and 6.724 from each of Bednarik and Kerry.

These Kerry votes awarded to Bush offset 6.724 of the 228.883 already awarded to Kerry above, so that kerry's total increases by the difference between these numbers, or 222.159 votes. Bush's total falls by 76.294 - 20.174 = 56.120. So correcting the "caterpillar" miscount in cluster 171 would reduce Bush's "lead" by 222.159 + 56.120 = 278 votes.

[L. Coyote]

67. kick n/t

[Carolab]

68. The one thing that stood out for me

in posts 30 and 39 was that there were no wrong Bush votes defaulting to Kerry but there was one wrong Kerry vote defaulting to Bush.