Do these Cleveland precinct results look right to you?

[Will F.D. People]

It looks to me like this problem is caused by the fix to another problem. Presumably rotating the names in the different precincts is done to correct some perceived ill. And it turns out that this may cause a bigger problem than it fixes. This is just a lesson to keep in mind when trying to address other perceived problems. 

[elcorazon]

why does all this remind me of the bcs?

I actually suspect that these types of problems tend to adversely impact the dems.  The more densely populated the area is, the more likely multiple precincts will vote together.  Dems do better in more densely populated areas.  Therefore, votes most likely to be for dems occassionally get shifted due to some sort of error, usually on the part of the poll workers.

I don't think this was enough to alter the results, but it could have been.

They should have all the ballots at one voting place be the same to avoid this error.

[jimrtex]

I've looked at the results some more and have identified a two cases where Kerry votes were credited to Bush and "Nader".  But I also looked at some down-ballot races to see whether ballot rotation combined with ballot mixup would indicate that the problem was whole ballots punched on devices set up for another precinct. 

Note: the guide booklets used for the punching devices are made page by page, with each page attached to a separate spindle.  The spindles are separated by the distance between rows on the punch card.  When the pages are opened flat, the left (even) page has the candidate names, the space between the spindles exposes the hole guides for punching the row of holes for the races on the left page, and the right (odd) page is typically blank or has an instruction to turn the page for more races (Palm BeachCounty, Florida in 2000 was exceptional when the right page was used for listing candidate names as well for the butterfly ballot). 

If the guide booklet for one of the voting devices included a single erroneous page, then only some of the races would be messed up (it would be like if the voter went to a punching device for the wrong precinct, punched his presidential choice, and was then told to switch over to the correct device for his race, and used that device to punch out the rest of the ballot).

This does not appear to be case in Cuyahoga County.  The mixup is also detectable on down ballot races.

3B Kerry is 1st, 3I Peroutka is 1st.
3B Badnarik is 4th, 3I Kerry is 45h.

3I Peroutka 70 votes.
3B Badnarik 41 votes.

The ballot order for the Senate race between Voinovich and Fingerhut is reversed between these two precincts.  The effect is that some votes for Fingerhut in 3I would be punched for Voinovich (corresponding to the 3I Peroutka votes, but with some reduction due to fewer people voting in the Senate race - falloff was around 10 to 15%).  But it would also mean that some votes for Voinovich would be recorded for Fingerhut.

But the same thing would happen in 3B (votes being swapped between the two).  Let's say that that the voters intended to split 70-30 between the two candidates in both precincts, but that 10% of the voters used the wrong voting device.  The effect would be to convert 7% of all votes from the leader to trailer, while also switching 3% of all votes the other way.  The net result would switch the 70-30 margin to a 66-34 margin.  In addition, both precincts would be affect by a similar amount (unless the wrong voting device was used disproportionately by voters in one precinct rather than the other.  The Peroutka vote in 3I was 13% of the total, while the Badnarik vote in 3B was 9% of the total).

Given that Voinovich had reasonable support even in a highly Democrat area (he is from Cleveland, and was Governor before becoming Senator), it would be almost impossible to detect a ballot mixup.  There is little net effect between the two precincts - the pro-Voinovich effect would be only visible comparing against other precincts, and would still be small.

The result in 3B: Fingerhut 303, Voinovich 156.  In 3I Fingerhut 304, Voinovich 163.

But there was another down-ballot race where the effect is visible.  There was a non-partisan election for the State Board of Education that attracted 3 candidates, Brown, Corrigan, and Lesnick.  Brown was a relatively easy winner, with Corrigan following and Lesnick finishing 3rd.

Based on the ballot order you would expect the 3B results to show
Brown -> Lesnick (largest effect, Brown had most votes to start)
Corrigan -> Brown (medium effect)
Lesnick -> Corrigan (smallest effect)

Based on the ballot order you would expect the 3I results to show
Brown -> Corrigan (largest effect, Brown had most votes to start)
Corrigan -> Lesnick (medium effect)
Lesnick -> Brown (smallest effect)

Because Peroutka received more votes in 3I than Badnarik in 3B, the effect in 3I should somewhat larger.

The results:
3B: Brown 296, Corrigan 79, Lesnick 68
3I: Brown 344, Corrigan 101, Lesnick 32

These are neighboring precincts, which had 30 votes for Bush between them, and had very similar results in the Senate race, yet voted dramatically different in a SBOE race.
4F Kerry is 1st, 4N Badnarik is 1st.
4F Peroutka is 3rd, 4N Kerry is 3rd.

4N Bandarik 163 votes.
4F Peroutka 215 votes.

In these two precincts, the order of the Senate candidates was the same, so that mixing up the punching devices would have no effect.   
The results were:

4F Fingerhut 338, Voinovich 154
4N Fingerhut 331, Voinovich 122

But in the SBOE race, the ballot order was different.

Based on the ballot order you would expect the 4F results to show
Brown -> Corrigan (largest effect, Brown had most votes to start)
Corrigan -> Lesnick (medium effect)
Lesnick -> Brown (smallest effect)

Based on the ballot order you would expect the 4N results to show
Brown -> Lesnick (largest effect, Brown had most votes to start)
Corrigan -> Brown (medium effect)
Lesnick -> Corrigan (smallest effect)

The results:
4F: Brown 228, Corrigan 183, Lesnick 54
4N: Brown 227, Corrigan 59, Lesnick 123

Peroutka received 41% of the 4F vote, while Badnarik received 33% of the Peroutka 4N vote, so you would expect a dramatic effect here, and you do.  Corrigan nearly catches Brown in 4F, while in 4N they badly lag.
6L and 6M.  This is an example of where Bush and "Nader" benefited from mispunched Kerry votes.  I detected it by looking for precincts where it appeared that Lesnick had received a relatively high share of the vote in the SBOE race. 

Of the 15 or so precincts that I checked, this is the only one that I was able to find that had an effect in other races.  Some of the precincts I checked were in the western, white ethnic, part of Cleveland, where Lesnick's name may have been helpful (this is the area represented by Dennis Kucinich).  In these areas, higher support for Lesnick could represent actual voter support.

Ralph Nader had originally qualified for the presidential ballot in Ohio, but was then ordered off the ballot.  His space appeared on the ballot as "Disqualified Candidate".  In the alphabetical ballot order used in Ohio, "Disqualified Candidate" appeared after Kerry and before Peroutka, just where you would expect Nader to appear.  In the canvass, the vote for "Disqualified Candidate" is shown as 0.

If voters in 6L had used a 6M punching device, Kerry votes would be punched for Bush.  If voters in 6M had used a 6L punching device, Kerry votes would be punched for "Disqualified Candidate".  The results:

6L: Badnarik 1, Bush 82, Kerry 325, DQ 0, Peroutka 0
6M: Badnarik 0, Bush 9, Kerry 367, DQ 0, Peroutka 1.

While this might not look too odd, in this area the 20% vote for Bush in 6L was extremely high. 

In addition, in 6M, 467 votes were cast, but only 377 were counted for any of the 5 candidates.  In other words, an undervote of 90 or 19.2%.
While I suspect that a few people did abstain, and others failed to successful dislodge a chad, and others may have voted for "Disqualified Candidate" as a protest vote, in other precincts the undervote appears to be only a few percent.

Further evidence of a problem here is the presidential vote.  As in 3B/3I the two candidates are swapped between the two precincts.  And as in 3B/3I we would expect the difference to be masked.  The results:

6L Fingerhut 203, Voinovich 133
6M Fingerhut 235, Voinovich 143

This is not too unusual other than in 6L where Voinovich ran 20% ahead of Bush (40% vs. 20%), and in 6M where Voinovich ran 36% ahead of Bush (38% vs 2%) indicating how anomalous the Bush vote was in 6L.

In addition if we look at the votes cast, vs. the votes counted for each race:

6L: 425 cast, 408 for President, 336 for Senate.
6M: 467 cast, 377 for President, 378 for Senate.

That is, in 6M the undervote for senate was one less than it had been in president, while in 6L the undercount increased from 4% to 21%.  This indicates that about 20% of the 6M voters used a 6L voting device to vote, with Kerry votes being redirected to "Disqualified Candidate", which were not counted.

[Andrew]

It occurs to me that it's probably more likely that the machines were set up before they were placed around the room, and that some improperly set-up machines were in the wrong places (rather than people going to the wrong places to vote).  That's just a guess, though.

It could be a schmuck who didn't think about what he was doing, or a worker who wasn't told by his supervisor that the machines were different, or someone trying to cause votes to be shifted from Kerry (in these strongly democratic precincts, Kerry was the only candidate who stood to lose significant votes).

[jimrtex]

It occurs to me that it's probably more likely that the machines were set up before they were placed around the room, and that some improperly set-up machines were in the wrong places (rather than people going to the wrong places to vote).  That's just a guess, though.

It could be a schmuck who didn't think about what he was doing, or a worker who wasn't told by his supervisor that the machines were different, or someone trying to cause votes to be shifted from Kerry (in these strongly democratic precincts, Kerry was the only candidate who stood to lose significant votes).

Perhaps, though the varying magnitudes argue against that.  There appear to be around 800 ballots in Cleveland that were misdirected.  380 of these were in 4F/4N (Cleveland has 21 wards and on the order of 200 polling places), so almost 1/2 the problem was in 0.5% of the polling places.  Almost 40% of the the votes were misdirected there.

The two next worst polling places, the problem is around 10 to 15%.

There were a couple of instances where the precincts were of disimilar sizes.  If one precinct has 380 voters and another 70, would it be likely that they both have the same number of polling booths, with one getting swapped from each set?

And 8G/8H/8I appears to be a 3-way mixup, with some showing up as Badnarik and Peroutka votes, but also a big difference in the Bush vote.

It would be interesting to know how the polling places were set up.  Whether it was practical to go to the wrong punchng booth.

What happens if you catch the problem at 10 AM on election day?  Or what if it happens while some poll workers are taking their lunch break, and the replacement didn't get complete instructions.  Are you going to call up the press and say that you may have just cost Kerry the election?  Or do you correct the problem as best you can.

Another possible mixup: some voters place their ballot in the wrong ballot box.  I've heard of that happening in Houston, where it was found because the box count didn't match the register count.

Now imagine if the poll workers decide to fix this up.  Let's say that for some reason some voters didn't see the box for their precinct.  They were directed to the correct device, but then when they finished, they weren't being watched as closely.  Too many ballots for 3B go in the 3I box.  You discover the problem.  How do you correct the problem?  You pull some of the ballots from the 3I box and stick them in the 3B box.  Your numbers will come out right at least.

[Andrew]

Does anyone else think it's ironic that if you go to the wrong precinct, they will give you a provisional ballot that will later be disallowed, but . . .

They will put multiple precincts in one building, so that voters can vote in the wrong precinct under official supervision, screwing up the results?