Diebold DRE Analysis -- Cuyahoga County, OH 2006 Primary (ESI)

[Bill Bored]

Long report folks, but the bottom line is that Diebold even found a way to screw up PAPER ballots!

<http://www.cuyahogacounty.us/bocc/GSC/pdf/esi_cuyahoga_final.pdf>

Ballots

Issues encountered in reading ballots were physical problems endemic to the paper medium. Issues encountered either in the course of trying to read a specific ballot or in the course of reviewing the entire VVPAT were manifold. ESI also asked the recount team to record and describe anomalies found on the VVPATs. Some examples of such issues and anomalies include:

• Blank VVPAT – that is, VVPATs with nothing printed on them

• Accordion-style crumpling of the VVPAT

• Inexplicable long blank spaces

• Torn VVPAT and VVPATs joined together with tape

• Printing anomalies (faded ink or irregular ink distribution)

• Text missing from VVPATs

In each case, the physical integrity of the ballot is compromised. As, under Ohio law, the paper receipt is considered the official ballot (see appendix), the fact that the physical ballots were in so many cases destroyed means that ballots would be lost in a recount or contest.
Specifically, the manual count team reported the following:

• Six VVPATs (1.4 percent) of the total VVPAT Cartridges were blank. In one of these cases, the transfer case that was supposed to be used to deliver the ballots on Election Nights contained one empty canister.

• Forty-three (10.4 percent) of the VVPATs were physically compromised in any of the following ways:

smeared print, torn paper, crumpled or folded paper, paper taped together, blank spaces or printer anomalies Printer problems were not evenly distributed throughout but rather were clustered in particular vote centers. For example, 18 vote centers experienced 100 percent of the printer errors (4 vote centers experienced 46% of the printer errors).

-snip-

[Bill Bored]

1. Comparing the Count

Key Finding: Discrepancies were found across vote counts stored on different mediums across the election system.

ESI conducted an exhaustive analysis of regular voted ballots from onboard machine memory compared to manual counts of paper ballots, official results, and other interim and election reports. The comparisons revealed a wide range of discrepancies. Some discrepancies may also reflect ESI’s errors in processing.

• A lack of inventory controls and gaps in the chain of custody of mission critical assets, such as DRE memory cards, DRE units, and VVPAT cartridges, resulted in a significant amount of missing data. Because of the missing data, ESI is unable to give a definitive opinion of the accuracy of the Diebold TSX system.

• Due to limits in the data, software computational abnormality contributing to the count inaccuracies cannot be ruled out. Computational abnormality could be the result of a failure to adequately test the
voting equipment before the election or to manage the various databases appropriately.

• In multi-precinct polling places, voters could vote on machines located in other precincts. Accordingly, ballots from a number of precincts appeared on the same VVPAT tape. VVPAT ballots, however, lack a header identifying the precinct. Without this information, it is not possible to conduct a precinct-level tally of the VVPAT ballots.

• While discrepancies between the VVPAT summaries and VVPAT ballots themselves were relatively small, discrepancies between paper record and the electronic record were considerably larger and more pervasive. For the most part the discrepancies can be characterized as the DRE memory cards and Election Archives registering more votes than were produced by the VVPAT summaries or by hand counts of the VVPAT ballots.

• These conclusions are based on the data ESI was able to obtain from the May 2 election; however, data could not be retrieved for 13 VVPAT summaries, 87 VVPAT cartridges, 53 election archives and 3 DRE memory cards, which were used to tablulate the official vote count.

[bleever]

2. That's not a flaw: that's a Feature!

Spoilage will be critical in the fall elections. The ability of Diebold machines to contribute in this regard is a genuine "value-added" feature with scalable benefits!

[Kolesar]

3. Voter Verified Paper Actuated Tapes?

[OnTheOtherHand]

4. Voter Verified Paper Audit Trail

sometimes rendered Voter Verifiable etc., which is probably more accurate, since no one can force the voter actually to verify it. But officially under Ohio law it's Voter Verified.

[Kolesar]

5. TY. Well, those descriptions in the O/P are horrifying

Blank paper tapes? Crumpled and faded? Did you know Cuy Cty is planning to buy hundreds more of the machines for November? Arrggh!

[OnTheOtherHand]

6. yeah, I couldn't decide whether you were kidding above --

hmmm... how about we dub the Diebold/Cuyahoga implementation "Vapor or Visibly Pathetic Attempt to Tranquilize"?

I guess better pollworker training will help, although I'm not sure how much. If you thought the OP was horrifying, try plowing through the middle of the full report. The other big report on the May election is appalling, too.

[Bill Bored]

7. We need a summary of how the Electronic Totals don't match either. nt

[OnTheOtherHand]

8. still trying to wrap my head around that, but...

OK, so they tried to trace every machine used in 50 vote centers and copy their internal archives to a database; get their hands on all the original data cards from those 50 vote centers and copy those to another database; and then compare. (They also tried to count every individual audit-trail ballot, and every VVPAT summary.)

Their general finding is that the discrepancies between the two different DRE numbers (internal archives and cards) are "relatively small," as are the discrepancies between ballots and summaries -- but the DRE totals run over 1000 votes higher than the paper totals. As for the electronic totals not matching each other, well, they match exactly in 37 of 50 vote centers, they are within 5 votes in 8 more, and at the other extreme, there were 2 or 3 (the charts seem to disagree, and I haven't stared at the DB yet, if it has even been posted) that differed by over 25 votes. Blackwell does a bit worse on DRE than VVPAT (go figure), while Strickland does a bit better on memory cards than on archives (his margin appears to be 0.4 points larger).

ESI drolly comments (p. 117, PDF p. 122), "The question these data raise is why there should be any discrepancy whatsoever between the two electronic media." Indeed. Not that there "should" be any discrepancy with the paper, either, but they sorta knew some of that would be missing or unusable.

[Melissa G]

9. Hey BB... Do you know if these VVPAT were of the toilet paper

variety? Those temperature sensitive thermal paper ones that I am sure we are considering as a 'fix up idea' down here in currently 104 degree Texas?....

[Bill Bored]

10. Yes I do and yes they were! nt

[Melissa G]

11. Thanks, I 'll bookmark this and use it for evidence at the next hearing

I hope activists cite this all over the country!