Problems with accuracy of voter verification using VVPATs (SOS-HAVA-01-06-0001)

[Mary Eberle <m.eberle@xxxxxxxxxxxx>]

Dear Mr. Gardner,

Thank you for speaking with me after the May 24 hearing concluded. You
asked two specific questions, and I hope these answers will be helpful. 
They should not be considered comprehensive, however. In your position 
as Voting Systems Specialist for the State of Colorado, you surely have 
access to more information from the vendors themselves. It seems that 
they could be required to monitor and disclose VVPAT accuracy failures.

Please add this e-mail to the 7 April 2006 HAVA complaint documents, 
Tracking Number SOS-HAVA-01-06-0001.

1. You wanted me to give you a documented case in which the VVPAT did
not match the voter's computer-based input during a real election, not
an LAT, etc.

______________________________________________________________________

1.a. Please see this Web site:
http://www.votersunite.org/info/Dieboldinthenews.pdf. Below are two
pertinent items taken from the site. I have "highlighted" two VVPAT
problems by surrounding them with asterisks.

(1.a.1) Page 11: July, 2004 AccuVote TS DeKalb and Chatham Counties,
Georgia. In what Georgia Secretary of State Cathy Cox called "a very
ordinary primary election day," voting problems ranged from incorrect
summary pages ...  Ziyadah Sabir said she's concerned the touch-screen
machine didn't properly record her vote. **********The summary page,
which allows voters to review their choices before casting their
ballots, failed to show some of her choices and showed incorrect choices
for others, Sabir said. Poll workers could not fix the problem, she
said.*********** "That's not very reassuring," said Sabir of DeKalb
County, who was voting for the first time on the machines.

(1.a.2) Page 19: November 2004 AccuVote TS Maryland. On election day,
TrueVoteMD registered 383 reports involving 531 incidents of problems
encountered by voters. Many voters reported votes switching on the
screens.42 These problems ranged in severity from moderate inconvenience
to outright voter disenfranchisement. The significance of these
complaints increases dramatically, however, when one considers that
Maryland had a total of 1,787 precincts this year, and TrueVoteMD was
only able to cover a fraction of these?some 6%. ... Most incidents that
were reported to TrueVoteMD fell into the following categories:
? Lost votes due to incomplete ballots that were missing candidates or
entire races
? Lost votes due to machines crashing or freezing before the voter cast
a ballot
? Lost votes due to ?smart card? and encoder failures
? Lost votes due to delayed poll openings because of machine boot-up
failures
? Lost votes due to voter abandonment because of unacceptably long waits
? Lost votes resulting from touch screen failures that included vote
switching, review screen malfunctions, unintended selections and
submission of ballots before voters had made selections,
hypersensitivity to touch that caused voters to complain that ?it was
out of control and I have no idea who my votes were cast for,? screens
going blank
? Lost votes from unreadable voting machine hard drives (PCMCIA cards)
? Lack of privacy because machines were oriented so that the screens
were visible by those waiting on line to vote
********? Discrepancies between electronic vote tallies and manual vote
tallies**********
? No offer of provisional ballot or wrongful denial of provisional ballot
? Disenfranchised voters due to failures in the Motor Vehicle
Administration registration process.
? Inadequate staffing and insufficient training of election judges

____________________________________________________________________

1.b. Even though the following example is not from an actual election, I
think that it has valuable lessons for us in the current hearing on Hart
DREs with VVPATs. The Web site
http://www.vote.caltech.edu/events/2006/VTPsummit/Pandolfo_pres.pdf
shows voters' experiences with ballot verification in an experimental
setting. The Web site describes a test of voter use of two kinds of
voter verification systems that had built-in inaccuracies; many voters
failed to detect these inaccuracies. Another conclusion: "VVPAT can
themselves be used as part of a hacking scheme."

_________________________________________________________________________

1.c. The following article was posted May 26. The Web site is
http://www.helena-arkansas.com/articles/2006/05/25/news/news1.prt. The
terminology is different from what we are using, but the description
fits what we are calling VVPATs.

Phillips County's election woes continue
By MICHELE PAGE
World Staff Writer

Candidates, spectators, Phillips County election officials, and Kevin
Berry, of Election Systems & Software met Thursday at the Phillips
County Courthouse in an attempt to sort out some of the election misery
the county is suffering.

[snip]

Hours later, other problems popped up, as the paper ballots inside the
Ivotronic machine, and those stored in the Supervisor Personal Election
Ballot, failed to match.

[snip; the preceding refers to matching VVPATs to electronic voting
records, so this statement represents a failure of VVPATS.]

... Jill Friedman-Wilson, spokesperson for ES & S of Omaha, Neb., spoke
with the Daily World about the problems of the system.

[snip]

Wilson said that the problem stemmed from the Ivotronic and the old
Legacy 150 equipment not "matching up." The teams of election officials
are not able to merge a consolidated report for the county. She added
that the company regrets the delay and it was specific to the two
systems. Wilson reported that ES&S had received reports that all the
votes in Phillips County had been counted but that they could not
generate the final report.

[snip]

The commission plans to start a manual hand count at noon today. The
Phillips County Election Commission reported that representatives from
the Republican and Democratic parties from Little Rock will be present
at the hand count.

________________________________________________________________________

2. You wanted me to give you background on the Hart failures in Texas in
March 2006. The worst problem was in Tarrant County, Texas. Here is the
Web site: http://www.dfw.com/mld/dfw/news/local/14056065.htm.

_________________________________________________________________________

Thank you for taking the time to examine these data. I hope that these
data will encourage you to support decertification of the Hart eSlate
system for Colorado. Paper ballots will keep the elections verifiable,
and there are systems that allow the paper ballots to be marked
anonymously by blind voters. No system is perfect, but the Hart eSlate
system and its DRE ilk are fatally flawed.

Sincerely,
Mary

Mary Eberle
1520 Cress Court
Boulder, CO 80304
(303) 442-2164