The ES&S M100 is used in some or all counties in a long list of states. Voter marked paper ballots, even if the machines that count them are inaccurate,are more auditable because accurate vote counts can always be recovered by hand-counting the paper ballots,unlike with touchscreen voting systems,but machine counts are not to be

::::::::

ES&S Opti-Scans Found Miscounting by 4% (8% margin-swing)

Dear Friends,

I do not have all the details, but this story needs immediate
investigation before calling winners tomorrow in any state using ES&S
M-100s to count its votes whenever the reported margin is closer than
8%.

The paper ballot precinct-based optical scan voting system used in a
large proportion of US jurisdictions is producing wildly inaccurate
counts of ballots that are off by as much as 4% per precinct - thus
altering the vote margin possibly by as much as 8% by shifting votes
from one candidate to another - depending on how the errors occur.

The ES&S M100 is used in at least some or all counties in the
following states: AL, AR, AZ, CA, CO, FL, IA, IL, IN, KS, ME, MI, MN,
MO, MS, MT, NC, ND, NE, NM, OH, PA, RI, SD,TN, TX, VA, WA, WI, WV, and
WY.

The story, with links and copies of at least one original document is
forwarded in the email below.

Voter marked paper ballots, even if the machines that count them are
inaccurate, are more auditable because accurate vote counts can always
be recovered by hand-counting the paper ballots, unlike with
touchscreen voting systems, but machine counts are not to be blindly
trusted - and this is a good example why not.  It is easier to detect
errors produced by paper ballot optical-scan systems than by
touchscreen e-ballot systems.

The usage distribution of the ES&S M-100 scanners in the U.S. can be
found here:

http://verifiedvoting.org/verifier/index.php

This is yet another reason why States should be conducting routine
independent manual post-election audits like those described in this
proposal:  How to Audit Election Outcome Accuracy
http://electionarchive.org/ucvAnalysis/US/paper-audits/VoteCountAuditBillRequest.pdf


---------- Forwarded message ----------
From: Charlie S.
Date: Mon, Nov 3, 2008 at 9:19 PM
Subject: ES&S opscans found miscounting by 4%
To: Verfied NM Voting <vvnm@yahoogroups.com>


There has been a grave concern raised about the accuracy of ES&S M-100
opscans (ES&S's primary precinct model)  raised  10 days ago.

A Michigan Clerk wrote to the (federal-ish) Election Assistance
Commission on october  24 to tell them she has been given a
substantial  reason to doubt their accuracy.  In Logic and Accuracy
testing for the the clerk reports that  some of her  Communities
reported "the same ballots run through the same machines, yielded
different results each time"

http://blog.wired.com/27bstroke6/files/m100_issue_letter_10.24.08.pdf

How big was the error?  Wired extends the story:

"According to news stories, a race in the August Republican primary in
one Michigan township did have a discrepancy in tallies that were
counted by hand and by ES&S optical-scan machines. The clerks race in
Plymouth Township was recounted after the losing candidate requested
it. The initial machine count had showed Joe Bridgman defeating Mary
Ann Prchlik by 1,920 to 1,170. But the hand count
(http://www.journalgroup.com/Plymouth/8342/canvassers-uphold-decision-in-plymouth-township-recount)
-  narrowed the margin to 1,885 to 1,727."

http://blog.wired.com/27bstroke6/2008/11/michigan-electi.html

That is, the totals shifted nearly 4% of one candidates 1170 votes to
his opponent: that reduced the margin of victory in that township by
8%!!  (a very uncomfortable error)

It was both ironic and especially disconcerting since that the race in
question was for the township's clerk and the winning candidate was
the Deputy Clerk.

Note that this is one township.  And also note that the L&A tests were
 off in 4% of the communities running the L&A tests. It does not
report how many machines in each community failes, but if follows the
common pattern their probably is 1 or at most 2 machines in a typical
precinct.   So the impact on statewide raises is presumably somewhat
damped over the impact on local races but is still absurdly large.

Conversely, the problem is attributed to dust buildup blocking the
sensors and possibly interfering with the electronics.  The concern is
that the next election is going to dwarf the past one in the amount of
dust generated.

"In Oakland county when officials there met with ES&S to discuss the
errors encountered during logic and accuracy testing, ES&S maintained
that the problem was dust and debris build-up on the sensors inside
the machines.

"This has impacted the Digital to Analog Converter (DAC) settings for
the two Contact Image Sensors (CIS)," Johnson wrote the EAC."

Now presumbaly, a lot of the dust comes from the kind of ballot stock
used so we don't know if this will occur nationwide.  But since these
are spec-ed by ES&S it's plausible that this will be ubiquitous.

Like NM, Michigan also is forbidden from doing their own "cleaning" by
ES&S on pain of voiding the warrantee (and possibly the
certification).

I have advocated before that NM urgently needs to get a larger IT
staff in the SOS office to get this situation under control.  NM needs
to have a credible staffing level in part to give it leverage to
renegotiate these maintenance contracts that are bankrupting the
clerks, or causing them not to maintain machines.

Charlie

Authors Website: http://ElectionArchive.org

Authors Bio:
Founder and President of US Count Votes, dba The National Election Data Archive and volunteer for honest, accurately counted elections since 2003. Masters degree in mathematics with emphasis on computer science. Has written numerous academic and scientific papers with computer scientists, statisticians, and mathematicians on election integrity topics, inluding how to calculate minimum manual audit amounts necessary to ensure election outcome integrity.