July 19, 2006
originally to Kaitlyn O'Hara at Representative Udall's office
Kaitlyn
Thanks for speaking to me. I think Representative Udall may be interested in my
following remarks which get at the issues upon which much of the accuracy of our
elections rest. In particular a technical understanding of the accuracy of
voting machines is critical. It is of great importance to use a technical
criterion in determining the applicability of the whole voting system to
correctly represent the will of the voting public as the measured vote spread
becomes narrow as it so often seems to do. It would be best to have NIST advise
the EAC on this very topic and both should advise local and state governments on
the correct thresholds to use in making their recounts, and on the necessity of
depending on human counts rather than machine counts when the elections are
closer than the sampling error of the tests used to check the voting system
accuracy.
It would be extremely useful in contrast to be aware of the Swiss system
http://www.swissvs.org/ for paper voting which I refer to below. This is a
counterexample to the too-often heard rejoinder that human counting is not
accurate as machine counting. Too many clerks believe this. There are techniques
which can be used (such as human sorting) in which the human count just gets
more accurate with repeated effort. This can not be said for the machine
implemented voting systems which have no innate ability to understand anything.
I hope that the unjustified self-confidence and expert elitism of election
officials and their promotion of undeserved confidence in electronic voting
systems can be pierced by the questioning of Representative Udall and our other
Representatives on the Science Committee.
Harvie Branscomb
El Jebel, CO 81623
hbranscomb@eagledems.org
Rep. Udall:
I am the Chair of Eagle County Democrats and an experienced
election evangelist for Eagle County. I think the VVPAT is not sufficient to
insure quality elections, and a true hand marked paper ballot is preferable. We
must encourage hand counting wherever possible, and even though I am a
technologist, for voting we must reduce the interference of mechanical and
electronic processes as much as possible in the voting system (as is done in
Switzerland, for example where the entire nationwide election, up to 6 times per
year, is hand counted on paper within 4 hours). The VVPAT which is required in
Colorado will only work if the following apply:
1) the VVPAT printer is available with the DRE, working throughout the election
and supplied with paper and kept secure;
2) the VVPAT printer functions correctly - we have evidence otherwise for the
Hart eSlate where the printers self test gave various print results during the
acceptance test;
3) each voter actually verifies the VVPAT- with Hart eSlate, the blind voter can
not verify, and the normal voter is unlikely to verify because the lcd screen
with larger brighter font is also showing the ballot information at the same
time as the voter is expected to verify the smaller, darker, less visible paper
to the left of the machine;
4) if the paper roll for the VVPAT is actually legally defined as "the ballot"
for purposes of any recount;
5) if any recount instructions provided by an attorney general or secretary of
state actually requires that the paper roll be read by hand and not by
machine... and not by reading the bar code which is printed on every paper
record for this purpose (the bar code is not verifiable by the voter), and if
there is a reliable way to hand count the paper roll;
6) if every court with jurisdiction determines that the paper record is to be
treated as the ballot after considering the actual experience of actual voters;
7) if a recount is run in case the spread of the election either within the
local jurisdiction or across the entire district applying to the race is within
the margin of error defined by the Logic and Accuracy test. The LAT in Colorado
calls for a minimum of 25 ballots per party plus enough for each ballot style,
or perhaps only 75 ballots to be tested. This gives an accuracy for the test of
only greater than 1%, hence if the vote spread were to approach 1%, say less
than 2%, a hand recount should be conducted to overcome any uncertainty about
the ability of the machine to count votes. If multiple counties are involved in
the district, the margin of error defined by the LAT for each county should be
summed and compared to the spread of the votes on that race in the district. If
the sum of the margins for error exceeds the spread of the vote in that race, a
hand recount must be called for. The LAT is the only proof that the voting
system is functioning correctly, but it’s accuracy is severely limited by the
number of ballots provided for the test. There is a significant likelihood that
the errors contributed by individual counties would not average out, and no
proof that they would. The law regarding recounts is very insufficient in this
regard.
In addition, a post election audit is laudable, but it must be conducted as a
true audit, with truly random choice of ballots and machines and races, and any
irregularity must be logged and made available to the public and kept in a
repository of election irregularities. It is important that the machinery should
offer the opportunity of a true audit, and not a recount of a random set of
ballots, using a separate process at a separate time.
Finally each state must have a bipartisan oversight of the functions of the
secretary of state and each county must have bipartisan oversight of the
functions of the county clerk in regards to elections.
Harvie Branscomb Chair, Eagle County Democrats hbranscomb@eagledems.org
