Re: [ElectionIntegrity] Re: Conference call on federal legislation?

From: Kathy Dopp <kathy_dot_dopp_at_gmail_dot_com>
Date: Sat Nov 18 2006 - 15:53:48 CST

Barbara,

OK. You're amazing. I'm very glad to know that you have contacts with
Common Cause too. Please ask CC to SLOW down their pace because the Holt
audit provision currently is unverifiable, not transparent, has no teeth, is
insufficient, and probably unconstitutional - and thus would only provide
false assurance, if anything. Thank you.

On 11/18/06, Barbara Simons <simons@acm.org> wrote:

There are many questions still
> that would need to be answered,

Most of your questions are answered in this paper (unfortunately it is a
little long, so I'll try to pull information from it to answer your
questions below):

http://vote.nist.gov/ElectionIntegrityAudit.pdf

such as how to make sure that the
> selections for the audits are truly random,

This list of items re. audits from VerifiedVoting is very good and includes
the suggestion that if any program or spreadsheet are used for the audit,
information be made publicly available on it:
http://www.verifiedvotingfoundation.org/downloads/AuditQ-Machine.pdf

Also, here is an excellent short paper by Cordero, Wagner, & Dill, and on
how to make random selections transparently:
 http://www.cs.berkeley.edu/~daw/papers/dice-wote06.pdf

determining who is in
> charge of the audits,

State or county auditors, citizen vote count and audit committees
required to have qualified people on them with advances degrees in computer
science/math/statistics and election integrity activists and nonvoting
election officials, auditors could be election judges or poll workers of
multiple parties, or independent accounting firms.

deciding what percentage should be audited
> (possibly a function of the size of the vote),

Barbara, If we agree that the point of elections is to ensure that correctly
elected candidates are sworn into office, then you MUST read this paper to
know how to correctly calculate the minimum audit size that is necessary to
ensure that any outcome-altering vote miscount is detected. It is, as you
say, a function of the number of total vote counts, but is also dependent on
the margin between candidates.
http://vote.nist.gov/ElectionIntegrityAudit.pdf

The margin between candidates must be used because the smaller the margin,
the smaller the amount of miscount that could wrongly alter the race, and
the larger the audit sample size must be.

There is also a paper by Wand of Stanford that I don't have the URL for that
talks about the problem with vote count unit size differences, which means
that DRE machine counts and optical scan precinct counts should be
separately audited, with precinct count size variation adjusted for; or the
entire pile of counts adjusted for count size variation, as our paper
describes how to do. But legislation could simply say "require the audit
amount to be sufficient to detect any outcome-altering vote miscount".

and determining how much
> of a discrepancy should trigger a total manual recount. (A one vote
> discrepancy among 500,000 ballots cast strikes me as too demanding a
> threshold, but I don't know what an appropriate threshold might be).

What should trigger expanded audits or a 100% manual count should be
based on whether the amount of discrepancy found could wrongly alter the
race in question, given the original design of the audit sample size, IMO.

I believe that a similar thing is happening with DREs. If you try
> head-on to eliminate them, you'll get tremendous push-back. But it
> becomes harder for the opposition to fight if you make demands, such as
> public software, that are easy to defend.

I completely agree. If we could fund the Holt bill's requirement for VVPBs
and sufficient, transparent, verifiable, independent, random audits; then
why not only fund open source optical scan systems and not DREs w/ paper
rolls since the DREs w/ VVPAT systems are so fundamentally flawed that they
can be used for vote fraud that is undetectable via manual audits? Do you
think that would be possible?

  If the
> Holt bill does not pass, I believe that in 2008 a large percent of the
> country will be voting on DREs in states in which there is no mandatory
> audit or, even worse, in which there are significant obstacles put in
> the way of doing an audit.

Yes, certainly true, although as currently written, the Holt audit
provisions are not verifiable, transparent, or sufficient to detect
outcome-altering vote fraud and could interfere with any good state audits
like the one being planned for NM, and have no teeth. i.e. the current Holt
audit provisions could be a total bust.

I am trying to find time to revise our Utah audit proposal to make it
conform to what Utah election officials put in their written audit procedure
with improvements to make their procedure verifiable, transparent,
sufficient, etc.

Please read this
http://vote.nist.gov/ElectionIntegrityAudit.pdfrecommendations on p.
14

Our paper, which is based on some prior work by myself and Ron Baiman and
the Brennan Center, calculates the MINIMUM amount of electronic vote counts
that MUST be audited if ensuring electronic election outcome integrity is
the goal of the manual audit.

A more conservative approach for calculating vote count audit amounts by
Ronald Rivest of MIT which he released simultaneously with our paper, uses a
formula that overestimates the amount that our election integrity audit
method calculates:

http://theory.csail.mit.edu/%7Erivest/Rivest-OnEstimatingTheSizeOfAStatisticalAudit.pdf

If you look at Rivest's appendix, he compares what he terms the "optimal
audit size" calculated by our numerical solution, with his own "estimated
audit size" that is calculated by Ron Rivest's clever formula. Roy
Saltman's work also agrees with our audit sample size calculation. Please
read our paper because it is the definitive work on the topic of correct
vote count audit sample sizes today, although we do call for further
research.

There is also a 30 year old paper by Roy Saltman, the father of federal
engineering standards for voting systems, whose work is in agreement with
ours, even back then. Barbara, the answer to the correct vote count audit
sample size is irrefutably given here:

http://vote.nist.gov/ElectionIntegrityAudit.pdf

There is no other way to calculate vote count audit sample sizes IF the goal
is to ensure election outcome integrity. It is just simple mathematics and
a numerical program since the equations could not be solved analytically.

Best,

Kathy

_______________________________________________
OVC-discuss mailing list
OVC-discuss@listman.sonic.net
http://lists.sonic.net/mailman/listinfo/ovc-discuss

==================================================================
= The content of this message, with the exception of any external
= quotations under fair use, are released to the Public Domain
==================================================================
Received on Thu Nov 30 23:17:09 2006

This archive was generated by hypermail 2.1.8 : Thu Nov 30 2006 - 23:17:19 CST