[Election-Methods] A New Measure of Election Auditing Discrepancy

[Kathy Dopp]

These election auditing methods work no matter how many candidates or
winners in a race. However, they would not yet work for any rank order
voting system. You guys might want to work on extending the
mathematics of election auditing so that they would.  Kathy
---

March 30, 2008
A New Measure of Election Auditing Discrepancy

Dear Friends,

Good news.  University of Berkeley Statistics Professor Philip Stark,
submitted a new paper (still a draft) describing how to analyze
election audit discrepancies to decide whether to certify an election
outcome or to expand the audit sample size -  to "The Annals of
Applied Statistics" a couple of days ago.

We now know not only the correct minimum audit sample sizes for
verifying election outcomes, but  also we have a procedure for
calculating the confidence that an incorrect outcome was detected -
given any sample size, knowing the actual discrepancies in the audit.
In other words, Stark's new method for analyzing discrepancies also
let's us know if the audit sample size was adequate or not.

Stark's method for analyzing post-election audit discrepancies uses
the maximum signed pair-wise relative discrepancy between all
winner-loser pairs in each precinct (or other vote count) to calculate
a test statistic; and he uses the same upper margin error bound [for
each winner-loser pair-wise margin] that I have been recommending for
months now.  So Stark's new method is conservative because if anything
it would err a little on the side of not certifying (i.e. on expanding
an occasional audit unnecessarily) in the case of a correct election
outcome.

Stark's method for analyzing election auditing discrepancies is simple
and elegant. However Stark's papers can be difficult to comprehend for
persons who are not mathematicians or statisticians. As soon as I
catch up on critical chores and paperwork, I plan to add easier
descriptions of his latest work to my papers (below) on election
auditing mathematics and procedures.

Mandatory Vote Count Audit
http://electionarchive.org/ucvAnalysis/US/paper-audits/legislative/VoteCountAuditBillRequest.pdf

The fundamentals of scientific election auditing described in lay
persons' terms (as much as possible): Post-Election Vote Count Audits
-- Probability Proportional to Margin Error Bound (PPMEB) Method
http://electionarchive.org/ucvAnalysis/US/paper-audits/VoteCountAudits-PPMEB.pdf

Stark's new paper on analyzing post-election audit discrepancies is posted here:
Maximum Pairwise Relative Margin Overstatement: A Sharper Discrepancy
Measure for Post-Election Audits
http://statistics.berkeley.edu/~stark/Preprints/pairwise08.pdf

Stark will be incorporating his new work (above) into his prior paper
(below) to provide a discussion which includes calculating election
auditing sample sizes: Election audits by sampling with probability
proportional to an error bound: dealing with discrepancies
http://statistics.berkeley.edu/~stark/Preprints/ppebwrwd08.pdf

--------

Difficult steps come next - getting state election officials &/or
state legislatures to adopt these new valid, mathematically correct,
election auditing procedures for verifying the accuracy of unofficial
election outcomes before the outcomes are declared official - to a
desired confidence-level.

Election auditing procedures currently used by the 20+ states which
audit elections are virtually all inadequate, internal, and/or employ
procedures which do not measure of the accuracy of unofficial election
results.

Some state legislatures will need to increase canvass periods to 28
days to allow enough time to manually audit mail-in and provisional
ballots as well as the Election Day and early voting results.  Also
most states need to implement public oversight of, and participation
in, ballot security and ballot reconciliation procedures - required so
that election audits provide the public a way to verify election
outcomes in an observable way.

There is still some work remaining - to figure out the best
mathematical approaches to the complexities of two-stage audits (stage
one would be auditing the early and election day counts; stage two,
the provisional and mail-in ballots that are often counted within 14
days after Election Day.) and to figure out the nitty-gritty of the
best methods for making publicly transparent random selections of vote
counts to audit.

Stark's latest work is a big step forward because it means that the
mathematics have been worked out now both for calculating election
auditing sample sizes and for evaluating the amount of audit
discrepancies to determine whether election outcomes should be
certified to a desired confidence-level or rather the audit samples
expanded.

U.S. citizens can help promote election auditing by asking their US
Representative (http://www.house.gov) to vote "Yes" on H.R. 5036. A
copy of the final House Admin Substitute version of H.R. 5036 is
posted here:  http://electionarchive.org/ucvInfo/US/legislation/HouseAdminHR5036-MAR28.pdf

Thank You!

Cheers,

Kathy Dopp

About the National Election Data Archive: The National Election Data
Archive has been organized for educational and scientific purposes of
promoting fair and accurate elections by promoting public access to
election records and data and by developing technology and
mathematical methods to detect any voter disenfranchisement or vote
count inaccuracy.  Such methods include independent manual vote count
audits, exit poll discrepancy analysis, and the public release and
scientific analysis of election data along with public release of
election records necessary to verify the integrity of elections. NEDA
is a completely non-profit organization that relies heavily on the
donation of time by its volunteers, many of whom are mathematicians.

P.O. Box 680192
Park City, UT 84068
phone 435-658-4657

http://utahcountvotes.org
http://electionmathematics.org
http://electionarchive.org

History of Confidence Election Auditing Development & Overview of
Election Auditing Fundamentals
http://electionarchive.org/ucvAnalysis/US/paper-audits/History-of-Election-Auditing-Development.pdf

Vote Yes on HR5036
http://electionarchive.org/ucvInfo/US/legislation/SummaryFlyer5036.pdf

Voters Have Reason to Worry
http://utahcountvotes.org/UT/UtahCountVotes-ThadHall-Response.pdf

"Enlighten the people generally, and tyranny and oppressions of body
and mind will vanish like evil spirits at the dawn of day," wrote
Thomas Jefferson in 1816