[EM] STV and weighted positional methods

[Kathy Dopp]

On Sun, Feb 1, 2009 at 4:30 PM, Raph Frank <raphfrk at gmail.com> wrote:

> Not 100% of programmers are trustworthy.  However, there would be a  flood of people pointing out that there is a problem with the count ... (assuming they release
> the ballot data).

Yes. Let's hope there would be sufficient capable knowledgable
programmers to check. However, most jurisdictions do not release the
ballot data in the US. In some states unbelievably they even keep the
machine counts secret from the public and refuse to release them. And
in Utah they even mak the precinct-level counts hard to obtain. Either
obtain them from all counties individually if you can (no requirement
for the counties to give them to you) or pay $25 and wait two weeks
until after the election outcome is certified and only then will they
mail you a CD with the precinct vote counts on them!!  Many states in
the U.S. are entirely susceptible to undetectable outcome-changing
vote miscount.

>
> I don't support electronic voting machines.  IMO, voting should be done with paper ballots.

Even when paper ballots are used, the counts are virtually always
electronic done by trade secret software compiled into machine
language in the US and in most states the publicly reported counts are
never checked for accuracy after the election.

>
> We do PR-STV using paper ballots in Ireland.  The counting is done in public.  Representatives from the media, political parties and other groups are all present watching the counters do the manual counting.

That is far superior a method to that used in most US states where
ballots are secretly counted by private companies and the counts never
checked at a level that would assure accurate election outcomes, or at
all in most states.


>
> So, you mean take 5% of the ballots at random and just recount those ones.

No, not ballots, because randomly selecting ballots and recounting
them with a machine and manually does nothing to assure that the prior
election counts were accurately reported.

One must randomly select publicly reported vote counts that tally to
the total results and manually count 100% of the ballots for each
randomly selected vote count and compare it to the reported election
results count.

> It may be a reasonable method for determining if a recount
> needs to happen, but I am not sure that it is a good way to do a
> recount.

It isn't a recount and Yes, if the sample size is adequate and the
procedures are valid, then such a manual audit does determine whether
to certify the election or to expand the manual audit, perhaps to a
full recount.

>
> Also, I am not so sure that wouldn't work for PR-STV.  A
> representative sample should give the same result as counting all the ballots subject to random variation.

Well randomly selecting ballots is problematic. Obviously there are no
publicly reported vote counts to select so with IRV it has to be
ballots.  To do that you'd need entirely different voting systems that
would print humanly readable unique random numbers on the ballots
after voters cast them to preserve anonymity and prevent vote buying,
publish all the ballots and their identifying numbers and then make
the random selections and then rifle through all the ballots to find
the selected sample to compare with the published results which again
only programmers could show add up to the reported outcomes.  For a
method that is fundamentally unfair, going through all the expense and
hoops and buying the new election equipment necessary to audit like
this is not worth it, even if someone wanted to go through all the
expense, time and trouble.  Hence it is more practical just to do 100%
manual recounts as a method of checking STV vote counts.

Why not use a fairer voting method that is easier to check the machine
counts since most methods are precinct summable?


> Well, you could check a random few ballot images and make sure the
> official rankings associated with those ballots are correct.

That doesn't tell you that the total is correct though, so it requires
that anyone who wants to know if totals are correct has a programmer
he or she trusts.

100% manual counts are the only reasonably understandable way for the
public to know that the counts are correct with STV, unlike with other
precinct-summable methods.

> I don't think spreadsheets are the be all and end all of programming simplicity.

Yes. well it is interesting that STV can not be automated in a
spreadsheet that you could give to election officials or others to use
to count it, whereas other voting methods can be.  Many more people
can use spreadsheets than check a complex program that could count an
STV election and make sure that the executable was really of the
correct source code, etc.

>
> In plurality, how do you see voters actually making sure the count is correct?  It's not like they would actually handle the ballots themselves.

No, but they can add up all the precinct totals and see that they add
up correctly and then manually audit a random sample of precincts. It
is very very trivially simple compared to trying to check the trade
secret machine counts of an STV election.

-- 

Kathy Dopp

The material expressed herein is the informed  product of the author's
fact-finding and investigative efforts. Dopp is a Mathematician,
Expert in election audit mathematics and procedures; in exit poll
discrepancy analysis; and can be reached at

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

http://utahcountvotes.org
http://electionmathematics.org
http://kathydopp.com/serendipity/

Post-Election Vote Count Audit
A Short Legislative & Administrative Proposal
http://electionmathematics.org//ucvAnalysis/US/paper-audits/Vote-Count-Audit-Bill-2009.pdf

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

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