mandatory vote audit musings

From: Jerry Lobdill <lobdillj_at_charter_dot_net>
Date: Tue Jul 04 2006 - 18:09:40 CDT

More about my concerns re mandatory vote count audits.

I'm afraid I didn't communicate my concerns very well in my post on
June 26. So let me try again.

I'll use different examples this time.

In a county-wide race for County Chair in the Democratic primary
there were 13500 total votes cast. One candidate got 5940 votes and
the other got 7560 votes after votes of all kinds were tallied. There
were a total of 634 precincts in the county. One JP district in the
county contains 63 precincts. This district is historically notorious
for mail ballot fraud. There were 1650 mail-in votes from this JP
district. In addition, there were electronic ballot scanners used in
each precinct on voting day. The DRE vote was limited mostly to early
voting, and there were no VVPBs for these votes.

The vote margin, calculated from the foregoing data is 1620
votes. The average number of votes cast per precinct is 21, a
depressingly small turnout, but this was a mid-term primary election.
(These data are real, by the way). The total number of registered
voters in the county is about 890,000 and the number of registered
voters per precinct varies from less than 100 to more than 3000.

1. The mail ballots in this case are very much more likely to be
corrupted than any other votes.
2. The number of mail ballots from the JP district precincts exceeds
the vote margin in this race (for County Chair).
3. There are probably many ways in which 1620 votes could have been stolen.
4. If there was an attack on the E-scanner boxes it seems reasonable
that not all boxes would be involved and that those boxes involved
would be those where the most votes would be expected to be cast.

Now, in the absence of a law prescribing specific audit rules I would
not elect to do a random audit using the hypergeometric function with
all 634 precincts considered equally probable targets for tampering.
I would not compute the number of precinct vote counts assuming 64
precinct vote counts (10%) corrupted and a 0.95 minimum probability
that the random audit finds at least 1 corrupted precinct.

I would first select all 63 precincts in the suspect JP district for
audit. I would audit the 1650 mail ballots in the suspect JP
district precincts. If that audit proved mail fraud, that would
likely be (I would think) sufficient evidence to trigger a complete
audit of the race. (Proof positive would definitely be discovered if
there was fraud.) But no presently proposed law provides guidance
for such cases as this.

Suppose instead that we had used the rules that the Utah proposed law
suggests to conduct a mandatory audit of this race. We would use the
hypergeometric function with 634 total precincts, assume that 64
precincts are corrupted, and find the number of precincts to select
at random so that we have 0.95 or greater probability of finding one
corrupt precinct. That would yield a calculation of 28 precincts to
provide a 0.953 probability of finding one corrupt precinct.

So then how would we randomize the selection? We could use a lottery
or bingo style machine, and select 28 balls. That's arguably the best
way to make the selection. But we have ignored our best information
about the probable geographic distribution of corrupt precincts and
we have included, as assumed equally probable corrupt precincts, many
precincts that in fact, common sense tells us are unlikely prospects
for corruption. So our population from which we select 64 precincts
is known NOT to be a population of 634 precincts all of which have
equal probability of corruption. This makes our law into an arbitrary
process in which we delude ourselves that we are conducting a proper
statistical test for corrupted vote counts when we, in fact, do not
know what the probability is of finding at least one corrupt precinct
when we follow the process.

Another potential problem: The 12th Congressional District of Texas
encompasses parts of four counties. This is not an unusual situation.
When the mandatory vote count audits are tied independently to
counties, there is no audit coordination between counties for races
like this. Should the discovery of tampering in the 12th
Congressional District race in one county trigger a full recount of
the race in all its counties? The proposed law is silent on this.

Happy Independence Day!

Jerry Lobdill


OVC-discuss mailing list
= The content of this message, with the exception of any external
= quotations under fair use, are released to the Public Domain
Received on Mon Jul 31 23:17:03 2006

This archive was generated by hypermail 2.1.8 : Mon Jul 31 2006 - 23:17:09 CDT