Post-Election Audits for DREs require JBC tally report

[Neal McBurnett <neal@xxxxxxxxxxxxxxxxx>]

Josh Liss and Shelley Bailey:

It appears that you are not implementing an audit of the eSlate
results.  This is required by both the statues and the Secretary of
State rule, and is separate from the issue we've discussed before
about the need to do an audit of the scanner results (BallotNow).
A partial recount is not adequate for either type of ballot.

The Election Rules (8 CCR 1505-1, as adopted June 14, 2006) says:
 http://www.elections.colorado.gov/DDefault.aspx?tid=184

 11.5.4.6 For Direct Electronic Devices (DREs) that do meet the
 requirement of section 1-5-802, C.R.S., used for any function of
 counting ballots in an election, after the close of the polls, the
 designated election official will manually verify all of the voter
 verified paper record produced with the report generated for that
 specific devices, which contains the election summary report. The
 Secretary of State shall randomly select two races to be manually
 verified.

As noted during the training for the Machine Judges, and in the eSlate
Desk Reference manual we got, each JBC can produce a Tally report all
by itself at the end of the election day ("Print tally" on page 26).
But for some reason you have disabled this function in our system.

Without generating this Print Tally before the machines to be audited
are randomly selected, there are no results available to be audited.
Taking the MBB out of the JBC and using the separate tally system to
produce results for it on a separate Election database amounts to a
recount, which is not an audit, does not comply with the statute, and
does not lend the same level of confidence in the final results.

The Print Tally function should be re-enabled, and tested during a
LAT before the election begins.

Below is an update of the issue I brought up before the 2005 election
about the need for a BallotNow audit.

Please give me a call so we can talk about this.

Thank you,

Neal McBurnett                 http://mcburnett.org/neal/
303-494-6493

----

Summary:

 I worked with the legislature in 2005 to get a requirement for audits
 in our law.  Unfortunately, the Secretary of State's rules did not
 properly implement the law, as detailed below, and true audits were
 not in fact done - they did recounts instead.  Without a true audit,
 the public has much less reason to trust the system.  The RFP
 Evaluation Team recommended that true audits be implemented.  Please
 help us fix this problem.

The Post-Election Audits Rule, 11.5.4, from the Secretary of State is
out of compliance with the statutes, for the case of central-count
election machines.  This affected the 2005 elections in nearly all
Colorado counties, since there is a large central-count component in
most counties due to early, absentee and mail-in voting.  The problem
will get worse in 2006 since we just learned that, e.g., the Hart
Intercivic eSlate system cannot produce an auditable report of the
votes on each DRE (electronic voting machine) either, as detailed at

 http://www.coloradovoter.net/moin.cgi/ManualCountAudit

Please help us get the Secretary of State's office to fix the rule as
suggested below, and work with the county elections officials to make
sure they follow the intent of the law.  The changes made in June of
2006 increased the percentage of audits for some results, but did not
address either of the legal issues below.

There are two legal issues:

 1) The rule does not implement an _audit_ of the original
 tabulation.  It calls instead for a _recount_, and is thus only
 another, weak, _test_ of a single machine.  The purpose of an audit
 is to ensure that the device worked during the actual counting
 process, not that it can be made to work afterwords.  The Logic and
 Accuracy Test already addresses testing.  This distinction is clearly
 a part of the legislative history of the audit requirement.

 2) The rule does not call for a "specified percentage" of the ballots
 to be counted, as required by law.  Rather, it limits the a tally to a
 maximum of an absolute number of ballots: 500.  For all other
 equipment, the goal of the statute is to audit about five percent of
 the ballots.  Looking at only 500 ballots would be a fraction of
 a percent of a common turnout in Boulder - totally inadequate.

For precinct-count systems and DREs, the law and rule are consistent
in calling for a five percent manual tally, and for audits of actual
election results.  The rule for central count machines should call for
no less.

In particular, as recommended during testimony before the legislature
last spring, for the Hart Intercivic equipment the rule should select
a significant number of batches of votes, and a significant total
number of ballots, for manual tallies.  See this document for details:

 http://www.coloradovoter.net/moin.cgi/ManualCountAudit

Here is the relevant text from the statute:

 1-7-514 (1) (a) (I) ... where a central count voting device is in use
 in the county, the rules promulgated by the secretary pursuant to
 subsection (5) of this section shall require an audit of a specified
 percentage of ballots counted within the county.

and here is the relevant rule:

 11.5.4.4 For Optical Scanners used for the purpose of counting
 ballots in a Central Count/vote center environment as defined herein,
 the designated election official shall randomly select five (5)
 percent but not more than five hundred (500) ballots of all the
 ballots counted on the specific audited device. If the amount of
 ballots is less than five hundred (500) on the audited device, then a
 minimum of twenty percent (20%) of the ballots counted on the device
 will be manually verified. The public counter for that voting device
 shall be reset to zero, and the ballots shall be recounted on the
 voting device. A new report will be generated from the electronic
 count of the ballots and shall be manually verified. The ballots and
 a copy of the report shall be sealed in a separate container and
 secured with the remainder of the official election records for the
 election. The Secretary of State shall randomly select two races to
 be manually verified.

The solution to the first problem requires care during the original
vote counting process.  In order to preserve the principle of random
selection of the audited results, it is necessary to obtain tallies
for small batches of ballots.  During the audit, the elections
official then selects random batches, and compares manual tallies for
each batch.with the original machine tallies, as is done with other
types of equipment.

Hopefully the second problem can be fixed by just substituting the
word "less" for "more", and making sure the percentage refers to all
votes centrally counted:

 the designated election official shall randomly select five (5)
 percent but not _less_ than five hundred (500) ballots of all the
 ballots counted on _all the Central Count/vote center equipment_.


The rule also has to spell out the need for an auditable report as
clarified at

 http://www.coloradovoter.net/moin.cgi/ManualCountAudit

and that needs to be required of voting equipment before it is
certified.

In addition, the random selection of machines to test is currently
scheduled to happen "Within twenty-four (24) hours of the close of
polls".  I the case of a delayed count or delayed reporting, this
should not happen until the machines are done counting and a tally has
been obtained and made known.

Thank you,

Neal McBurnett                 http://bcn.boulder.co.us/~neal/
303-494-6493