Colorado considers verifying vote counts.

January 24, 2005

Representative Alice Madden
Senator Shawn Mitchell

RE: SB05-079_01 http://www.leg.state.co.us/

We are forever grateful to you both for taking the first step to address major flaws in Colorado’s voting system. We hope that you will accept the following recommendations as constructive.

1. Only verified votes count. The draft bill contemplates the possibility of two ballots being cast by each elector -- a “permanent paper record” and an “electronic record of vote”. The notion of two ballots per elector should be eliminated. There should be only one ballot, and it should be the permanent paper record that was verified and cast by the eligible elector. Counting votes on an unverified ballot is not meaningful.

2. Adequate verification of count. The draft bill contemplates a random verification of the vote counting processes. This is highly desirable, but the method proposed in the bill is inadequate.

a. The random sample of twenty precincts is not sufficient. Since each county uses a unique counting process (procedure, personnel and equipment), the sample should include precincts from each county.

b. In the draft bill, it is not clear which ballots in each selected precinct are to be included in the sample. The votes for every contest on all of the ballots cast in a selected precinct should be included in the sample including Election Day ballots cast in the precinct, absentee, mail-ballot, early, provisional, emergency, and any other eligible ballots.

c. The draft bill should require that the number of ballots counted be verified against poll books and batch control sheets for each voting method to ensure that ballots have not been lost or added during the original count.

d. In the draft bill, it is not clear what will be reported following the verification. The original and the verification count for each contest should publicly report: the number of eligible voters, the under votes, over votes, the number of ballots cast for each method of voting, and the votes for each contestant for each method of voting - Election Day ballots cast in the precinct, absentee, mail-ballot, early, provisional, emergency, and any others.

e. To achieve a high quality count, the people counting the votes for the Secretary of State must not know the earlier vote tally results for the precinct(s). They must report their written findings before learning earlier results.

f. Suggestion. Ballots that were rejected during the original process should be subject to reconsideration during the verification count.

3. Transparent verification. In the draft bill, the Secretary of State is not required to conduct the verification in a public and transparent manner. The draft bill should explicitly require that every step in the verification count, including the random selection of precincts, must be open to authorized poll watchers who must be able to actually see, read, and hear everything that is happening as it is happening.

The bill should establish a process for the public to challenge the verification process and results.

Before counting the votes on the sampled paper ballots, the Secretary of State should make available to poll watchers a file for each selected precinct containing a scanned image of each ballot surface. Poll watchers should be permitted to independently count the votes to increase confidence in the process. They should be permitted to load the file into one or more computers. A means should be established to uniquely identify each ballot surface so that people can refer to each specific surface.

4. Defective count consequences. The draft bill should specify what is to occur when any count total differs from the original count.

The bill should state that each difference between the original and verification count is significant, and that differences will not be “averaged out”. For example, a different number of early votes (say +27) and an equally different number of absentee votes (say – 27) for the same candidate must be considered 54 errors, since the number of votes is linked to the poll book for each voting method.

Under what circumstances is the counting system (procedure, personnel and equipment) considered to be unreliable? Under what circumstances will all of the votes in all of the precincts in a county need to be manually recounted? Under what circumstances will all of the votes counted in Colorado using a particular vendor’s vote counting equipment need to be manually recounted?

5. Timing. The draft bill should establish time limits and time relationships for completing the verification of the random sample, resolving public input, completing any candidate requested recount, and declaring official results.

6. When is remote access to machines permitted? We do not understand PG 3, LN 22 – 27. It suggests that a voting machine is permitted to be remotely accessed under some circumstances. The circumstances should be specified.

7. Public access to electronic election data and source codes. PG 4, LN 1 – 11 should provide for public access to the source codes including BIOS. It must also include a requirement that the version and level of all software must be tamper proof and verifiable. Also, the section should provide for public access to copies of any electronic election data created during the course of an election. Sample files should be made available at least three weeks before the election. The production data files should be made available before verification counting begins.

We hope that you will incorporate changes to the draft bill to accommodate these suggested improvements. We are available to assist in any way that we can.

Sincerely,

Al Kolwicz
Executive Director

CAMBER
Citizens for Accurate Mail Ballot Election Results
303-494-1540
AlKolwicz@qwest.net
www.ColoradoVoter.blogspot.com

Boulder County vote counting system is not reliable

DATE: January 11, 2005

TO: Boulder City Council

FROM: Bill Eckert, Al Kolwicz

SUBJECT: Analysis of Boulder County Vote Counting System – Revision to January 4th report.


The City of Boulder intends to use Boulder County’s vote counting system for the March 2005 City Election. Because of the problems encountered by the County during the November 2004 election, it is risky to use the system for the City election.

One way to estimate the reliability of a vote counting system is to count the votes twice. If the results are identical, the system is considered to be reliable. (A reliable system is not necessarily accurate since both results could be incorrect.)

Fortunately, a recount was required for the St. Vrain School District contest. The difference between YES and NO in the original four-county election was less than ½ of one percent – 127 votes.

Recount results and the following analysis shows that the Boulder County vote counting system is not reliable.

• Only 22.8 percent of the 79 precincts achieved identical results.
• There are 759 differences in “ballots counted” section of the reports.
• There are 34 differences in the “votes counted” section of the reports.

Clerical error might explain some, but not all of the differences. Counting equipment error might explain some of the differences. Intentional changes could explain some, but not all of the changes. For example, in precinct 4171107003 there are no differences in the number of ballots counted, yet there are 4 more YES and 1 more NO votes counted.

No matter how the differences are explained, it is the case that the results counted by the Boulder County Vote Counting System are not reliable enough to be used by the City of Boulder for its upcoming March 2005 election.

Potential consequences of using this unreliable system for the City’s March election are:

• The wrong candidate might get elected.
• A recount might be required (possibly counted using the unreliable system).
• The election can be challenged and a new election required.

The Boulder County vote counting system should not be used for the March City election.

====================

REPORT SUMMARY
January 5, 2005

January 4th, Boulder County released a copy of the original and recount reports for the St. Vrain School District contest. Each county report shows, for each precinct, the number of ballots cast for each voting method (absentee, early, and Election Day). See attached ORIGINAL and RECOUNT reports.

Also attached is a third report showing: (a) the differences in the number of ballots counted by voting method and precinct between the original count and the recount, (b) the differences in the number of votes counted by precinct in the original count and the recount, and (c) the differences between the individual precinct reports.. Notice the following:

The differences between the ballots counted in the original and recount were:

179 Absentee
207 Early
373 Election Day
759 Total

The differences between the votes counted in the original and recount were:

14 Yes
20 No
34 Total

The differences between precinct reports in the original and recount were:

A. Precincts with the same quantity of votes:

18 Same quantity of ballots and, same voting methods
41 Same quantity of ballots and different voting methods
59 Total

B. Precincts with a different quantity of votes

03 Same quantity of ballots and same voting methods
09 Same quantity of ballots and different voting methods
02 More ballots
06 Fewer ballots
20 Total

Only 18 (22.8 percent) of the precinct reports were identical. The remaining 77.2 percent were different.

No matter how the differences are explained, it is the case that the results counted by the Boulder County Vote Counting System are not reliable enough to be used by the City of Boulder for its upcoming March 2005 election.