Report taking shape
By RICHARD VALENTY Colorado Daily   11 April 2005

The Boulder County Election Review Committee is getting closer to releasing its report detailing problems with the 2004 county election process, but it appears a complete draft could be at least several weeks away from completion.

On Friday, the ERC drew up preliminary draft material dealing with voting system options, and the group is likely to suggest the county should use the existing Hart InterCivic "Ballot Now" system for the Nov. 2005 municipal elections. The group will submit the completed report to the Board of Boulder County Commissioners for consideration.

The ERC is charged with creating recommendations to avoid a repeat of the slow 2004 election results, when full precinct tallies from the Nov. 2 election were not released until Nov. 5. The all-volunteer committee has heard hours of testimony on technical issues, including improperly printed ballots, print specifications and paper conditioning or quality, and the ERC report is likely to include recommendations for future scrutiny instead of definitive judgment on all problems.

"We are going to use a lot of 'coulds' and 'woulds' and 'mays' in this report," said ERC chair Richard Lyons Friday.

The majority of the ERC agreed the Hart system was "ideally suited" for mail-ballot elections, although it has not been officially decided that the county will use mail ballots in November. Election officials were able to begin processing ballots 10 days before Election Day in the March 8 City Council mail ballot election. 2005 citywide voter turnout was less than 10 percent of 2004 countywide turnout and staff had more complete experience in using the Hart system, and complete results were released by 9:15 p.m. on Election Night.
 

However, the ERC is likely to recommend changes, not necessarily to replace the Hart system, for the 2006 mid-term election, which will probably see high voter turnout. The county processed at least 160,000 ballots in 2004 using eight scanners located at the Boulder County Clerk's office, and ERC members heard almost-weekly testimony suggesting the scanners and available staff could only process small percentages of the ballots hourly, even at peak efficiency.

The ERC could recommend renting or purchasing precinct scanners for the 2006 election. The group suggested the existing Hart system will work well in processing early and absentee ballots, but also said processing Election Day ballots at the precinct would speed the process significantly.

A number of voting activists have suggested doing away with electronic systems altogether and switching to hand-counting. The ERC is likely to list hand-counting as a "considered but not recommended" option in its report, but the group could recommend a hand-counted ballot audit to test the accuracy of the machine count.

Paul Walmsley, a local software development professional, computer systems analyst and election judge, gave a presentation Friday detailing his plans for a statistically significant hand count audit.

Walmsley said the county could ensure that voting machines are interpreting "99.5 percent" of ballots correctly in "99 out of 100" elections by manually reviewing 921 live ballots randomly out of an election where 150,000 or more ballots were cast.

In the Walmsley method, the election system would need to be able to generate a ballot interpretation report based on ballots with distinct serial numbers. For example, the report might say ballot number 0017 voted "yes" on state Amendment 34 while ballot number 0023 voted "no."

The hand counters would compare machine interpretation of voter intent with their own. City and county staff, along with temporary workers, performed a different type of hand audit after the March election. City Clerk Alisa Lewis chose three batches of 200 live ballots, the temps manually counted them and matched their tallies with the machine tallies, and the totals matched perfectly.

Walmsley and fellow voting activist Neal McBurnett have both suggested a test would be more transparent if the ballots were truly ed at random. Walmsley said Friday it would also be better to examine individual voter choices rather than aggregate tallies, but said a post-election test is necessary to enhance voter confidence that the machines performed properly.

"I think at least doing some kind of an audit is better than no audit at all," said Walmsley.