Panel directs ire toward voting system

By Brad Turner    The Longmont Daily Times-Call    18 June 2005

BOULDER ? The manufacturer of Boulder County?s new $1.4 million voting system made unrealistic claims about the machines? capabilities both before and after the messy 2004 election, according to an investigative panel and government records.

Neil McClure, vice president of Hart InterCivic, estimated the system could tally the November 2004 election in nine hours when he spoke to county commissioners in January 2004 as they were mulling whether to purchase the equipment.

When election week came, printing problems and improperly trained volunteers contributed to a 68-hour ballot tally, according to the Election Review Committee report released earlier this week. County commissioners appointed the nine-member team in December to investigate the causes of the tumultuous tally.

In testimony before the ERC, McClure said the equipment should have taken 22.4 hours to tally the election ? less than a third of the time it actually took.

Pre-election projection

?This is based on real empirical data from the field,? McClure assured county commissioners on Jan. 29, 2004, when he projected it would take nine hours to count a 90,000-ballot election in Boulder County.

McClure made his sales pitch with strong backing from County Clerk Linda Salas, who called Hart?s BallotNow system ?more flexible, more public and more easily verified? than other ballot-counting equipment.

?I am confident that I am recommending the best solution for Boulder County?s current voting-system needs,? she said.

Commissioner Tom Mayer questioned whether Hart?s time projection was too short because the assumptions were based on counting a one-page, two-sided, 11-inch by 17-inch ballot. Voters in some parts of the county had to fill in five-sided ballots because of numerous state and local issues.

At the time, McClure assured the commissioners the Hart system could ?absolutely? handle a longer ballot.

?There?s really no ceiling as to how long a ballot can get, and the system can accommodate that,? he said.

The count-time assumptions presented at the Jan. 29, 2004, meeting were based on the use of four Hart machines, priced at $55,000 each, to scan ballots.

The county ultimately purchased eight scanners in hopes of counting ballots in about seven hours.

But on Election Day, miniscule printing errors by an outside contractor caused 13,000 of the roughly 90,000 ballots cast to be rejected by the Hart system as flawed or unreadable.

Those ballots had to be tallied manually by volunteers, and the count dragged on until Friday night of election week.

The county board would not have purchased a system that took days to tally an election, Mayer said Monday.

New figure in hindsight

Amid harsh criticism from voting activists, McClure defended his system to the ERC earlier this year.

Even with the thousands of damaged ballots and many five-sided ballots, the election should have been counted in 22.4 hours, McClure said when he testified before the ERC after the election.

The ballots would be scanned in nine hours, and ?resolution? of damaged or unreadable votes could be completed in another 13 hours, he said.

The ERC dismissed those claims in its report.

?Hart?s estimate is clearly too conservative,? the ERC wrote.

The panel?s analysis describes a chaotic scene in the room where ballots were tallied on eight Hart scanners during election week.

?There was no time when all of the scanners were operating simultaneously,? the committee wrote. ?Most of the time, there were no more than two or three scanners being used. There were not enough machine operators and the training of these operators was minimal.?

McClure?s projections supposed that all eight machines would be used simultaneously, and each damaged ballot would be ?resolved? by election volunteers in 15 seconds.

That pace ?would be difficult to achieve, much less sustain over a 22-hour counting time,? the ERC wrote.

Later, the ERC altered the variables in Hart?s estimate to reach a more realistic projection.

They gave election volunteers 30 seconds, instead of 15, to resolve each rejected ballot. They also noted that protests from voting activists kept officials tied up in the days before the election, so early voting and absentee ballots were not tallied until Election Day.

Using the new figures, the ERC estimated that ballots should have been tallied in 55.4 hours ? ?similar to the actual time? spent counting, but still shorter, the panel wrote.

Speed or accuracy

Hart InterCivic chairman David Hart said Thursday he stands by all of the estimates presented by McClure before and after the election.

Boulder County?s election was bound to be relatively slow because officials had made ?extremely unusual? demands for the voting system they wanted to purchase, he said.

?We were told from the get-go that speed doesn?t matter and that accuracy does,? David Hart said. ?This system was ed because it had a unique capability to track every ballot and provide an audit trail.?

But tracking every ballot takes time, he said.

Mayer on Monday questioned Hart InterCivic?s argument that a rapid count was never in the cards, especially considering McClure?s sales pitch last year.

?I?m dumfounded that now no one says it was realistic or possible,? the county commissioner said.

Hart systems performed rapid tallies in the past. Officials in Orange County, Calif., counted 250,000 ballots ? with 135 candidates on each? for the 2003 gubernatorial recall in just 17 minutes on Hart equipment, David Hart said.

However, many ballots with stray marks and other flaws were left unresolved in that election, he said.

Boulder County officials need to decide whether they prefer a speedy tally or an accurate one, he said.

Officials will understand the system better next time and run a smoother election, he said.

?It?s not going to happen again,? he said of the slow count. ?Every customer we?ve ever had gets better and better as they refine the process.?

ERC: Hart OK for now

Despite headaches in November, the ERC did not recommend scrapping the Hart system.

The scanners performed flawlessly in March, tallying 13,721 mail ballots for the Boulder municipal election in less than three hours.

In turn, the committee endorsed using the existing system for this November?s mail-ballot election. That advice was irrelevant because the county commissioners approved mail ballots for November at a meeting earlier this month.

For the 2006 mid-term election, officials should consider purchasing optical scanners to count votes at the precinct level to increase efficiency and cut counting times, the report said.

Hart is seeking federal approval for a new line of precinct-level scanners. At about $4,500 apiece, the units cost far less than the larger scanners purchased for the 2004 election but dozens of them would be needed at precincts or vote centers throughout the county.

The ERC did not comment on whether the machines for precinct counting should be purchased from Hart.