Voters short changed at the polls

State's rush to use untested electronic machines results in many failures, loss of franchise

By Ian Hoffman, STAFF WRITER

When Katherine Shao went to vote Tuesday morning in Emeryville, all she found was a row of powerless, blank touchscreen voting machines, soon joined by a single, harried poll worker.

In a rush herself, Shao helped by signing herself in as a voter and booting up the machines.

The screens glowed in welcoming colors. Local elections officials had touted them as faster, more accurate and "as easy to use as an ATM." They paid for a video beckoning voters to "Touch the Future."

But no electronic votes could be cast that morning at Anna Yates Elementary School: To vote, Shao needed a digital ballot, and the code for her ballot was locked inside yet another machine. The device, a Precinct Control Module model 500, stubbornly resisted entreaties to come to life. No code, no ballot, no voting.

Shao had to leave.

 She never got a say on paying an extra dollar to cross the Bay's bridges, on forcing Sacramento to deliver balanced budgets, on paying the highest sales tax in California to rescue local public-health clinics, on shouldering $27 billion of state bond debt for the next decade or who besides George W. Bush should be leader of the free world.

"My mistrust of the workings of this ethereal, dual-boot system really makes me wonder whose votes end up counting," she said.

Same for Kathy Nikkelson and other voters in Newark. At least 50 Pleasanton voters were turned away, a lot of them Republicans, just like hundreds of Republicans who ran into the same problems in San Diego County.

Technology is nonpartisan. And so was the failure of voter-card encoders Tuesday in Newark, San Leandro, Emeryville, Oakland, Pleasanton and across San Diego County.

The culprit wasn't a hacker or bribed programmer. The blame lies with elections offi-cials and Diebold Election Systems.

In spite of state rules and purchase contracts requiring the use of well-tested and nationally qualified voting equipment, they chose just days before the March 2 primary to use a Diebold device never tested against Federal Elections Commission standards and never approved by the National Association of State Elections Directors.

It wasn't the first time that voting systems not fully tested or approved were used in California elections. An audit last fall revealed that Diebold had uncertified software running in all 17 counties that it served at the time. State officials also gave one-time approval recently for Diebold's latest touchscreen machine, even though it had not been fully tested and showed potential vote-tabulation "anomalies."

The reason: four counties had spent $35 million on the new machines and wanted to use them Tuesday.

But it took Tuesday's voting problems in Alameda and San Diego counties to expose the ramshackle nature of voting system testing and approval as California and other states rush to embrace electronic voting.

"The whole certification process has become plastic in order to accommodate the fundamental necessity of holding elections. You would bring the state into constitutional crisis if you didn't," said computer scientist David Jefferson, a member of a state task force on touchscreen voting.

Critical device

The PCM 500 happened to be the one device that, if it failed, could cripple electronic voting. The encoder's proper function was essential to the operation of more than 17,000 voting machines in California on Tuesday, most of them supplied to Alameda and San Diego counties for $38 million, plus annual maintenance fees.

"In the long run, it says the certification process is not worthy of the confidence that has been given to it," Jefferson said. "The certification process is broken. It's not working and was not designed for the era of software-driven elections."

Nationwide, Diebold expects about $150 million in 2004 revenue on its e-voting machines and other election services. A tenth of that money would come from the sale of its new precinct control modules and other new services: computerized voter registration records and poll books.

At 1:45 p.m. Tuesday, technicians and poll workers had restarted electronic voting at most of 200 affected precincts in Alameda County. News reporters were clamoring for an explanation of the breakdowns. A Diebold public relations officer approached Brad Clark, the county's registrar of voters, and suggested he put off attributing the problems to her company's encoders.

Clark nodded his agreement that the full magnitude of the encoder problem wasn't absolutely clear.

"Is there any other way in which I can support you today?" Diebold's Ann Sinclair asked.

As Sinclair walked away, a reporter asked whether she was with Diebold.

"No," she said. "I'm just here to help Brad out."

She begged off further conversation, saying she had a cold.

It's not clear to e-voting skeptics what exactly Sinclair and other Diebold PR staff in California are helping the integrity of elections or Diebold's bottom line.

In Solano County, as activists asked county officials recently to not use Diebold TSx machines until they were qualified under federal voting-system standards, county voter registrar Laura Winslow deflected reporters to a Diebold PR subcontractor, Jo Murray. Murray lately identifies herself as spokeswoman for the Solano County Registrar of Voters. She says her services are covered under the "voter education" portion of Solano County's $3.4 million contract with Diebold.

When Congress approved $1.5 billion for states and counties to buy new voting systems after Florida's chad-filled debacle of 2000, lawmakers probably didn't intend the spending of "voter education" funds on corporate PR, says Kim Alexander, president of the California Voter Foundation.

E-voting critics say elections officials are too cozy with e-voting vendors. Fact is, electronic voting joins elections officials and vendors such as Diebold at the hip.

Reliant on Diebold

Unlike paper and mechanical voting systems, the highly technical nature of electronic voting has made elections officials reliant on Diebold to handle everything from software and hardware troubleshooting to fixing bad vote tallies, as a Diebold technician did in Alameda County for absentee ballot totals in the October recall election.

E-voting systems are proprietary and designed as single functioning units, so elections officials have nowhere else to turn for software and hardware than the original vendor, unless they are willing to discard the multi-million-dollar contracts on which they staked their credibility.

It's one reason that California and other states are left approving voter use of hardware and vote-counting software that hasn't been fully tested or approved under federal standards.

After 12 years, the Federal Election Commission revamped its voting-system standards in 2002 partly to add new requirements for accuracy and security.

The standards and the testing are the touchstone of voter confidence. Elections officials and e-voting vendors regularly assure voters that the new equipment is accurate, secure and trustworthy because it has been federally tested and certified.

"Certification is the last line of defense," Alexander said. The tests and approvals are especially valuable when electronic voting systems offer no paper backup records to assure voters and elections officials of accurate vote recording.

"It's not enough even if it were all happening, and it's not," Alexander said.

Lack of certification

Yet not a single piece of the electronic voting equipment used by 43 percent of California voters on Tuesday, nor the software that counted those votes, has been certified as meeting the 2002 federal standards.

In fact, no e-voting equipment in the nation meets the new standards. Most hardware and software is certified to 1990 standards geared more toward earlier voting technologies, such as optical scanning systems.

None of the voter-card encoders for Diebold voting systems has been certified to any standard, state or federal.

Older versions, which look like a hand calculator hollowed out to accept a voter-access card, were produced by a Diebold subcontractor and used without detectable problems in the October recall election and two others.

But the devices lacked enough memory to meet California's complicated voting rules for modified-open primaries.

They could not provide codes for eight political parties, plus unaffiliated voters who wanted a Republican, Democratic or nonpartisan ballot.

That complexity required encoders that could call up 2,838 different ballots in Alameda County alone. But then poll workers would have to handle at least two of the old encoders. The only encoders that could handle the task in a single device were Diebold's PCM 100 and PCM 500.

Diebold submitted new, more capable encoders late in 2003, too late for pre-primary testing under federal voting standards. No other e-voting vendor in California has so consistently pushed deadlines and pressed for state certification without full testing, according to elections officials.

"Diebold has a habit of dragging their feet, doing things late," said Tony Miller, legal counsel for the elections division of California's Office of the Secretary of State. "With each failure, their credibility declines in the eyes of everybody."

A single, software-testing laboratory, Huntsville, Ala.-based Ciber Inc., agreed to a quick, limited test of basic function. It ran five counties' worth of ballot codes through the devices. But it did no further testing for reliability and durability.

Eight days before Tuesday's primary, California state elections officials rush-approved a one-time use of the new encoders for the election.

Miller said there was little other choice.

Lack of training

Technically, poll workers could have used a touchscreen machine in each polling place to encode voter cards. But they'd never been trained to do it, and local elections officials feared that they couldn't spare the touchscreens.

"We decided these are not practical alternatives," Miller said. "The (county) elections officials were very concerned about how they could even conduct the election without these devices, and we took that to heart."

On Tuesday, dozens of the encoders booted up into a Windows screen that no poll worker had seen in training. It took Diebold technicians to realize the machines were behaving differently because of low battery charge.

Other encoders appeared to have loose memory modules, and technicians tried talking poll workers through re-seating the modules on election day. Some encoders also appeared to have insufficient memory: technicians guided poll workers on the phone through deleting backup files inside the devices.

All told, encoder problems delayed or halted voting in as many as one in five Alameda County precincts and one in seven San Diego County precincts. In Alameda County, voting problems injected extra uncertainty into election results. With more than 10,000 absentee and provisional ballots still being counted, the results of close races could change over the next 25 days.

The secretary of state's office is investigating.

"We are taking extraordinary steps to find out what did go wrong," Miller said. "We will take steps to prevent it from reoccurring, this (encoder problem) and other issues that disenfranchise voters. This is the highest possible priority of this office."

As for Katherine Shao, she says her next vote will be on a paper absentee ballot.

Contact Ian Hoffman at ihoffman@angnewspapers.com