11 years, most states have relied on voting systems tested to minimal federal standards, the results withheld from public scrutiny and given the green light by a nongovernmental agency working on a shoestring budget.
The era of approving tools of democracy on the cheap is coming to an end, and judging by talk at a national gathering of voting experts here this week, few will be sorry to see it go.
Carnegie Mellon University computer expert Michael Shamos, a state voting-systems certification official for Pennsylvania, is one of the staunchest advocates for new, fully computerized electronic voting systems.
But judging by what he has seen emerge from secretive, private labs known as "independent testing authorities" and approved by the National Association of State Elections Directors, Shamos said, "There's stuff in there that's so horrible, I can't understand it."
He found a quarter of the voting systems presented to Pennsylvania unsuitable for elections, with such "glaring failures" as an inability to tally votes correctly. A recent study led by the University of Maryland showed all of six voting systems tested did not record 3 to 4 percent of the votes. What does pass state muster often can break down.
"I have good reason to believe that 10 percent of systems are failing on Election Day. That's an unbelievable number," Shamos told an assemblage of voting-system makers, elections officials and scientists. "Why are we
"Things are getting through the certification process that really shouldn't," said software architect Eric Lazarus of DecisionSmith, a voting-systems consultant for the Brennan Center for Justice at New York University.
But if voting systems testing is as broken as scientists and voting advocates say, there is wide-ranging debate about what the future should look like. For the next year or two, the way that voting systems reach U.S. counties and voters is not likely to change.
The U.S. Elections Assistance Commission is working on new standards for voting systems, with more and tighter rules on wireless communications, ease of use and backup paper records for electronic voting machines. The commission and another federal agency also will be taking over approval of voting systems from the national election-officials group, as well as choosing labs to perform the testing.
But most of those changes will take 12 to 18 months. And commission Director Tom Wilkey, who had a hand in setting up the current method of voting-systems testing and approval, suggested this week that testing labs will be chosen in much the same way as they have been.
Given typical delays in approval and purchase of new voting systems, that means the next Congress and president will be chosen in 2006 and 2008 on voting equipment tested and approved under the current national system or something very like it.
But states such as California are moving toward more rigorous testing, perhaps in league with other states. To explore those ideas, California's Secretary of State Bruce McPherson hosted a conference on voting-systems testing that was the largest West Coast gathering of major players in the debate on voting technology in at least two years.
"California needs to assert leadership in this," said Stanford computer-science professor and VerifiedVoting.org founder David Dill. "I think we need to move more quickly than the federal government in this area."
Voting-system makers are wary of tougher demands by the states. The four largest vendors cautiously endorsed the idea of multistate testing, as long as it could be done at one time and place.
"It would be similar to taking your car to an inspection site and seeing if it met the emissions requirements for Kansas, Colorado and Texas," said Ian Piper, compliance officer for Diebold.
But vendors say testing already is rigorous enough, redundant and expensive. More testing will take time and money that could be used making voting systems better, they said.
Vendors can spend several hundred thousand dollars getting a state's OK. Many states rely on the national testing alone and start buying approved voting systems almost immediately. Florida and Georgia rely almost exclusively on their own testing.
California uses both the national testing and its own, which under McPherson has grown in rigor.
His office now requires every voting-system maker to supply dozens of machines for a massive, mock election to ferret out manufacturing or reliability problems. McPherson also has agreed to let a computer expert try hacking into a Diebold system, and state officials are weighing whether to require the same kind of security testing for all voting systems.
Diebold and some local elections officials are frustrated that it has taken more than two years to get the firm's latest touch-screen approved for sale and use in California.
Partly because of the testing and Diebold's own delays, a quarter of California's counties, including Alameda, Marin and San Joaquin, probably will miss Jan. 1 federal and state deadlines for offering new, handicapped-accessible voting machines with a paper record. Local elections officials want several months of lead time getting used to the new machines before the June primary.
"He's risking another huge meltdown for counties," said Los Angeles County Registrar Conny McCormack, head of the state association of local elections officials. "It's a setup for failure."
But the testing has disclosed security holes, as well as a bug that caused one in five Diebold touch screens to crash if a voter slid a finger on the screen, and forced Diebold to fix frequent paper jams in a printer for ballot records. Fixing the "sliding finger" bug alone cost the firm at least $250,000.
McPherson said this week that deadlines will have to take a back seat to making sure voting systems are secure, accurate and reliable.
"There is no compromise where election integrity is concerned," he said. "I cannot in good conscience certify systems that are not fully tested. This is a one-time show, and we're going to do it carefully."
Contact Ian Hoffman at email@example.com.