E-Vote Machines Drop More Ballots

By Kim Zetter
for Wired.com

Feb. 09, 2004 PT

Six electronic voting machines used in two North Carolina counties lost 436 absentee ballot votes in the 2002 general election because of a software problem, raising increasing doubts about the accuracy and integrity of voting equipment in a presidential election year.

Election Systems & Software said problems with the firmware of its iVotronic touch-screen machines, used in a trial run, lost ballots in two North Carolina precincts during the state's early voting in 2002. ES&S, the largest U.S. maker of election equipment, is also the focus of attention into lost votes last month in Florida during a special election.

ES&S, Diebold Election Systems and other electronic voting-machine makers are coming under increasing scrutiny about the accuracy of their devices. The manufacturers claim their machines work fine and don't need auditing mechanisms -- like printers that would give voters receipts confirming their choices. But opponents are finding more and more anecdotal evidence of discrepancies and anomalies. In addition, more election officials and losing candidates are raising concerns as all U.S. counties move to replace mechanical devices with computer-controlled touch screens.

"All of this just underscores the need for voting machines to have a paper trail," said Stanford University computer science professor David Dill, who runs Verified Voting, a group that is pushing election officials and legislators to mandate voter-verified paper ballots that provide a way to audit.

In the North Carolina case, a software glitch made the ES&S machines falsely sense that their memories were full and an error message displayed, said company spokeswoman Meghan McCormick in an e-mail to Wired News.

"Because the memory-full message appeared very quickly, some voters did not realize that their absentee vote had not been recorded," she added. "On election day, the election results were tabulated correctly, and the technical issue did not affect the outcome of the election."

"If this happened with one version of the firmware, how can we be sure that it didn't happen with other versions of the firmware?" asked Dill. "How can we be sure that other counties didn't lose votes that they didn't catch?"

ES&S found the error in October 2002 during absentee voting in Jackson County, North Carolina, and fixed the problem there. But Cherie Poucher, director of the Wake County Board of Elections in North Carolina, said ES&S didn't notify her of the problem, even though the two counties were using the same firmware version.

During Wake County's early election, absentee voters filled out an absentee ballot and submitted them to poll workers. Each application had a computer-generated tracking number, and before the voters cast their ballot on one of six touch-screen machines, a poll worker keyed the number into the machines. But when poll workers later compared the number of votes on machines to the number of absentee applications, the figures didn't match.

Poucher said her county immediately took the machines out of service and called an ES&S official. Only then did ES&S say that Jackson County, which used the same version of firmware, had experienced a similar problem days earlier. Poucher then asked ES&S for an audit trail to track down the lost votes. An ES&S technician visited Wake County and tried to retrieve the data for the audit report, but had to download the information onto a PC card and ship it back to ES&S central offices. Two days later, ES&S e-mailed the audit report to Poucher.

Using that information, Poucher's office tracked down the 436 early voters, notified them that their votes had been lost and asked them to revote for the Nov. 5 election. All but 78 recast their vote -- a number, she emphasized, that would not have changed the outcome of the election.

"This was one week before a major election," Poucher said. "Our office is not used to having something major to deal with at that time. It was a lot of added responsibility. We are viewed in this area with high integrity; I had to work hard to get everybody to understand that we were doing everything we could to keep that integrity."

Poucher decided not to use the trial touch-screen machines in the general election. Instead, she replaced them with the ES&S optical-scan system the county has been using since 1992. Because optical-scan voting involves a paper ballot that gets scanned through a machine, all votes would have a paper trail. Poucher sent ES&S a bill for nearly $6,000 to cover her office's scrambling.

The same ES&S iVotronic machines were implicated in election votes lost last month in Florida. In that case, 134 votes went uncounted in Broward County in a race to elect a state house representative. The lost votes belong to voters who signed in and cast ballots, but the ballots were recorded as blank. The figure represented 1.3 percent of more than 10,000 residents who voted in Broward.

In a contest between seven Republican candidates, the victor won by only 12 votes. To account for the lost votes, some speculated that Democratic voters came to the polls, realized that no Democratic candidates were on the ballot, and simply decided to cast blank ballots without choosing a candidate.

But Chas Brady, a spokesman for one of the losing Republican candidates, told the St. Petersburg Times he found it hard to believe that nearly 2 percent of voters took the trouble to go to the polls but didn't vote.

In the end, what really happened in Broward can only be speculation, because the voting machines left no paper trail to determine whether the machines lost the votes. Unlike the early voters in Wake County, who filled out an application with a tracking number, voters who cast their ballot on touch-screen machines on election day do not fill out any forms.

Several Florida county officials have asked the state to mandate a voter-verified paper trail in the state. But Gov. Jeb Bush and state legislators have said they're not interested.



In accordance with Title 17 U.S.C. Section 107, any copyrighted work in this message is distributed under fair use without profit or payment for non-profit research and educational purposes only. [Ref. http://www.law.cornell.edu/uscode/17/107.shtml]

Back to Current Edition Citizen Review Archive LINKS Search This Site