Florida e-voting data lost, then found

Florida officials first lost then found archived data in Miami-Dade County from the 2002 gubernatorial primary election, adding another example to the growing list of cases that electronic voting opponents cite when criticizing touch-screen voting machines.

In this case, electronic machines recorded fewer votes in the election than the number of voters who had signed in at polling locations, according to a survey of 31 voting precincts. But when a citizens' group, the Miami-Dade Election Reform Coalition, requested electronic ballot image data that the county was supposed to have stored, county officials found that system crashes had apparently wiped out the data.

But July 30, the data turned up on a CD that had been tucked away in a conference room near the election supervisor's office, said Seth Kaplan, executive assistant to the supervisor of elections for the county.

County officials added a tape backup system last year to ensure that data doesn't get lost again, he said. They also instituted a policy in January 2003 that requires all election data to be burned onto CDs and stored in addition to the tape backups, Kaplan said.

Florida law requires that officials retain federal election data for 22 months, and state or local election data for one year, said a spokesperson for the Florida Secretary of State's office.

Kaplan said his office will make the data available to the coalition.

Last year, officials from the Florida chapter of American Civil Liberties Union examined 31 voting precincts and found that at least 1,544 more voters had gone to the polls than there were votes cast. The difference between voters and votes in the precincts studies averaged 8.2 percent.

About half of the missing votes were cast by African Americans, ACLU officials said. In the 2000 election, when minority disenfranchisement was an issue, studies shows that about 4.4 percent of voters using punch card ballots did not have their votes recorded.

The coalition wants to use the digital ballot images to continue investigating the discrepancies, said member Martha Mahoney, a law professor at the University of Miami. An image for each ballot is recorded, and the total number of images should match the total number of votes on the summary report that the machine generates on election night.

Harris Miller, president of the Information Technology Association of America and a supporter of electronic voting, disputed the assertion that the loss of information indicates a problem with e-voting. Instead, he said, it shows a flaw in the procedures that election officials used to manage data.

"It's about backing up," he said. "It's like saying, 'We wrote down the election results, and someone put them in the incinerator.'"

Miller also said that a discrepancy between the number of voters and the number of votes does not mean votes went unrecorded. "In any election, there are always a small number of people who come out and [do] not vote," he said.

Skeptics of e-voting have advocated measures that would increase election officials' ability to conduct audits after an election. The most common proposal is to require that the machines generate a receipt that voters verify. Should a challenge arise, or if officials simply want to audit the accuracy of the machines, they could count the paper receipts by hand and compare the results to the machine-generated totals.

Will Doherty, executive director of Verified Voting.org Inc., a group that advocates a paper trail and other security measures, said the permanent disappearance of data would have made the question of the election's accuracy unanswerable.

Without the data, "there is no record to check against, neither a voter-verified paper ballot nor even an admittedly insufficient electronic record," he said.

Even though officials found the election data on the CD, the situation is not better, he added. "It's not acceptable for election records to be lost, even if temporarily," Doherty said.

***

Lost votes?

In a study of the 2002 gubernatorial primary election in Florida, the state chapter of the American Civil Liberties Union examined 31 precincts and found evidence that some votes may not have been counted.

Voters who signed the rolls: 18,752

Votes recorded: 17,208

Number unaccounted for: 1,544 or 8.2 percent

Similar undervote in 2000 on punch card ballots: 4.4 percent

Estimated minimum number of possibly lost 2002 votes in 198 precincts with reported problems: 8,500

Highest loss rate at some precincts:21.5 percent

Source: American Civil Liberties Union of Florida