Voting mystery stirs call for paper trail
Audit printers, a possible solution for electronic "undervotes," would be costly and can't be ready by Nov. 2.
By JEFF TESTERMAN, St. Petersburg Times

October 4, 2004

TAMPA - In the Aug. 31 primary, the population of a small town - 12,498 voters - appeared at the polls in Hillsborough County and apparently decided not to vote in the race for state attorney.

No one is sure why those voters didn't vote, or if they did, what might have happened to their votes.

Local residents declined to vote in even greater numbers in Hillsborough races for property appraiser, School Board and some judges. In fact, the "undervote" - the phenomenon in which citizens push an activation card into a touch screen voting machine but have no vote tallied for one or more races - was significantly higher in Hillsborough than in many other large Florida counties.

Palm Beach County, which uses the same Sequoia Voting Systems machines as Hillsborough, saw an undervote ranging from 3.1 percent to 8.4 percent of voter turnout, depending on the race. Broward, which uses optical scan voting machines, had an undervote of 6.4 percent on the one race open to all voters.

Hillsborough's undervote on countywide contests ranged from 9 percent in the race that re-elected State Attorney Mark Ober, to a whopping 17.5 percent in the race in which Charles "Ed" Bergmann was elected circuit judge.

In Pinellas County, which uses Sequoia machines, the undervote on countywide contests ranged from 9.1 percent in a School Board race to 13 percent in a judicial contest.

The question is: Why?

Are voters refusing to vote? Or are votes not registering on electronic machines?

Florida counties scurried to buy the voting machines after the Legislature outlawed punch-card balloting in the wake of the hanging chad controversy of the 2000 presidential election.

"I have always been concerned about the undervote on electronic machines," said Rebecca Mercuri, a computer expert at Harvard University who has written extensively about voting issues. "We don't know what happens with the votes because there is no real audit of the machines."

Mercuri maintains that fully electronic machines provide no means for the voter to confirm that the ballot cast corresponds to what is later tabulated. She believes that in some cases an unscrupulous programmer could even write code that records one result on the screen and another in a tabulation record.

Mercuri also has a feeling that the new electronic systems simply confuse voters, especially the elderly.

Hillsborough's tabulation of votes in August was beset by two significant problems: A computer server slowed to a crawl and delayed election night results; and 245 votes were never counted because a machine at an early voting site was not activated properly.

Buddy Johnson, Hillsborough's supervisor of elections, attributed the glitches to human error and believes voters should have complete confidence in the touch screen machines.

Johnson believes undervoting occurs when voters decline to vote because they know little about a race, or when they simply want to send a message that they don't care for any of the candidates on a particular slate.

"People just undervote," Johnson said.

Rob MacKenna, an Eckerd Corp. computer programmer who is the Democratic challenger to Johnson, strongly disagrees. To address the undervote question, he says, local government must pay for a paper-trail audit system as both a check on the electronic voting system and a boost to voter confidence.

MacKenna's skepticism about the cause of the undervote stems from the presidential preference primary last March, where a single question was listed on the ballot but where 255 Hillsborough voters, or 0.76 percent of the turnout, had no vote tabulated. Had 255 residents driven to the polls, signed in, walked to the touch screen machine, then decided to abandon the whole idea?

MacKenna refuses to believe it.

"The big question is, are we losing people's votes?" MacKenna said. "That's what the preliminary data seems to suggest. And that's the No. 1 thing we can't have."

Last month, MacKenna flew to a vendors show in Las Vegas to see audit printers manufactured by Sequoia. The shoe box-sized devices fit on the side of a voting machine and display a roll of audit paper under a fixed pane of glass.

The printer has cash register-style paper, easily visible to the voter, which displays the vote. It also generates a bar code that can be used with an optical scanner for recounts and election audits. Surprisingly, MacKenna said, voters who use machines with printers vote faster, apparently because they readily see their vote result and don't try to backtrack.

MacKenna's solution would not be cheap. The county would need about 4,000 printers, and at $700 each that amounts to $2.8-million. MacKenna thinks federal dollars allocated in the Help America Vote Act might be available to defray the cost.

The drive to ensure a paper audit of touch screen voting machines got new impetus last week when a federal appeals court revived a lawsuit by Rep. Robert Wexler, D-Boca Raton, challenging the electronic machines. Wexler believes a paper audit is necessary for a recount of votes on touch screen devices.

But don't look for the audit printers at the Nov. 2 general election. The printers haven't been certified by the state of Florida and can't be ready by election day.

Johnson shrugs off the paper audit issue as moot for the time being. "We push for every good innovation we can get," he said. "If this thing works and gives voters greater confidence, I absolutely will look at it.

"But if I can't do it by November, why try? It's just a political issue."

Mercuri, the computer science expert, says there is a silver lining around all the doubt being cast on touch screen and other electronic voting machines. More people are voting by paper absentee ballots, she said, and that means their votes are less likely to end up being undervotes.