http://www.wired.com/print/science/discoveries/news/2006/10/71999

Wired

Ohio Election Portends Trouble

Kim Zetter,10.31.06 | 2:00 AM

Six years ago the world watched dumbfounded as the Florida 2000 fiasco exposed the messy underbelly of U.S. election administration. Since then states have spent hundreds of millions of dollars on new electronic voting equipment to ensure that the nation would never experience such mishaps again.

But two recent and lengthy reports examining this year's May primary in Cuyahoga County, Ohio -- a pivotal state where the electoral votes gave President Bush his second win in 2004 -- make it clear that Florida-like fiascos are far from behind us.

The reports, totaling more than 500 pages, paint a disturbing picture of how million-dollar equipment and security safeguards can quickly be undone by poor product design, improper election procedures and inadequate training. From destroyed ballots and vote totals that didn't add up to lost equipment and breaches in security protocols, Cuyahoga's primary is a perfect study in how not to run an election.

The findings have ominous national implications. Cuyahoga County could play an important role in deciding two races in next week's election that will help decide which party controls the Senate and House. But one of the reports concluded that problems in the county were so extensive that meaningful improvements likely could not be achieved before that election, or even before the 2008 presidential election.

Moreover, few voting activists and election experts believe the problems are unique to Cuyahoga.

"I suspect that Cuyahoga County may be below average (in terms of how well it ran its election), but if you lift up the rock and look at election administration across the country, you'll see the same thing elsewhere," says David Dill, Stanford computer scientist and founder of VerifiedVoting.org, a proponent of paper-verified elections.

The problems in Cuyahoga County, a heavily Democratic county that encompasses Cleveland, began with a glitch in the design of the optical-scan ballots -- black lines separating sections of the ballot were too thick, and the fill-in bubbles were in the wrong place, preventing scanners from reading them.

The gaffe caused a six-day delay in election results, but wasn't insurmountable. The county hired an army of office temp workers to recast the votes from the 18,000 optical-scan ballots onto touch-screen machines -- both types of voting machines, made by Diebold Election Systems, were used in the county for the first time in the primary.

On the surface, that seemed to solve the problem. But a post-election audit of votes and election procedures revealed deeper issues in the way the state's most populous county ran its primary -- problems that, if not fixed, could open the state to serious legal challenges from candidates and voters in a close election.

Among the findings in one report (.pdf) prepared by the Cuyahoga Election Review Panel:

  • Due to poor chain of custody for supplies and equipment, 812 voter-access cards (which voters place in touch-screen machines to cast their ballot) were lost, along with 215 card encoders, which program the voter-access cards. Three hundred thirteen keys to the voting machines' memory-card compartments, where votes are stored, also went missing.
  • Officials set up two user accounts on the computer running vote-tabulation software, then assigned one password to both accounts and allowed multiple people to use them, thwarting any effort to identify individuals who might access and alter the system.
  • Sixty Board of Election employees took touch-screen machines home a weekend before the election to test a procedure for transmitting data on election night.
  • The election board hired 69 taxis to transport observers to precincts to collect memory cards and paper rolls on election night. But many cab drivers ended up gathering the materials themselves, and about half the cabs returned to the warehouse with election data, but no observer.
  • In at least 79 precincts the number of voters who signed the poll books didn't match the number of ballots cast. At least eight precincts had more ballots cast than registered voters. Because some polling places served several precincts, some of the discrepancies are explained by voters being directed to the wrong machines, an error that did not result in uncounted votes. But even when investigators tallied ballots and signatures for all precincts in a polling place, 15 locations still had mismatches. In one case investigators found 342 more voters than ballots.

There were also problems with tracking the voting machines themselves, according to a second report (.pdf) produced by the California-based Election Science Institute. ESI was hired before the election by Cuyahoga's Board of Commissioners to audit the election and measure the accuracy, reliability and usability of the county's new voting machines.

Out of 467 touch-screen machines assigned to 145 precincts that ESI audited, officials couldn't locate 29 machines after the election, despite days of searching. And 24 machines that were found had no data on them. "All their paperwork says (the machines) were deployed to polling locations but we can't figure out why there's no election data on them," says ESI founder Steve Hertzberg.

It's possible the machines were sent to polls but never used -- perhaps they malfunctioned or voter turnout was so low the machines weren't needed. But so far no explanation has been forthcoming. Cuyahoga County election director Michael Vu responded to initial questions from Wired News, but did not respond to follow-up questions about the missing machines and data

Experts say the chaos in Cuyahoga is a bellwether of broader election-administration problems nationwide.

"It should be a general wake-up call for all states and localities to make sure they have addressed these problems in their statutes and procedures to make sure that they don't run into the same problems," says Thad Hall, a political science professor at the University of Utah who assisted with the ESI report. "People should be taking all of this very seriously."

Beyond sloppy procedures, the ESI report found technological problems with the printers installed on the county's 5,000 new Diebold touch-screen machines. The printers produce the voter-verified paper audit trail, or VVPAT, mandated by a new Ohio law.

The printer problems turned up when the ESI team set out to examine the accuracy of the touch-screen machines. The team compared four sets of vote data from a sampling of 145 precincts. The data included electronic votes recorded on removable memory cards (used to tally the official count); electronic votes on flash memory inside the touch-screen machines; individual ballots on the paper-audit trail rolls; and a summary total of those ballots printed at the end of the paper roll.

Although the majority of the paper rolls were easy to read, 40 rolls contained ballots that were physically compromised in some way: Rolls were crumpled accordion-style due to paper jams; ballots were printed atop one another, making them illegible; rolls were torn and taped. Eighty-seven rolls were missing entirely.

The compromised ballots and missing paper rolls demonstrate that the paper trails voting activists spent three years pressuring states to mandate could be useless in a recount. Per Ohio's recount law, and laws in 14 other states that mandate paper trails, the paper roll is the official ballot in a recount.

But the law doesn't address what to do when paper records are illegible. An administrative rule from the secretary of state says the official ballot would revert to the electronic record if the paper trail were illegible or destroyed, which, in theory, would get every vote counted.

VerifiedVoting.org's Dill cautions that if just 10 percent of an election's paper rolls are compromised, the purpose of the paper is defeated, and a door is once again open for someone to rig an election.

"If someone wanted to fix an election they could program the machine to produce a certain number of faulty paper ballots so that (officials) had to count the electronic ones instead," Dill says.

The machines should be designed to shut down if the printer jams or poll workers load the paper improperly, says Dill.

Dan Tokaji, assistant professor of law at Ohio State University, says the issue of illegible ballots and conflicting recount laws clearly weren't well thought out in advance by Ohio lawmakers.

"It's quite possible we could see litigation over this issue in the event of a close election," he says, noting that the same problem could arise in other states. "We know about the problems with paper trails in Cuyahoga, but in other counties both in and out of Ohio we really haven't done a careful assessment. So I don't think we know how widespread the problems with the VVPAT system are."

And unfortunately, he says, most people don't worry about the problems or fix them until after there's a blowup similar to what occurred in Florida.

Vote discrepancies also turn up in the ESI audit, and voting activists seized on this aspect of the investigation as evidence that there were problems with Diebold's software. On about 80 paper rolls, the votes totaled from the individual ballots didn't match the summary of votes at the end of the roll. In most cases the count was off by one to five votes, but in some it was off by more than 25.

Here, too, paper jams were the culprit, says Gary Smith, director of elections in Forsyth County, Georgia, who led the paper count for ESI. Printer problems led to missing and illegible ballots, causing the final numbers to be higher than the sum of the individual ballots. When his team printed fresh paper records from the memory cards, the totals from individual ballots matched the summaries at the end of the rolls.

The ESI audit also uncovered mismatches between paper-vote totals and the digital totals from the memory cards, and 13 cases in which the removable memory cards and the flash memory inside the machines didn't match up.

Diebold said there are innocent explanations for why memory card totals didn't match flash memory totals. Eleven of the discrepancies were a side effect of election workers' transference of botched optical-scan votes onto touch-screen machines. Because they failed to also add the optical-scan votes to the flash-memory data from those machines, the two sets of data didn't match when ESI compared them.

The other two discrepancies occurred when poll workers placed memory cards in the wrong machines, then moved the cards to the correct machines after voters had already cast a few ballots on the machines -- a switch that did not affect the voting tally, but seriously complicated the audit afterward. When election staff gave ESI data for those machines, they were unaware that some votes on these memory cards were backed up on the flash memory of other machines.

Cuyahoga County election director Vu says his staff has verified Diebold's conclusions but ESI has not had a chance to independently verify the explanations. However, all these issues point to the most important finding from Cuyahoga County, says Hertzberg: it was so difficult for ESI to get accurate and complete data to conduct its recount.

ESI had been assured by Vu's office that the data ESI was auditing did not include optical-scan votes, says Hertzberg. In fact, ESI had to request data from Vu's office six times because it kept getting the wrong data, or data that was tainted with optical-scan votes.

Vu's staff also had a dozen different versions of the same vote data but no file-management system or naming conventions to keep track of them or explain why so many versions existed, says Hertzberg. "So these guys were sitting there trying to search for which data set to give us and they couldn't figure it out," he says.

Political scientist Michael Alvarez of the Caltech/MIT Voting Project, who helped with the ESI report, agrees: Although many people focused on the vote discrepancies, the fact that ESI found it so difficult to conduct an audit is the biggest problem, and it doesn't bode well for what could happen in November or in a close presidential race.

"An independent entity, be they ESI or be they anybody else ... should be able to walk in and very easily replicate that election outcome," Alvarez says. "That's the sort of thing that is what induces voter and candidate and media confidence in the process."

Vu says his office was overwhelmed with the task of adjusting to new procedures and equipment. His staff also never anticipated that poll workers would swap memory cards between machines; they're addressing this issue in poll worker training for November. "For the most part," he says, "the majority of the county did relatively well."

Hertzberg worries that things could be worse Nov. 7 because turnout is expected to be larger than in the primary. Cuyahoga has 1 million registered voters. Hertzberg urged Vu to develop and practice a manual count procedure before Election Day, but says there's been no movement.

"It's going to take time," Vu says. "You can't go from having problems in one election to perfection in the next election." Vu says he has made some changes for November, but mentions only issues with poll worker training, pay and recruitment.

The county Board of Elections and Board of Commissioners did recently appoint The Center for Election Integrity at Cleveland State University to monitor progress in the implementation of reforms recommended by the two reports, and the group will monitor the county's actual conduct during elections.

See Also:

Building a Better Voting Machine

Paper Trail Urged as E-Voting Fix

An Introduction to E-Voting

E-Vote Guidelines Need Work

More Questions for Florida