Florida Primary Recount Surfaces Grave Voting Problems One Month Before Presidential Election (FL)

Kim Zetter    Wired    07 October 2008

A month of primary recounts in the election battleground of Palm Beach County, Florida, have twice flipped the winner in a local judicial race and revealed grave problems in the county's election infrastructure, including thousands of misplaced ballots and vote tabulation machines that are literally unable to produce the same results twice.

Experts say the brew of administrative bungling and mysterious technological failures raise new and troubling questions about the county that played a crucial role in the 2000 presidential election debacle, and is one of a handful of counties considered pivotal in the upcoming presidential election. Voting advocates are fearful that problems here and perhaps in other election hot spots could trigger a replay of the disputed 2000 election.

"It doesn't get any more swing than that swing state," said Pamela Smith, president of election-integrity group VerifiedVoting, "and that's a major county. This is going to be a very high-turnout election. In any election you should be able to have justifiable confidence in the outcome. If you're having different results every time you count the ballots, that's not going to create confidence."

At issue is an August 26 primary election in which officials discovered, during a recount of a close judicial race, that more than 3,400 ballots had mysteriously disappeared after they were initially counted on election day. The recount a week later, minus the missing ballots, flipped the results of the race to a different winner.

The county eventually found the missing ballots after a prolonged hunt. But it also turned up an additional 200 or so ballots that officials never knew were missing and that were never counted in the original tabulation of the race. A recently-completed recount with all of the ballots has restored victory to the original winner. But the month-long saga has left voters and state officials exasperated and distrustful of the ability of county officials to run a competent federal election in November. More important, it's also uncovered perplexing problems in some of the county's high-speed optical-scan tabulation machines, made by Sequoia Voting Systems, which flunked accuracy tests prompted by the recount producing different results for the same batch of ballots.

The probe began after Florida's August primary, which involved a variety of local races in different districts. In Palm Beach County, the one-page ballot included a race for a Circuit Court judge seat, which resulted in attorney William Abramson beating incumbent Judge Richard Wennet by just 17 votes.

Florida election law requires a recount when a margin of victory is one half of one percent or less. But when officials conducted the recount a week after the primary, they discovered they had 3,478 fewer ballots than when they'd counted them on election day. Undeterred, they proceeded with the recount, which resulted in Abramson losing to Wennet this time, by 60 votes.

The county planned to certify the recount results, despite the absence of nearly 3,500 ballots and protests from Abramson, until state election officials stepped in and said they would not accept the results in that race until the county found the missing ballots.

A hunt for the ballots ensued, and was so successful that officials found an additional 227 ballots that were never counted on election day. All of the ballots were discovered in boxes in the county's tabulation center. Officials blamed the overlooked ballots on the disorderly way in which the recount was conducted, and the high number of ballots cast in the election.

On that last point, it's worth noting that about 100,000 ballots were cast in the county for the primary. The county is expecting more than half-a-million ballots to be cast on November 4th. The ballot in November will also be two pages long, as opposed to the one-page ballot used in the primary, increasing concerns that the county could become ground zero for an election meltdown in the presidential race.

After the missing and new ballots were discovered, a court ordered a second machine recount of the judicial race. The second recount confirmed the initial election results, that Abramson was the winner. But this time his margin of victory was 115 votes, up from the 17 votes by which he'd originally beaten Wennet.

But even that wasn't the final score.

As mentioned above, Florida law requires a recount if the margin of victory in an election is one half of one percent or less. That recount is only a machine recount, not a manual recount. However, if that recount results in a margin of victory that is one quarter of one percent or less, then county officials must manually examine ballots that were spit out by the recount machines for having an undervote or overvote in the disputed race, to determine if the machine missed legitimate votes or if the voter's intent was clear enough that the ballot should be counted anyway (say, for example, if a voter failed to follow directions on the ballot and circled a candidate's name instead of filling in a space next to the candidate's name as the ballot instructed).

So Palm Beach County proceeded to do a manual examination of some 12,000 ballots that the optical-scan machines had rejected. In the wake of that count, Abramson was still the winner, but now his margin of victory had gone down from 115 votes to 58 votes. Then a new surprise emerged.

Election officials discovered that an additional 159 ballots from 54 precincts may have had valid votes on them that never made it into the tabulation.

They determined this by looking at reports that each voting precinct produced on election day. Those reports indicated the number of ballots that the precinct-based optical-scan machines had flagged for undervotes or overvotes. Officials discovered that the numbers of undervotes and overvotes didn't add up to the total numbers that county election officials had calculated after the ballots were run through the county-based high-speed optical-scan machines.

This raised questions about whether optical-scan machines were erroneously rejecting legitimate votes as undervotes, or whether election officials had simply misrecorded the numbers or mistakenly placed legitimate ballots into piles of undervote ballots.

So county officials decided to ask a judge to let them conduct a third machine recount of some 3,000 ballots to see if they could find the mysterious 159 ballots that were throwing off the numbers. Additional legitimate votes were found in this round; at the end, Abramson was still the winner, but his margin of victory was now 61 votes, nearly the identical margin by which Wennet had won in the first recount.

"I'm overwhelmed," Abramson told the South Florida Sun-Sentinel. "This is why we're the greatest country ... We were faced with adversity and we all rose to the challenge."

Wennet had a different view.

He said Florida law allowed for only one recount and that the additional recounts were invalid. He asked a court to throw out any results after the initial recount that had him winning by 60 votes, even though that initial recount was conducted without some 3,500 missing ballots.

Barring that, Wennet asked for a do-over election on November 4th.

As part of his complaint, Wennet cited the unreliability of the optical-scan machines that counted the ballots. He asked the court to order a test of the high-speed optical-scan machines that were used for the recount.

Palm Beach County was using new optical-scan machines that it recently purchased from Sequoia Voting Systems for $5.5 million. The machines replaced paperless touch-screen machines that the county had purchased in 2002, which were bought to replace punch card machines that were involved in the 2000 election debacle. Florida outlawed paperless voting machines last year, leaving Palm Beach and 14 other Florida counties to change their voting systems for the second time in eight years. (All Florida voting machines, including touch-screen machines, must now produce a paper trail). The county used one model of Sequoia's scanners at precincts on election day, but used different high-speed scanners in the election office headquarters to conduct the recount.

So on Wednesday last week, the county conducted a test on a random sampling of its eight high-speed machines. It scanned about 262 ballots that had previously been rejected by machines for having undervotes or overvotes, and that had then been examined by hand to determine if the machine's reading of the ballots was accurate.

Only two of the county's eight high-speed machines were tested. That's because last Thursday the county was scheduled to begin programming its optical-scan machines for the November 4th election.

Officials expected the machines would reject the same ballots again. But that didn't happen. During a first test of 160 ballots, the machines accepted three of them. In a second test of 102 ballots, the machines accepted 13 of them, and rejected the others. When the same ballots were run through the machines again, 90 of the ballots were accepted. (All of these numbers come from the Sun-Sentinel story about the test. Election officials could not be reached to confirm the numbers.)

Wennet's attorney, Gerald Richman, who witnessed the tests firsthand, provided Threat Level with more details about the test, which, if accurate, clearly endorse his belief that the machines are highly unreliable.

Richman told Threat Level that four ballots that had previously been rejected by a high-speed machine were examined by a county canvassing board and deemed to have clearly-marked legitimate votes that should have been read by the machines. When the four ballots were run through each of the two high-speed machines again, three of the ballots were accepted and read this time by both machines, but the fourth ballot was again rejected.

Richman said the county then re-scanned two batches of 51 ballots each that had initially been rejected for having no vote cast in the judicial race, but that were found in a manual examination to contain legitimate votes for one candidate or the other. The first batch of 51 ballots were found to have legitimate votes for Abramson. The second batch of 51 ballots were found to have legitimate votes for Wennet.

In the first batch of 51 ballots containing votes for Abramson, 11 of the ballots that had previously been rejected as undervotes were now accepted by one of the machines as having legitimate votes, and the remaining 40 ballots were rejected as having no votes. In the second batch of 51 ballots containing votes for Wennet, the same machine accepted 2 ballots and rejected 49.

The same two batches of ballots were then run through the second high-speed optical-scan machine. This time in the first batch of Abramson's 51 ballots, the machine accepted 41 of the ballots as having legitimate votes (up from 11 on the other machine) and rejected 10 others. In the second batch of Wennet's 51 ballots, the machine accepted 49 of the ballots and rejected 2 the exact opposite of the results from the first machine.

"We just sat there with our heads spinning," Richman said. "It was unbelievable. Nobody has been able to explain it."

Richman said some of the ballots that were correctly marked were rejected, while other ballots that the machines read in the test should have been rejected by the scanners. These ballots were marked with a check or "X" instead of the voter filling in a broken arrow next to the candidate's name as the ballot instructed.

Because of the problems in these tests, a second round of testing was conducted last Friday on six precinct-based optical scan machines out of a thousand machines the state uses (these are machines that counted votes at precincts on election day, not the eight high-speed machines that counted ballots in the recount).

According to the Sun-Sentinel, the six machines functioned properly in that test. But Richman said the test only involved re-scanning ten ballots from two precincts to see if the ten votes that were initially recorded in the judicial race at those precincts remained the same in the test.

"We had no time to do more," Richman explained. "Those machines had to be serviced the following day and reset for the [November] election."

Neither Palm Beach County officials nor Florida's secretary of state have responded to several calls for comment. Sequoia Voting Systems also has not responded to a call for comment to explain why some of its high-speed machines rejected legitimate ballots. Sequoia has likely been preoccupied with another issue involving its optical-scan machines in Washington, DC, where hearings are being held to examine why a Sequoia system mysteriously added 1,500 phantom write-in votes to the September 9 primary election totals in one precinct and calculated a total of 4,759 votes cast in the precinct where only 326 were actually cast. The company has blamed the issue on static discharge or mishandling of a memory cartridge by poll workers.

Back in Florida, Richman has compiled an affidavit (.pdf) describing the results of the tests conducted on Palm Beach County machines to submit in Wennet's suit. But Richman, who is on Senator Barack Obama's legal team, says the problem isn't only how the machines malfunctioned in the recount and the tests, but how the problem might not have been uncovered at all if the judicial race hadn't been so close that it required a manual examination of the undervote ballots.

"[Ordinarily] nobody looks at ballots that have been counted by the machine," Richman said. "Normally there is a presumption that whatever the machine does is correct, and you only look at the overvotes and undervotes in a hand re-count. Which is a problem, because if [the margin of victory] is not within a quarter of 1 percent, you never get to a hand count. If there were mistakes involving other races, we'll never know what the mistakes were."

Douglas Jones, professor of computer science at the University of Iowa who has consulted with a number of states on voting machine issues, said the problem with the machines is likely inconsistent calibration among machines.

Jones blamed the federal voting system standards by which voting systems are tested and certified. He says the federal standards don't set a threshold for what should be an acceptable number of scanning mistakes and calibration decisions are thereby left to the companies that make them.

"This is an area where our voting system standards are virtually silent," he said. "The voting system standards only require perfect counting of perfectly marked ballots. They don't have anything to say about how the machine counts ballots marked by real people in real elections. The standards don't govern one of the most important things about the machines."

Jones also blamed election officials who fail to properly test machines before elections or who leave pre-election testing and set up of machines to vendors to do, rather than doing it themselves.

"That's the norm, that the vendor sets up the machines," Jones said. "And my experience is that counties that contract with the vendor to operate the voting system generally don't do anything to monitor the performance of that contract."

In 2004 in Napa County, California, optical-scan machines made by Sequoia failed to count more than 6,000 votes because the Sequoia employee who set up the machine failed to calibrate it to read certain kinds of pens that voters used to mark their ballots.

Florida is expected to be a fierce battleground in the presidential race. If Senator Obama or Senator McCain wins with a large enough margin, there will never be reason for a recount that might uncover the kinds of mistakes that were revealed in the primary aftermath.

If there is reason for a recount, the time frame to conduct it could be a problem, considering how long it took to complete multiple rounds of recounts in the judicial race. Florida state law gives counties only 12 days after a general election to certify election results (counties have only 7 days after a primary to certify results), which provides little time for a complicated or problematic recount to be resolved, that is, unless the state canvassing board refuses to accept problematic recount results, as it did in the judicial race this year.

In the 2000 presidential race, state officials weren't as accommodating. At the time, Florida's law allowed only 7 days to certify results from a general election, and then-Secretary of State Katherine Harris refused to extend the deadline to allow counties to conduct a manual recount of the paper ballots that were cast in the disputed presidential race. The state Supreme Court overruled her and extended the deadline by twelve days. But when Palm Beach County requested an additional extension to finish counting the last 1,000 ballots, Harris refused and certified George Bush the winner by fewer than 600 votes.

Smith, whose VerifiedVoting group led the national charge to replace paperless touch-screen machines with optical-scan machines, said it's not enough to have a paper ballots, Palm Beach and other counties using optical-scan machines need to have robust pre-election and post-election testing of machines to make certain they're calibrated correctly and that they're reading a wide range of pencils and pens that voters use to mark their ballots. They also need robust manual audits to randomly compare a sample of cast ballots with the machine tabulation rather than waiting on a close race and a recount to catch problems with the machines.

It's unclear if Palm Beach County plans to test its systems in the manner Smith describes. According to a news report, the county has asked Sequoia to examine all eight of its high-speed optical scanners next week.