Palm Beach Post
Poll workers ignored flaws in pre-vote machine tests
By Joel Engelhardt and Scott McCabe
In the dawn hours of the Election Day that would forever change Palm Beach County, Carl Cummis, age 83, and his crew of poll workers struggled feverishly to prepare.
They had one hour to turn the county library west of Boca Raton into a polling place, complete with 15 tested and ready voting booths. Cummis and his crew scurried through the rigorous procedures taught by Elections Supervisor Theresa LePore and repeated on countless election mornings.
But when it came time to put the machines through their paces -- to punch out every chad to make sure nothing could block voters from registering their choice for president and two dozen other elected offices ...
Somehow, they dimpled ballots. Somehow, they left hanging chads. Somehow, they didn't punch holes they were supposed to punch.
The failure of the machine tests illuminates another dark side of the election that thrust Palm Beach County into the glare of the national spotlight: Even poll workers struggled to punch chads the morning of Nov. 7, 2000. And yet, they did nothing about it, letting voters cast ballots all day on those same machines.
In the county's 531 polling places, 261 had machines that registered errors on the tests conducted before the polls opened at 7 a.m. If that happens, poll workers are supposed to seek a replacement machine. None did.
"That's cute. They didn't even notice?" said University of Iowa Associate Professor Doug Jones, an election equipment expert. "In the morning test, if they have a machine that leads them to a dimple, then they were plain careless."
In all, 11.6 percent of the 4,867 test ballots completed that morning recorded errors in the presidential column. The most common error? Skipping a chad, or several chads. In 65 cases, a poll worker didn't punch out a single one of the 10 chads representing the presidential candidates.
Thirty times poll workers skipped the right side of the ballot, punching out chads only for the six presidential candidates on the left side and providing more potential evidence of the confusion wrought by the two-page butterfly ballot.
The problems with test ballots also help deflate another Election 2000 claim: that it's impossible to dimple a ballot when trying to vote. After the election, one election worker even purposely stuffed a machine with chads and said he still couldn't produce a dimple.
But the poll workers could.
In fact, in 3 percent of the test ballots, poll workers made the kind of errors that made Palm Beach County voters synonymous with voting ineptitude.
Why? Was there something wrong with the machines that made it impossible to punch some chads? Were the machines too stuffed with chads? Was the ballot misaligned? Did the poll workers not punch hard enough? No one knows. Election officials never examined the test ballots and never reexamined those machines.
The rate of error on the test ballots surpassed that recorded by the contested election's end, when 10,311 Palm Beach County ballots -- 2.2 percent -- went uncounted because the voter recorded no choice for president. A March Palm Beach Post review of those ballots showed that more than 5,000 contained a dimple of some sort.
No calls about ballots
If the problems had been detected, the machines could have been taken out of service.
That's what poll workers are taught. But no poll worker called that morning to complain about mistakes on test ballots, said Tony Enos, the county's voting systems manager.
But workers in precincts with the most errors on test ballots said they saw nothing amiss, so they had no reason to question the machinery.
LePore said she reviews test ballots only to fend off challenges from unhappy candidates. The test ballot, she said, can prove that the machines were working. Otherwise, there's no reason to look at them once the election is over.
She said she explains all that to workers, spending five minutes during training sessions on the importance of the test.
No mention of chads
She gives workers a manual, which spells out the procedures: "Punch every possible voting position on every recorder to be sure all are in proper working order. . . . Initial it, and put card in TEST BALLOT ENVELOPE for return."
The manual doesn't specifically instruct workers to check the test ballots for hanging or pregnant chads. It doesn't advise workers to call for a replacement if a machine isn't working properly.
The errors indicate a problem with training, which will be important no matter what kind of machines voters use, said Jones, the Iowa elections expert.
"If you train that this is a ritual, it becomes a formality," Jones said. "If they followed the procedure as they were trained to follow it, then the question is, what exactly is the procedure they were trained to do and how much purpose were they told?"
Procedures required poll workers to call the elections warehouse and ask for replacement machines, which could have been delivered that morning. Or they could have gotten help -- for example, someone could have cleaned out the machine.
Some polling places had extra machines sitting right there, waiting to be used in event of an emergency. Cummis, the clerk responsible for checking the test ballots in Precinct 206D, said he had no reason to use his polling place's two extra Data Punch machines.
However, an examination of his precinct's test ballots showed it tallied more erroneous test ballots than any other, with 11. Dimples and other problems relating to poking out a chad could be seen on five of those ballots, the most in the county.
Cummis, of suburban Boca Raton, said he saw nothing amiss.
"There were no chads in mine," Cummis said. "Most of my people were experienced people. They were doing this for years."
One of his poll workers, Winifred James, recalls carefully punching the holes for every candidate but can't recall examining the test ballot, which she initialed. Instead, she simply handed it to Cummis, who slipped the test ballot into an envelope, sealed it and signed the back.
However, the 10 precincts with the most under-votes did not have repeated test ballot problems, indicating that the early morning tests alone could not have averted the high numbers of under-votes.
All machines not equal
The county used two types of punch-card voting machines: Data Punch and Votomatic. Most of the under-votes recorded on Election Day were cast on Data Punch machines, even though those machines were used by less than one-third of the voters.
Poll workers also had trouble with Data Punch machines. Test ballots in 26 of the 76 Data Punch precincts had two or more dimple-related errors while just eight of 455 Votomatic precincts recorded two or more dimple-related errors.
On election night, the sealed test ballots are sent to the elections office. Most remained sealed until reviewed by The Palm Beach Post over a period of several months.
Staff writer Meghan Meyer and database editor Christine Stapleton contributed to this story.