Top vote-getter? We may never truly know

By Andrew Garber   05 December 2004
Seattle Times staff reporter

 Washington may never know for sure whether more people voted for Republican Dino Rossi or Democrat Christine Gregoire, even after the hand recount that starts this week.

From Spokane to Seattle, elections offices are about to begin the staggering and unprecedented task of hand counting 2.8 million ballots. But will the hand count in the governor's race really be any more accurate than the two previous counts?

Even experts can't agree. It's not clear whether humans or machines are better at vote counting.

But in the end, the question may not matter. No election system is precise enough to determine who won a race this close, they say. Only 42 votes separate Rossi and Gregoire, out of the millions cast.

"It's closer than the technology and our capacity as humans to decipher," said Jeffery Mondak, a political-science professor at Florida State University. "You folks would do as well to flip a coin as to try to determine who actually won."

Process could be lengthy

The closeness of the election provides a tremendous incentive for candidates to prolong the process if they're losing, because the winner could change with each recount.

 
Secretary of State Sam Reed worries that although state law allows only one more recount, lawsuits could drag out the uncertainty for weeks longer.

Republican Party Chairman Chris Vance wouldn't speculate on what his party will do if Gregoire wins the second recount. But he isn't in a conciliatory mood. "We're not going to sit back passively and allow this election to be taken from us," he said. "Dino Rossi won."

What's happening in Washington could have occurred almost anywhere in the country, said Doug Chapin, director of electionline.org, a clearinghouse on election-reform news and information. "There are 49 other states who are mopping their brows and saying, 'Phew, it could have been us.' "

Why? Elections work fine when candidates win by a large margin. When victory comes down to roughly the capacity of a Metro bus, small errors ? stray marks on ballots, punch cards that weren't punched properly and human mistakes ? can cloud the final vote tally.

Like survey polls that try to show what people are thinking, elections have what statisticians call a margin of error.

"There is a margin of error in connection with any measurement system, whether we're counting fish in a lake or counting votes for a governor," said Kirk Wolter, a statistics professor at the University of Chicago who did research on what happened in Florida during the 2000 presidential election.

In an election, Wolter said, "there are millions of human interactions, and we're all human and we all make random mistakes."

11,000 ballots per worker

It's not known precisely what the margin of error is in Washington's elections system. But that's not surprising, given that people can't even agree about the accuracy of counting votes by hand vs. machine.

The coming hand count is a gargantuan effort, and King County will be responsible for a lion's share ? counting about 900,000 optical-scan ballots.

The county elections office plans to take over rooms on two floors of an office building at 9010 E. Marginal Way, near Boeing Field. Tables will be set up for 80 teams of three people to count the ballots. Two people will do the counting and the third person will write down the results.

The county also will hire security guards, observers and others to make everything work, about 300 workers total. Additional people from the parties and the public are expected to show up as well.

Ultimately, each person counting is expected to go through several hundred ballots a day, or more than 11,000 during the entire process. They'll work every day, with half shifts on weekends so people will get a day off.

The county's recount is expected to cost up to $400,000 and conclude by Dec. 22. Democrats have estimated a statewide recount will cost them about $1.5 million. State law requires whoever requests the recount to pay the bill. If the election is overturned, they get the money back.

"It will be mind-numbing work," said Bill Huennekens, King County's elections superintendent.

Like many experts, he believes hand counts run the risk of human error. "If there's a stack of 500 pieces of paper, one day someone can count them and it will be 499, and another day it will be 501," he said.

The county has safeguards, such as double counting the ballots. Still, he said, "There's a reason why we do it with machines in the first place. You're talking about 900,000 pieces of paper."

Florida tally: "pure hell"

Theresa LePore, the elections supervisor in Palm Beach County, Fla., agrees. Her office had to hand count about a half-million ballots in the 2000 presidential election between George Bush and Al Gore.

"A hand recount ? I don't care what system you have ? is less accurate than a machine count," she said. "The longer you sit there and look at stuff, and the tension and everything else going on, the more chance of error you've got. It doesn't matter what you're hand counting, whether it's punch cards or optical scan, whether it's the audit logs coming off touch screens."

In Florida, it took 37 days to count about a half-million ballots, she said. The county had 100 two-person teams counting ballots and observers from each party watching every team. There were members of the public watching the process, constant challenges from political observers, and protests on the street outside.

LePore, who worked 21-hour days during the hand count, described it as "pure hell."

Secretary of State Reed says he doesn't consider either machine or hand counts to be more accurate than the other.

Tim Likness, elections supervisor for Clark County, the state's largest county that has punch-card ballots, said he expects a hand recount there to be as accurate as a machine count.

On the other end of the spectrum, the Democratic Party and Gregoire's campaign contend a hand recount is more accurate because of the extra scrutiny ballots will get. It's a view supported by Lance deHaven-Smith, an elections expert with an upcoming book about the 2000 election, "The Battle for Florida," who contends human recounts are more accurate than machine ones.

"When you do a machine recount, you're going to ... miss things like people marking the ballots by pencil and writing things on them. There's amazingly more of that than you would imagine," deHaven-Smith said.

Even so, he said, Washington's race is so close, a hand recount may not deliver a clear winner.

"You would probably never get the same count if you did them by hand," he said. "The reality is you'd look at it one time and see something and the next time you wouldn't, simply because the ballots themselves change. The chads fall out."

Reed says that given the inherent uncertainty in a close tally, he would have been comfortable stopping after the last recount. "I'm not sure what the value is in doing another recount."

That said, the candidates have the right under state law to call for a second recount, he said.

He just hopes it ends there. "By having another recount, they will have exhausted the alternatives under Washington state law. To continue to pursue this, trying to get court rulings to give them another shot and delay the election longer, is not going to accomplish anything positive for the state of Washington."