Man or machine? Rossi-Gregoire recount prompts dispute over accuracy
12/11/2004, 9:34 a.m. PT
By DAVID AMMONS
The Associated Press    

OLYMPIA, Wash. (AP) ? Humans versus machines? 
No, it's not an old Schwarzenegger sci-fi flick, but a high-stakes dispute in the often strange world of Washington politics.

With the governor's race hanging in the balance, an unprecedented third vote count is now underway in the state's 39 counties ? by real live human beings this time. The two previous tallies, both by machine, showed underdog Republican Dino Rossi to be the winner by the tiniest margin on record, 42 votes at last count.

Democrat Christine Gregoire's allies plunked down $730,000 to secure an unprecedented second statewide recount, by hand this time.

All of which raises the timely question: Which method is more reliable and accurate? Human or machine?

And, ultimately, who really won? Can we trust the results? As one newspaper headline put it: "Top vote-getter? We may never truly know."

The experts tend to favor the machine recount, although some key election-watchers, including the secretary of state, find compelling arguments on both sides and acknowledge that any vote-counting system has a tiny built in error rate.

That means any election is an approximation ? usually reliable enough to keep candidates and voters happy. But in a contest this ultra-close ? both got 48.87 percent ? even the method of counting has become grist for debate.

The two sides in the governor's battle-from-hell are quite opinionated. Republicans see human counters as error prone ? and possibly hanky-panky prone. Democrats believe the two previous counts have been flawed and that only human judgment can truly determine voter intent.

There is agreement, though, that the country's largest hand count is a hellacious low-tech task and that auditors are doing their darndest to get it right.

"This is something that happens only once every 40 or 50 years in this country," says Dick Smolka, editor of Election Administration Reports in Washington, D.C. "It doesn't matter how many times you count, you'd probably always get a different answer. It's very difficult to count that many votes and be definitive if they're that close. Luckily, Washington election officials are as good as you're going to find in this country."

?__

THE PARTISAN DISPUTE...

Washington moved away from hand counting paper ballots decades ago, and has used machines and computers to tabulate the votes that are cast in a variety of ways ? by punchcards, fill-in-the-oval optical scan, or ATM-style touch-screen.

The first count took several weeks because of heavy use of absentee ballots. At the end of the machine tally, Rossi was ahead by 261 votes ? enough to trigger an automatic machine recount, but not quite close enough to mandate a hand recount.

The first recount ended with an amazingly close 42-vote gap. Rossi earned the title of governor-elect-for-now.

State law allows the losing side to call for a third count, and the Democrats did just that, ponying up $730,000 to defray costs.

The Democrats have an another unprecedented move up their sleeve. They're asking the state Supreme Court on Monday to require election officials to reconsider several thousand questioned ballots that were disqualified in the earlier rounds.

If Republicans are leery of the reliability of the hand count, they're livid about what they perceive as changing the rules of the election in midcourse.

Democrats say it's all about counting every ballot, and that the hand count is the better method.

"Only through a human audit will errors ... be flushed out," says Democratic state Chairman Paul Berendt. "We're not fearful there was fraud out there, but there were errors, we believe.

"When human hands get to examine those ballots, people get to see if they should be counted. Only human eyes are able to figure out voter intent sometimes. Machines aren't capable of deciphering."

Further, the machine manufacturers acknowledge a small, but very real, error rate, Berendt says.

"Machines really have taken over the world if we think they can do better than human beings in counting ballots," he says.

Republicans counter that technology, training and consistency of rules have vastly improved in the last decade, making it a bit more reliable than a hand count, particularly with such a large number of ballots.

Former Secretary of State Ralph Munro says hand-counting, be it cars on the freeway, birds and whales, or votes, inevitably leads to mistakes.

"Take a ream of paper home, take some pages off the top and then pass it around the dinner table and have everyone take turns counting the stack," Munro says. "I'll bet you'll get four or five different answers. People lose track, pages stick together, people get bored or tired."

Smolka says people at his card game don't always agree that there are 52 cards in a particular deck.

Former Republican Gov. Dan Evans adds, "Can you imagine 300 newly hired, ill-trained, overworked people counting by hand with people looking over their shoulders and getting accurate counts? It's ludicrous to think that this hand count will be anywhere near as accurate as the machine count just concluded."

?__

THE EXPERTS ...

The state's chief elections official, Republican Secretary of State Sam Reed, says both sides make good points, and that one method isn't provably, consistently more reliable to a statistically significant degree.

"Machines have the advantage of being totally objective ? they are not Democrat or Republican machines," he says. "They don't have a bias to make sure a person's vote count."

Machines carry a small error rate, generally about a hundredth of a percentage point, Reed says. That means out of 2.8 million ballots, they would be accurate to within 280 votes, ordinarily pretty close to perfection, he says.

Hand counting also carries a small error rate, probably never calculated, he says.

"Two people can count and come up with the identical wrong answer," he says with a laugh.

"But hand counting has some advantages over a machine, such as being able to see when a punchcard chad falls back into place and a machine couldn't `see' that vote. Humans can see things that machines can't in the area of voter intent."

A hand count, double-checked against each precinct's tally from the earlier recount, should be trustworthy, Reed says.

"Human beings are undoubtedly more prone to mistakes, but a hand recount may not be that much different if both are done well," agrees elections consultant Don Whiting.

Many, but not all, county election directors tend to believe machines are a bit more accurate, says Corky Mattingly, Yakima County's Democratic auditor and head of the state auditors' association.

"We all know machines make fewer errors than humans ? and they don't need food or water, either, or a break!" she says. Machine counts became the default system for counting ballots for good reason, she said.

"Do we really need this third count? I think we have to come to the point where we accept the computers' count."

King County elections director Dean Logan says the hand count is daunting because so many people will be handling the county's 900,000 ballots, applying fallible human judgment.

Political scientists say they haven't solved the debate.

Smolka says either system should be fairly accurate, but that the consensus is "you would bet on machine over hand recount. The error rate would probably be less. But that's no guarantee. And congressional researchers say every system is capable of doing a good job if used as intended."