Archive for Thursday, November 30, 2000

Experts calling Florida ‘a draw’

Statisticians say Bush’s 537 vote ‘lead’ truly renders race too close to call

November 30, 2000


And the winner in Florida is ... we'll never really know.

"For all intents and purposes, this puppy's a draw," said Contra Costa County Clerk Steve Weir.

Even if vote counters made a mistake just once every 10,000 ballots an accuracy rate of 99.99 percent their occasional goofs would still add up to more than Bush's current margin.

That edge of 537 votes is less than 1/100th of 1 percent of Florida's total vote for president, according to the results certified by Florida Secretary of State Katherine Harris.

And given the vagaries of an imperfect world, experts say that almost nothing has an accuracy rate of better than 99.99 percent. The U.S. Census is not that exact, not even paternity tests are required to be that precise.

The gap between Bush and Gore is so thin that even random chance should have produced a bigger difference.

If every Floridian who voted for one of the two main candidates had made the decision by simply flipping a coinheads for Bush, tails for Gorethere is an 82 percent chance that the margin would have been greater than the current 537 votes, according to calculations UC Berkeley statistics professor Philip Stark performed for Knight Ridder Newspapers.

Even the manufacturers of current vote-counting machines admit their creations are not good enough to measure such a close election.

"The vendors themselves have been saying they're 99.9 percent accurate," said Kimball Brace, president of Election Data Services, which helps advise local governments on voting equipment purchases.

One manufacturer told the New York Times his machines' accuracy rate can be as high as 99.99 percent still too vague to peg a true winner in Florida.

"However it comes out, it's sort of completely accidental and dependent on happenstance," said Don Ylvisaker, a professor of statistics at UCLA who often works with large amounts of data. "Six million (votes) is just a ton of stuff. There are obviously going to be errors in there of one sort or another."

Numbers, say the folks who work with them, are almost never perfect.

The U.S. Census, which determines the distribution of billions of dollars in federal money and the allotment of House seats among states, is thought to be about 98 percent accurate.

California law generally allows paternity to be established with DNA testing that shows a 99 percent probability.

"In general people talk about data having about 5 percent errors in it," said Jessica Utts, a statistician at UC-Davis.

Even if voter errors like "dimpled chads" and twice-punched cards are ignored, mistakes in the tabulation of that many votes are almost unavoidable, experts say.

Clerks might transpose a set of numbers as they read a counting machine. Two punch-cards might stick together as they slide through the gadget. A lone ballot might be dropped, no matter how honest the workers.

"There is no perfectly accurate system out there," Brace said. "Every system has its pluses and every system has its minuses."

About half the nation's voters pull levers or punch computer cards, a system that has caused so many problems in Florida this month.

Another quarter of the voters use pens or pencils to mark ballots that are scanned by optical equipment. Considered slightly more accurate, even they have potential for error. Improperly aligned cards or marks that are too light can hide a voter's pick.

Beth Miller, a spokeswoman for California Secretary of State Bill Jones, said people who tally the votes know there is room to improve.

"There is a general consensus among election officials that it's well past the time that we need to update the system," Miller said.

Many local election officials have their eyes on touch-screen computer technology. It takes the subjectivity out of the vote-counting process, said Weir.

"There's no issue about voter intent, there's no issue about dimples," he said.

But Brace said that even touch-screen technology has yet to be proven more accurate, partly because of the risk of programming error.

"You could make the case that you have less human interference to cause problems like hanging chads," he said. "You could make the claim that electronics would be a cleaner way. But we don't have any instances of testing that out, of making sure that the votes were accurately cast for the candidates of the voter's choice."

Types of voting systems

Punch cards (used by 34 percent of U.S. voters)

Voter punches hole in computer card

Pro: Low cost. Paper ballot available for recount.

Con: Hard for voter to confirm correct hole is punched. Punched "chad" may not come out cleanly and can block original hole. Rubber strips that grab punched chad can fail.

Optical scan (27 percent)

Voter places mark that is read by computer.

Pro: Low cost. Paper ballot available for recount. Option of in-precinct scanners that can announce improperly marked ballots.

Con: High printing costs for ballots which much be very precise. Lightly marked ballots can be missed by machine count.

Paper ballots (2 percent)

Voter marks hand-counted ballots.

Pro: Low cost for small communities. Paper ballot available for recount.

Con: Takes a long time to count.

Lever system (19 percent)

Voter pulls lever connected to counter.

Pro: Can prevent voting for two candidates in same race.

Con: No paper ballot for recount. Machines are no longer manufactured. Repair parts are hard to find. Limited supply can mean long lines. Counters have been known to malfunction.

Touch-screen (9 percent)

Voter touches computer screen

Pro: Can prevent improperly marked ballots. Eliminates cost of paper ballots.

Con: No paper ballot for recount. Expensive.

Source: Knight Ridder Newspapers interviews, Election Data Services.

Note: 9 percent of U.S. voters live in counties that use a mix of systems.

Commenting has been disabled for this item.