|
|
|
| |
|
|
Recount Updates Certified Florida results show Bush leading Gore by 300 votes. Hand recounts still possible. • Post headlines • Humorous ballots • See the PBC ballot • WPB rally photos • Post your feedback • Vote in polls • NEWS 12 video
High school football
Our Century
Weekly Sections
More Post News
Past issues
Post services
Announcements
Post departments |
|
Expert told officials: Don't use punch-card ballots
By Paul Reid, Palm Beach Post Staff Writer Under normal circumstances, computer punch-card experts are in demand nowadays about as often as locomotive and vacuum tube experts. But Matthew Bressler's phone is ringing today in -- where else -- Palm Beach County. Matthew Bressler knows punch cards and, as everybody knows, the next president of the United States likely will be the man who holds the winning hand of Palm Beach County ballot punch cards. Bressler, 70, of Jupiter, went to work for IBM in 1953 and was part of the U.S. Standards group that developed punch-card standards, as well as the American Standard Code for information interchange. He worked for RCA, Burroughs and IBM in his 40-year career and was recognized by the industry as one of its leading experts on punch-card technology. When Palm Beach County election supervisors asked his advice on punch-card ballots about 20 years ago, he readily offered it: Stop using them. "Cards are not as simple as you think," said Bressler. "They expand and contract under varying conditions of heat and humidity. Don't forget, they're made of wood pulp products -- cardboard. They're high quality wood fiber products, but still they absorb moisture. When they absorb moisture, or give it up, they expand or contract. "Our U.S. standards called for cards that expand no more than 0.050-inch in one direction and 0.025-inch in another. That's only a couple of dozen thousandths of an inch, but it's enough movement to cause a misreading of data, or votes in this case. The holes on the cards are just 0.0875 of an inch, so you can see that if the card expands, say 0.050-inch, you could misread the card." Who would have thought a cardboard card about as big as a dollar bill packed so much punch? "Here's what I suggested to election officials years ago," said Bressler. "First, I suggested they get rid of the card system. Cards are not as good as digital data processing. Period. "If they can't or won't get rid of the cards, I suggested they always make sure the cards are registered in the counting machine the same way they were registered in the voting machine. What's registered mean? Well, to punch the card properly, you have to line it up on two edges -- top and side, bottom or side, left or right, doesn't matter. What matters is you align the card the same way in the reading machine as it was in the voting machine. If not, you will get a margin of error unacceptable in rocket science and maybe in election results." How much of an error? "At least 1 percent, maybe more," Bressler said. "Not enough to make a difference in a lot of computations, and remember, back in the 1950s, cards advanced data processing a thousandfold. But if you need that last 1 percent of data, well, cards not properly registered will deny you your data." What about "chads," those little flecks of punched paper that hang on the back of the card? "Chads shouldn't be relevant if the ballots are run through the reading machines twice," says Bressler. "Obviously, that's been done now, with the recount. But we told the county years ago: Always run the cards through twice. This is critical. The first run allows the cards to reach the same ambient temperature and humidity as the air in the room. The second run is when you count the votes. You will always find a discrepancy between the first and second run. Just by coincidence, the test run should shake off the chads. So, they shouldn't be a problem." And if you don't run the cards twice? "A rocket aimed at the moon using punch cards would probably miss if the cards had been run only once," Bressler said. "Maybe an error rate of 1 to 2 percent. Cards run twice will, or should, show an accuracy rate near 99.9 percent." Bressler also recommends card users, including election officials, make up a sample batch of cards for test purposes. When the sample batch is run, the officials can compare the results with the known data on the cards. If a variation shows up, it means the cards are probably acting up. "Cards are fine unless you need greater accuracy than you usually need in an election," he said. "If the results are closer than one tenth of 1 percent, and you're using cards, you'll never know. The problem is not the confusion of box three or box four or arrows or whatever is on the ballot, it's the ballot material. It's in the cards. They're the problem." His conclusion: If you demand certainty in the information age, don't use punch cards.
paul_reid@pbpost.com
|
|||||||||||||
|
|
|
|||||||||||||
| |||
 | |||
By using Palm Beach Interactive, you accept the terms of our visitor agreement. Please read it. |
