November 19, 2000, Sunday
National Desk


New Focus on Punch-Card System

At a time when the corner store has a touch-screen machine capable of dispensing $20 bills while deducting them from your bank account, how can millions of American votes still be cast using relics from the time of Sputnik, and be counted by equipment that is almost guaranteed to get the numbers wrong?

The simple answer is money.

A combination of events -- a razor-thin election with huge stakes in a state where counties count ballots with four technologies, each with its own margin of error -- has put punch-card voting systems at the heart of historic uncertainty over the presidential election.

Yet the controversy presents no new issues about a technology that has always had serious shortcomings.

As a technological revolution, the punch-card voting system was a putsch, rapid ly supplanting rival technologies in the late 1960's, shortly after it was introduced. And state and local governments have continued to use punch-card systems even though alternatives have surpassed them in ease of use and accuracy.

Punch-card voting systems are cheap. They are fast. And, before the words ''dimpled chad'' became a part of the national lexicon, elections officials thought they were accurate enough.

A lever-operated machine, the technology most commonly used in the early 1960's and still used by counties in New York State and elsewhere, cost $5,000 then and weighed nearly half a ton. By contrast, the first punch-card system, used in 1964, cost about $800 and weighed only about 20 pounds, saving money for transportation and storage.

And the card readers that came with the voting machines could count thousands of votes in mere minutes, saving untold amounts in poll-worker overtime over the hand-counted paper ballots.

''If you looked at market penetration of the punch card, it first came into states where they paid poll workers by the hour,'' said Todd Urosevich, vice president for customer service at Election Systems and Software, which markets voting systems using punch cards, optical readers and computer touch screens. ''By the early 70's, it was being used by half of the voting jurisdictions.''

Rebecca Mercuri, the president of the consulting firm Notable Software and an authority on electronic voting tabulation, said: ''There's such pressure to get the returns quickly. You can run the punch cards through at a high rate of speed, print out the computerized report, assume it to be correct, and get the results broadcast on the 11 o'clock news.''

For that speed, the machines sacrificed accuracy. ''With any marginal card, the card reader says, 'I'm going to throw that out,' '' Ms. Mercuri said. It does not mean that a voter did not vote a certain way, just that the machine cannot be sure, whether because of chad -- as the paper punch squares are known -- or some other technical issue. ''It's a false negative and it needs to be relooked at manually,'' she said.

The fact that every run through a punch-card machine results in a different count, as incompletely detached chads fall out, is troubling, but it has seldom affected the outcome of an election. Election administrators were relieved that while counts were always wrong, they were wrong in a consistent way.

''It's almost a certainty that every recount you do that both candidates will p ick up votes,'' said Tony Sirvello, elections administrator for Harris County, Tex., which includes Houston. ''I've done 150 or 200, and I've never had a recount turn the results around.''

Newer systems, especially those using optical scanning technology, are probably more accurate, and now quite cheap. An analysis of the 1996 presidential election by The New York Times found that counties using punch cards reported hundreds of thousands more unvoted ballots -- a potential indicator of counting problems -- than those using technologies like optical scanning and lever machines.

In fact, if counties using punch cards had had the same rate of ballots on which no vote for president was officially recorded as counties using lever machines , there might have been 200,000 more ballots cast for president that year, roughly the margin of Al Gore's popular vote plurality this year.

The 14 counties in Florida that use punch cards showed especially large rates of ballots with uncounted presidential votes this year -- 3.9 percent of the ballots in those counties, compared with 1.5 percent in the counties using optical scanning. Some of that could be attributable to confusing ballot design, some to voter confusion and some to machines' counting irregularities.

Election administrators across the country have grown disgusted with the punch-card systems. ''Nobody's migrated to a new punch-card system in the last decade, '' Mr. Urosevich said.

Mr. Sirvello said: ''We've been wanting to get rid of the punch-card system because we don't want it anymore. We were happy for the first 12 or 15 years, but we've started looking at new technology.''

But the systems have been difficult to replace, and money is the reason. The touch-screen computer systems, called direct-recording electronic, which are the most modern and potentially accurate systems, cost $3,500 per unit. Replacing a punch-card unit costs only about $225.

Optical-scanning systems, which read penciled boxes on a paper ballot, are as cheap as punch-card systems and probably more accurate. But resistance to spending any money at all on a new system can be strong.

After the courts overturned a 1996 Congressional primary result in Massachusetts because of chad, Secretary of State William Galvin tried to ban the use of punch-card ballots, but was opposed by the state's older mill cities, which said they did not have the money to buy optical scanning equipment. Mr. Galvin had to persuade the Legislature to arrange a $3 million revolving loan program before the systems were replaced.

Copyright 2000 The New York Times Company
May not be reproduced or transmitted without permission.