The Voting-Machine Problem

It was fitting that when Y2K disaster finally struck, it did so in a way nobody had foreseen—and in a way nobody even recognized as Y2K disaster. But that was what the electoral mess in Florida really was. It was a breakdown in the ability of obsolete but essential computerized systems to accurately meet the needs of the society, and though it didn’t specifically involve trouble handling the number 2000, in every other essential it was exactly what people had been predicting a year before. It arose from inherent flaws in a 1960s-era computer system’s way of handling information with punch cards and from the fact that that computer system was unnecessarily still in use decades later; it was triggered by the system’s inability to cope with a specific situation arising in the year 2000; and it generated a level of anarchy that government itself seemed at a loss for a while to contain.

The only difference was that it involved not the two-digit dates originally designed to fit on punch cards but the punch cards themselves. And for all that has been said about how we avoided Y2K crises, how we didn’t avoid this one is just as illuminating.

The use of modern voting machines began in 1892, in Lockport, New York. Secret election was still a fairly new concept, and only in the previous half-century had paper ballots come to prevail nationally over “open voting,” in which the voter simply said out loud whom he preferred. As paper ballots proliferated, their counting grew more cumbersome, and Jacob H. Myers, of Rochester, New York, was the first inventor to come up with an effective replacement that could both mechanize tabulation and impede the fraud that paper ballots invited. He was a maker of safes, and his initial voting machine was a kind of a huge walk-in safe, 10 feet square, with two doors. You went in one door, locked it behind you, punched keys to pick your candidate, and exited by the other door. The keys you punched were mechanically linked to counters, so at the end of the day polling officials had merely to read out the results.

The machines were terrifically expensive—each one cost $600, the equivalent of about $12,000 today—but they caught on nonetheless. By 1920 most of upstate New York was using them, and by 1960 most of the United States was recording votes mechanically. Myers’s invention survives today, in more compact form, in the mechanicallever machines still used by about 20 percent of the electorate. Nobody has manufactured them, though, or even manufactured replacement parts for them, for decades; most of the tens of thousands of them in use are between a quarter- and a half-century old.

The next important generation of voting machine was the punchcard device, the machine that caused the trouble in Florida. It didn’t come along until 1964, more than 70 years after Myers’s invention and very late in the era of the punch card. Punch cards had been around since the beginning of the nineteenth century, when a French inventor named Joseph-Marie Jacquard used cards with holes in them to control the way a loom wove patterns in fabric, an innovation that swept the textile industry. Within the century, an American named Herman Hollerith developed punchcard technology to conduct the 1890 census, the first use of the cards for information processing. He was so successful that in 1896 he was able to found the World Tabulating Corporation; by 1924, after several mergers and acquisitions, it became known as IBM.

By then, railroads and insurance companies had taken to using the cards and the machines that ran them for their record keeping; by the 1940s, government offices and libraries were commonly employing them too; and when computers became popular for business use in the 1950s, punch cards became ubiquitous, arriving in the mail with almost every bill. A political science professor at Berkeley named Joseph Harris had the idea of adapting them for voting.

He was inspired by IBM’s introduction of a device called the Port-o-Punch, a penlike tool for poking out the holes in a card, and his goal was not improved accuracy in vote counting but speed and economy. He reasoned that punch-card ballots would require relatively little investment in machinery yet could be almost instantly tallied. To develop the technology, he turned to a Berkeley engineering professor named William Rouverol.

The assemblage that Rouverol devised has not changed, and it is used not only in Florida but in 37 percent of all counties in the country, more than any other technology for voting. (Myers-type machines and newer optical-scan and computer technologies both cover about 20 percent of voters.) The voter receives a punch card and places it in a stand together with a booklet that indicates which holes represent which candidate; he or she uses the Port-o-Punch to poke out the appropriate holes and then drops the card into a bin.

The concept provided obvious benefits. When introduced, it cost about $800 per voting station and weighed about 20 pounds, versus $5,000 and nearly half a ton for a traditional voting machine. Localities that still used paper ballots could suddenly afford machines. The cards provided a clear and simple paper trail. The readers that came with the machines could count votes in minutes, whereas paper ballots would have taken hours or days. Furthermore, since election officials could afford to buy many of the machines, lines at the polls got shorter.

When the Votomatic, as it was called, entered use in California in 1963, Gov. Edmund G. Brown proclaimed it would “revolutionize the system of voting.” Curiously, punch cards were playing a part in a very different kind of revolution in California at the same moment. While the 1964 election was going on, the radical Free Speech movement was being born, at Harris and Rouverol’s own university, Berkeley. The movement’s leaders chose the punch card as their emblem for the depersonalization and enforced conformity they felt they were fighting. A student pinned a sign to his chest that said: I AM A UC STUDENT. PLEASE DON’T BEND, FOLD, SPINDLE, OR MUTILATE ME. Others began wearing cards as nametags, with holes punched out to form the letters FSM or the word STRIKE. They began burning the punch cards with which they had registered for school along with their draft cards. By 1970, the symbolism was so pervasive that Earth Day was popularly represented by a poster that showed the planet over the words DO NOT FOLD, SPINDLE, OR MUTILATE .

Meanwhile Rouverol’s invention caught on. And it revealed its shortcomings. To begin with, it sacrificed accuracy for speed. As Rebecca Mercuri, an expert on electronic vote counting, put it, “With any marginal card, the card reader says, ‘I’m going to throw it out.’” When cards were run through the machines a second time, in a recount, incompletely detached chads would fall out, so usually every candidate would pick up votes. In 1968 in Detroit, a rainstorm drenched a load of ballots on their way to be counted. In 1989 the Virginia gubernatorial election couldn’t be finalized until mid-December as recounting went on amid the same disagreements about chads heard last year in Florida. In 1996, Massachusetts held a recount that continued for months and included ballots dimpled by dull styluses. By 2000, virtually everything complained of in Florida had happened before, and one state, Iowa, had banned the machines as far back as 1984.

Rouverol, who is 83, testified at hearings in Florida last November, and true to his calling as an inventor (he holds more than 200 patents, mostly having to do with gearing), he blames all the problems on the expiration of his patent and his own control over the invention, in 1982. After that, he said, “there came on the market machines that weren’t quite as good. … Most of what’s gone on and what has created this messy situation developed a number of years after.” Specifically, during voting the punch card sits against a rubber backing, and if the backing isn’t maintained or is replaced with a cheaper material, the stylus will have a harder time making a complete hole.

Needless to say, the manufacturers of the later versions of the machines disagree completely with Rouverol’s diagnosis. In any event, it is a delicate technology. In Palm Beach County, after every election three workers empty every one of 5,000 machines of chads, apply a protective silicone spray to the rubber backings, check for problems, and then put the machines in storage in climate-controlled rooms. Before the next election, they test every machine with copies of the actual ballots that will be used.

Why such a temperamental old technology for such a momentous job as electing a President? For two simple reasons: because it’s affordable and because it’s not clearly worse than anything else available. In New York City, where the earlier mechanical machines still serve, 70 repairmen roam the city on election day from 5:00 A.M. until midnight, trying to fix breakdowns. They never catch them all. The two newer technologies that have been establishing themselves in the last couple of decades—optical-scan devices, where the voter marks a ballot with a special pen whose writing can be read by a computer, and ATM-style direct-to-computer voting—have problems too.

In 1998, Hawaii tried a new optical system and found voters using their own pens and pencils instead of the ones supplied. The counting machines got clogged with ink and lead and couldn’t read the ballots. Arizona tried online voting for its Democratic primary last year and ran into various technical problems. The chief danger in computerized voting has been well summed up by Jim Adler, founder of votehere.net, which runs online elections: “Keeping the ballots secret and at. the same time making sure that the ballots can be audited is a daunting task.” Finally, a punch-card machine can be replaced for about $200, whereas a touch-screen computer is likely to cost $3,500 and an optical system $6,000. That’s an insurmountable difference for many hard-strapped municipalities contemplating machines to use maybe twice a year.

The bottom line, in the words of Deborah Phillips, chairman of the Voting Integrity Project, in Arlington, Virginia, is that “in voting technology, there are no quick fixes or magic bullets.” It may be hard to accept the idea that no matter how we vote there will always be error, but it’s also hard to accept the idea that as the statisticians kept telling us, Florida in 2000 was a statistical dead heat anyway, so close that the margin of victory or loss lay within the margin of error of any method of counting. There was no way any technology could produce an unchallengeable winner. The fluke was not the trouble with the technology but the closeness of the race. It multiplied the importance of even the minutest inaccuracy.

No vote-counting method can ever be perfect, because no technology can ever be perfect—and that’s because technology is but the continuation of human activity by other means. William Rouverol came up with an especially elegant solution in 1963. It was cheap, efficient, portable, and fast, and it solved far more problems than it introduced. But like any technology, it had its flaws: its tendency to clog with chad if not repeatedly emptied and to partially punch ballots if not perfectly maintained. That hardly makes it an amateurish invention, the work of some local Andy Hardy thinking, “Let’s go out to the barn and make a voting machine.”

Although, in this case, it just happens that the inventor actually is Andy Hardy. When Rouverol was a boy, his mother wrote a play titled Skidding , about a Judge Hardy and his family. She modeled the judge’s son, Andy, after her own son, and when the play was turned into the 1937 film A Family Affair, the part was played, for the first of many times, by Mickey Rooney.

William Rouverol, the all-American boy, is no less the all-American inventor. Asked by a reporter last November what he was working on now, his answer was curt and simple: “I’ll tell you when it’s patented.”