Punch card ballots and the 1966 Georgia gubernatorial election.

During the summer of 1966, CBS News, in a joint venture with the polling firm Louis Harris & Associates and IBM, launched the most ambitious effort ever to forecast national elections on a state-by-state basis. CBS hoped to scoop the other networks on election night by being first to call the winner in every race for Governor or Senator across the country and, if the experiment was successful, to apply the methodology to the 1968 presidential elections.

In each state holding a senatorial or gubernatorial election, Harris selected a “representative” sample of voting precincts--some 5,700 nationwide. In each selected precinct, CBS hired a local resident, preferably someone known to--or even a relative of--an election official, and supplied him or her with credentials as a CBS News correspondent for one night to cover the counting of the ballots and call in the results as soon as they were announced. Operators in New York would feed their tallies into an IBM mainframe computer which was programmed to project the winner based on the returns from the sample precincts.

The recruitment operation was located in a nearly empty floor of a New York skyscraper, which CBS News had recently vacated for their new Broadcast Center on West 57th Street. As a college student majoring in political science, I jumped at an entry-level opportunity to work there, put in long hours on the job and was eventually rewarded with a supervisory position.

Telephone polls and telemarketing were not yet the blight they have become and we had little experience to draw on. Our recruitment task was complicated by Harris's use of quota, rather than probability sampling, which led him to frequently change his selections in order to adjust for some demographic factor or another. I have a vivid memory of Peter D. Hart, then the Harris liaison to the recruitment operation, coming up to me late on the Friday night before the election and telling me that we would have to find a half dozen new recruits over the weekend because “Lou feels we need more black precincts in North Carolina.”

Shortly after the project got underway, Louis Harris gave a rousing speech to the recruiters, telling them that they had to do their very best because, unlike commercial marketing research, in election forecasting everyone would know the next day if we were off by a few percentage points. As it was, the CBS effort was mostly successful. The only failure was in the Maryland Governor's race, where a little-known Republican county executive named Spiro Agnew, with the support of many liberal Democrats, had defeated George Mahoney, a perennial candidate who ran under the racially charged slogan “Your Home is Your Castle.”

But most of the media attention that year was focused on the race for Governor of Georgia, where the Democratic candidate was Lester Maddox, famous for handing out pickax handles to fight integration at his restaurant, then closing it rather than allow it to be desegregated. Maddox had defeated Jimmy Carter and former governor Ellis Arnall in a bitter primary to face Republican Howard “Bo” Calloway in the general election. The race was considered close from the start, but it became utterly unpredictable when Arnall, who detested Maddox and everything he stood for, launched an independent write-in campaign in an attempt to stop him.

Calloway’s support was strongest in the suburbs of Atlanta, particularly in DeKalb County, where a new computerized voting system from IBM was being used for the first time. The Votomatic used pre-perforated Hollerith cards placed into a flip-over template displaying the names of the candidates next to holes through which a voter would push a stylus to record his or her vote. Unlike voting machines and paper ballots, which can be read and tallied on the spot as soon as the polls close, the punch cards had to be collected from all the precincts and taken to a central mainframe computer to be tabulated.

After the polls closed in the 1966 election, DeKalb County Sheriff's deputies dutifully gathered up the ballots, sealed them in heavy burlap bags and tossed them into their pickup trucks to be taken to the county courthouse. The deputies had no doubt been thoroughly instructed not to fold, spindle or mutilate the ballots, but it seems that no one had thought to warn them about the effects of moisture on punch cards.

As it happened, there was a heavy rain in the Atlanta area that evening, and many of the bags containing the ballots were soaked through by the time they reached their destination. It would be days before the last of the punch cards was dry enough to be fed into the computer, and so no results from DeKalb County were included in the projections when CBS News called Lester Maddox the winner in Georgia that night.

Calloway did well in DeKalb County, and the heavy turnout there ended up putting him ahead of Maddox by just over 3,000 votes, out of nearly a million cast statewide. But Ellis Arnall’s write-in effort had captured over 5% of the vote, and as a result, Calloway received only 46.5% of the total vote to Maddox's 46.2%. Georgia law required an absolute majority of the votes cast to win the election, and so the matter was sent to the Georgia House of Representatives, which was overwhelmingly Democratic.

When, as expected, the state legislators declared Lester Maddox the winner some weeks later, CBS newsman Harry Reasoner quipped that the Georgia House of Representatives had once again made an honest man of Walter Cronkite.

A direct consequence of the calls in Maryland and Georgia was that CBS fired Louis Harris and established a permanent unit within the News Division to handle election forecasting, hiring a sampling expert named Warren Mitofsky from the Bureau of the Census to run it. Under his direction, the CBS Election Unit contributed many major advances in survey research over the next 30 years, including the exit poll.

The Votomatic system was reasonably successful and adopted in many jurisdictions, but after it featured in scandals over alleged vote rigging in California in 1969, IBM decided that the election business would be an endless source of problems and sold the division to a group of management employees. There have been many well-documented complaints about problems with punch card voting systems since then, but these have never had the high visibility they had in Florida in 2000 and have been largely ignored in the national press.

The fact remains that machines provide an imperfect means for counting ballots. Errors can often be reduced by using more advanced technology, but not eliminated altogether, and their nature is not always predictable. This was illustrated most recently by the Orlando Sentinel when they examined undercounts in 16 Florida counties that used optically scanned paper ballots in the 2000 election. They found many ballots that had been correctly marked by the voter, yet were not counted because the ink used did not have the precise reflectivity required by the equipment. Unlike punch card ballots, there were no questionable decisions to be made about hanging or dimpled chads, but had those ballots been counted, the outcome of the presidential election would have been different.

Problems with computerized voting systems have sometimes led to vocal criticism from within the data processing profession, not so much because of difficulties with punch cards, but because many feel that such systems can never be made secure enough. In 1980, a Votomatic programming error caused thousands of votes to be incorrectly tabulated in Orange County, CA, after which a data processing trade publication ran a contest for computer programmers soliciting undetectable ways to throw an election. They received thousands of successful entries.

We now see a rush of software and computer vendors hoping to capitalize on the sudden notoriety of the punch card ballot by proposing new (and, they hope, lucrative) high-tech alternatives, such as touch screens or online voting. But any system that does not provide both an unalterable physical record of each vote cast and the means for voters to verify that their votes were correctly recorded presents an open invitation to tampering. In many ways, the technology for manipulating electronic voting is similar to that used for computer worms or Internet hacks and without a physical trail, there can be no possibility of a recount.

Much of this was spelled out in a 1988 report from the National Bureau of Standards titled Accuracy, Integrity and Security in Computerized Vote-Tallying. The author, Roy G. Saltman, called urgently for the elimination of pre-scored punch card ballots such as those used in Florida in the 2000 election. Unfortunately, the administration of elections is usually paid for at the local level in this country, so it is nearly impossible to kill off bad technology that saves money. That could change now, but I certainly wouldn't bet on it.

(Copyright 2001 Jan Werner)

______

You may also wish to read a fine essay by Lorrie Faith Cranor of AT&T describing drawbacks of various voting systems in light of the 2000 Florida election.