Electronic voting machines redeux
It is important to make every effort to assure that the electoral process is above suspicion, and try to at least try to reach the 1 to 1.5 percent error figure believed to be inherent in even the best voting system. Without moving in that direction people will lose confidence in the system. Losing confidence in the system could be a precursor to failure of the system.
In looking at voting machines in the January 2007 issue of ÆGIS, we came to the inescapable conclusion that error rates with punch cards, lever machines, and DRE (direct recording electronic) devices are so significantly higher than paper ballots that their use should be rejected out of hand. At that time we felt that the best choice would be tallying well designed paper ballots using optical scanners.
This conclusion was based on two assumptions that appeared to be reasonable, yet have turned out not always to be true. That being the case, we felt that the integrity of the electoral process is so important that this merited another examination.
The first assumption was that optical scanners were simple, and could reliably validate ballots, spiting back bad ballots uniformly across machines, thus allowing the bad ballots to be replaced if needed. It appears that in practice this is not always true. In the Florida presidential election, for example, the machines tended to work properly in white precincts, so that relatively few spoiled ballots that went uncounted. In non-white precincts, however, this feature did not work quite so well, so that many spoiled votes ended up going uncounted. This is bad.
The second assumption was that the tracking of the machine-read votes would be more reliable than other computerized or mechanical systems. While optical scans of paper ballots are far better than anything other than hand-counted ballots, scanners are still computers. Even for something as simple as counting ballots, computerized optical scanners, like other computers, are susceptible to error and tampering. This, too, is bad. It is important to remember that even if the scan is accurate, and even if there is no voting fraud, bad ballot design can cause serious problems. It is estimated that faulty equipment and confusing ballots cost us 1.5 million to 2 million lost votes.
There are two obvious ways to deal with the problems inherent in using optical scanners to tally paper ballots. The first way is precinct-level hand counting. While there are more precincts in New York or California than in rural areas where hand counting is more prevalent, there are also more hands available to do the work. The advantages of hand counting are:
(1) Counting of ballots can be done publicly, observed and recorded by the press and everyday citizens who are registered voters in the precinct where the counting takes place.
(2) Although paper ballots, as with computers, lend themselves to fraud and tampering, security safeguards are much more easily put in place to protect against tampering.
(3) The cost of hand-counted paper ballots is far less than buying machines. While hand counting is slower than machine counting, if it prevents recounts it can provide a good alternative to optical scanning.
The second way of dealing with the weaknesses of optical scanners is to recognize the potential problems in their use, and implement measures to overcome these potential problems. In both hand counting and optical scanning you will get better results with well-designed ballots than with confusing, poorly designed ballots.
How do we try to overcome these potential problems? First, ballots should be designed by competent designers, rather than ward heelers, so that what is read reflects the intent of the voter.
Second, optical scanners should be chosen by a competent technician with the goal of picking one that will work, rather than one accompanied by large political contributions.
Third, we need to implement procedures to deal with less-than-perfect marking. As an example, suppose you make a mark, change your mind, put a big X through the bad mark, and then mark the candidate you really want, drawing an arrow to it. A dumb machine will bounce this as a bad vote, while a smarter human will immediately interpret the intent of the voter. So, ALL questionable votes need to be examined by a person. Will there still be bad ballots with lost votes? Sure, even with well designed ballots there will be spoiled votes, but many fewer of them.
Fourth, we need to deal with the reality that the machines can go down. In this case the ballot can simply be accepted for subsequent counting. After all, it is the ballot which is the vote, not the calculated total by the machine. Finally we need to deal with the problem of machine error and tampering. This can be dealt with by random audit. We are assured that a statistician can tell us how this auditing needs to be structured in order to detect and obviate error and tampering.
Other issues
Besides voting machine issues, here are three other sources of lost votes that must also be addressed. The least significant of these (though we have no clue as to the scale of this problem), are losses associated with handling of absentee ballots.
More significant are losses associated with polling place operations. We have seen cases in which someone can’t find a key to open the polling place, or unlock a piece of equipment. As well as cases in which voting has been delayed by hours in precincts when machines went down, or provisional ballots were not available. And cases in which already registered voters’ names were missing from the rolls. Whether deliberate or accidental, it is estimated that at least a million votes are lost due to polling place operations. The most significant problem is the registration process. It is estimated that between 1.5 million and 3 million votes are lost because of registration issues. In some cases this is because of unintentional error, and in other cases it is deliberate. We know one voting activist in New York State who was threatened with arrest during a registration drive. Fortunately, she was accompanied by an attorney and a journalist, thus easing the problem (and making the front page of the local paper, which hopefully helped further diminish the problem). While voter intimidation, exclusion, and fraud might have been acceptable in the past, they shouldn’t be acceptable today.
How meaningful are these problems in real-world terms? Well, in the 2004 presidential election 121,480,019 votes were collectively registered as having been cast for Mr. Bush, Mr. Kerry, and Mr. Nader. With 62,040,606 votes, Mr. Bush had a lead of 3,012,497 over Mr. Kerry’s 59,028,109 votes.
Imagine that in this election there were no registration, polling place, absentee ballot, or voting fraud losses. Instead, we simply had to take cognizance of the best-case system error of 1.5 percent in an entirely clean election. To have won the popular vote in the last presidential election on the basis of actual votes cast (rather than system error), one would have had to have more than half the votes plus 1.5 percent. Put another way, to win the 2004 presidential popular vote by actual rather, rather than system error, you would need to have more than 62,350,388 votes. Which means that with 62,040,606 votes, Mr. Bush’s margin fell within the 1.5 percent system error. If we throw in the more realistic 6 million lost votes, it is impossible to say which candidate received more cast votes.
Do we think a lot of voting fraud is taking place? Well we spend much of our lives dealing with fraudsters in the world of business. When you consider that the line between politics and business has blurred to the point of invisibility, we would have to be blind, deaf, and dumb – combined with just having fallen off the turnip truck – to assume that the world of politics, money, and power has less fraud than the world of money and power. Since votes these days tend to be so close, a little tampering goes a long way toward fixing an election.
