The great American e-voting experiment

[RedCell]

Democratic elections are supposed to be decided by the will of the people. That principle was called into question by the 2000 election for the president of the United States of America, which was famously determined by just 537 votes in Florida and one Supreme Court decision. In November, it may be under scrutiny again, as another close presidential election could be decided by the accuracy of a raft of new voting technologies.

Across the US on 2 November, while some voters register their intent using traditional paper ballots, punched cards and lever machines, others will be using less tried-and-tested systems such as optical scanners and electronic touch-screen voting machines. These new systems are supposed to count votes more accurately.

But concerns are being voiced that electronic technologies can just as easily mean that more votes will go missing or be miscounted. They might even be used to commit election fraud. And unlike conventional voting systems, many electronic systems leave no paper trail to allow results to be double-checked.

The unprecedented use of these novel technologies will make the coming election a huge experiment in electronic voting. “If the presidential election is decided by electronic voting in some swing state, one could imagine a bitter fight with no way to resolve it,” says David Dill, a computer scientist at Stanford University in California. “We could be entering uncharted waters with this election. People will not take a funny electronic result lying down: there will be challenges.”


Pregnant chads


In principle, of course, elections are decided by counting up the votes electors have cast. In practice, it has never quite been that simple. Voters can never know for sure that the vote they cast will actually be counted in favour of their chosen candidate. Most notoriously, in 2000 in Florida, 40-year-old voting machines produced punched cards with “hanging”, “dimpled” and “pregnant” chads that ultimately determined who became president.

That, along with poorly designed ballots across the state, led to calls to deploy new voting technologies throughout the US, a move enshrined in the 2002 Help America Vote Act. Nearly $4 billion of funding has been allocated to pay for thousands of new machines to be rolled out across the country. Around 1 in 5 counties will be embracing a new voting technology next month, many of them for the first time.

The accuracy of a voting system is often assessed by what is called the “residual vote”. This is the difference between the number of voters who turn up at polling stations and the total number of votes allocated to the candidates. Voters can still choose to spoil their ballots, but if one system regularly produces a higher residual vote than another its accuracy may be questioned.

According to this criterion, the most accurate way to record votes is to use optical scanning machines. These work in a similar way to photocopiers, and register a voter’s pencil mark on the ballot by the amount of light it absorbs. These systems produced a low average residual vote of around 2.1% during presidential elections from 1988 to 2000, according to a study to appear in the Journal of Politics by Stephen Ansolabehere and Charles Stewart of the Caltech-Massachusetts Institute of Technology Voting Technology Project.


Annoying interface


Touch-screen voting machines have a far higher residual vote of 3.0%. These machines register a residual vote when a voter activates the machine but then fails to cast a vote. Experts attribute the high residual vote on these machines to their sometimes confusing or annoying interface, which require voters to navigate a menu and touch the screen to register their vote for their preferred candidate. Punched cards have a residual vote of 2.9%.

The residual vote is not the only measure of the success of a voting technology. Of equal concern is whether voting machines ever allocate votes to the wrong candidate or can facilitate an election fraud. This is where many new voting technologies have attracted most criticism.

Four independent studies in the past 18 months have identified problems with voting machines that could lead to vote tallies being mistakenly altered or deliberately tampered with (see New Scientist print edition, 14 February). The flaws affect both the hardware and software of machines made by two companies: Diebold Election Systems of McKinney, Texas, and Sequoia Voting Systems of Oakland, California. Concerns have also been raised about machines made by other manufacturers.

“When the margins in an election are wide, these problems are inconsequential, but when the margins are narrow, as in Florida in 2000, these problems dominate the news,” says Douglas Jones, a computer scientist at the University of Iowa who has written numerous reports on the security of e-voting systems.


Inexperienced staff


Such concerns have led to calls for every vote cast electronically to be recorded on paper as well. That would allow the vote tallies of machines to be audited if the vote is close, the machines malfunction or are used incorrectly by inexperienced staff, or if an election fraud is suspected, Dill says. Indeed, he is so concerned about the problem that he has helped set up a non-profit organisation called Verified Voting that campaigns for the introduction of paper trails.

Will Doherty, president of Verified Voting, recommends that 2% to 3% of all electronic votes are validated against their paper equivalent to ensure that the vote-tallying software has not been corrupted. “Then you have a way of determining when an error happened and the opportunity to correct it,” agrees Charlie Strauss, a computer scientist at Los Alamos National Laboratory in New Mexico.

Optical scanning machines and punched card machines allow this to be done, as they retain the marked paper ballot. But lever machines, where voters pull a lever that corresponds to their preferred candidate’s name, do not. Nor do most of the current touch-screen electronic voting machines designed by Sequoia, Diebold and Election Systems & Software of Omaha, Nebraska. Yet, together, these machines will be used in more than 1 in 4 counties across the US.

In Florida, for instance, paperless machines will be used in 15 of the state’s 67 counties. Congressman Robert Wexler has already filed a lawsuit arguing that this in unconstitutional, as such machines rule out the possibility of the vote being recounted by hand.


Shaken confidence


And mistakes do happen. In 2002, touch-screen machines made by Sequoia Voting Systems were implicated in a fiasco in a local election in Bernalillo county, New Mexico. The system registered only 36,000 votes out of the 48,000 that had been cast. It turned out that the error occurred after votes were downloaded from individual machine’s memory cards to a central tabulator: a software bug told the tabulator to ignore all votes cast above a certain threshold.

The inconsistency in the voting tally was only spotted shortly before the vote was certified. The bug was later found, admitted to, and corrected by Sequoia technicians who retrieved the missing votes stored on back-up memory cards in individual machines. “It shakes one’s confidence to know there was a problem this basic in the system,” says Dill.

Strauss argues that an electronic recount of this kind itself raises an ethical problem. Only company technicians, rather than election officials, are qualified to fix the software, but then any changes they make are not then re-certified. “A paper recount would have made everyone less queasy,” he says.

This should not be a problem in Nevada, where all the electronic voting machines used on 2 November will be required to produce a paper record of each vote. Voters will then be able to check this paper ballot and drop it in a ballot box. Nevada is the only state to require this, but California and Ohio, among others, will be watching the outcome closely as they intend to move to the same model for elections in 2006.


Destroying votes


But even a paper trail will not guarantee that a system is foolproof or fraud-proof. Ted Selker of MIT, who visited several precincts during state elections in 2002 to watch Nevada’s system being deployed for the first time, says he saw several things that worried him about the set-up.

They included unsecured printer cables that could have been pulled out, the use of thermal printing paper that blackens easily on a hot day, and several paper jams. “I saw three people looking at the paper trail on election day,” he says. In one case, he reports seeing an election official open a printer, take out a roll of jammed votes and cut them up with scissors.

A spokesman for Nevada’s secretary of state says he knows of no such problems, and points to an audit of 5981 votes that found a 100% correspondence between the electronic tally and the paper trail. However, counting paper ballots by hand is itself notoriously unreliable, and one expert contacted by New Scientist laughed out loud when asked if such an exact match is plausible.

Stewart also points to possible legal disputes if paper records conflict with the electronic tally. What will be the legal status of paper records? The law does not say. There could be disputes, for example, over whether any disparity is due to paper jams, miscounting of paper ballots or a software bug.


Crying foul


But most people still feel more comfortable with a paper trail. For instance, in Georgia, in the 2002 elections to elect a governor and state senators, votes were recorded on a $54 million suite of new touch-screen machines. When the election resulted in a 10% swing towards the Republicans from the latest opinion polls, the media cried foul despite their being no evidence of any fraud or software glitch.

And in California, state officials agreed this year to allow concerned voters in 10 counties using touch-screen machines the option to cast a traditional paper ballot instead.

Despite such perceptions, “the paper trail is much more prone to fraud than the electronic trail”, Selker says. He argues that the electronic result is often more likely to be reliable.

Yet public trust in electronic voting is a major issue, one that has not always been helped by the machines’ manufacturers. Diebold Election Systems allegedly told California state officials that its Accuvote-TSX machines were federally “qualified”, when they were not, and the machines were used in state elections in four counties in March 2004. Now the state attorney-general Bill Lockyer is suing Diebold for false claims about its products after California’s secretary of state Kevin Shelley decertified the machines in April.


Conflict of interest


Mistrust also stems from the unsatisfactory process by which machines are certified. Machines must first be “qualified” by an Independent Test Authority (ITA) which checks that they comply with federal standards, and then certified by the state. But the ITAs are paid by the machine vendors themselves. “We see that as a conflict of interest that works against the public,” Doherty says.

Avi Rubin, a computer scientist at Johns Hopkins University in Baltimore, Maryland, says that ensuring an election is fair is so important that security checks should be carried out on the programmers. “How do they compare to what the government requires for the people who handle nuclear weapons?” he asks. “I know the answer.” Rubin spearheaded the attack on voting software with a report in July 2003 that pointed out numerous security vulnerabilities in Diebold’s Accuvote-TS voting machine.

The outcome of America’s experiment with electronic voting will be studied by dozens of other democracies that are rethinking how to record their people’s votes. Watching will be officials from Australia, India, Venezuela, Brazil, Japan and Europe.

Ironically, all eyes may finally fall on the swing state of Ohio, where the secretary of state, Kenneth Blackwell, has delayed the introduction of Diebold electronic machines until software glitches can be definitively ruled out. So next month, the citizens of Ohio, where George Bush and John Kerry are running neck and neck, and pollsters predict the outcome could be as close as that in Florida in 2000, will be voting on punched card machines. Could history be about to repeat itself?


---------------------------------------------
http://www.newscientist.com/news/news.jsp?id=ns99996523