Nashville's Voting Booth Blues

After the recount debacle in Florida eight years ago, one would hope that election irregularities would be a thing of the past. With the media circus that ensued, it seemed reasonable to expect that election officials would keep a close on future elections out of the fear of potential national embarrassment, if for no other reason. Unfortunately, elections have become even more susceptible to corruption as the democratic process goes digital. While old-fashioned voter fraud still persists, newer, even more dangerous forms of election tampering have emerged across the nation.

Back in December, thieves broke into the Davidson County Election Commission office in Nashville, reports the Tennessean. The computers taken held not only the Social Security numbers of Nashville voters, but also held other personal information. In order to protect the identities of the voters, the metropolitan government offered one free year of credit monitoring to all 337,000 Nashville voters at an estimated cost of $800,000. While Metro has insurance to cover most of that cost, it has also launched a lawsuit against the security firm it retained to secure the election commission office, arguing that a lapse in security led to the theft of the computers.

Of course, plain theft of data, while dangerous, pales in comparison to the wholesale hacking of elections. A study done by the Center for Information Technology Policy at Princeton University in 2006 revealed a number of security flaws in the most common models of electronic voting machines, the Diebold AccuVote-TS and the Diebold AccuVote-TSx. According to the authors of the Princeton study, professor of computer science Dr. Edward Felton and graduate students Joseph Calandrino and Alex Halderman, these machines were scheduled to record the votes of up to 10% of all registered voters nationwide. Dr. Felton and his colleagues demonstrated the ease with which someone might hack a machine: someone armed with nothing but a screwdriver, a USB flash drive, and physical access to the voting machine could in less than a minute upload malicious software to alter vote totals or otherwise corrupt the democratic process. Diebold, which has since changed its name to Premier, vehemently denied the findings of the Princeton group, claiming that its machines used the latest physical and software security protocols.

The recent New Hampshire primary, though, contained a number of rather suspicious discrepancies in voting patterns between jurisdictions that used Diebold machines and those that used hand-counted paper ballots. The website checkthevotes.com performed a thorough examination of this phenomenon. The study assumed that a candidate would take a roughly equal percentage of the vote from both hand-counted and machine-counted polling stations. For example, it was assumed that were a candidate to take 30% of all hand-counted votes, then that candidate should have also taken 30% of all machine-counted votes.

This was not the case, however. For example, Mitt Romney took 25.5% of the vote across all hand-counted ballots, but took roughly 33% of all machine-counted ballots. In the Democratic primary, Hillary Clinton took 34.7% of all hand-counted ballots, but took 40.1% of all machine-counted ballots. The biggest “losers” in the primaries—those candidates who took a significantly higher percentage of hand-counted ballots than the percentage they took of machine-counted ballots—were Barack Obama, Ron Paul, Mike Huckabee, and John McCain. Of course, there are any number of explanations for these surprising results other than election fraud. Farhad Manjoo of Salon.com notes that urban polling stations, which predominantly count their ballots by machine, tended to have voters that favored Hillary Clinton, for example.

Even a raw probabilistic analysis does little to suggest foul play. Assuming that Hillary was expected to take the same proportion of hand-counted and machine-counted votes, the Markov inequality puts an upper bound on the probability that she would take at least the actual 40.1% of the machine-counted votes at 0.865, with probability 0 corresponding to an event that never happens and probability 1 corresponding to a certain event. This upper bound of a 0.865 probability essentially states that the probability that Hillary would have equaled or bettered her 40.1% of machine-counted ballots in a repeated New Hampshire primary is less than slightly less than certain.

What does all of that math mean?  It means that I’m a huge math geek. It also means that foul play with voting machines was probably not a factor in Hillary’s surprising win in New Hampshire. However, something else worrisome happened in the New Hampshire primary: many machine-cast ballots were not hand-counted at all. Dennis Kucinich requested a recount to verify the totals, and is right to do so. Voting without any paper trail is like depositing money at an ATM without getting a receipt, as several voting-rights advocates have argued.

As Manjoo writes, “[a]fter every election, officials should randomly count some number of ballots to double-check the machines' results. It is amazing that this is not a standard procedure across the country; it is a disgrace that election officials aren't rushing to implement such procedures now.” While wide scale election fraud has not occurred yet, election officials would do well to eliminate the pending threat of electronic vote tampering by manually verifying vote totals.