Friday, November 8, 2013

Why I Hate Electronic Voting Machines


Today’s battle over the “missing” Fairfax County 8th Congressional District absentee votes has given me flashbacks.  

In 2009, I was chairman of the Fairfax County Democratic Committee.  Gerry Connolly was elected to Congress in 2008.  In January, 2009, Sharon Bulova was elected Chairman of the Fairfax County Board of Supervisors in a special election. 

Upon her resignation, a special election was held for Braddock District Supervisor.  John Cook and Illryong Moon faced off on who would represent Braddock District's 100,000 citizens.   As we waited for the votes to come in, everyone was waiting for the votes to come in for the largest (and Democratic leaning) precinct in the district – Fairview – while margin between the candidates stood under at 70 votes.  If Fairview Precinct gave Mr. Moon the same margin it gave President Obama in 2008, it would change the result of the election.

While I was waiting at the victory party, I received a phone call from one of my former Mt. Vernon District Precinct Captains, Michael Gropman, who had moved to Fairview and was inside the polling place.

He told me there were two touch screen voting machines (technically called "Direct Recording Electronic Machines" or "DRE’s") in the precinct and that the “master” machine that communicates wirelessly with all machines in the precinct and tabulates the entire precinct's vote was spitting out tapes that were causing the precinct vote totals to be nearly double the total number of voters that had passed through the precinct that day.

The machines were examined.  After it was determined they could not produce accurate results, they were turned off, sealed and taken to the Fairfax County Government Center.  Mike met me at the victory party.  We went into a booth and I took a recorded statement from him about what he had observed.  I was concerned litigation was coming. 

We arrived the following day for the Canvass.
  
First, we engaged in a debate about whether to turn the machine on.  No one knew whether turning the machine on would destroy the forensic value of the data on the machine.  After communicating with a company that services the machines, The Board learned that turning on the machines would cause the machines to upload data files and rewrite them to add new information from being turned on.

The Electoral Board then decided that it was necessary to turn the machines on to do anything, time was of the essence and the Board would save the thumb drive back ups as the forensic copies.

The County Registrar's employees then turned the machines on and asked them to print their result tapes.  One of the failsafes, the "predictive counter," was off and the “master” machine printed out a tape that again suggested hundreds of "extra" people had voted on top of the 689 people who had been manually checked in by officials at the front table.
   
The Board attempted to print the results separately from each machine.  The tape from the “non-master” machine it suggested about half of the 689 votes had been cast on that one machine.  That seemed right. 

Then, we asked the “master” machine to print a separate tape and it produced a tape suggested 600+/- votes had been cast on it alone.  Adding the two tapes together produced a result that was even different than the “master” tape. 

No mathematical analysis could produce any logic behind the machine’s “master” results – e.g. it was not doubling votes, adding tapes together, etc.  The "master" machine's individual tape and "master tape" results were completely illogical. 

Next, the Electoral Board turned to the Code of Virginia to determine what to do next.  The Code said the Board should follow the machine’s manual.  After I cross examined the Registrar's Office employees, it was conceded that the manual did not say what to do if the machine was doing bad math (clearly because the machine should not have been programmed to incorrectly add results). 

The Code then said to follow the manufacturer’s instructions.  We were advised that the manufacturer had gone bankrupt and was no longer in business.  Calling the manufacturer was a dead end.

The County had a contract with a company to service these machines, but they were different from the manufacturer.  They did not program the original source code.  The Board decided that they were not referenced in the Code.

The Board then elected to print what are called the “ballot images” – a listing of each voter’s individual vote.  Apparently, this had never been done before in Virginia because a machine had never malfunctioned like this.  

The ballot "images" were printed out.   Each tape was probably 100 yards long. 

The Fairfax County Registrar read off all 600+ votes one at a time while each Electoral Board member, campaign representatives, and party volunteers each made marks on paper – “Moon…Cook….Moon….Cook, etc.”  After an hour of counting, everyone added up their chalk marks and the results matched the manual count from the front table exactly to the vote.  

The Electoral Board decided that must have been the vote for that precinct.

Adding that up to the totals, Mr. Moon had lost the election by 89 votes out of 13,133 that were cast.

Mr. Moon thought about requesting a recount.  Given that the Board had to invent procedures to deal with the situation, there might have been cause to raise issues, but given that that manual count matched the number of voters who were manually counted, he decided that a court would probably find that while the results suspicious, they were likely accurate and that he should not put taxpayers through that expense.

As of today, no one has ever explained why the voting machine produced that result.  The algorythms in these machines are very simple totalization software.  The odds of a software bug are very low.  I asked the State Board of Elections to do a forensic analysis of the disputed machines.  No one followed up.

Every since that day, I have insisted on a paper ballot if one is available.  At least we can look at the paper and, as Sarah Conner warned us in The Terminator, we don't have to trust the machines. 

1 comment:

  1. Scott, good article. FYI, I did a forensic analysis of the memory cards from the failed machine - I understand how the incorrect calculations totaled up, but still can't figure out how the incorrect numbers got in there. No one knows - and these machines are still used throughout much of Virginia. Thankfully, Fairfax has largely phased them out, but a significant fraction of Virginia's votes are still totaled on these machines that have an undiagnosed failure - and no reliable way to recover.

    ReplyDelete