The problem with the VVPAT

by Christopher Wilson
August 31st, 2006

Much hay has been made over the ESI report on Cuyahoga County and their problems in the recent primary.  While there are a number of interesting findings, I want to focus just a bit on the issue of the VVPAT in Ohio .

When the VVPAT was proposed in Ohio there was never any specificity regarding HOW the VVPAT would be constructed.  There were some general specifications, but there were no detailed requirements.  Part of this was to allow the vendors to develop their own systems and to allow various implementations of the VVPAT. 

Almost immediately the proposed solution from vendors was the thin strip of paper on a large roll (the so called “Toilet Paper VVPAT”).  This had the advantage that a large number of voters could be accommodated by a large roll along with a lengthy ballot.  Elections administrators for the most part simply accepted these solutions.  I had raised concerns immediately about such a system and did not like the thin strip on a large roll.

However, legislators were comforted now by the fact that the VVPAT was law.  Election activists were comforted as well.  As an election administrator I was not comforted.  The fact is that elections administrators have the duty to actually see that these systems work where the proverbial rubber-meets-the-road.  Activists, academics and legislators do not.   I took the new requirement as a red flag because the ‘devil is in the details’ and no one was concerned about the details.

When Diebold presented their machine to the Ohio Board of Voting Machine examiners, questions were raised about the VVPAT and its size and how it would actually be used.  Their response was that it wasn’t a requirement of theirs to state how the VVPAT would actually be used for recounts.  They had provided it as required and they had met the letter of the law.

The VVPAT had a bar code and the code could be read by any 3^rd party bar code scanner and software of any design.  Diebold stated that they did not have an automated reader for it, but one was in the works.  But counties were on their own in terms of how they would actually deal with the paper.  I found this to rather shocking that something as basic as this would be overlooked and not required of vendors.

When I attempted to find if Diebold had made public the format for the bar code I was told it wasn’t available.  Naturally, if one simply scanned the bar code something would come up in some sort of a text file format.  But exactly what the specifications were for that remained unknown.  As a result it wasn’t clear how a 3^rd party would design software to read the bar codes.  Nor could I figure out how scanning and reading the bar codes of the official ballots could be done via software that was not certified nor required to be certified, and could be created by just about anyone. 

In essence, Diebold was going to meet the letter of the law, and not one bit more.  It was frustrating to realize, over 2 years ago, that the VVPAT would be an administrative nightmare and no one seemed to care.  True enough we see that has come to realization in the Cuyahoga report by ESI.

At the Ohio OAEO meeting in Cleveland I saw that Diebold now does have a certified piece of software that reads and totals the barcodes.  They also have a not yet fully finished reel to reel bar code reader. 

The whole bar code issue gives fits to many people as well. After all, if I can hook up the reel to reel tape and read the entire roll, how do I know that that software tabulated it correctly?  Doesn’t the whole bar code scan simply push the issue back one more step?  How do we ensure or test the accuracy of the bar code scan/software?  Well, by hand counting I suppose.

The roll of paper is the WORST possible design for actual hand counting of ballots.   How do you deal with it?  Do you simply try to leave the roll of paper intact, or do you cut it where each new ballot appears?  If you cut it how do you deal with and secure all these thin curling strips of paper?  If you leave the roll intact and try to scroll it somehow, how do you know where you stopped and started?  Plus this only allows one operator to work on a roll at a time.  If the roll were even to be cut in half, you could have two people working on it at the same time.

 All of which leads us to the conclusion that the VVPAT, while addressing a real concern (the use of purely unaudited DREs) creates a whole host of administrative issues that have yet to be fully addressed.

 Keep in mind that when people criticize the use and role of the VVPAT they are not suggesting that we simply trust DREs without paper trails as some have suggested.  Nothing could be further from the truth.  But what elections administrators want is a workable solution, driven by real world professionals; not by legislators, academics or vendors.  We can all work together on this, but we must not forget that the sole responsibility to manage and deal with VVPATs rests with elections administrators.  The rest are on-lookers to some extent.