Why Complexity? Look to the Balkans
John Sebes
Many thanks to what Dave Barry would call “alert reader Brandon F.” for posing a question that comes up a lot concerning digital voting. To paraphrase slightly: why not specify and standardize on ballot paper, ballot layout, ballot marking locations in the layout, and scanning systems to automate counting? Why is everyone making this so complex?
At the risk of making the answers more complex, I’ll set aside (for another day) some other questions that are hidden inside Brandon’s, such as: Why are elections officials so keen to reduce the number of paper ballots, in favor of electronically cast ballots? What is the problem with relying mainly on manual counting? Are these election officials’ preferences mainly to get the work done faster/better/cheaper, or is there some public good that accrues from these preferences?
Instead, for today, we’ll both look back and look ahead at elections procedures based on optical scan of manually marked ballots. Until recently, this was a fairly common elections procedure, so it’s a good question – why weren’t these standardized? Or, looking ahead, would it really be hard to standardize them?
Part of both these questions’ answers lie in the (for lack of a better word) balkanized elections landscape in the U.S. The Federal government delegates to states the responsibility for conducting Federal elections. Each state has its own separate standards for the production of paper ballots; what works in one state won’t be acceptable in some other state. Each states delegates to its counties the responsibility for conducting elections within the state’s guidelines. Each county contains multiple different jurisdictions for local officials and ballot measures, school district, harbor district, coastal commission, and more. In most densely populated areas, most precincts have their own distinct ballot “style” (list of races and measures), and some precincts even have two – and poll workers have to decide which ballot a voter is entitled to based on what side of the street they live on.
That’s what I mean by balkanized. And to take California as an example, ballots must fit on the front and back of one sheet of paper, without resorting to fine print. A ballot is literally a “tall order” that each county must fill several times for each election.
Looking back, the barrier to standardization was a commercial one. Opscan machine vendors simply had a lot of work to meet the needs of enough states to support a reasonable business, and the extent of conformance with state regulations was a competitive advantage that was valuable in direct proportion to the amount of effort, expense, and opportunity cost required to achieve and maintain state certification.
Looking ahead, such barriers can be irrelevant to a non-profit organization like OSDV working in the public good and funded in part by voters. Sorry if that sounds like a plug, but in fact part of OSDV’s work is to create the technology and processes to enable a future like the one Brandon asked about. But that work involves more than meets the eye, and for today I’ll just list some of the ingredients – most of which were not part of the “ecosystem” of opscan systems of days past, some of which are beginning to exist in useful form, and some of which will require considerable effort going forward. Here’s the partial list: a data representation standard for ballots; ballot definition tools; ballot design tools; ballot layout standards; system specifications for trustworthy computerized opscan systems; trustworthy opscan system products; means for independent assessment of trust; opscan systems with features sufficient to meet the requirements of all states; opscan systems that can be tailored to exclude features not wanted by some states; assessment of each state-defined variant system; certification in each state.
Though logistically feasible and without real technical difficulty, that’s still a load of work go create something that doesn’t exist today: a system trusted to do its job correctly without security or reliability issues, and that’s ready to head off to a wild ride in the Balkans.
Exec Summary:
NO custom paper sizes – standard 20# 8.5 x 11 inch paper
Standard font sizes – how about courier 10
Yes, use more than one sheet per ballot when needed.
DETAILS:
Of course, I think controlling access to paper ballots is only reasonable.
Paper ballots (standard 20# 8.5×11 inch) with pencil markings -standardized (courier, for example – keep it simple – not custom paper sizes, font sizes, you get the idea) location of “bullet markings”- would seem that scanners (standard 8.5×11) have little risk to damage ballots (and considering copier products, the hardware is cheap) – BUT your point is of course very important – not to allow willful damage/destruction of ballots.
The idea is that most anyone/organization -which qualifies to get access to ballots- could afford to bring computer and a scanner to their local polling place to help audit/double-check results. Keep the costs low. AND if these automation products are too expensive, just do it the old fashion manual count – especially for small voting sites.
I’m glad to read you are aware of other countries election procedures. I would also think that Carter Center and others:
http://en.wikipedia.org/wiki/Election_monitoring
might well have good ideas/experiences, especially on audit techniques.
side note: Traditionally exit polls done by professional pollsters were used to very accurately predict end results – and therefore good to spot “suspect results”. I have never heard why these pollsters lost their accuracy except that people started lying to pollsters in 2000. very suspicious to my mind – i.e. what is really going on? break down of techniques? too much cost savings? something more sinister?
Although parts of Brandon’s excellent suggestions are not immediately compatible with the practices (and in some cases regulations) of today’s county elections offices, you can bet that OSDV public demonstrations of reference systems will be as simple as that. And as they evolve to support existing practices, they will continue to be operable in as simple a manner.