One of the special elections held yesterday in Virginia had a result so close that there are enough votes to change the outcome on one touch-screen from which the officials have not yet been able to extract the voting data. Even if election officials are eventually able to retrieve data they are confident enought to certify, the resulting election vote counts will certainly fall well within the limits for a state-financed recount. And if there is a recount, and a legal contest, it's unlikely that the loser will fail to challenge that confidence in the votes from the defective machine.
In the Commonwealth of Virginia, we have six or seven elections a year, whether we need them or not. Our election laws call for most vacant offices to be filled by special election within a month.
In the case in question, Gerry Connolly's victory in the race for Tom Davis's old seat (VA-11) left open his old job as Chairman of the Board of Supervisors for Fairfax County. In a special election held last month to fill that position, Sharon Bulova, the Democrat, won narrowly. But, since she was the Supervisor for Braddock District (VA counties are divided into "districts", not "wards", or "townships"), her victory in turn opened up her old seat for another special election, which we held yesterday. This election pitted a current schoolboard member, Ilryong Moon, as our candidate, running against the Republican, John Cook, over who would succeed Bulova as the Braddock supervisor.
We knew it was going to be close, because it is always a struggle for us to get decent turnout in these special elections, and low turnout generally helps the other side. But we didn't bargain on it being within 69 votes, or 0.58%, with 24 out of 25 precincts reporting.
By itself, that would be within recount range, a problem, but probably not of national interest. What makes this worthy of your attention is that 25th precinct that hasn't reported yet. It couldn't report because the election officials couldn't get one of the two touch screens used in that precinct yesterday to print out any results. County technical support personnel were called to the scene last night, but they couldn't get the errant machine to give up its secrets either. Right now, the canvass of votes that produces the official reslts is taking place, as usual for the day after an election, but with this unusual addition that they are also trying to get any count at all for the precinct in question, which was not able to give an unofficial count last night.
It has been reported, presumably based on numbers checked off the poll books, that this precinct voted 707 people last night. If it splits 57-43 our way, as it did in the special election for Bulova last month, that would net Moon 101 votes, and we win. In fact, Moon wins if the split is as far down as 55-45, because he just needs a 70 vote advantage form this precinct. We don't yet have results from the other, good touch-screen that did not malfunction last night, and those results, because they should match the results in the broken machine, will take away most of the guesswork. But if we never do get good data from the broken machine, its results would be a matter of inference, however reasonable, and not an actual count.
It is extremely unlikely that the results, as reported so far (Fairfax County website), plus what we get from the operable machine in that hold-out precinct, would leave either candidate ahead by more than the votes we know lie hidden in the broken machine. If, for example, the above assumptions are correct, and Moon gets 57% from this precinct, then the good machine will give him a 50 vote gain on Cook, and his 69 vote deficit will become 19. But there will still be roughly 350 votes locked up in the broken machine. At that point, the race would be a 0.16% squeaker, and thus well within the recount range, the state-pays-for-it recount range.
We can't yet say whether technicians will be able to read the hard drive on the broken machine with enough technical confidence that they, and the county, will report such results as part of an official result. But even if they do, will results obtained from a system that failed be trusted by the court of law that will almost certainly be hearing this case? This is a Diebold machine, almost needless to say. Is their software such that failure in the tabulating and reporting functions can be definitively separated from failure to register votes accurately as they are cast? Even if that's technically true, I suspect that we may soon see the first case that cracks open the black box, that forces the courts to at least consider standards for the legal system trusting or rejecting a whole class of technical assertions about election software.