Coverage of the Takoma Park election continues, with a good article in Wired. And so does the audit!
Some people who showed up on election day couldn’t be verified as registered voters. Thanks to one of the useful HAVA provisions, they got to vote provisionally, meaning their ballot was set aside in an envelope labeled with their name, and their eligibility was checked later. A number of folks did turn out to be eligible, so their ballots need to be tallied. The Scantegrity team has scanned those ballots, and has thus updated its D (shuffle) and R (results) tables which now include all of the old plus new ballots.
So it’s time to run my audit code again, which I had to update for this new development (I didn’t realize we were counting provisional ballots, too!). It looks like all of the confirmation code reveals check out.
Though the vote totals did change, the low number of provisional ballots means that no race came close to having a new winner because of provisional ballots.
once upon a time, you mentioned something about vote secrecy…
The secrecy of provisional ballots is much less than that of normal ballots, since you can obviously see which candidates gained a small number of votes, and thus you can often tell how some provisional voters voted. I hear that Josh Benaloh has proposed a slightly different approach, which I think is a very nice twist: count *all* of the ballots by default, and then exclude the provisional ballots that fail. This would mean that only the privacy of the unqualified voters would be at stake, not that of the qualified voters, and that’s much nicer. Next time, maybe!
We wait for complaints from voters who might not have been able to verify their confirmation codes. Then, once the complaint period ends, the Scantegrity team will meet again to do two last audits: the left-vs-right opening of the shuffle tables, and then the full opening of the spoiled ballots. The tallies already produced won’t change, but those meetings will confirm our confidence in those tallies.