Paper trails

[womzilla]

Avedon is still guest-blogging on Eschaton, which is a good thing. She deserves the audience.

One of her posts today repeats a point she's made before, which is important and worthwhile and unfortunately not quite correct:

And that's where the real problem lies, because a paper trail is meaningless if no one ever looks at it. The initial count of optical-scan ballots is done by machine, and if you fiddle the machine count - which you obviously can - so that no race is close enough to require a recount, no one will ever know. There is ample evidence that exactly this may have happened in 2004. Note that the graph I reproduced here shows that, while hand-counts produced discrepancies with exit poll results of well under a percentage point, optical scanners gave us around 5% - better than for other machine-dependent methods, but still not so good, and definitely enough to throw an election. Certainly it's all you need to cast a victory as being by a wide enough margin that no one asks for a recount.

See, if all those ballots from optical-scan machines were actually reviewed, we might find that there was no difference between the exit polls and the actual votes - that is, that the machines had been tweaked to give a false result.

But since no one ever demanded a hand-count of those ballots, it's unlikely that we'll ever know.

She's absolutely right that a paper trail is useless if it is never examined, and that relying completely on machine counts to determine if an recount is triggered is insufficient. This is a point that needs to be made again and again.

But she's wrong that the only way to guarantee a non-cooked count is hand-counting. Starting in 2000, I've seen many people suggest that all election systems be auditable and that audits be conducted at random as well as when the election is close. Hand recounts of 5% of all precincts should provide more than enough opportunities to spot cooked results, as long as the recounts are conducted randomly. In fact, I'd go so far as to suggest that in any election, any candidate on the ballot can request recounts of up to 5% of the precincts for no reason whatsoever, at no cost to the candidate. More recounts can be requested at no cost in cases where the election is close (say, the winning candidate receives less than 2% more of the vote than the next-place finisher) or when any credible evidence of miscounting can be presented, even if that evidence is purely statistical. This would be cheaper and faster than total hand recounting but would, I think, be an adequate guard against election theft by ballot miscount.

Comments

[lavendertook.livejournal.com]

I think one point she is making is that if you can fiddle with the program so that what gets printed is not what the voter actually puts in, auditing doesn't help.

I share this concern. Though I'm helping push for auditable machines over ones that leave no paper trail, I'm leaning toward the hand-count camp, myself. And I think most other nations do, too. Democracy is worth the time and expense.

[womzilla.livejournal.com]

From the comment I quoted and from her earlier posts, it's clear that she thinks that both the paper ballot and the hand-counting are absolutely necessary. I think that universal hand-counting is not necessary as a first step, if and only if there are mechanisms in place to guarantee that hand-counting happens often enough to make criminality impossible.

The mechanisms I proposed for random and easy recounts should be available regardless of the initial steps for counting. Other reforms--open-source software for the balloting machines, for instance--are also necessary.

One of the reasons I don't favor universal hand-counting is that it is expensive. American elections are very complex, and an honest vote-counting machine makes them managable.

[nancylebov.livejournal.com]

I'm suggesting again that there be paper ballots, and those ballots be publically posted with video cameras pointing at them so that anyone in the world can do a count.

I believe that there can be sufficient surveillance that it's possible to be sufficiently sure that the ballots on the wall are the same as the ballots from the voting booths.