[EM] danger of coercion (Re: First U.S. Scientific Election Audit...)

[Abd ul-Rahman Lomax]

At 08:50 AM 5/11/2007, peter barath wrote:
> >This is presumably about public ballot imaging. I'm only
> >aware of one major drawback, which is an alleged increased
> >opportunity for vote-buying or coercion. While there are
> >ways to interdict this, I would note that the danger of
> >vote-buying is overblown; both vote-buying and coercion
> >must be fairly widespread to be seriously effective,
>
>I don't think coercion must be widespread to be seriously
>effective. Even the possibility that in some time someone
>can identify my ballot can alter my vote toward the more
>feared party.

Two comments on this:

(1) In order for the ballot to be identifiable, the voter must 
cooperate in marking it in a way that identifies it. Thus 
vote-buying, with the purchaser demanding that the voter make the 
ballot identifiable, is much more practical than obtaining the 
cooperation of a coerced voter who is not voluntarily cooperating. 
Nobody would be able to identify your vote unless you make it 
identifiable. Techniques for doing this have included punching a 
small hole in the ballot with a pin. Generally, rules prohibit 
extraneous marks on a ballot. I have also considered that, where 
coerced cooperation is a reasonable possibility, a certain percentage 
of ballots could be extracted and separately counted under closed 
conditions. Images of these ballots would not be made public.

A coercer would not be able to know for sure that a voter had failed 
to cooperate.

(2) Direct democracy generally requires open voting. Coercion seems 
to be rare; in fact, I've never heard of it. (I lived in a Town 
Meeting town until last year, for about five years.) I'm sure that 
people do sometimes alter their votes because they think they will 
not be popular; but actual coercion is another matter.

The fear described above is misplaced. Nobody would be able to 
identify your ballot unless you make it identifiable. You could write 
your own name in as a candidate for some office, something I would 
not recommend to anyone who is unable or unwilling to risk exposure!

I have suggested Asset Voting for "difficult" circumstances, and, 
under those circumstances, I would not suggest that anyone register 
as a candidate who could not manage the security necessary. Asset 
makes huge fields of candidates possible, but it is very difficult to 
provide close security for large numbers of people, so difficult 
conditions would necessarily reduce candidate fields, though they 
could still be substantial without harm.

>In many countries, if the votes were not secret, this
>fact itself would change the results, even without any
>actual coercion or threat of coercion.

That's correct. But this does not apply to what we are proposing. It 
takes actual coercion to force voters to identify their own ballots. 
And procedures can specially handle ballots which have identifying marks.

Again, if conditions were such that ballot imagine would cause 
serious security risks, in spite of what I've written, an imaging 
program could be suspended. There are other, albeit more expensive, 
ways of confirming that ballots have been properly counted.

The serialization which I proposed would be done on batches of 
ballots which have been shuffled, and the serialization only requires 
that each ballot have a unique number. The numbers could be in random 
sequence, as long as they are unique. So, for example, one could 
print labels with unique serial numbers randomized, and then, when 
the ballots are removed from the box where they have been deposited, 
the labels could be applied. Another way to add the serial numbers 
would be to run the ballots through a printer. But a sheet of crack 
and peel labels is pretty low-tech; they can be prepared in advance, 
and if you ruin a few labels, no problem. All that matters is that 
the ballot images be uniquely identifiable. The numbers could be 
handwritten in a box on the ballot, for that matter. They do not 
identify the voter, only the ballot.

(Serialization is necessary so that independent counts can be 
combined and compared. Comparing sums is a *lousy* way to confirm 
counts, unless the batches being compared are small. I would have the 
counters in Florida work with relatively small batches, counted by 
independent teams. Where they agreed, the batches would be packaged 
and labeled and set aside. Where not, more detailed procedures would 
be followed, counting with smaller batches, down to examining 
individual ballots. The procedure would have identified and 
segregated all truly controversial ballots for judicial review. The 
idea that final counting procedures needed to be settled before 
recounting could proceed was ... insane. You don't know what counting 
procedures are necessary until you know what ballots and actual 
ambiguity problems you have! Further, specifically documenting the 
exact ballot ambiguities could have made it possible to determine the 
range of legitimate controversy; it is quite possible that this range 
did not reduce the victory margin of the true winner below one vote. 
But the process was successfully obfuscated by those who did not want 
a detailed and thorough count to take place ...)