[EM] IRV ballot pile count (proof of closed form)

[Abd ul-Rahman Lomax]

On this topic, something gets missed in the technical details of 
transmitting precinct data. First of all, there is no doubt but that 
with scanning techniques, it's possible to transmit all ballot data 
from a precinct to a central counting facility, with IRV or any other 
ballot, without requiring sophisticated equipment at the precinct. 
I've suggested public ballot imaging, which would be literal images, 
so that any member of the public can see what an election observer 
could see. A simple form of this would be fax transmission; with 
various simple forms of optimization, this could be reasonably efficient.

If there is local analysis of ballots then the necessary data 
transmission requirements are reduced, but, then, this opens up a 
door to various manipulations unless the analysis is open and visible 
to observers. Traditional IRV counting, as in Australia, avoids this 
by manual counting, the ballots are sorted into piles, the pile 
counts are transmitted, and then when central tabulation has the 
counts from all precincts, totals are made and then instructions go 
back to the precinct to distribute the ballots of eliminated 
candidates and transmit back the additions to each remaining stack.

But how does one audit an election? This is the concern of voting 
security people like Kathy Dopp. With precinct-summable methods, 
sampling techniques can be used, and a sample is like a precinct. 
However, IRV's chaotic behavior with eliminations means that samples 
can't predict election results, it's actually necessary (or at least 
necessary with considerable frequency) to count all the ballots. 
That's only necessary with summable methods when the result is 
extremely close, i.e., a near-tie, but IRV creates multiple 
opportunities for near-ties that can affect the outcome of a necessary round.

The real problem, then, is the chaotic behavior of IRV, easily 
visible in Yee diagrams. The voting security problem is simply one 
consequence of that chaos. FairVote tries to cover it up with 
arguments about "core support," but in real elections when there is a 
real problem, the Core Support Criterion that they invented, just so 
they'd have a criterion to satisfy, is really irrelevant. There is no 
violation of that Criterion in those elections, i.e., a winner with 
no first preference votes. Rather, a candidate is eliminated who 
merely was in third place in higher preference votes, and by only a 
small margin.

Plurality "eliminates" all candidates not in first place in first 
preference votes, thus requiring strategic voting for voters who 
support other than one of the top two. IRV really does the same, if 
there are three viable candidates, as in Burlington, Vermont, in the 
2009 mayoral race. It fixes only the "first-order" spoiler effect, 
where the third place candidate or other minor spoiler candidates is 
so far below being electable that bullet votes are truly wasted.

Given that much better methods exist, have been tried and worked, and 
are much easier to canvass, WTF?