[EM] Condorcet How?

[Raph Frank]

On Tue, Mar 23, 2010 at 2:46 AM, robert bristow-johnson
<rbj at audioimagination.com> wrote:
> and, i'm not sure who, but someone introduced a measure in the state
> legislature to elected the governor by IRV (there is a perennial Prog
> candidate that doesn't get any traction because Vermont is not all like
> Burlington or Brattleboro).  but we know (and Kathy won't let us forget)
> that IRV is not "precinct summable" and that would be a ridiculous mess for
> a statewide election (they would have to transmit via internet, individual
> ballot data to the capitol for tabulation and then securely bring up a disk
> or thumb drive (and the original paper ballots) with the ballot data up for
> verification on a later date.

This isn't strictly true.  An alternative to central counting would be
for additional communication from the central location.

Something like

- Each local area counts first choices and sends its result to the
central location
- central location figures out if anyone has a majority, if not it
declares a candidate eliminated
- local count centers eliminate that candidate and send updated totals
- (and repeat)

It isn't as fast as each center being able to do it at its own pace.
Also, each round takes as long as the slowest local center.

Also, in principle, the central count could issue instructions like
"provisionally eliminate X" after it has only received part of the
count from a specific round.  If the remainder of the votes mean that
X isn't eliminated, then it could be canceled.  In most cases, a
provisional command based on 50%+ of the ballots it likely to be
correct, especially if the margin is small enough.