[EM] IRV in San Francisco

[Justin Sampson]

On Mon, 15 Nov 2004, Eric Gorr wrote:

> The only tie-breaker that I can think of in common usage (it gets
> written in to proposed laws) which would cause the implementation of IRV
> to fail ICC would be what I have called "deterministic"...where all
> candidates tied for least votes are eliminated at the same time.  With
> this tie breaker, if the clones have > 50% support from the voters, it
> is possible to develop a case where they split the vote between them
> equally and are are eliminated causing a clear minority candidate to win
> the election.

Interesting. In San Francisco, multiple candidates can be eliminated in a
single round but only if their combined votes are less than the next
higher candidate, whether or not they're tied. So if the current round
votes are 2000, 1500, 900, 400, 100, then the lowest three candidates
would be eliminated all at once. However, this is identical to eliminating
them one at a time since the votes transferred from each one can't
possible help the others not get eliminated. A true tie like 2000, 1500,
900, 900, would be resolved randomly by pulling slips of paper from a
bowl.

Justin