[EM] i also liked what FairVote says about IRV and monotonicity.

Kristofer Munsterhjelm km_elmet at lavabit.com
Tue Oct 11 08:41:23 PDT 2011


robert bristow-johnson wrote:
> 
> 
> 
> "Instant Runoff (IRV): monotonicity criterion - Both two-election
> runoffs and IRV can fail the monotonicity criterion because voters
> who shift to this otherwise winning candidate may shift their votes
> away from the candidate who would otherwise be in the runoff,
> resulting in a different, and stronger opponent in the final runoff,
> who may defeat the otherwise winner. Many election experts dismiss
> this criterion as having no real world impact (for more details see
> www.fairvote.org/monotonicity, and Austen-Smith and Banks [1998])."
> 
> i don't know how you can definitively cite an example of failure,
> since we would have to turn back the clock, rerun the election and
> show that somehow, the legitimate winner would have won with fewer
> votes for him/her.  what we *can* show (and this was exactly the case
> for Burlington 2009) is that if some voters had, on their way to the
> polls, changed their mind and changed their first-choice vote from
> someone who had lost, to the candidate who had actually won, that
> this previously winning candidate will lose (not to the previous
> first choice who lost, but to the Condorcet winner who did not make
> it to the final IRV round).  this is clearly the case in the
> Burlington 2009 election as Warren Smith had pointed out in
> http://rangevoting.org/Burlington.html .  how can FairVote deny
> non-monotonicity in IRV?

I like the "hypothetical dual" argument of Warren's, here. Consider two 
  elections. In election vA, A wins. Then A is raised to make election 
vB,  where X wins. Then either vA is wrong because lowering A (with 
respect to vB) made A win, or  vB is wrong because raising A (with 
respect to vA) made A lose. However, we cannot know, from  the ballots 
alone, which is the case, and therefore the outcome under either vA or 
vB is suspect.

That argument also handles the counterargument of strategy - that 
monotonicity doesn't matter because nobody could gather enough voters to 
  "force" A to lose through raising A. In the hypothetical dual 
argument,  any pair of elections exhibiting monotonicity failure casts 
doubt on the  outcome of either of the pair. Therefore, strategy doesn't 
enter into it.

In the case of the Burlington pair, I'd say that the suspect election is 
the real one, rather than the one where Kiss was raised yet didn't win.




More information about the Election-Methods mailing list