[EM] IMHO, IRV superior to approval

[Adam Tarr]

Just to review - this started as a discussion of 1P-1V.  I pointed out that 
the criteria seems to be based more on the way the votes are read, than the 
results.  To drive the point home, I presented a sequential way of working 
through the ballots so that both IRV and Approval eliminate candidates by 
taking two candidates, pairing them off, and eliminating one of them.

Stephane Rouillon wrote:

>Adam Tarr a écrit :
>
> > Steph wrote:
>
> > >If your ballot is exhausted it is because you decided not to provide a
> > >full ranking.
> > >It is your choice not a fact inherent to the electoral method.
> >
> > Approving of both or neither of the final two candidates in approval is a
> > choice not inherent in the method, as well.
>
>With IRV, a voter can make sure its preference will count when only two
>candidates are left at the last round.  With Olli's sequential equivalent
>version of approval, voters cannot be sure they will participate in that 
>last selection because they do not know in advance who will be the last 
>contenders.

True.  But as I said before, what is so special about the last 
elimination?  The vast majority of the IRV voters will NOT participate in 
every elimination round.  Each round eliminates one candidate.  Why is the 
last elimination more important?  What if your second choice is eliminated 
earlier, and you never got to participate in that contest, and now you're 
choosing between your fourth and fifth choices?

>In my eye, approving of both or neither of the final two candidates in 
>approval
>is not a choice, it's a guess. And that is inherent to approval. With 
>approval, the
>voter needs to guess in which pairwise comparison its participation will 
>most make a difference (and puts the cut-off there).

And this is analogous to the question of whether to bury your favorite to 
protect your compromise in IRV.  That, too, is a guess.  If you bury and 
guessed wrong, you end up electing the compromise you didn't need (just 
like approving an un-needed second choice).  If you don't bury and guessed 
wrong, you end up indirectly electing your least favorite (just like 
failing to approve a needed second choice).

IRV does not solve this guessing game any better than approval.  In fact, 
as Mike Ossipoff has pointed out many times, approval is more forgiving to 
poor strategy in several of these cases.

Only Condorcet allows you to vote sincerely without fear in this situation 
(ignoring for the moment the more remote possibility of a circular tie).