[EM] IRV & Spoiler Effect

[Forest Simmons]

On Wed, 12 May 2004, wclark at xoom.org wrote:

> Eric Gorr wrote:
>
> > If this is true, no ranked ballot method can be free from the spoiler
> > effect, but do we not generally claim that the better Condorcet
> > methods are spoiler free?
>
> Condorcet is sometimes claimed to be spoiler free because it can satisfy
> what RobLa called "relaxed versions of IIAC, such as Local IIAC."
>
> Full IIAC (which does indeed capture the spirit of the spoiler effect) is
> argued to give Condorcet trouble only when there is an actual ambiguity in
> the will of the electorate, e.g. cycles.  In all other circumstances
> (characterized by "Local IIAC" or its variants) Condorcet actually passes.
>
> Whether or not you concider cycles in Condorcet to be an artifact of the
> system, or a representation of actual ambiguity, obviously plays a big
> role here.


It seems to me that any method that fails the FBC suffers from the
"spoiler effect" in some degree, because if voting your favorite candidate
A over compromise C is likely to make despised candidate B win instead of
C, then A can be considered a spoiler for C, and A might even claim that C
spoils A's chances of winning.

In other words, if there is an incentive to vote compromise over favorite,
then favorite can claim that compromise spoiled his chances (because of
those voters that gave in to the incentive) and compromise can claim that
favorite spoiled his chances (because of those voters that did not give in
to the incentive).

I think that this is one reason some of us put such a high premium on the
FBC.

Approval satisfies the weak FBC perfectly and economically, but completely
fails the strong FBC.  Condorcet doesn't satisfy the weak FBC, but almost
satisfies both it and the strong FBC.

Candidate Proxy economically ameliorates the spoiler effect without
completely satisfying either the weak or strong FBC. The degree of
compliance varies with the version of Candidate Proxy under consideration.

IRV ameliorates the spoiler effect, perhaps better than Candidate Proxy,
but at greater expense. IRV's spoiler effect seems to be particularly bad
at the point a third party starts to become viable, i.e. at the most
crucial stage of evolution of the political landscape.

Mike Ossipoff is working with the AERLO option and his recent Stepwise
AERLO option in an attempt to bring Condorcet (and other) methods closer
to compliance with various versions of the FBC.

These AERLO enhanced methods require ballots with check boxes for the
options and ways of specifying AERLO levels, so they are not strictly
ranked ballot methods.  In fact, you could think of them as Declared
Strategy Voting (DSV)  methods, although they are not as elaborate as most
of the other DSV methods that have been proposed.


Forest

>
> ----
> Election-methods mailing list - see http://electorama.com/em for list info
>