IRV and competing methods. (fwd)

[Forest Simmons]

Xander,

I want to elaborate a little on one point that I passed over rather
glibly.

Forest

Xander wrote (in part):


> I see the potential failings of IRV, but it does allow us to show our
> support and greatly reduces the odds, if not eliminate them altogether,
> of
> spoiling.  My question about IRV is whether the failings you point out
> are
> as glaring in practice as they are in theory.  I don't know enough about
> how IRV has worked in the real world to even venture a guess, but it
> seems
> to me that the kind of examples that reveal it's faults are based on
> some
> unlikely assumptions, such as that all first choice Dems would rank R
> second. 

I want to point out that I used the example only as a mathematician would
use a picture of a triangle during a mathematical proof of the sum of the
internal angles being equal to 180 degrees.  The picture is not really
needed in the proof.  All of the general statements and supporting
reasons stand on their own. The picture is just to illustrate the meaning
of the words in the proof for the inexperienced reader.

In the context of a three way race, a centrist could be defined as a
candidate who has fewer last place votes than first place votes when each
of the other two candidates has as many last place votes as first place
votes. 

The existence of this kind of centrist is more the rule than the exception
when there are three strong candidates.

If the centrist has fewer first place votes than either of the other
candidates, then the supporters of the weaker of the other two
candidates will benefit by falsifying their first choice (often making no
difference which of the other two candidates they place above their
favorite on the ballot).

This result is quite general, not limited to the illustrative example that
I gave.

Note again the asymmetry of IRV: On the one hand IRV says that the
candidate with the fewest first place votes is not good enough to be kept
in the race. If IRV recognized that last place votes are just as
informative as first place votes, the rules would also say that the
candidate with the least number of last place votes is too good to
be eliminated immediately.

An example of a symmetrical rule that would never eliminate the Condorcet
winner (in a three way contest) is this: Eliminate the candidate with the
greatest difference in the number of last place votes and first place
votes. 

Under this rule, no candidate with fewer last place votes than first place
votes would ever be eliminated.

Now what's wrong with a method that requires voters to falsify their
preferences in order to defend themselves from the candidate they consider
to be the worst?

(1) Successful use of these "insincere" strategies requires high quality
information about the sizes of the different factions and their likely
preferences. In this era of corporate control of information 
and disinformation, this requirement is dangerous.

(2) As I mentioned before, in a runoff simulation (like IRV) results from
previous rounds are not available to inform strategies for later rounds.

(3) There are psychological stresses involved in falsifying preferences,
even when the voter doesn't consider such strategies to be immoral.

(4) The outcome of an election in which many of the ballot preferences
were falsified (in self defense) can be misleading, giving false mandates,
etc.

(5) The spoiler effect is inextricably connected with this lesser evil
effect.  A growing minority party has an artificial downward pressure from
this effect.  That's good if you consider the two party system to be
the only safe or sacred system.

(6) The only good thing about falsifying preferences is that sometimes it
can help make the most of a bad system. But this is no excuse for keeping
a bad system. There are excellent systems that give no strategic incentive
for falsifying preferences. There is no spoiler effect or lesser evil
dilemma in these excellent systems.

Even the simple symmetrical elimination rule I mentioned above eliminates
the spoiler problem in any race with only three strong candidates, i.e.
any typical single winner race in a district that hasn't been Balkanized
by repeated elimination of centrist candidates.

Forest