[EM] Re: Issues, Condorcet, and IRV (was: IRV vs. plurality)

[Forest Simmons]

On Wed, 6 Aug 2003, Alex Small wrote:

>  Adam Tarr said:
> > PR, if adopted in the USA with large districts, would surely lead to
> > some  Libertarian representatives.
>
> I agree.  Hopefully it would also lead to a more moderate party in the
> fiscally conservative/socially liberal mold.  For that matter, hopefully
> it would also lead to a fiscally liberal/socially conservative party.
> Having four political sectors instead of the same old left-right would be
> far more interesting.

Even now "lefties" say that the media have a right wing bias, while
"right wingers" say the media have a left leaning bias.

This difference in perception is partly from different vantage points, but
mainly from concern about different issue axes.

The media are fairly liberal when it comes to identity politics, abortion
rights, gun control, etc. fairly middle of the road on environmental
issues, social spending, etc.  But when it comes to foreign policy, etc.
they definitely side with the big bucks.

In fact, they seem to side with the big bucks on every issue that the big
bucks really care about, i.e. any issue that has a potential for a
dramatic influence on climate for big business profits.

That makes sense because the media are largely monopolized by large
corporations with all kinds of big business interests.

Right now we have two big business parties that differ only on issues
that don't bother big business very much, when at all.  These two parties
try to focus on these differences and magnify their importance so that we
think we have a big choice to make when we go to the polls.

Note that both Democratic and Republican candidates BOTH get huge backing
from virtually ALL of the large corporations.

The US populace knows that it doesn't make much difference on the big
issues whether a Democrat or Republican gets in, so they stay home and
don't vote.

Gore is more of a war hawk than Bush.

Gore paid lip service to the environment, but when he was in a position to
do something, he sold out just as quickly as Bush.

Neither the Democrats nor the Republicans are anywhere near the center of
the majority of the US populace.  What's more, the line connecting them
doesn't come anywhere near that center.  The situation looks like this:



        Democrat Elite  <----  one meter  ---->  Republican Elite

                                  /\
                                  |
                                  |

                                 one
                                 mile

                                  |
                                  |  Libertarians
                                  \/
                          90 percent of
                         the US electorate

                        Greens



The vertical axis represents issues where the interests of big money
conflict directly with the interests of the common people.

The horizontal axis (blown out of proportion as it is by the
politicians and corporate media) represents all other issues lumped
together.

As long as big money controls the elections and has the lobbying power
(money) to control the elected politicians, the situation won't change.

I think that Approval and Condorcet would exert much more pressure towards
the (generalized) median position (about half way between the Greens and
the Libertarians) than would IRV.

The Kemeny order is the permutation of the candidates that minimizes the
average distance from voter permutations (i.e. rankings on their ballots).

Since (where they are defined) medians minimize average distances, we can
say that the Kemeny order is a generalized voter median of candidate
permutation space (i.e. "social order" space).

Note that the Kemeny order always ranks the CW first, when there is one;
indeed Kemeny's method is a Condorcet method.

It follows that (to the degree social order space is an accurate
reflection of issue space) the CW is the candidate who minimizes the
average distance from the voters in issue space.

[In this connection, remember that medians are more stable under
transformations, for example the transformation from social order space to
issue space, than are means (which minimize average SQUARED distances).
Borda minimizes the average squared distance in social order space, but
with a different metric.]

If there is a CW, any other candidate has a greater average distance to
the voters.  How can that other candidate claim to better represent the
voters?

Perhaps he can by using a different metric to measure the distance, but I
haven't seen any evidence from IRV supporters that there exists such a
metric, let alone any claim that Kemeny's metric is inferior to another
(like Borda's).

In other words, there doesn't seem to be any variational (i.e.
optimization) principle to support IRV.

Instant Runoff (i.e. IRV) is a simulation of what would probably happen
under actual standard runoff.

In order to carry out the simulation, you have to acquire the voter
preference orders.

But if you are going to require the voters to provide all of that
information for runoff, why not use it to simulate a (more thorough) round
robin tournament? (for which it is completely adequate)

If you buy a truck to carry the corn to your storage silo, why not go
ahead and use it to carry the corn to market, too?  If you buy a windmill
to pump water from a well, why not use it to generate electricity, too?
[especially when the generator system and electric pump come with the
windmill at no additional cost]

If there is a CW, and the CW is not the IRV winner, then a majority of the
ballots will show a preference of the CW over the IRV winner, so how can
IRV claim to represent the majority in that case?

Suppose there is no CW, and suppose further that a majority of the ballots
show a preference of the IRV winner over the Ranked Pairs winner. Does
this mean that a majority actually prefers the IRV winner?

Perhaps, IF the IRV ballots are sincere preferences, which is unlikely
when IRV and a good Condorcet method disagree.

That's why few if any on this EM list-serve believe IRV completed
Condorcet to be worth considering seriously.

IRV applied to the Smith set I would be willing to consider (as a way of
pseudo randomizing a choice from the Smith set) :-)

Forest