[EM] voter strategy & 2-party domination under IRV voting

[Rob Lanphier]

On Sat, 2005-08-13 at 18:48 -0400, Dave Ketchum wrote:
> NOT at all clear that 2-party domination is as evil as some claim.

This is a really good point to consider.  We probably need to discuss
the specific characteristics of the two-party system we rail about in
order to judge if a system is bad or good.

In the case of plurality, the character of the two party system is such
that candidates are motivated to optimize for 50%+1, alienating the
other 49.whatever percent of the population that makes a good scapegoat.
That leads to a two-hump distribution of candidates along a largely
unidirectional political spectrum.  The most polarizing issues are used
as tools to divide the electorate into two halves (e.g. abortion, gay
rights, separation of church and state, business owners/managers vs
labor, private vs public sector).

In the case of IRV, it arguably gets worse, because it becomes possible
for weak candidates to run without being spoilers, but strong candidates
are in just as much danger to split the vote.  Hence, the pressure will
still be on to cluster into two idealogical camps.  Even when the threat
is mild, voters would still be nervous.  Hence, I think that's why
burying is common tactic in Australia.  My understanding of the
simulations is that you still have two humps in the idealogical
distribution curve, but that they are further apart, so the zig-zag
characteristic of switching between the two parties is actually
exaggerated under IRV.

However, two parties clustered near the center of the spectrum may not
be such a bad thing, so long as those parties don't have too many unfair
advantages like the ones we see today (e.g. ballot access barriers,
"safe" districts, etc).  Those two parties should be in genuine fear of
their competition, much in the same way that things work in a healthy
free market (e.g. Dell and HP in the PC market, FedEx and UPS in private
package delivery).  There are only so many brands that people can keep
in their heads, so having two dominant "brands"/parties of candidates
may not be so bad.  That said, free markets don't typically settle long
on just two players, and when they do, there very well could be some
unfair advantages at work.

William Riker wrote a paper titled "The Two-party System and Duverger's
Law: An Essay on the History of Political Science" in the American
Political Science Review in December, 1982 which covers the topic.  It's
a good coverage of some of the more insidious aspects of the system;
including the disincentives to smart candidates who don't fit in one of
the two dominate parties.  It's a good reminder of why a two-party
duopoly is usually bad.

Regardless, I have yet to see a convincing case that Condorcet methods
naturally lead to two-party duopoly.

Rob