"Better" IRV?

Thanks to Warren, Paul, Dan, Bart, Abd, et al for the sudden flurry
of very interesting posts related to promoting better voting methods!

There is a voting method which has not been discussed very much, but
which I think has a good chance of being accepted by voting reformers
and by the general public.  I wonder what y'all think of this method
- is it a good method to promote?

This method was briefly called the "orphan method" on the EM list.  I
suggest the name Bottom Two Runoff IRV, or BTR-IRV.  Feel free to
pronounce "BTR" as "better" :-)

BTR-IRV is the same as standard IRV, except that when a candidate
must be eliminated, the candidate is chosen by an "instant runoff"
between the two candidates who have the fewest ballots assigned to
them.  The runoff examines _all_ of the ballots to determine which of
the bottom two candidates is least preferred by the voters.

Features of BTR-IRV:

1. BTR-IRV is actually a Condorcet method, because it will choose the
Condorcet Winner, if there is one, as the winner of the election.  I
have found that most IRV supporters like regular IRV because it
(supposedly) chooses the candidate who has "the support of the
majority of the voters".  Thus, many IRV supporters show some
interest in Condorcet methods.

(Their reasons for preferring IRV over Condorcet-SSD and the like are
usually: a) Condorcet is "too complicated", hard to explain in 30
seconds, etc. b) IRV has "momentum", has been used for many years in
Australia, and is getting a toe-hold in the US, whereas Condorcet
(and Approval Voting) haven't been used in public elections at all.
c) Changing course and supporting a different method, or multiple
methods, would lose that momentum and weaken their chances of
implementing any method that is better than Plurality. d) They belong
to some organization or committee that is committed to promoting IRV,
specifically.  They don't have the flexibility within their
organizations to promote other methods.)

BTR-IRV provides the advantages of Condorcet in IRV "packaging", thus
perhaps allowing IRV supporters to support BTR-IRV as "a type of" IRV.

2. BTR-IRV is a simple variation of IRV.  Thus, IRV supporters and
the general public can understand it as well as they can understand
standard IRV.

3. Since BTR-IRV appears to be a simple variation of IRV, IRV
supporters might be willing to shift their support to BTR-IRV.  Thus,
we could channel the energy of IRV supporters to support this voting
method that is much better than IRV.  I have "sold" BTR-IRV to one
IRV supporter already.

Problems with BTR-IRV:

1. Doesn't pass the Favorite Betrayal Criterion.  Thus, there is
still some pressure on voters to falsify their vote, choose the
lesser evil, etc.  However, I believe this pressure exists only when
there is no Condorcet Winner.

2. Who wins when there is no Condorcet Winner?  The winner appears to
be whoever ends up with a majority of the ballots as the result of
the BTR-IRV elimination process.  The choice seems almost random. 
The winner doesn't seem to be chosen according to any easily-stated
principle.  Is there any simpler way to describe it?

Both of these problems are relavent only when there is no Condorcet
Winner.  How often would this happen?  How serious are these
problems?  When there is no Condorcet Winner, would that usually
indicate that the three or more circularly tied candidates are about
equally preferred by most voters?

3. BTR-IRV, IRV, Condorcet, etc. are all more complicated to
understand, and more costly and difficult to implement than Approval
Voting.

4. In the worst case (no Condorcet Winner) BTR-IRV is probably as bad
as IRV regarding the amount of information that has to be transmitted
to a central vote counting location, in order to determine the
winner.  (The problem with having to transmit a large amount of data
is that it becomes much more difficult to audit the election and
verify/insure that the data was not tampered with between the
precincts and the central counting location.)

I am moderately concerned by each of those four problems.  But,
overall, I think that BTR-IRV is much, much better than standard IRV,
and really in about the same league with Approval Voting and "good"
Condorcet methods.  And since it is a variation of IRV, it should
appeal to IRV supporters.  BTR-IRV seems like an excellent compromise
between the IRV and the anti-IRV camps.

What do you think?

Cheers,
- Jan