[EM] book recommendation

[Bart Ingles]

"John B. Hodges" wrote:
> 
> Greetings- I am still "educating myself" about all these issues. I
> asked for book recommendations and got websites instead, which is OK
> but I find that websites are shallower than books, almost by
> necessity since as a practical matter they are much shorter. At any
> rate, I have found exactly the sort of book I was looking for, and
> thought I would mention it.
> 
> MAKING MULTICANDIDATE ELECTIONS MORE DEMOCRATIC, by Samuel Merrill
> III, Princeton University Press, 1988, our library has it catalogued
> at JF 1001 M47 1988 (there were many  other books in this general
> subject area at JF 1001).

Excellent choice, I recommend it as well.

> This list, if I may be so bold, seems to have been established as
> part of a drive to defeat one particular reform that seems to be
> gaining political momentum, the Instant Runoff method (Hare, or
> "Australian" ballot.) The commentary here seems to put down IRV with
> extreme predjudice.

Not entirely, but I agree that most seem to support approval voting
and/or Condorcet methods.  On the other hand, I would venture that most
of the people here started out as IRV supporters.  My initial posts to
this list were pro-IRV.

Some of the pro-IRV posters on this list in the past have been, shall we
say, less-than-genuine.  This probably hasn't helped the disposition of
others toward that method.


> Merrill finds that IRV-Hare is a major
> improvement over Plurality on all measures.

Not all measures, although IRV was competitive with approval voting in
terms of Condorcet efficiency, and midway between plurality and
approval/Condorcet/Borda in terms of social utility efficiency, at least
when there were not too many candidates, and the candidates were not
"bunched up" with similar platforms.

One of the most important findings of the book was the sensitivity of
IRV to multiple candidates with similar policy positions.  If you look
at the graphs of Condorcet efficiency on page 26, and of social utility
efficiency on page page 36, you can see that reducing the candidate
dispersion has a devastating effect on the performance of IRV and
top-two runoff.  In the bottom graphs on those two pages, IRV and
top-two weren't much better than plurality.

It took a while for me to grasp the implications of this-- originally I
took this to mean merely that IRV wasn't good at picking centrists in
some situations.  But what it really shows is that the spoiler effect is
still there, enough to reinforce the two-party system almost as much as
plurality.

> Fans of the Condorcet
> criterion, which is basically a measure of the tendency of the method
> to select centrists, should note that IRV seems to do that quite a
> bit better than Plurality. Whether IRV is the best method possible is
> open to debate; but it ain't all THAT bad. C'mon, people.

It depends on your goals.  If you believe in a strong two-party system,
and are mainly interested in reducing the nuisance value of smaller
parties, then IRV is the way to go.  If you want to see more than two
competitive parties, or if you want new parties to be able to evolve to
displace old ones as they become obsolete, then IRV is not going to be
much of a change from the current system.

For those who consider IRV to be only a modest improvement, and its
adoption as an impediment to implementing better methods, I think it's
understandable that some consider IRV to be a step in the wrong
direction.  It depends on the situation, though.  For places where it's
already implemented, the "step backwards" argument is moot.

> Merill's analysis of the effects of strategic or "insincere" voting,
> IMHO, eliminates Borda and IRV-Coombs as serious contenders. IRV-Hare
> seems resistant to manipulation.

I agree, at least where it concerns strategy on the part of voters.  But
the effect of Hare on parties and candidates is to reduce or limit the
number of viable candidates, which could also be considered a form of
manipulation.

> Approval voting gets a thorough look. It is vulnerable to
> manipulation, different rules having different effects; in assorted

If you're talking about two stage procedures in Chapter 7, it's well
known that runoff-approval hybrids are vulnerable to voter strategies
such as collusion.  But those hybrids aren't approval voting.

> scenarios it usually but not always ranks better than IRV on
> Condorcet efficiency. I haven't yet finished reading those sections,
> which get rather technical.

> The two-party system is obsolete and dysfunctional.
> Better forms of democracy: www.fairvote.org

...for a two-party system which is less obsolete and dysfunctional?  :-/

Bart