[EM] Re: Suter on Election-methods Digest, Vol 6, Issue 13

[Warren D. Smith]

> From: RLSuter at aol.com
> Subject: [EM] Is range voting the panacea we need?
> To: election-methods-electorama.com at electorama.com
> Message-ID: <12d.51cf9406.2ef25020 at aol.com>
> Content-Type: text/plain; charset="US-ASCII"
> 
> Will someone on the list who has studied range voting and compared it to 
> Condorcet, approval, and other methods please comment on Doug Greene's
> paper? He  appears to be saying that range voting is superior to all other single
> winner methods. Are there good arguments against this conclusion? Does range
> voting have serious flaws? If so, could someone briefly summarize them?
>  Thanks,
> Ralph Suter

--Reply:  I am one of Greene's coauthors.
There are senses in which range voting is superior, perhaps not to all,
but certainly to all the common, single-winner voting methods.
To learn about that, see my 2000 "range voting" paper 
and associated material at
    http://math.temple.edu/~wds/homepage/works.html  #56.

The senses of superiority are both theoretical, and experimental
via computer simulations of millions of elections.  I personally
find the computer evidence more convincing than theoretical reasoning.
Specifically, computer shows range voting has smaller "Bayesian regret"
than all other
voting methods I tried, robustly across a very large number of different
scenarios,
different numbers of voters, numbers of candidates, honesty/strategy 
behavior of voters, ignorance/knowledge of voters, different "utility 
generators", etc.  Bayesian regret is argued to be clearly the uniquely
best 
statistical yardstick for measuring single winner voting system quality.

Warning: this 2000 paper is not perfect and I am planning on revising
it, 
including some rewrites and fixes of the theoretical parts and
updating the computer study to include even more voting systems,
including
many discussed on "election-methods".   (If anybody wants to help, let
me know...)  I have in fact already partially done that and so far the
conclusion
of range's superiority has not altered.

The present 2004 paper by Smith, Quintal and Greene is about a
real-world
study involving human voters.  It learned some real-world lessons
about range and approval and other voting and is recommedned reading as
a 
sanity check for those of you who may be getting a bit too theoretical.

-- 
  Warren D. Smith
  wdsmith at fastmail.fm