[EM] Voting Systems Study of the League of Women Voters of Oregon

[Kristofer Munsterhjelm]

Markus Schulze wrote:
> Hallo,
> 
> in September 2008, the League of Women Voters of Oregon
> published its Voting Systems Study:
> 
> http://www.lwvor.org/documents/ElectionMethods2008.pdf
> 
> This is what they write on Condorcet methods (page 16):
> 
>> A Condorcet ballot lists every pair of candidates with
>> voters being asked to designate their preference in
>> each pairing. The confusing ballot and difficulty of
>> tallying votes made this system impractical. Now,
>> however, voters can rank candidates with computers
>> used to determine the winner. Some online communities
>> use Condorcet voting, but for government elections
>> new voting equipment would be required. There are
>> also infrequent cases where the Condorcet system
>> produces a circular outcome. "Candidate A may beat B,
>> B may beat C, and C may beat A."
> 
> Interestingly, in the table on page 12, they claim
> that Condorcet methods are not majority methods.
> 

This seems completely wrong. First, voters don't have to care about the 
innards of the system. As the very paper they link to "for more details" 
(by the LoWV of Minnesota) says, "[f]rom the point of view of voters, 
the Condorcet system is another ranked system". Even in a manual 
implementation of Condorcet, voters wouldn't be exposed to the Condorcet 
matrix directly (and shouldn't, because then they can submit 
intransitive votes); rather, those counting the votes would construct 
the matrices.

Second, they claim that Condorcet methods are not majority methods. I 
can understand this with regard to top two runoff (since it's possible 
for a candidate to have less than 50% support whenever there are more 
than two), but not for Instant Runoff. If instant runoff is a majority 
method and Condorcet is not, then the perverse "eliminate the candidate 
ranked first by most until only two remain, then pick the winner" is 
also a majority method, and even if we do say that IRV is a majority 
method, then presumably BTR-IRV, which is Condorcet, also is a majority 
method. Perhaps even the Borda-elimination methods would be majority 
methods.