methods proposed to sw Committee

[Steve Eppley]

Mike O wrote:
>[This is just a list, and _not_ a ranking]
>1. MPV
>2. plain Condorcet
>3. Smith Condorcet
>4. Saumur's random solution of circular ties
>5. Shugart's modification of Runoff
>6. Approval
>7. Regular Champion
>8. Any 1 or 2 methods that Demorep wants to pick from the
>   ones that he's mentioned, to define them precisely and
>   add them to this list

I've mentioned a couple of methods which use rated (aka weighted)
ballots:

1. Rated ballots tallied by Condorcet's method, ignoring the amount
of difference between the two candidates in each pair by converting
each pairvote to A>B, A=B, or A<B.  I think I called this method
"WB/Condorcet" to indicate it uses weighted ballots.  (If we include
this in an EM vote on sw methods, I'd suggest we also include
WB/Smith//Condorcet.)  

Example ballot: {A=10, B=9, C=5, D=0, E=-10}
Example tally:
  46 vote: D=100, C=-100, N=-100
  20 vote: C= 75, D=   0, N=   0
  34 vote: N= 80, C= -20, D= -80
  The first step is to convert the ratings to rankings:
  46 vote: D>C=N
  20 vote: C>D=N
  34 vote: N>C>D
  The rest proceeds as in Condorcet (or Smith//Condorcet, etc.).
    ==>  Clinton wins the Condorcet tie-breaker.

WB/Condorcet shares the good properties of Condorcet's method.
Voting strategies are just as ineffective, so it should do an equally
good job of electing the candidate which beats all others pairwise if
there is one.  And no candidate which is less preferred to some other
by a majority can win if there's a candidate which doesn't have a
majority against.  Voters might have some indirect incentives to
stretch their relative preferences, such as to influence the apparent
"mandate" (or lack of "mandate") of the winner, but arguably this
problem may be less important than not having any voter rating info.

My commentary when I proposed WB/Condorcet included its potential
downside of sour grapes when the method doesn't elect the candidate
who would have won if the same votes had been simply summed.
  46 vote: D=100, C=-100, N=-100
  20 vote: C= 75, D=   0, N=   0
  34 vote: N= 80, C= -20, D= -80
  D =  100*46 +  0*20 + -80*34 =  1880
  C = -100*46 + 75*20 + -20*34 = -3780
  N = -100*46 +  0*20 +  80*34 = -1880
     ==>  Candidate D wins by simple sum and screams bloody murder.
          Other voters yawn, noting the extreme voting by D supporters.

But the "No sour grapes" standard for public elections certainly
doesn't have to disqualify WB/Condorcet, WB/Smith//Condorcet, or
WB/Schwartz//Condorcet from being used in EM for any sw voting we do
here, so perhaps we shouldn't use ranked balloting if/when we vote 
here.

2. York's method: rated ballots tallied using renormalization and 
elimination.  (WB/RE, which is my acronym for this but not York's.)

Since I posted York's method, he proposed some other related methods
which he claims are better than that one.  I'm not going to describe
them here, though.  Their definitions look very complex, so I think
it would be a hopeless task to persuade electoral reformers to work
for them even if they're the cat's pajamas.  (York appears not to
have a strong understanding of how undesirable are methods containing
incentives to strategize, so I doubt he's really found the cat's
pajamas.)  I expect it will be left to small groups experimenting
with online voting systems to make or break his methods, and if he
really has something, then his reforms can evolve out of the
success of the groups which use his methods. 

* *

If memory serves, I didn't comment when Rob proposed using Mike's
single-winner document as our draft FAQ.  I think it would take
considerable work to hammer that into something really good, so like
it or not we'd probably need to have many little votes on our
revision decisions.  (If we can't work up a critical mass to try 
standards-based group writing, I'll take another look at Mike's paper 
and decide whether I want to get involved in editing it.)

I didn't cast a vote about a month ago after Mike cast his vote on
which standards to discuss first.  I would have voted, since I want
to try that standards-driven group writing idea, but I thought it
would be better to wait until more people agreed to try the idea.  
My vote would have been to approve discussing "Simplicity" and
"Decisiveness" (Matthew Shugart's two standards) to encourage him
not to drop out of our discussion, or at least obtain his most
significant contributions before he departed.  I believe Matthew 
has since unsubscribed, unfortunately.

Rob opposed that group writing idea.  We don't need his okay, but we
would need to have a critical mass of participants, I think.  As I
recall, three of us (Kevin, Mike, and I) wanted to try it.  Anyone
else want to try it?  

If we don't, I could vote right now my preference for the
Condorcet(0) family of methods, but I'm not convinced that the ER
group will really be moved by the decision of a few self-selected
individuals if we don't have the standards-based documentation to
justify our votes. 

Another issue we might want to take up here is whether reformers
ought to be spending more time working for single-winner reforms. 
Campaign finance reform and prop rep get most of their attention.  
Is that optimal?  If not, what can we say or do to correct the
imbalance? 

---Steve     (Steve Eppley    seppley at alumni.caltech.edu)