[EM] Re: Condorcet package

[Dave Ketchum]

On Wed, 23 Feb 2005 08:49:11 -0800 Ted Stern wrote:

> On 23 Feb 2005 at 01:00 PST, Dave Ketchum wrote:
> 
>>Counting votes:
>>     (wv) seems the appropriate choice.  If two voters rank a pair of 
>>candidates (a=b) as equal, then (a>b) and (b>a) should each get one count.
>>     The Condorcet array for each precinct, the whole district, and any 
>>meaningful subset of the district, should be public.
>>     Cycles CAN happen, and should be recognized as near ties.  How to 
>>resolve cycles must be solved before system is ready for voting.
>>     Write-ins counted:  Yes, as if one more candidate - each of which 
>>can be voted for by more than one voter (could demand advance notice but, 
>>right now, no notice makes sense to me).
>>
> 
> Hi Dave,
> 
> Nice post -- I'm with you on most of your points, but have some questions
> about these last few.
> 
> I agree that wv appears to be best for public elections, but disagree about
> how to count a=b.  The whole point of ranking two candidates equally should be
> that the voter is saying he doesn't care about that pairwise contest.  So he
> doesn't want the race between the two candidates to be ranked as highly as it
> would have been with one person voting for one and another voting against.


He is ranking them as being tied, as in x>a=b>y.  He likes each better 
than y, but not as well as x, but a exactly as well as b.  Does not 
matter, except:
      I want the array public, and think the counts will be more 
reasonable as above.
      In resolving cycles I think this makes a difference, and WANT the 
above count to avoid pushing voters away from voting a=b.

> 
> If you prefer to go that route, I would be more comfortable if approval
> weighted pairwise were used as the first sort key (with Ranked Pairs, for
> example), with wv as the tiebreaker second sort key, followed by margin, then
> random ballot.


You are suggesting more complexity to have to explain to all, and maybe 
introduce opportunities for strategy, than I even want to think about. 
For example, wont bringing in approval push voters toward ranking multiple 
candidates when, for many, ranking just their major party preference is 
all that is meaningful in most elections.

So, I want ONLY Condorcet.

> 
> Cycles should be quite rare, but yes, they could occur.  Current practice with
> near-ties is to have an automatic recount.  That should be the case with
> cycles also.  If the cycle doesn't resolve, and Ranked Pairs isn't
> satisfactory, what about a runoff using the rule I suggested?  Smith set,
> followed by higher approved candidates than minimum approval within Smith set,
> followed by approval runner-up; Concorcet completion resolved by wv-RP.


Agreed that when the counts are close, a recount is appropriate.

Cycles would be rare with a normal distribution of voters.  However, get 
the right collection of voter opinions, and you can have lots of cycles 
with 3 or more members.  Given 3 opinions each backed by 1/3 of the 
voters, and each group's second choice being the next one on the list - 
lots of cycles.

While I think a significant percentage of cycles are not close enough to 
justify a recount, it is little expense and simpler to recount all cycles.

I insist cycle resolution do its thing with ONLY the array as data - no 
approval or other complexity.

NO reruns.  If we get a true tie, flip a coin or equivalent.  Close to a 
tie has found a winner.
      Exception - if recount determines that what happened should not 
count as an election, do a new one.

> 
> How would you suggest introducing legislation for single-seat election
> changes, using single-issue initiatives for example?


-MY first shot would be that this is worth passing in one step via 
legislation, without an initiative.

-MY next try is that elections are a single issue for initiative, doing 
away with primaries and using Condorcet in the general.  Takes some 
thought as to effect of the various changes, but campaigns are simpler, 
and doing one election is cheaper for voters, system, and candidates.

-MY next try is to fix general elections, leaving primaries for 
legislation or a second initiative.

With Condorcet you treat true enemies differently from competitors that 
think much like you:
      Do not want true enemies to get votes.
      Want yourself and competitors to get more votes than enemies - just 
want to be liked a bit better than all others.
      Condorcet's array is valuable information - compares EACH pair of 
candidates liking; with less puzzles than Plurality's spoilers.

> 
> Ted 

-- 
  davek at clarityconnect.com    people.clarityconnect.com/webpages3/davek
  Dave Ketchum   108 Halstead Ave, Owego, NY  13827-1708   607-687-5026
            Do to no one what you would not want done to you.
                  If you want peace, work for justice.