Electoral College reform (was Re: Addition to earlier p

[Steve Eppley]

Don wrote:
>Don: I have changed my two candidate concept somewhat. My position
>now is that if two or more candidates are in reach of a win then we
>drop all the other candidates and use a single-winner method to
>obtain a winner in the reformed states.
-snip-

I agree that this is what to do when there are two or more who can
win an electoral majority, except when the reformed states have
solidly voted "to defeat" one or more of those who can win.  (See 
below.)

>Don: In the cases in which there is only one or no candidate within
>reach of a win the policy shall be that we drop no candidates. The
>policy is changed because in these cases the reformed states are not
>in the position of being the deciding power between two of more
>candidates that are in reach of a win. In other words the reformed
>states cannot play the King Maker. Therefore it is best to let the
>single-winner method decide who gets the electoral votes of the
>reformed states.

I disagree somewhat with this part of the algorithm, in the case
where exactly one candidate can reach an electoral college majority.
The reformed states have only two possibilities here: they can elect
that candidate or they can throw the election to the House of
Representatives.  I think there are obvious situations where they'd
rather elect that one even if s/he isn't the leader in the reformed
states--s/he might have been a close second, etc.  And I think there 
are obvious situations where they'd rather let the House decide, such 
as when the House is likely to pick a candidate higher in the 
reformed states' order than that one.

Perhaps there's a way the method can guess which one would be 
picked by the House, such as by using the party affiliations of 
the candidates and the parties' seats in the House.  If the one
candidate who could win an electoral college majority is more
preferred in the reformed states than the candidate(s) belonging to
the party which has a majority of House state delegations, perhaps
the reformed states should award their delegates to the one.

>>Steve:just use the results of the sw method in the reform states
>>after eliminating the candidates who can't possibly win from the
>>ballots which were already cast.
>
>Don: What I like here is the order - you are using the
>single-winner method after dropping candidates (carve it in stone).
>This will work. I was under the impression that your policy was to
>work the single winner method FIRST in the reformed states and then
>eliminate candidates. 

I wrote it that other way initially, and when I rewrote it I assumed
the two were mathematically equivalent, in that they'd produce the
same result.  Thanks, Don, for pointing out the possibility that
changing the order of the steps may sometimes lead to different
results.  

But it's unclear whether it's better to determine the reformed 
states' collective candidate order before or after eliminating the 
candidates who can't win nationally.  Why do you think it's so much 
better to eliminate the national can't_wins before tallying the order 
in the reformed states?  It looks like this choice makes a much bigger 
difference if the method is IR than if it's Condorcet, so this issue 
may need to be considered in the light of the sw method.

At the moment, I see no compelling reason not to delete the 
can't_wins from the ballots before tallying, so perhaps we can 
quickly agree to carve this in stone as you request.

>Question: What do you do with the votes of the dropped candidates in
>the reformed states? - I say these votes should be salvaged and
>reassigned to the remaining candidates per the other selections of
>the voters.

That question is specific to Instant Runoff, since pairwise methods
don't waste votes.

I agree: if the method is IR, delete the dropped candidates from the 
voters' rankings.  For example, if candidate X can't win nationally
then the ballot X>Y>Z becomes Y>Z, and the ballot Y>X>Z becomes Y>Z.

>>Steve:I also agree with Mike's suggestion that the candidate who
>>finishes last in the combined reformed states should also be
>>eliminated from contention.
>
>Don: I am having trouble understanding this rule. This last place
>finisher, is he one of the candidates in reach of win if he gets the
>reformed states votes? If so I say he should not be eliminated. He
>has a chance to win. The voters in the reformed states should decide
>if he is to win or lose.

Finishing last in the reformed states means that, to the voters in
the reformed states, s/he is the most despised candidate.  I think
this means they've decided s/he should lose, and want him/her to get
zero delegates even if s/he is the only one who can win nationally.

>Or: if this last place finisher is one of the minor candidates with
>the lowest amount of popular votes in the reformed states - he is
>last - he is most likely out of reach of winning anyway and would be
>eliminated by your statement above. I do not see the value of this
>rule??

Again your terminology ("lowest amount of popular votes") is more
appropriate for IR than for a pairwise method.

You're right that it would be unlikely that a candidate who finishes 
dead last in the reformed states would be the only one who could win 
nationally.  Which means that the value of this rule is small enough 
that it may not be worth the additional complexity.  What would be 
more worthwhile is a way for the reformed states to let the House 
decide in more cases than this unlikely one, such as when the 
reformed states disapprove of the one who can win nationally.  
Is this a job for the "None of the Rest" choice?

>>Steve: There's no point in awarding delegates to the most despised
>>candidate just because s/he's the only one who can win a majority
>>of the electoral college.  That would be worse than letting the
>>House of Representatives pick a winner.
>
>Don: This is not only good - it is GREAT!  I want it carved in
>stone (my words). 

I'm not clear on where you stand on this, since above you said you 
see no value in it.

>This supports my contention that we must allow the voters of the
>reformed states to dump the despised candidate when there is only
>one who can win a majority of the electoral college. We do this by
>not dropping any candidates. 

You seem to be overlooking the case where the one who can win is 
neither the last-place finisher nor the leader in the reformed states.
What if the only one who can win nationally finishes 2nd in the 
reformed states?  By not dropping the can't-wins, you'd let the House 
decide rather than elect the candidate who finished 2nd??

>We work the single-winner method to the conclusion of a single
>candidate no matter who wins the electoral votes of the reformed
>states (what will be - will be).

It "won't be" if all but the one who can win are dropped...

>Steve:>
>>Single-winner methods do more than pick a winner--they also sort all
>>the candidates in a "collective preference" order if the following
>>algorithm is applied:
>>
>>To find which candidate finishes in nth place in the collective
>>order, eliminate the 1st thru (n-1)th place finishers from the
>>ballots, retally the remaining candidates using the sw method,
>>and select the "winner" of the remaining candidates.
>>
>>For example, given the 3-candidate race with the familiar ballots:
>>   46: Dole>Clinton>Nader
>>   20: Clinton
>>   34: Nader>Clinton>Dole
>>
>>If the method is Condorcet then the collective order is:
>>   1.Clinton 2.Dole 3.Nader
>>That's because Clinton is the sw winner and therefore the 1st place
>>finisher in the collective order.  After eliminating Clinton from
>>the ballots, what's left is:
>>   46: Dole>Nader
>>   20: wasted
>>   34: Nader>Dole
>>Dole wins this by Condorcet and gets the 2nd place spot in the
>>collective order.  That leaves the 3rd place spot to Nader.
>>
>>If the method is Instant Runoff then the collective order is:
>>   1.Dole 2.Clinton 3.Nader    (Details omitted.)
>
>Don: Slight Correction: Using Instant Runoff, Nader should be
>second - unless you have dropped Nader because he is not in reach of
>national win.

I disagree with your "correction."  Follow the algorithm I provided 
above:  To find the 2nd place finisher, drop the 1st place finisher 
Dole from the ballots and retally:
   46: Clinton>Nader
   20: Clinton
   34: Nader>Clinton
Clinton wins this, so Clinton is 2nd and Nader is 3rd.  
The IR collective order is:  Dole, Clinton, Nader

>Steve:
>>So, the current proposal (which perhaps Don, Mike and I can all
>>agree on) is:
>>   1. Determine the collective order in the combined reformed states.
>>     (The sw method isn't specified here, but all the reformed states
>>     would have to agree on which one.)
>
>Don: I agree that we should determine which candidates will be
>running in the single-winner race, but we have to yet to agree how
>to determine the selection. I do not know what you mean by
>collective order.

Right, we haven't agreed to 1, and the proposal which appears to be 
leading is to not determine the collective order until the can't_wins
are dropped from the ballots.

I defined collective order above by providing the algorithm for
calculating which candidate finishes in each position of the order. 
It's a "recursive" definition, since to find the nth finisher you 
first find the 1st, 2nd, ..., n-1th finishers.  Then you drop them 
from the ballots and see who "wins".  This "winner" is the nth place 
finisher.

>>Steve: 2. Eliminate from the collective order the last place finisher.
>Don: I still do not know the value of this rule as I have stated
>earlier in this post.
>
>>Steve: >   3. (optional) Eliminate candidates beaten by NOTR.
>
>Don: I must pass on comment here - NOTR is out of my knowledge.

NOTR stands for None of the Rest.  (Also called None of the Above or 
None of the Below.)  It is a "choice" each voter can insert anywhere 
in his/her rankings, and it's interpreted as the dividing line 
between the candidates approved by the voter and the candidates 
disapproved by the voter.

Example ballot:   1. Dole  2. Clinton  3. NOTR  4. Nader
This voter approves Dole and Clinton, but disapproves of Nader.

Even if the method is IR, the info in the ballots allows pairwise
tallying of each candidate versus NOTR so it's possible to find which
candidates are disapproved (beaten pairwise by NOTR).  In #3 above,
I was suggesting that no electoral delegates would be given to any 
candidates beaten pairwise by NOTR even if this would throw the 
election to the House.

>>Steve: 4. Eliminate from the collective order all candidates
>>who can't win a majority of the electoral college.
>
>Don: I agree with this as far as it goes. I would like to make additions.

I assume that by additions you mean what you mentioned above: that if
only one can win a majority then none are eliminated.

>>Steve: 5. If at least one candidate remains in the collective order,
>>       award all the reformed states' delegates to the collectively
>>       most preferred of the remaining candidates.
>>       Else award all the delegates to the candidate who was first in
>>       the collective order before any eliminations were made (i.e.,
>>       first in the order determined in step 1.)
>
>Don: I cannot comment because I do not know what collective order means.

Do you have enough to go on now?

>>Steve: Example:
>>Suppose the reformed states have 120 delegates to award.
>>Suppose Dole and Clinton each won at least 150 delegates in the
>>unreformed states but Nader won less than 150 delegates.
>>Suppose the collective order in the reformed states is:
>>   1.Dole 2.Clinton 3.Nader
>>Step 2: Eliminate the last place finisher Nader:
>>   1.Dole 2.Clinton
>>Step 3: not applicable: NOTR wasn't used.
>>Step 4: Eliminate Nader (again) since he can't win 270 delegates:
>>   1.Dole 2.Clinton
>>Step 5: Award the 120 delegates to Dole, who wins the office.
>
>Don: Steve, I must study this example before I can comment on it
>(forgive us).
>        But offhand I question the opening numbers.
>                Reformed States   120
>                           Dole   150+
>                        Clinton   150+
>                          Nader   150-
>                      -------------------
>                          Total   570 +or-

No contradiction.  Nader could have 50...
What matters is that he can't win nationally.


>Don: What follows is my plan (as of now) in order of operation.
>
>One: In the electon the voters will be allowed to make more than
>one selection.

I don't like this terminology.  It sounds too much like the Approval 
method, which instead of using ranked ballots just lets the voters 
indicate more than one Yes.  I suggest you make it clear that you're 
talking about ranked ballots.

>Two: After the election we check the electoral vote counts of all the
>candidates in the UNreformed states. If two or more would be in reach of a
>win, provided they were to win the electoral votes of the reformed states,
>we drop all other candidates from the reformed states tally.
>
>Three: The votes of these dropped candidates are reassigned to the
>remaining candidates per the other selections of the voters.

Three doesn't apply to pairwise methods, which don't "assign" votes
to one candidate at a time.  To use a more universal wording, how
about saying that the dropped candidates are dropped from the
ballots?  For example, dropping Y from the ballot X>Y>Z produces 
the ballot X>Z no matter what sw method is used.

>Four: If only one or no candidate is in reach of a win, we do not
>drop any candidates. All candidates will run in the single-winner
>race coming up next.

For simplicity I might go along with this.  However, to be even 
simpler why not just eliminate this rule and simplify the "two or 
more" in Two to "any".  So any candidates who can't win a majority 
are dropped, even if this means dropping all of them (throwing it 
to the House) or electing the only one who can win (even if s/he 
is despised in the reformed states--highly unlikely).

Here's a possible wording for a revised Two (eliminating Four):

  Two: After the election we check the electoral vote counts of all
  the candidates in the UNreformed states.  Any candidates which 
  can't possibly reach a majority of the electoral college, even if
  they also win all the electoral votes of all the reformed states, 
  are dropped from the reformed states' ballots.

That last phrase also negates the need for a separate Three.

>Five:-snip-
>Six:-snip-

I also don't like the "vote-sum" terminology in your later message.

Since it's not necessary to describe the single-winner method in the
clauses about handling the electoral college, I suggest dropping all
of those sw details to keep this electoral college proposal simple 
and "modular".

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