Single Winner Balloting

[Steve Eppley]

Donald Eric Davison wrote:
>          I cast my vote in the Single Winner Balloting for the - -
>                I N S T A N T   R U N O F F   M E T H O D

If you want, you can cast a more detailed ballot.  Perhaps you'd like 
your ballot to be interpreted as:

  Instant Runoff    = +100
  all other methods = -100

or

  1. Instant Runoff
  2. None of the rest

There's still time to amend your ballot to rank or rate more than one 
method.  (I find it quite odd that after all this discussion, you'd 
cast a vote-for-only-one ballot.)

>     At this time I would like to ask all members of this list the
>following question: Do you believe one hundred percent that
>Condorcet has made its case? If not - then I urge you to also vote
>for Instant Runoff. 

How about some symmetry?  Has a 100% case been made for Instant
Runoff?  If not, why not urge us to vote against Instant Runoff too?

Even Rob Ritchie of CV&D, who only talks about Instant Runoff, admits
when asked that Condorcet is technically superior.  His concern with
Condorcet is that its definition may be too complex for the
electorate to understand, which could make it more difficult to win
the initiative campaign for it.  In other words, he prefers Instant 
Runoff to Condorcet on the standard "Simplicity of Definition."

Here's a quote of a message Rob Ritchie posted in the ER list:

  >Date:          19 Mar 96 10:21:50 EST
  >From:          Robert D Richie <75377.1623 at compuserve.com>
  >Subject:       Re: [ER] Re: Third parties &
  >To:            Elections_reform <Elections-reform at igc.apc.org>
  -snip-
  >I want to stress that I would never oppose a Condorcet method, if it
  >was proposed as a bill or on the ballot.  But I wouldn't expect it to
  >succeed or really even get off the ground for any election of
  >consequence until we have a transitional reform like the instant
  >run-off.

I think Rob's making a mistake about it being significantly easier
to win the Instant Runoff reform.  Also, it may be harder to make 
the transition from Instant Runoff to Condorcet, since so much money 
will have been spent on the first (unnecessary) step, and the voters 
will not like being asked to make another change--they may lose faith 
in the reformers who would then be saying to them, in effect, "we 
would have proposed Condorcet earlier but we felt you were too 
dim-witted to understand that Condorcet is better than Plurality."

>With your votes we can sweep this election.  Right now it is a three
>way tie between Instant Runoff, Steve and Mike.

Depends on how our ballots are tallied.  Are you suggesting tallying 
them using Plurality??  

By the way, Steve and Mike are people, not methods.  :-)

>     I'm content that Condorcet has not made the case on why
>Condorcet is superior to Instant Runoff. Condorcet requires us to
>make a giant leap of faith in order for us to go from Instant Runoff
>to Condorcet.

I don't believe I've made a leap of faith about its superiority. 

Instant Runoff suffers from the fallacy that being ranked second is
a poor ranking.  The truth is that many voters may like their second
ranked choice (or third, etc.) very much, and the more candidates are
freed to run by a good nonfratricidal electoral method the more
likely this is.

Here's the example I posted in March after Dole sewed up the
Republican nomination.  All we know for sure are the vote-for-only-
one ballots that were cast.  But based on what I'd been reading it
appeared to me there was a strong possibility that Alexander would
have beaten Dole if only the two had been competing, and Dole's "V8"
group of advisors were worried about this too--they convinced the
Dole campaign to target Alexander heavily with negative ads to try to
ensure that Alexander never outpolled Dole in any state.

Let's speculate the following Republican voters' rankings:
  30 vote: 1 Dole       2 Alexander
  26 vote: 1 Buchanan   2 Alexander
  24 vote: 1 Forbes     2 Alexander
  12 vote: 1 Alexander  2 Dole
   8 vote: 1 Alexander  2 Forbes    3 Dole
 ----
 100

  The six pairings:
  Buchanan beaten by Dole        26 to 50
  Forbes beaten by Dole          32 to 42
  Buchanan beaten by Forbes      26 to 32
  Dole beaten by Alexander       30 to 70
  Buchanan beaten by Alexander   26 to 74
  Forbes beaten by Alexander     24 to 76

  Alexander crushes all the other candidates in head-to-head 
  pairings, by virtue of his second place standing in the speculated
  ballots.

The plurality and runoff systems would have eliminated Alexander
from contention right away--which is exactly what happened to him in
the vote-for-only-one Republican primaries even though his "favorable
rating" was higher than any other candidate.  The system didn't ask
the voters to express their second choice; the system gagged them and
confronted them with dilemmas.

What would Instant Runoff (MPV) do with the same ballots?  The MPV
algorithm is more widely known among reformers, since it's the
trivial case of the popular STV proportional representation
algorithm.  It works by using the voters' rankings to simulate a
sequence of runoffs, with each runoff round eliminating the weakest
one of the remaining candidates (attrition).  In each round, a
voter's vote goes to the one of the remaining candidates who the
voter ranked highest.

Here's the same example, tallied with MPV:
   Round 1:  
      Dole      30  
      Buchanan  26  
      Forbes    24  
      Alexander 20 (12+8)
      ==> Alexander eliminated, since he has fewest first rank votes.
   Round 2: Votes of voters who ranked Alexander 1st are counted for
   their second choices (Dole or Forbes in this example):
      Dole      42 (30+12)
      Buchanan  26 
      Forbes    32 (24+8)
      ==> Buchanan eliminated.
   Round 3: Since Buchanan voters left unranked the remaining
   two candidates in this example, their votes are wasted, not
   transferred:
      Dole      42  
      Forbes    32 
      ==> Forbes eliminated.
   Dole is declared the winner.  Alexander is declared a big loser.

This illustrates that the MPV algorithm has some of the same
drawbacks as plurality and runoff.  And because it can fail to carry
out the will of the people, it's also subject to voter dilemmas,
though not as badly as plurality and runoff.  The Pairwise method
works much better than MPV.

>     Condorcet is a belief - you must believe - like a cult.

I guess I didn't realize what I was doing when I decided to play 
devil's advocate.  :-)

>     I have read all information that has been made available on
>this list and the CVD web site. The arguments do not compute. Simply
>put Condorcet argues that if neither of the two lead candidates has
>a majority then the third candidate must be the winner.  This poor
>logic can be found on the CVD web site under Condorcet's Method via
>Alternative Methods.

CVD's website is atrocious, and your statement that "the third 
candidate must be the winner" indicates you don't understand how it 
works.  Can we test this?  Would you please describe how both Instant 
Runoff and Condorcet work?

It's true that in the popular Dole/Clinton/Nader example, Clinton
has fewer first-ranks than Dole or Nader.  But that's isn't what
gives Clinton the win.  What gives him the win *in that example* is
that if he were paired against Dole in a runoff he'd beat Dole and
if he were paired against Nader in a runoff he'd beat Nader:

  45: Dole, Clinton, Nader
  20: Clinton
  35: Nader, Clinton, Dole

  The Clinton vs Dole "runoff":  Clinton 55, Dole 45
  The Clinton vs Nader "runoff": Clinton 65, Nader 35
  He doesn't win because he's the 3rd candidate; he wins because the 
  voters prefer him.

Now try tallying those ballots with Instant Runoff:
  Round 1:  Clinton is eliminated with only 20 first-rank votes.
  The ballots become:
     45: Dole, Nader
     20: thrown away
     35: Nader, Dole
  Round 2:  Nader is eliminated.  Dole wins.

Is Instant Runoff really the method electoral reformers ought to 
unite around?

>     Sometimes Condorcet alludes to the term True Selection of the
>People as if that is its standard - but there is no standard. If
>there were a standard we would have the right to ask - Why not use
>the standard as our election method and drop both Condorcet and
>Instant Runoff?

I'll assume you're referring to the Majority Rule standard, which 
states that a majority should outweigh a minority.

I don't understand what you mean by using a standard as a method.
Care to elaborate?

>     The bottom line is that Condorcet is using the results of the
>Condorcet Method as its standard and proof that Condorcet is the
>best.  Condorcet position is that the Condorcet method must be the
>best because it yields results that look more like Condorcet than
>anything else. This logic is similar to saying that a hippopotamus
>is called a hippopotamus because it looks more like a hippopotamus
>than anything else.

I encountered that phenomenon when I was trying to come up with 
wording for a rigorous criterion for the "lesser of evils" dilemma 
from the perspective of the voter.  I wasn't able to come up with a 
wording which didn't appear to me to arbitrarily favor a method.  
(That doesn't mean it's not possible, but I failed after a couple 
nights thinking about it.)  So I didn't bother to post the criterion.

There have been some other criteria posted which may superficially 
appear to stack the deck in favor of Condorcet.  I don't believe 
that's what is entailed, though.  I suggest that Donald get more 
specific and provide quotes to show what he is objecting to.

>     For the sake of making another point let us assume that
>Condorcet does yet prove it is the best - like some of you believe.
>Do you think that the people are going to buy into the Numerology of
>Condorcet?  I Think Not!  That would be too tough a sell.

Why?  Proof would be proof.  Why do you think they'd sooner buy into
a worse method instead of the best?

>     Therefore I urge both sides of this debate to vote for Instant Runoff
>and let us get on with Election Reform.

After all that, I don't see any clue why you prefer Instant Runoff.

Even the method's new name is bad.  For one thing, it ought to be
plural--Instant Runoffs--since it's an iterative method eliminating
one at a time.  For another, I've been told by various people that
the name Instant Runoffs makes more sense for pairwise methods like
Condorcet, since a runoff usually connotes a 2-candidate runoff and
pairwise methods calculate all the possible 2-candidate runoffs.

>Signed: Donald Eric Davison of The Election Reform Gospel According to New
>Democracy at http://www.mich.com/~donald

--Steve

P.S.-  Here's an old message from the ER list.  To the best of my
recollection there was no reply from Don until now.  Maybe now we can 
resume the dialog and find out why our understanding of Condorcet 
(and maybe Instant Runoff?) is so dramatically different.

-----------Forwarded message follows-------------
Date:          Mon, 25 Mar 1996 10:43:58 -800
From:          "Steve Eppley" <seppley at alumni.caltech.edu>
Subject:       Re: [ER] Runoff vs Condorcet etc.
To:            elections-reform at igc.apc.org

Don Davison wrote about Condorcet (and Approval):
[snip]
>Assumption Number One: These two methods depend on the assumption
>that all selections have as much weight as the first selection. I
>say "No-No" that is not true.  My one vote is separate from my list
>of selections. My first selection gets the entire weight of my one
>vote. My second selection has no weight - zero - nothing.
[snip]

So you want your other preferences to be ignored by the tallying
method.  Not I.  I recognize that when Condorcet's method is used,
it won't hurt my first choice at all if I specify my lower ranked
preferences too.  My first choice gets my full strength vote.

Example ballot:   {1 Nader  2 Clinton  3 Dole}

This ballot is a full strength vote for Nader over Clinton and for
Nader over Dole.  It's no less a vote for Nader than the truncated
ballot {1 Nader} or the vote-for-only-one ballot {Nader}.

It's also a vote for Clinton over Dole.  Using Condorcet's method,
this helps Clinton beat Dole but doesn't help Clinton beat Nader.

Your assumption #1 is true of Approval, but not Condorcet's method.

>     The right to vote means the right to differentiate. When you
>regard all my selections as being equal you take away my right to
>differentiate - you take away my right to vote.

You misunderstand how Condorcet's method works.  It does just the
opposite of your claim.  It lets you differentiate and makes your
expression matter.

>Assumption Number Two: These two methods assume and are depended
>upon the voter making more than one selection - or must do so - or
>shall be commanded to do so. If I ever find myself voting in some
>election in which one of these two methods will be used to compute
>the vote I will only make one selection. (we won't dance - you can't
>make us)  When everyone realizes that it is not to their advantage to
>make more than one selection these two methods will be rendered
>ineffectual - a non event - moot.

Again, you're incorrect.  Condorcet's method does just the opposite.
It will be to the voter's advantage to rank more than one candidate.
The key point you need to grasp is that it will not risk your
favorite's chances of winning if you include additional choices
ranked afterward.

The second key point is that it won't increase the risk of your
lesser-of-evils choice getting beaten by the worse-of-evils if you
choose to rank your real favorite first.

>                     _Automatic_Runoff_Feature_
>                    /                            \
>                   /                              \
>                  /                                \
>                 /                                  \
>                /                                    \
>               /                                      \
>  ____Current_/                                        \_Condorcet____
>         <--  <--  <--  <--  <--  <--  <--  <--  <--  <--  <--
>                        +---------------------+
>                        |  Go Back to Start   |
>                        |   Do Not Pass Go    |
>                        | Do Not Collect $200 |
>                        +---------------------+

Your chart has incorrectly switched the positions of Condorcet and
Automatic Runoff.

>These two methods are based on false assumptions - they are not
>improvements - they are not '_ _ more sophisticated systems _ _".
>Foil: (At best they are misinformation)

The misinformation is yours.  Let's continue this dialogue to clear
it up.

>Just because more people selected Alexander as the Best Man for this
>concocted wedding does not mean that he should get to go on the
>Honeymoon.

A bad analogy.  You're using "best man" in two different ways.  It
would be more precise to say that Alexander is selected by the people
to be the Bridegroom, yet gets cuckolded.

--Steve