NOTR in a circular tie

Steve Eppley seppley at alumni.caltech.edu
Thu Dec 19 16:34:20 PST 1996


Demorep wrote:
>I would suggest that NOTA cannot logically be relatively ranked
>against any specific candidate.  A NOTA vote means NO candidate is
>wanted.  

That's not what I meant by NOTA.  I'm referring to a token which can
be transitively ranked by the voter to delimit approved candidates
from disapproved candidates:   BAC/ED

Perhaps I should have stuck with the term NOTB or NOTR, but I thought
the context would make it clear that by NOTA I didn't mean a voter
shorthand which indicates NO candidate is approved.

So I'll repeat my point that approval and disapproval are really
*relative comparisons", not absolute ratings:  Approval of a
candidate is a *relative pairwise* preference that it's better for
the candidate to be elected than for a new election to be held, and
disapproval of a candidate is a *relative pairwise* preference that
it's worse for the candidate to be elected than for a new election to
be held.  Neither is an absolute value of a candidate's utility.

>I would suggest that that the average voter would not think it is
>not logical to relatively rank NO candidate versus ANY candidate in
>any sort of combination such as X candidate > NO candidate > Y
>candidate.   

It's hard to parse that grammar, but I think I understand the gist.
Demorep appears to be suggesting that many voters are too stupid to
be able to insert a dividing line between the candidates they approve
and the candidates they disapprove.  I agree it would be confusing
if the dividing line were labeled "NO candidate", but that's not a
label I'd suggest using.  

Maybe Demorep would acknowledge the following would be more
intuitive ("logical") for voters:

        X
        Z
  --none of those below--
        Y

Maybe Demorep is right about "challenged" voters.  Maybe there's no
label which would work well; maybe the instructions would be too
hard for voters to understand.  But maybe it would be more confusing
for voters to have to vote twice in each race (once to rank candidates 
and once to approve/disapprove them) as he proposes.  Perhaps the 
voters should simply be asked to rank the candidates.  :-)

>NOTA is shorthand for voting an absolute NO on each candidate.  

I don't think we ever agreed to this, except in the context of a
"vote for only one" method.  But if this is going to be standard
language here, then change my recent references to NOTA to be NOTR
(none of the rest).  I think then Demorep may need to reread my last
message with that in mind.

>If a majority of the voters vote NO on each candidate, then NO
>candidate should be elected (see 1-1-98 example at bottom).

I agree that if the Approved//Condorcet method is used, the winner in 
that example below should be None.  

The questions for reformers are whether the best methods allow no
winner, and, if so, whether it's best to combine that with other
single-winner reform proposals or have the voters decide on the
reforms separately.

>Mr. Eppley wrote:
>>Sincere:                       
>>46: AB/                         
>>20: BA/                       
>>34: /BA                         
>>B would beat A.
>>B would beat NOTA.              
>>B would be elected.  

>>Order-Reversal:
>>46: A/B           <-- reversal
>>20: BA/
>>34: /BA
>>B beats A.
>>NOTA beats B.
>>A is elected.  Reversal succeeds.

>[D comment- 66 voters approve A.  20 voters approve B.  A wins
>without having to go head to head with B.  

Agreed.  Using Demorep's Approved//Condorcet method, reversal by A's
supporters succeeded in electing A.

>I note that the 20 BA/ voters could insincerely vote 20 B/A so that
>neither candidate gets majority approval- 46 A, 20 B.  I also note
>that the 34 /BA voters could insincerely vote 34 B/A causing B to
>win 54 to 46.]

Yes, Demorep's preferred method is a strategic mess for all the voters.  

>D- According to Mr. Arrow there are strategy problems with all
>nondictatorial election methods.  Does Mr. Eppley agree with 
>Mr. Arrow's observation ?

Yes.  Ken Arrow was a pretty sharp fellow.

But the strategy problems in some methods are significantly 
worse than in other methods.  Does Mr. Demorep agree with my
observation?

>Order reversal obviously can occur with plain Condorcet.  I would
>rather have strategy problems with majority approved candidates
>than minority approved candidates.  

Because of the strategy problems over whether to approve or
disapprove, it's false to think candidates *voted* majority-approved
are *sincerely* majority-approved, or that candidates *voted*
minority-approved are *sincerely* minority-approved.

If Demorep's Approved//Condorcet method is used, I'd expect a lot of
voters would vote to approve only their favorite and vote to disapprove
all the other candidates, especially if their candidate is well-
funded so s/he can compete successfully in a rematch.  Why settle 
for the compromise candidate if you can win by attrition?

Order reversal is unlikely to occur with Condorcet.  Condorcet's
method deters reversal because the tactic is likely to *backfire*. 
Saying it "obviously can occur" because examples can be constructed
is an overstatement of the (non)problem because of the implausible
assumptions behind the examples.  This has been discussed here many
times.  

>Mr. Eppley would apparently see nothing wrong with a sincere vote
>of-- 
>1   BA/
>1   AB/
>98  /BA 
>(i.e. 98 percent of the voters do not want either candidate but B
>wins by a Great Leader super mandate of 99 to 1 using plain
>Condorcet).  Once again, I thank Mr. Eppley for causing me to
>produce the 1-1-98 example.

Demorep is welcome.  He's misrepresenting my views, however, if he
think I'd say that B should win.  Since the Condorcet winner is '/',
no one should win.

(Round and round we go.)  Demorep thinks that acceptable candidates
won't rush to enter an important race when all the other candidates
are despised.  Not likely.  

Here's Demorep's scare scenario:

  1 Hitler /
  1 Stalin /
 98 / Hitler, Stalin
 After this vote, a new election is held and an acceptable candidate
 (say, Eisenhower) competes and is elected:
  1 Hitler /
  1 Stalin /
 98 Eisenhower / Hitler, Stalin

I don't think it's reasonable to believe an Eisenhower wouldn't have
run in the original election, if only Hitler and Stalin were 
contending.

I don't consider this unlikely scare scenario to be a good reason to
be placid about the strategy problems Approved//Condorcet would have,
and the "vacuum of power" scare tactic could undermine the passage 
of the important Condorcet reform.

>Using 2 candidates is not real since there will obviously be more
>than 2 candidates in real Condorcet elections. 

True.  I don't see Demorep's point though.  If anything, this point
should alleviate one's concern that the voters will be offered only
bad choices, none of which could defeat Hitler and Stalin pairwise.

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




More information about the Election-Methods mailing list