replying to Demorep, 4/14/96

[Mike Ossipoff]

DEMOREP1 at aol.com writes:
> 
> Recent postings indicate the importance of getting a majority rule winner in
> single winner elections using the Condorcet method.

If you mean a majority winner, there usually isn't one. But Condorcet's
method does, as you say, carry out majority rule better than other
methods. And it does so even if there isn't a separate disapproval
vote. I have no objection to a disapproval vote, or to a NOTB alternative,
but I claim that Condorcet already would be very unlikely to elect someone
who could get disapproved by a majority.

> Namely, (a) a majority disapproval vote will get rid of candidates not wanted
> by an overall majority of the voters, and 

As I said, Condorcet does an excellent job of doing that anyway, due to
its count of "votes-against".

> (b) a candidate who is defeated head to head with each other candidate should
> lose (i.e. is defeated by all head to head majorities) (the recent 3

Some say that. "Should" is a strong word. I have no objection to
your rule to exclude Condorcet losers, and of course it would be nice
to do so as long as it's done without losing Condorcet's other important
properties. I'd include that Condorcet Loser exclusion rule if it's
important for public acceptance, or to avoid criticism.

> Apples-Chocolate example).
> There may be some overlapping in the above obviously but (b) should exist as
> a second majority defense.
> 
> There are problems with strategic truncated voting.

Oh yes there indeed are. In Copeland's method there certainly are. But
in Condorcet's method there aren't. Thank you for re-emphasizing that
distinguishing feature of Copeland.

> Extreme Example
> 499 A, B
> 2 B
> 500 C

In your extreme example, as in less extreme ones, with Condorcet's
method, the truncation doesn't accomplish anything for its perpetrators.
C doesn't win. In fact, the Condorcet winner B wins. So where's the
problem, Demorep, unless you really are talking about a serious problem
of Copeland's method. Of course in Regular-Champion, the version of
Copeland that has been proposed for use as a public proposal, by somoene
on this list, C wins--With Regular-Champion the truncation works.

So Demorep & other critics of Pairwise methods, are quite right to say
that truncation is a problem for most Pairwise methods. But not for
Condorcet. In particular, as Demorep's example so well shows, 
truncation is a problem for Regular Champion. Actually it's a problem
for other Copeland versions too.


> (The general extreme case is N, N-1 and 2 votes- a total of 2N + 1 votes)
> Note- If there is a disapproval vote first, then -
> the A and B voters combined could/would defeat C 501 to 500,

In Condorcet, the A & B voters combine to put C out of the running.

> the B and C voters combined could/would defeat A 502 to 499 and
> the C votes could/would not defeat B (due to second choice votes of A voters)
> 500 to 501 (i.e. B wins).

No, this doesn't require a disapproval vote--it's a natural & obvious
consequence of Condorcet's votes-against count.

> 
> Note that the B voters do not vote for A (a one sided coalition).


What's your point?

> Note that if 2 A voters truncate their ballots, that even B could/would be
> defeated 500 t0 499.

As I've so often told Demorep, if the extreme voters who need the compromise
refuse to vote for the compromise, then no method can help them.

> 
> If there is not a disapproval vote first, then
> A beats B 499 to 2 margin 497
> B beats C 501 to 500 margin 1
> C beats A 500 to 499 margin 1
> or a circular tie A>B>C>A.

What method are you talking about, Demorep? With Condorcet's method
there is indeed a circular tie, & the Condorcet winner, B, is then
Chosen by Condorcet's circular tie solution.

> 
> Thus, C and A are each beat by 1 vote in their (only) defeats.

"Young's method" counts margins of defeat as you are doing. Condorcet's
method & Young's method are two completely different methods. If you're
trying to show that we shouldn't choose Young's method to promote to the
public then ok.

> C beats A (assuming that the candidates with the equal lowest margins go head
> to head again).

You assume a lot, Demorep. You make up lots of rules of your own, &
you end up with a completely original method. That's fine, to propose
an original method, but your're incorrect if you call it Condorcet's
method.

> 
> To discourage truncated voting (as in the above Extreme Example) I again
> suggest that after a disapproval vote that each tie breaking round using head
> to head Condorcet that the candidate with the lowest number of votes (first
> choices plus additional choices) should lose. 

Truncation needn't be discouraged in Condorcet's method, since it
gains nothing for the truncators anyway. But in any rank-balloting vote,
there'll be much truncation, so what we don't want is a method that's
vulnerable to truncation, as Copeland's method is.

> 
> If there is less truncated voting (i.e. more additional choices voted), then
> a candidate with a higher plurality/majority would win.

By an original proposal of yours?

>  
> In the example above, the first choice B voters (and even possibly the first
> choice C voters) might be encouraged to vote second choices (possibly causing
> A and C to each survive a majority disapproval test).
> 
> In a tie breaker situation among the candidates of a majority coalition, it
> would seem that the candidate of the coalition with the lowest number of
> votes should lose.

Here I take it that Demorep is proposing a Plurality tie-breaker. I've
proposed Plurality for solving the "pairwise ties" that can happen
in small committee elections, but not in public elections, but Plurality
has no place as a circular tie solution.

> 
> Note the above 2 against 1 example can be expanded- 3 against 2, 5 against 4,
> etc. 
> 
> I note again that one of the chief competitors to Condorcet is the top 2
> runoff method (top 2 in a primary election go to a runoff primary election
> (as used in several partisan elections in the U.S.) or to a general election
> (as used in most nonpartisan elections in the U.S.) which respectively
> produce a majority partisan nominee or a majority nonpartisan winner). 
> 
> Condorcet avoids the defect of the top 2 runoff method that a eliminated
> choice (third or lower) might beat each of the top 2 in head to head
> pairings).
> However, the Condorcet expanded remedy must try to maximize the possibility
> that the winner is a majority of all voters winner.

Very few methods would fail to elect a candidate who is ranked #1 by

a majority. Condorcet, Copeland, Runoff, Shugart's Modified Runoff,
would all elect that candidate with a majority, if members of that
majority voted him alone in 1st place. The trouble is that with
Copeland & the Runoff methods, voters will sometimes feel forced
to _not_ vote their favorite alone in 1st place. And when that happens,
the favorite of a majority can lose in Copeland & Runoff. So any 
claim that the favorite of a majority will win in Copeland is based
on the false assumption that voters will rank him alone in 1st place.

> .-
> 


--