Bruce, Condorcet & Copeland
dfb at bbs.cruzio.com
Wed Apr 3 20:25:12 PST 1996
I couldn't include a copy of Bruce's letter with this reply, because
when I try to write in such a copy of those letters, the lines get
scrambled & duplicated, probably because of the messages' line-length,
which results in "wraparound", and probably is what causes my e-mail
editor to go haywire when replying in a copy of those letters. Bruce,
could you shorten your line-length?
But this is a brief reply anyway, to just a few comments.
Sure, we can first discuss the things that we're likely to agree on,
but, unless you want my statements to go unchallenged, on subjects
that we're less likely to agree on, it would of course be necessary
for you to send & reply to arguments on those subjects too. Because
until you do, those statements of mine will remain unchallenged, and
my arguments will remain unanswered.
You said that there's at least 1 method that's better & simpler
than Smith//Condorcet. Bit it's kind of noticeable that you neglected
to say what that method is...
So, if there's a method that's simpler & better than Smith//Condorcet,
would you mind telling us what it is?
There are simpler methods than Smith//Condorcet, and though I
haven't heard of any better methods proposed, I've talked about
extra refinements that could make that method better still (though
I wouldn't propose them to the public in an initial proposal). But
why don't you tell us what method is both better & simpler?
Of course I wouldn't object to Smith//Condorcet//Plurality. Plurality
is the tie-breaker that I'd add, if I were going to add a tie-breaker.
But I don't consider the addition of a tie-breaker to be necessary or
important, unless someone said that for some reason custom requires it.
Condorcet's method is virtually never going to return a tie in a public
election. Returning a tie in a publi election, and requiring a tie-breaker,
is a trick that's pretty much exclusive to Copeland (and some other methods
that none of us have proposed, except for an initial narrowing of
choices, as in Smith//Condorcet).
Though tie-breakers are very important to have in small committee
elections, it isn't really necessary for Smith//Condorcet to have
one in public elections, though I have no objection to
You defined "//" in terms of rank-balloting methods only. I've
used it with a slightly looser meaning, because I've said that,
by BeatsAll//Approval, I mean that if no 1 alternative beats each
one of the others, then a 2nd balloting should be held, by Approval.
Other than that, I use "//" with the same meaning that you do.
And I agree that it isn't good to call Smith//Condorcet "Smith-Condorcet",
because the hyphen, as you said, should be reserved for methods
devised by 2 people, such as the Simpson-Kramer method. I was just
avoiding the unfamiliar "//" notation on ER.
About the name of MPV, I've been calling it Instant Runoff, because
that's what its main proponents have begun calling it. I usually adopt
the names used by proponents. True, Hare (as I'll call it in this
letter) doesn't do exactly what Runoff does, but the name "MPV" is
especially inappropriate, considering what Hare does to majority rule.
I like the other name that you (maybe jokingly) proposed for Hare:
"NMPV" (Non-Monotonic Preferential Voting)
But I feel that we're stuck with the name used by Hare's proponents,
so we might as well get used to IRO.
"Quick & Dirty"? I _was_ being brief, because I hesitated to go
into it in detail on ER. Besides, & more important, I'd already
talked about some very basic standards on which all electoral reformers
agree, and which, judging by what they say & how they vote, are also
very important to progressive voterss. Standards that Condorcet meets,
and which Copeland fails.
I was talking about the desirability of getting rid of the lesser-of-
2-evils problem, and getting rid of the need for defensive strategy,
standards that the academic authors are strangers to.
I've talked about these standards on this list before, a lot, and
I hesitate to repeat it, for that reason. But maybe I have to now.
We'd like to be able to be fully counted against Dole, by voting
Clinton over him, even though we don't vote Clinton in 1st place.
We'd like to cast a reliable, full-strength vote for Clinton against
Dole, while also casting one for Nader over Clinton & over Dole.
That, Bruce, is what would get rid of the progressive voter's
"lesser-of-2-evils" problem. It's what would relieve the voter of
the need to use defensive strategy in order to defeat someone
I don't mean to be "quick & dirty", I resent the implication that
I haven't been forthcoming with details about the standards that
I'm referring to.
One way that I've worded this has been my "Generalized Majority Standard":
A group of voters consisting of a full majority of all the voters should
be able to get any result that they all want (electing someone or preventing
the election of 1 or more candidates), without having to rank a less-liked
alternative equal to or over a more-liked one.
And, under as broad a range of conditions as possible, we'd like
the members of that majority to be able to get what they all want
without any kind of defensive strategy at all, including truncation.
That could be made into the kind of yes/no test that could be
called a "criterion", by changing the 2nd paragraph to say:
And, unless every alternative has another alternative ranked over
it by a full majority of all the voters, then the majority group
referred to in the 1st paragraph should be able to get what they
all want without any kind of defensive strategy, including truncation.
I've already talked about how Condorcet's method satisifes the
2nd paragraph, and how it, for all practical purposes, satisfies
the 1st paragraph--and how a small modification could make it
impossible, even in principle, for it to fail the 1st paragraph.
I've also talked about why, in plusible situations where the
result really matters, every candidate won't have a full
majority against him--won't have another candidate ranked over
him by a full majority of all the voters.
Copeland doesn't have these majority-rule properties. Which means
that it doesn't do what we'd like it to, when it comes to
lesser-of-2-evils, defensive strategy, & majority rule.
And, since I'm guessing that Copeland is the method that Bruce
is going to say is simpler & better than Smith-Condorcet, I'd
like to say a little more about how Copeland picks its winner.
Though it of course uses votes to determine who beats whom, it
thereafter ignores votes, and instead counts candidates, counts
the number of candidates that a candidates beats & is beaten by.
Some of us have talked about how irrelevant that standard is.
We've described how, when voters rank candidates according to their
party, the winner of a circular tie involving several parties depends
entirely on how many candidates the various parties run. Does it
make sense to pick the winning party on that basis? It needn't be
parties, though--even in a nonpartisan race, candidates could pick
duplicate candidates to join the election, candidates with the
same policy proposals, etc. And then, if voters vote according
to, say, policy proposals, the outcome will again depend on how
many candidates of each type are running. Steve pointed out that
running more candidates costs more money; the winner of an election
shouldn't depend on which parties have the money to run more
Bruce hasn't answered that argument, or the related one that
supposes that we're going to vote by Copeland on what movie to go to,
when there are Westerns, crime movies & adventure movies. In a
circular tie, where people rank by genre, Copeland could pick a Western
for no reason other than that there are lots of crime movies &
few adventure movies. As I said, what kind of a way is that to
pick a movie genre? What kind of a way is that to pick a winning
political party, Bruce? How do you answer this?
I've also pointed out that, though Smith//Condorcet has very important
properties that Copeland doesn't have, when it comes to lesser-of-2-evils,
majority rule, & getting rid of need for defensive strategy,
Smith//Condorcet also meets every objective criterion & standard
met by Copeland. That means that, when it comes to standards, criteria
& properties, Smith//Condorcet has Copeland completely dominated.
You never answered that, Bruce. What do you say to that?
Realizing that Copeland can't be defended in a comparison with
Smith//Condorcet, I invite you to have a go at it.
And after I've talked about the academic criteria & Condorcet,
it will be evident that even plain Condorcet is better than
Before I quit for now, I'll just mention the example that I've
35%: Nader, Clinton
What does Regular-Champion do in this situation? It rewards the
truncation of the Dole voters by picking Dole, instead of
Condorcet winner Clinton.
Condorcet's method would pick Clinton. Copeland is vulnerable to
truncation; Condorcet isn't.
In the Regular-Champion system, the Dole voters know that if they
have a plurality, they can easily steal the election by making
a circular tie.
More information about the Election-Methods