Fairness, votes against, and Condorcet(x)

Mike Ossipoff dfb at bbs.cruzio.com
Fri Apr 19 10:41:18 PDT 1996

When you ask why not count the Clinton voters against Nader as well
as against Dole, in the circular tie solution, that means asking why
Condocet's method only looks at defeats. True, we don't have to do it
that way. The Simpson-Kramer method differs from Condorcet in just
that way:

If no 1 candidate beats each one of the others, then the winner
is the candidate over whom fewest voters have ranked the candidate
ranked over him by the most voters.

This differs from Condorcet in leaving out the words "...who beats him".

So, in that 46,20,34 example, by Simpson-Kramer, the candidates would
have the follwing scores:

Dole 54%
Clinton 46%
Nader 46%

So that count rule would be tie-prone.

And if this were an order-reversal situation, Nader might not win
that 46-46 tie, and there would be much less deterence for order-reversal.

Also, if the Dole voters only truncated, then the scores would be,
by the 2 methods:

              Condorcet          Simpson-Kramer

Dole            54                  54
Clinton         34                  46
Nader           46                  46

So, with Simpson-Kramer, Clinton doesn't necessarily win, as he
would in Condorcet. Condorcet, not Simpson-Kramer, is the one that
does the best job of electing the Condorcet winner in that
common situation.

Someone might say "So what if Clinton loses to Nader because the
Dole voters truncated? The Dole voters can't complain about that!"

No, but the Clinton voters are part of a 66% majority against Nader,
and they have a right to comlplain that they were wronged by what
happened due to the Dole voters' truncation. It would be nice if
a method, at least to the extent possible, made sure that the 
Condorcet winner wins. Why should the Clinton voters, part of
that Clinton over Nader majority, suffer because of someone else's
mistake, if it can be avoided. Besides, maybe only some of the Dole
voters would truncate, resulting in a Nader victory. Not only they,
but the non-truncating Dole voters, would suffer for that.

That's why I like Condorcet better than Simpson-Kramer. One reason
anyway. There's another reason. Say there's a big circular tie, and,
in accordance with what Mr. Condorcet suggested, we ignore the
smallest defeat, breaking that circular chain. Guess who's now at
the top of that newly-linear chain? The candidate whose defeat, the
smallest one, was discounted. For this reason Condorcet's method,
based on defeats, makes more sense.


Now, as for what the Clinton voters want, they apparently don't want
anything, when it comes to Dole vs Nader. So we needn't worry about
violating their wishes (unless we, as Bruce proposes, attribute to them
preferences that they never voted, when using a count method in which
that changes the election result).

So if their truncation is sincere, then they don't care.

But if their truncation is an anti-order-reversal counterstrategy,
then they are truncating in order to ensure that the Dole voters
can't make Nader be beaten by a majority (and the Nader voters couldn't
make Dole be beaten by a majority either). They're doing it to make
order-reversal by either extreme a reliable losing strategy.


So, in the sincere instance they don't care, and in the insincere
instance, they do care, and they want what Condorcet does.


But it isn't just in order-reversal situations where it matters--
it matters in the much more common trunction situations (It seems to
me that truncation took place in every rank-balloting election I've
conducted or participated in).

And Condorcet's way of counting is what keeps the Dole voters
from stealing the election, intentionally or otherwise, by their
truncation. As I said, Condorcet, not Simpson-Kramer, is the method
that better ensures that the Condorcet winner will win, avoiding
unnecessary dissatisfaction for members of majorities.


And, even if all the rankings are sincere, and the Dole voters really
like Nader better than Clinton (!), then it's still fair for Nader to
win, and it still makes sense, because Nader is the only candidate
who doesn't have a majority saying that another candidate is better.


You said that the Clinton voters can punish the order-reversal of
the Dole voters by insincerely ranking Nader over Dole. Right, any
method other than Condorcet's method makes it necessary to vote
insincerely in order to counter order-reversal--or even truncation,
which doesn't require _any_ counterstrategy in Condorcet. 


Why use x = 0 instead of .5 or 1?  Because to do otherwise is to
count people as voting preferences that they never voted, which is
bad news in methods where that can change the election result.
Someone said any method where that changes the outcome isn't a
good method. Oh? A method isn't good if its result changes when we
falsify people's preferences, counting un-expressed preferences?


But, as I said, it really comes down to what we want from a method.
As I said, I'd hope that we'd use a method that wouldn't make voters
feel forced to falsify their preferences, to vote insincerely, to
use defensive strategy, to abandon their favorite to save a lesser-evil.

If you agree about that goal, then you agree with me that Condorcet's
method is the best.




More information about the Election-Methods mailing list