[EM] Condorcet's strategy problem
nkklrp at hotmail.com
Tue Sep 13 20:13:06 PDT 2005
First, Condorcet's method, as the term is used here, is too broad a
category. It refers to all the pairwisei-count methods that solve circular
ties based on dropping weakest defeats or keeping strongest defeats. So I'd
rather just discuss the wv Condorcet methods. I'll refer to them as "wv".
As Warren said, wv's problems in public political elections are enough to
recommend other methods instead, for public political elections. WV's
problem in public political elections is its FBC failure. In public
political elections, only methods that meet FBC should be proposed.
There's been discussion here about how often wv fails FBC. But, you know, it
doesn't even matter how often. The mere fact that Worst can win because you
didn't bury Favorite under Compromise is enough to make people bury
Favorite. I've reached that conclusion after conversations with
In fact I myself would do favorite-burial in public political wv elections.
It's been some time since I posted about this, so let me know if I say
The WDSC compliance of all wv methods means that, with those methods, a
majority who prefer X to Y never need to reverse a preference in order to
make Y lose. But are you sure that a majority agree with you in preferring X
to Y? The trouble with the majority defensive strategy criteria is that
their guarantee depends on such a majority. Such a guarazntee just isn't
good enough for lesser-of-2-evils progressives. And, in public political
elections, it isn't good enough for me either.
Though I don't agree with the progressive lesser-of-2-evils voters about
what is acceptable, I'd protect my acceptable candidates by favorite-burial,
just as they do now, and would with wv.
Say that the candidates are in two sets such that the merit differences
within those two sets are negligible compared to the merit difference
between those two sets.
I've called that the "acceptable/unacceptale" situation. For me our
elections are an acceptable/unacceptable situation. For the
lesser-of-2-evils progressives that seems to likewise be the case.
The only difference is that, for me, the Democrats are among the
unacceptables, while, for the lesser-of-2-evils progressives, the Democrats
are (unaccountably) among the acceptables.
How do you vote in wv, in an acceptable/unacceptable situation?You vote in
whatever way maximizes the probability that no unnaceptable will win. Maybe
you should rank the acceptables in order of winnability, with the AERLO line
applied just below the acceptables.
The acceptable/unacceptable situation simplifies Condorcet strategy, but
maybe not as much as one might like. Ranking the acceptables in order of
winnability sounds right, but maybe it's an oversimplification, and there
should be some other (closely-related) measure by which the acceptables
should be ranked.
Maybe rank the unacceptables too, in order to increase their votes-against,
if the method is PC, or to disfavor them more with regard to the strength of
beatpaths against them, if the method is BeatpathWinner. And apply ATLO
between the acceptables and unacceptables. If the unacceptables are ranked,
they probably should be ranked in reverse order of winnability, though,
again, that might be an oversimplification.
Yes that's a lot of maybe. Anyway, the problem with that is that those who
think someone like Kerry or Dean is acceptable are likely to rank them alone
in 1st place, even though they prefer Nader. I don't want to make them have
to do that. So please, let's not propose an FBC-failing method for puiblic
Aside from all that, and as I've already said recently, RV is completely
familiar and liked. We've all been asked to rate things up to 10. There are
innumerable ways to count rank ballots, and most of them--I could probably
say nearly all of them--are no good. So, though there are some
newly-described rank methods that meet FBC, and I might rather have one of
them enacted than RV, I insist that RV is a much more winnable public
proposal. By far the most winnable of the few FBC-complying methods, and
much more winnable than any rank method.
As I said, I believe that BeatpathWinner or CSSD is great for committees and
organizations, where there's no frightening "greater-evil", so that FBC
isn't so necessary.
But then there's the question of whether you want the best method for that
application (your committee or organization), or whether you want to
demonstrate and build precedent for a method that could be winnable with the
public. If it's the latter, then RV is a better choice.
Of course one could do one's organizational elections with 2 ballotings--an
RV balloting and a BeatpathWinner balloting. The RV balloting demonstrates
RV, and gives it some precedent, while the BeatpathWinner balloting brings
SFC compliance, and its winner is the organization's actual choice.
I admit that using RV, but not going by its result, doesn't make for as
strong a precedent for RV. That would of course be a reason to use RV
exclusively, and let its winner be the organization's choice.
It depends on how you want to balance 1)The properties desired for your own
application vs 2)Setting precedent for a good and winnable publc method.
Express yourself instantly with MSN Messenger! Download today - it's FREE!
More information about the Election-Methods