Condorect sub-cycle rule
David Marsay
djmarsay at dra.hmg.gb
Mon Sep 29 08:06:29 PDT 1997
Dear Markus,
Thank you for your responses. You inspired me to look through
the archives more thoroughly. I still don't understand the sub-cycle
rule, though. First, do you have a reference to Tideman, MIIAC and
GITC?
I wont attempt the original example. Here's a simpler one.
Consider options A,B,C,D, where A>B>>C>>>A, D>B>>C>>>D, and A~D,
where more >s indicates a larger majority. Then ABCD is a cycle with
symmetric sub-cycles ABC, DBC. Mike's sub-cycle rule that you quote
allows to to pick either sub-cycle arbitrarily. You claim that this
makes no difference.
The way I see it, if I start with the sub-cycle ABC I have
A>B>>C>>>A, for which A>B is the weakest link, and the
Condorcet-winner is B. Thus I delete A and C, leaving:
D>B, so D is the overall winner.
Now I repeat, but with the other sub-cycle. By symmetry, A is now the
winner.
The only way around this that I can see is to look at the whole
cycle.
I understand from a DEMOREP1 posting of 26 July 1996 that Young
cites Condorcet as considering all cycles together. This gives B.
However, if one considers all cycles together, then a strong
sub-cycle A>>>B>>>C>>A will effectively destroy all information of a
lesser strength. This seems undesirable.
It seems to me that one needs to consider all sub-cycles of the same
cycle together. This notion is given a meaning as follows:
Note that if a relation has cycles, then there exists a unique set of
basis cycles. These are disjoint connected cycles, such that every
cycle lies within a basis cycle.
(To construct, choose any cycle and put it into the set. For each
remaining cycle, r, in turn, for each cycle s in the set that it
intersects, add s to r, and put r in the set).
Now, where the basic Condorcet procedure gives cycles, form the basis
cycles. These correspond to Condorcet's "impossible opinions". For
each basis cycle, find the least threshold T such that discounting
all members of the cycle with Condorcet-plurality of less than T
removes the cycles (creates possible opinions).
This sub-cycle rule is now rather like the one I
posted earlier.
--------------------------------------------------
Sorry folks, but apparently I have to do this. :-(
The views expressed above are entirely those of the writer
and do not represent the views, policy or understanding of
any other person or official body.
More information about the Election-Methods
mailing list