David Marsay djmarsay at
Wed Sep 24 05:49:04 PDT 1997

I am interested in multi-party decision making, but have no experience 
in elections.

It seems from the archives that if M is some candidate method (such 
as FPP) then Smith//M will be no worse, and will often reduce some 
of the main failings of M. Condorcet (where preference cycles are 
broken where there is a least strong majority) is an obvious 
replacement for Smith in this role. It has some problems, but 
I argue that they are acceptable. Is there some significant property 
of voting systems that I should consider?

In particular, Condorcet may declare X < Y, contradicting a pair-wise 
majority. However, it only does this where there is a chain of 
preferences from X to Y that are each stronger than the pair-wise 
preference of X over Y. If this is regarded as reasonable, then one 
must abandon Condorcet and look for a superior 'M'. (From the 
archives, this would seem to be unprofitable).

I have some ways of rationalising Condorcet. For now, let us suppose 
that above behaviour is accepted. Then it has to accepted that 
candidates that create such contrary preferences are 'relevant'. Thus 
when considering phenomena like 'stalking horses', one has to allow 
that some changes may be legitimate. I claim that the stalking-horse 
like phenomena for Condorcet are much less serious than those for 
FPP. Moreover,  considering ballots that have cyclic preferences 
shows that any method that breaks cycles must have similar problems 
to Condorcet.

Condorcet is responsive to single additional ballots. Thus cycles are 
unstable (unlike Smith). Tactical voting may break cycles, but this 
is not as serious as FPP. Moreover, I claim that tactical voting 
could easily back-fire. Considering ballots that have cyclic 
preferences shows that any method must have 
similar problems to Condorcet.

Thus Condorcet is not significantly more prone to these problems than 
other methods. Moreover, Condorcet provides some 
protection against vote-splitters and the rich party syndrome.

Condorcet//M appears entirely reasonable, M being relatively 
unimportant. It therefore seems preferable to both M and Smith//M.

>From this point of view, the answer has to be Condorcet//M, and it 
remains to find a suitable rationalisation for Condorcet.

I welcome comments, particularly on:
- other properties to consider;
- published rationalisations for Condorcet.

Also, is there a translation of Condorcet around?


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