Condorcet, Thin Ice, and Battle Cries

Hugh Tobin htobin at redstone.net
Sat Apr 26 22:45:47 PDT 1997


New Democracy wrote:
> 
> Dear members,
> 
> Hugh Tobin wrote:
>       It is not a matter of transferring votes, just a rule that if A
> is ranked higher than B on a ballot, then A gets one vote against B.
> 
> Don writes:
>      You are on thin ice Hugh - you are saying "by defination" it is not a
> matter of transferring votes.
> 

Actually, it seems I am swimming upstream, trying to convey a simple
definition of a Condorcet system.  We must first understand how a system
works, before we can compare its merits against another -- for example,
against an iterative system that truly involves transfer of votes, in
the sense that they go to only one place at a time and never come back. 
It is possible to describe the Condorcet system using the terminology of
transfers of votes, just as it is possible to describe the movements of
celestial bodies using an Earth-centric system with Ptolemaic epicycles.
In either case it is unnecessarily complex and confusing to do so. 

>      But the point is minor - the major problem with Condorcet is the same
> - voters are being asked to give support to lower preferences when their
> number one preference is still a contender - people will not like to do
> this - enough will rebel to render Condorcet moot.
> 

The voters are not being "asked" to do anything -- in Condorcet they are
being allowed the right to vote all of their true preferences, without
the pressure to vote for a lesser evil over their true favorite.  This
is hardly reason to "rebel" at Condorcet. Some voters may "truncate" in
Condorcet, perhaps because they are misinformed as to the effects of
doing so (e.g., by reading some postings on this list), and this will
reduce the voting power of those voters to an extent.  But it is
patronizing to the citizenry to reject a system on the supposition that
voters will irrationally fail to vote their true interests. I would not
assume that those voters who actually think about how the system works,
rather than simply follow instructions by ranking their preferences,
nonetheless are so dense that they cannot distinguish Condorcet from a
system in which one can give maximum support to one's first choice only
by failing to name a second choice (e.g., a system that give declining
numbers of points to each ranking and then awards victory to the highest
point total). 
If Condorcet is implemented and explained, most voters with preferences
will express them on the ballot, either because they will not consider
the algorithm at all or because they will understand that it is in their
interests to vote a full ranking, even though their first choice does
not have to be eliminated before the lower rankings can have any
significance.

But perhaps you mean that Condorcet cannot be sold because voters who
understand the system will be uneasy at the possibility, remote as it
may be in most cases, that a vote for a lower choice will adversely
affect the chances of a higher choice in Condorcet.  Those voters should
be able to see that their votes for second choice in Condorcet also
could help their first choice win a circular tie against a lower choice
-- so the fact that these votes are considered before that first choice
is eliminated can be a benefit.  More important, they should also be
able to understand the serious defects in the alternatives -- and to
rebel, for example, against run-off on the ground that it may force them
to choose the lesser of two evils for their first-place votes, even
though their favorite would beat each of those evils. 

>      I am thinking that I will adopt a battle cry:
> 
>          "Secret Ballots - Secret Preferences"
> 
>      If you believe in the secret ballot then maybe you should also believe
> in the secret lower preferences. This can be done by the computer software
> - the program can be written so that it will not reveal lower preferences
> until the first preference candidate is no longer a contender.
> 
>

Of course ballots should be secret.  There is no problem in maintaining
secrecy in Condorcet.  No individual's preference needs to be "revealed"
at any time.
But I presume by "not reveal" you mean that the computer should not
consider the lower choices.  The problems you create in implementing
this principle are (1) that you will sometimes eliminate the candidate
who should have won -- i.e., who beats each other candidate, and (2)
that you will provide incentives to voters to falsify even their first
choices.  I have previously shown a plausible example where, in "instant
run-off," one might rationally vote for one's last choice first.  Never
in Condorcet.

No system can guarantee that the voter will always do best by voting all
his true preferences, so the mere possibility in Condorcet that a true
second-place vote might impair the chances of the first choice is not a
basis to reject Condorcet.  This is particularly true because in the
general case for Condorcet, when either a circular tie is unlikely or
the voter has insufficient information about the votes of others in
order to make a tactical choice (or both), the voter's best interest
does lie in casting a full and true ranking.

You seem to assume there is some inherent value to a sequential process,
like a run-off, even though a candidate may be eliminated before enough
voting data has been evaluated to determine whether there is any other
candidate that is preferred by a majority of those who have a preference
between the two.  But process is not an end in itself.  What we need is
an algorithm that produces democratic results from the preferences
expressed by the voters.

You can think of Condorcet as a system in which lower preferences are
not considered unless and until the first choices have been tabulated
and it has been determined that no candidate has a majority of them.
Once is has been determined that there is no "beats all" winner, it is
"time" to consider all the pairwise races and see if there is a "beats
each" winner.

Battle cries about the mechanics of the voting process are not
substitutes for analysis of the consistency of outcomes with democratic
principles and of the incentives created by alternative systems. 

Regards,
-- Hugh Tobin

> 
> Donald Eric Davison of New Democracy at http://www.mich.com/~donald
> 
> - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



More information about the Election-Methods mailing list