Bad Condorcet winners?

Richard Moore rmoore4 at home.com
Wed Mar 21 22:08:40 PST 2001


DEMOREP1 at aol.com wrote:

> Mr. Simmons wrote in part--
>
> > Demorep said:
> >
> > >It is not *average* utilities that are important.
> >
>
> I think Demorep has a point here.  Instead of optimizing average SU
> someone might want to optimize median SU, or most likely SU, or minimize
> the likelihood of SU below some cutoff value, etc. especially if the short
> run outcome is very critical.

Well, this points out one "problem" with social utility. If you ask people
to rate candidates on a utility scale from 0 to 100, people will tend to
put their favorite candidate at 100 and their least favorite at 0. A utility
of 100 could actually be assigned by a voter to a candidate he believes
is mediocre, if only because all others are much worse. Another voter who
rates one candidate at 100 and another at 0 may in reality have very little at

stake in the election. So taking an average, or a median, or other statistical

measure doesn't quite represent the true aggregate utility. On the other
hand, this shouldn't be taken to completely invalidate utility metrics, since
at least in the aggregate the differences in scale between voters would
tend to cancel out.

Considering this, I would think the mean or the median would be better
metrics than the mode or the likelihood of SU falling below a cutoff --
especially the last option, since any chosen cutoff point will mean
different things to different voters. If everyone could agree that any
candidate below, say, 40 on the 0-to-100 scale is not worthy of the
office, then the cutoff criterion could work, but I don't think that
this would be acceptable to all voters.

 -- Richard




More information about the Election-Methods mailing list