[EM] David Gamble reply, 24/1/04 0815 GMT
nkklrp at hotmail.com
Sat Jan 24 01:33:10 PST 2004
Mike Ossipoff wrote:
>The CW is the social utility maximizer.
Not always. Please justify this statement.
Are you sure I said it that way, without any qualification. If so, it should
have been qualified. But for what day is that message listed in the
archives, the message in which I said that? I just want to check on whether
there was any qualifying sentence, or clause in that quoted sentence, that
you didn't quote.
When we discussed this some time ago, it seems to me that there was
agreement that, with a 1-dimensional political spectrum, in which disutility
is measured by distance, the CW is the SU maximizer.
We can re-check to find out if that's so. In that scenario, a candidate at
the voter median position is the CW. I believe that it was shown that the
voter median candidate is also the SU maximizer.
It was discussed in terms of a 1-dimensional issue-space because that's
simpler. No one said that that isn't true with more issue dimensions.
But it also seems to me that, in Merrill's spatial simulation studies, Borda
did _very slightly_ better than the pairwise-count method that was tested.
Rankings were assumed sincere. This assumption is useful because it shows
what happens to voters who vote sincerely.
But the main thing is that, in those studies, pairwise-count, which chose
the candidate who pairwise beat everyone else, did much better than IRV. As
I said, Approval also did significantly better than IRV.
In Approval, the simulated voters used the 0-info above-mean strategy.
With one dimension, say X is at the voter median and Y isn't. Which has a
greater summed distance to all the voters?
If we start where X is, and go toward where Y is, as soon as we've left the
voter median point, we're moving away from more voters than we're moving
toward. So we're increasing the summed distance from voters.
With 1 issue dimension, the CW is the SU maximizer, if disutility is
measured by distance.
>If the voter-median position is occupied only by that despised corrupt
>candidate, why isn't anyone else contesting that position? That seems
>suspicioiusly odd in a Condorcet election, which would encourage many more
>candidates of all persuasions to run. And yet you only have that one
>despised candidate at the voter-median position. Add a better one, and both
>sides will rank him over the despised corrupt middle candidate.
The Condorcet turkey was not a despised candidate, he/she was a non-entity,
nothing, Fluffy the dog type candidate.
No, that won't do.
If he's CW, and has lowest SU, it's just as I said: Someone without his
special disutility could join him at the voter median point, and easily beat
The scenario went like this. There are 3 candidates A, B and C. A and B hold
strong opinions that are in opposition to each other. Both A and B gain both
fervent supporters and vehement critics amongst the voters. C does nothing,
says nothing, pleases no-one but also offends no-one. In the Condorcet
the electors vote as follows:
C the turkey candidate wins simply by virtue of being inoffensive.
Your rankings don't show that C has the lowest SU.
I find it odd that you have a problem with people preferring an inoffensive
candidate to an offensive one, and with a majority of such voters giving C a
majority over someone whom they strongly dislike, thereby preventing the
strongly disliked candidate from winning.
You yourself might not like C, but if we elect A or B, then we're violating
expressed majority wishes.
Maybe you know better than the voters do about which candidate woiuld be
better, but most would agree that that choice should be up to the voters. Or
maybe you should just be allowed to appoint the best candidate as winner :-)
It's always possible for a candidate to dodge issues, dodge debates, lie
about what he's going to do after he's elected. That isn't the fault of the
voting system. Nor is it any less a problem when the method is IRV. Don't
blame it on the voting system. The voters just have to have the judgement to
determine whom to trust. If they don't, then no voting system will help
You also wrote:
>It's common knowledge here that no method is without any faults or
>vulnerabilities or strategy needs. In fact every method can give a need for
>defensive strategy. We can choose the one that creates the least drastic
>defensive strategy need. Approval & Condorcet both do much better than IRV.
This common knowledge is not often stated.
On the contrary, it's very often stated on EM.
A common way for somebody to
promote a 'pet' electoral method is to do the following:
1/ Find a set of criteria that your system meets and state them.
2/ Dismiss the criteria your method doesn't meet as irrelevant or even
just don't mention them.
During the time that EM has been in existence, we've had to answer this
abyssmal ignorance many times. There should be a FAQ with answers to all the
stupid questions like that. Or else David should check the archives better
before presuming to tell us how it is.
David apparently is using his ESP here. Because, without ESP, how does
David know that someone started by choosing a method, and then searched
around for criteria that the method would meet.
A method proponent presumably has some reason for why he prefers that
method. The proponent tells why he prefers that method. David, however, is
saying that that proponent is lying and that actually the proponent prefers
that method for some other unknown reason that he (and David) won't tell us.
And that, though the propoent doen't really like that method for the reasons
that he gives, he finds other advantages of that method, and lyingly tells
us that those are the reasons why he prefers the method.
I suggest that the obvious presumption is that the proponent prefers a
method for the reason why he says he does. If David wants to claim
otherwise, then the burden is on David to show that the proponent is lying
about why he prefers the method.
For instance, a number of authors discuss the strategy problems of
pairwise-count methods. I was taking part in discussions about strategy
problems of voting systems, especially pairwise-count methods, and I wanted
to find a voting system that didn't have the strategy
problems that most pairwise-count methods have. Truncation & order-reversal
are the things that can keep a CW winning in a pairwise-count method. So I
wanted a method in which those things could be prevented from stealing an
election, by minimally drastic defensive strategy.
I found a method that could do that. The method now known as PR (wv). I
later found that that that method is a close interpretation of one of
Condorcet's proposals, and named it Plain Condorcet.
I proposed Smith PC, for compliance with the Smith set, and the other
criteria whose compliance follows from Smith Criterion compliance, such as
Condorcet Loser, etc.
By the way, Smith PC was voted the best method, in EM's first poll on that
question, which means that I wasn't the only one who valued the properties
I was advocating wv before this list began, because it did the best job of
getting rid of the strategy problems that I wanted to get rid of.
Then other wv methods were suggested. BeatpathWinner, Ranked-Pairs, SSD,
David might claim that I'm lying, and that actually I prefer wv for some
other reasons that I'm not saying, and that I've just adopted the majority
defensive strategy criteria because they make Condorcet look good. But if he
wants to claim that, his claim is worthless unless he can demonstrate its
I didn't find the majority defensive strategy criteria. I wrote them. Except
that Steve Eppley wrote GSFC, which is really the best one of all to comply
3/ Make your selected criteria sound as important as possible (really build
Who, in particular has done that, David? I tell how the defensive strategy
criteria measure for the popular goals of majority rule and getting rid of
the lesser-of-2-evils problem. I tell in what sense the methods that comply
with those criteria minimize defensive strategy need and respect majority
So perhaps you'd be so good as to give us an example of someone making their
criteria sound as good as possible, and really building them up, by saying
something that you can show to be unjustified.
Or is it also bad to discuss a criterion's value even what one says is
4/ Describe your selection of criteria as 'objective' ones that have been
selected by 'experts'.
Again, who has said that their preferred criteria are the objective ones
that have been selected by experts?
It's very widely agreed on EM that the combination of Condorcet's Criterion
with resistance to truncation or order-reversal stealing a CW's victory is
important. No one has said that this standard was selected by experts, but
it's a popular one on EM. These goals the ones that led me to advocate wv--I
was interested in how those failures could be prevented with minimal
The majority defensive strategy criteria were written because they state in
precise terms what I was looking for in a voting system. They're also
precise statements of the popular EM goals of truncation resistance and
order-reversal resistance, but written in a way that's applicable to all
methods, including those that have worse problems instead of truncation &
order-reversal. Such methods can fail in the ways that we don't want
truncation and order-reversal to make pairwise-count methods fail.
5/ Prove that your method meets them and that other methods don't.
The best wv methods have been proven to meet all the majority defensive
strategy criteria. Approval has been proven to meet FBC & WDSC.
It's been proven that IRV meets none of those.
But saying that is only dishonest if the criteria were dishonestly chosen in
order to make a method look good, when that method was chosen for other
Talking of which is electionmethods.org anything to do with you ?
You mean other than the fact that it lists me as a contributor?
David, yoiur participation is making this mailing list have to act as an
elementary school. It would be better if you would start reading through the
list's archives, so that you wouldn't repeat elementary questions that have
been answered many times in the archives.
There are now three new levels of MSN Hotmail Extra Storage! Learn more.
More information about the Election-Methods