[Election-Methods] Election-Methods Digest, Vol 42, Issue 72

CLAY SHENTRUP clay at electopia.org
Mon Dec 31 03:53:23 PST 2007

> From: Dan Bishop <danbishop04 at gmail.com>
> Subject: [Election-Methods] Linear Spectrum MMPD analysis:
>         Single-Winner   Plurality
> In the coming weeks/months/whenever-I-feel-like-it, I will be performing
> simulations to evaluate the performance of multi-winner methods.  In
> order to do this, I will make the assumptions that:
> * There is a uniform linear political spectrum.  (Other models of voter
> behavior will be considered later.)
> * Candidates are uniformly-distributed random variables in.
> * All votes are sincere.  (i.e., a voter at position V votes A>B iff
> abs(A-V) < abs(B-V))

judging the utility of a multi-winner method is tricky, and i do not
think you can do it the way you describe here.  if you did that, then
simply using multi-winner range voting - which isn't proportional -
would do well according to this metric.  warren proposed looking at a
set of yes/no directions that candidates would vote on various issues,
and that might be better.  it's a very very hairy problem - how much
does a voter like a legislature comprised of X and Y, given that we
know how much he likes X and Y?

> From: Juho <juho4880 at yahoo.co.uk>
> Subject: Re: [Election-Methods] rcv ala tournament
> Ok, ranked methods are obviously not included in the list of methods
> that you find acceptable.

it's not a matter of what i think, it's a matter of utility
efficiency, ballot spoilage rates, cost of implementation, etc.  rv/av
just blow away ranked methods in this regard, hands down, no question.

> Ok, you seem to promote Range(ratings) and not Range(approval).

i promote it because it's better.  the important thing isn't that i
promote it, it's that it's better.

> One of the Democrats considers changing from approval to ratings style:
> - overall ("absolute"/"idealistic") social utility could go up or
> down or stay the same (depending on which party is "right")
> - personal utility would go down

well, no.  the average voter's personal utility is the social utility
divided by the number of voters.  so a greater social utility means
the average voter has a greater personal utility.

> - overall utility as this voter sees it would go down if this voter
> wants to improve social utility

i don't buy that argument.  i think a voter increases social utility
more by voting honestly (letting others have their way more than if he
had voted approval-style) more than by strategically voting
altruistically for the candidates he thinks will make other voters
happy.  mainly just because the former cedes power to the other voters
based on what they actually expressed on their ballots, whereas the
latter relies on the altruistic voters' best guess as to what other
voters want.  i'm not sure whether there's a way to concretely address
this question however, so it remains speculative as far as i can tell.

i think the best we can do is look at warren's social utility figures,
which show a large benefit for range over approval.

> If all would vote in (non-normalized) ratings style the (average
> personal) social utility could be better, but for individual voters'
> decisions on how to vote (or for groups) it doesn't make much sense
> to switch to ratings style.

i didn't say that it makes sense for an individual voter to use
ratings.  it makes sense for him to want his society to use ratings,
but for him to vote approval-style.

and we, as benevolent deities, would want our created societies to use
range voting instead of approval, as it would increase the average
voter satisfaction.

> People also tend to have their own understanding on what is best for
> the society

i think that's a very good point.  i tend to think ron paul would make
a lot of people better off who strongly disapprove of him.  i believe
that if they understood economics and foreign policy better, they'd
understand why his policy would actually be beneficial to them in the
long run.  i talk to people all the time who say they support people
like giuliani and clinton, and are blissfully unaware of their
connections to evil corporations, and their repugnant views on freedom
(giuliani says freedom is about authority, and clinton voted for the
flag desecration amendment).

but one thing i notice is that the more politically savvy people
support candidates other than the front-runners.  so here's an
interesting thing to consider.  do educated people, who are more
likely to know more about world affairs, tend to have their more
educated voices strengthened when using a voting method that is more
complex to game?  i know it sounds elitist, but it's a legitimate

> What if we look at Condorcet at the same level of recommending a
> strategy that works well in most cases. The recommended strategy
> could be: "list the candidates in the order of preference". That is
> quite simple and understandable and works in most situations for the
> benefit of the voter.

that's not the strategy.  the gist of the strategy is to start with
that ordering, then bury your favorite front-runner's strongest

> So, is Condorcet strategy more straight forward than Approval
> strategy? Approval strategy (e.g. "vote one of the frontrunners and
> candidates that are better") requires the voters to estimate the
> opinions of other voters while the described Condorcet strategy does
> not.

that's because you described honest condorcet voting and strategic
approval voting.  apples and oranges.  you could rectify that by going
back and either describing the burial aspect of condorcet, or by
describing the approval voting strategy as "vote for every candidate
you like more than average", which requires no knowledge of the
opinions of others voters.

c'mon juho...you know better than this.

> From: Kevin Venzke <stepjak at yahoo.fr>
> Subject: [Election-Methods] RE : Re:  RE : Re:  rcv ala tournament
> I certainly concede that Condorcet more frequently offers a useful strategy
> that doesn't require much thought

that's baloney.  saying "vote for the candidate you'd select in a
plurality election, and then all the ones you like better" is as
simple or simpler than saying "order the candidates by preference, and
then raise your favorite front-runner to the top, and bury the other

>, but to my mind that's not the issue I
> was considering. If you wanted to specify the "complete" Condorcet strategy
> on the same level that we can describe optimal Approval strategy, I guess
> the result would be extremely complex.

tre bon!

> From: Kevin Venzke <stepjak at yahoo.fr>
> Subject: [Election-Methods] RE : Re:  RE : Re: RE : Re: rcv ala
>         tournament

> Rob,
> ? May I assume a voter is allowed to bullet-vote? If not, how does your
> method resolve:
> 40 A>B>C
> 35 B>C>A
> 25 C>A>B


oh, that rob.

> Does your information allow rated information? It seems this should be no
> problem since you would not have any incentive to exaggerate your
> preferences when the method does its best to get you what you want.

do you call "preference reversal" an exaggeration?

> From: Kevin Venzke <stepjak at yahoo.fr>
> Subject: [Election-Methods] RE : Re:  rcv ala tournament
> You should remember from past discussions why the claim that "Condorcet
> voting ... is nowhere near as utilitarian as range voting" was criticized.
> Ask Warren.

i may not have been on the list at that time, but i think that is an
absurd notion.

> From: Juho <juho4880 at yahoo.co.uk>
> Subject: Re: [Election-Methods] RE : Re:  RE : Re:  rcv ala tournament

> Yes, I'm to some extent comparing apples to oranges, but I think the
> Approval strategies are also not quite "complete" since they rely on
> some estimate on how others are going to vote and that estimate is in
> real life always unreliable and incomplete. The Approval strategies
> that regular voters use will probably also rely on generalizations
> like identifying who are the "frontrunners" and making one's decision
> based on this. So, I was trying to compare an incomplete Condorcet
> strategy to an incomplete Approval strategy.

i don't follow.  condorcet strategy also relies on knowing who the
front-runners are.

> Message: 8
> From: Kevin Venzke <stepjak at yahoo.fr>
> Subject: Re: [Election-Methods] Brams and Sanver on rank methods
> Steven J. Brams and M. Remzi Sanver wrote this in a Dec 2005 paper:
> "A majority-preferred candidate is likely to have a more coherent point of
> view than an AV winner, who may be the most popular candidate because he or
> she is bland and inoffensive - a kind of lowest denominator who tries to
> appease everybody. *Not* choosing such a candidate makes PAV [a hybrid
> method proposed in this paper] *coherence-inducing for candidates* by
> giving an advantage to candidates who are principled but, nevertheless,
> command broad support."

i think this is kind of a stupid comment.  being inoffensive may not
cause you to _lose_ someone's vote, but if you don't do anything to
offend anyone, you probably haven't done anything to _earn_ someone's
vote.  so, for instance, say you have two clones, X and Y, who are
completely equal in approval voting.  if Y decides to diverge from X
by adopting some policy positions that turn some people off, but earn
more support than they lose, then Y can defeat X.  so when people say
things like "approval would elect colorless uninspiring candidates", i
just roll my eyes.  those candidates must have been better than some
more "coherent" alternatives out there if they got enough votes to
win! i mean, does anyone call a guy like gorge bush "coherent"?  pfft.
 what does "coherent" even mean in any objective sense anyway?

> In this paper ("Voting Systems That Combine Approval and Preference") they
> also "propose" Bucklin, understanding that all ranked candidates should be
> considered "approved."

i believe bucklin has a higher bayesian regret than approval, at least
under certain circumstances. and it's mild benefit may be countered by
its implementation cost compared to approval.

> >From the abstract and introduction: "Information on the rankings and
> information on the approval of candidates in an election ... are [both]
> important in the determination of social choices."

well, that's a question to be determined by bayesian regret
calculations, which i don't think support brams here.

More information about the Election-Methods mailing list