[EM] Fwd: Is Condorcet The Turkey?

[Dave Ketchum]

On Sun, 15 Jun 2003 10:26:30 -0700 Bart Ingles wrote:
> Kevin Venzke wrote:
> 
>> --- Bart Ingles <bartman at netgate.net> a écrit : >
>>
>>>Kevin Venzke wrote:
>>>
>>>>Yes, but I think we all still like Condorcet.
>>>
>>Bart wrote:
>>
>>>I'm not crazy about it.  In a ranked election to choose an election
>>>system I might rank it above "Irving", but in an approval election I
>>>would likely bullet vote for approval voting.
>>
>>Sorry for speaking for you (and anyone else).  I was just thinking that,
>>within the "turkey" thread, it didn't seem like anyone was actually
>>saying Condorcet was bad.
> 
> 
> I didn't say it was bad necessarily, I just said I wasn't crazy about
> it.  If nothing else, at least it appears to be one of the few systems
> (along with Approval) which are free from both Duverger's law and from
> catastrophic strategy incentives.  But some of my misgivings are noted
> below.
> 
> 
> 
>>I take it you prefer Approval because it requires its winners to be overtly
>>"approved" on ballots?
> 
> 
> That was my original reason, in that Approval appeared to be capable of
> dumping very low-utility/low-worth Condorcet winners when the number of
> candidates is small.

They could not get to be Condorcet winners without a BUNCH of voters 
preferring them over the competition (not necessarily ranking them first, 
just ahead of the competition - Condorcet easily disposes of candidates 
ranked first by a minority and ranked low by others).
> 
> Other reasons:
> 
> (1)  Approval works on existing voting equipment, while Condorcet
> requires a computer to tally the ballots.  Translating preference orders
> to an N x N matrix is an n-squared operation, which would not be
> possible on a public scale without computers.  So even Australia, which
> has been using ranked ballots for years, would likely need to retool to
> support Condorcet elections.  Manual recounts would be out of the
> question (for what that's worth).

Agreed doing ranked ballots in public elections requires computers, but 
there are MANY reasons for moving to computers, such as assisting voters 
who need various kinds of help (for example just being able to do a 
write-in without being both tall AND right-handed - presently required in 
upstate New York, and requiring ability to pick the correct write-in slot).

Part of the game is making sure the computers are able to do correct 
counts - something some laws seem not to demand ability to verify.  Here I 
mean that there NEEDS to be provision for extended poll watching 
provisions to verify whether the computers CAN record correctly. 
Separately, I am against printing any kind of paper to be given to voters 
until/unless someone can define a way to do this that will not permit 
destructive use of the paper.
> 
> (2)  The people I have talked to on this list and elsewhere who have
> actually been involved in conducting elections have been pretty adamant
> that the system must be simple to administer, participate in, and
> understand.

I claim Condorcet fits.  Fact that people cannot do the recording both 
fast and correctly does not make understanding how it is done that difficult.
> 
> (3)  Strategy concerns, not necessarily catastrophic but enough to
> complicate voting.  For example, under Condorcet voters have the ability
> to dump a low-utility Condorcet winner using a prisoner's dilemma-based
> strategy.  Some of us are quick to discount the possibility of voters
> using this type of strategy, but this phenomenon is actually quite
> common.  Civil society is built on individuals sacrificing to varying
> degrees for the common good.  We willingly forgo certain things without
> expecting immediate compensation, because we expect to be compensated in
> the "big picture".  Otherwise we'd all be out looting museums.  Or
> common courtesy-- take the simple act of holding a door open for a
> stranger carrying an armload of packages-- it's not like you expect a
> tip.  Rather, you expect to reap the benefits of living in a society
> where people hold doors open for strangers with packages.

Please explain "prisoner's dilemma".  I do not remember any suggestion of 
such.
> 
> In any case, whether to pursue this strategy or not is one more decision
> voters would have to face, in addition to ranking the candidates.  In
> some cases it could be the more difficult decision.
> 
> With Approval, the strategy is much simpler: decide where to set your
> threshold.  If you are undecided about a candidate who is right on the
> threshold, you can always toss a coin (or better yet, focus your
> research on that candidate, rather than waste your time on the
> "no-brainer" choices).

May seem simple to you as you think on it.  Gets difficult if A is your 
favorite, AND you would like to approve B if A cannot win, YET do not want 
to risk making A lose by also approving B.

To me Condorcet is about as simple as it gets - list the candidates in 
order, starting with your favorite and continuing until you care not which 
of the remainder wins (which can only happen if none of those you do list 
wins).  If there are some that are tied in your mind, just list them 
together in any order (you are only here because you do not care which 
should win if two or more of these are in contention - by placing the 
group you call them less liked than those you list first, and more liked 
than the remainder).
> 
> (4)  One Condorcet strategy situation occurs to me which apparently
> does not involve a prisoner's dilemma: when you really do have NO
> preference between some of the candidates.  In this case, it seems to
> me, you should always rank the indifferent candidates strategically
> relative to one another, presumably by burying the one(s) who present
> the greatest threat to your favorite.

By listing your favorite first you have already said as much as you can to 
protect your favorite.  If you like/dislike this collection equally you 
reasonably list them together, but the only reason to do this sublist 
other than randomly would be because you DO care which of them wins if 
they become part of the decision.
> 
> To expand this a bit, I can imagine a voter presented with a long list
> of candidates who ranks the top few sincerely, and then decides he
> doesn't care about the rest.  But instead of truncating, the voter
> should probably rank these candidate in inverse order of their threat
> potential, with any likely Condorcet winners ranked last.

As I thought I said above, your actual ranking placed every listed 
candidate above this "rest".  Listing one of these cannot help them 
against those you have already listed as better - only as preferred over 
those not listed before them.
> 
> In essence, if you have no preference between two or more candidates you
> should never show it sincerely on the ballot (unless they are running
> equal in the polls as well).

Not quite.  If there is a rotten lemon, then you list every better-liked 
candidate, to say that each of these others is better - even if you care 
not which of some subset of these wins.

Not clear to me how "equal in the polls" matters, even if you believe the 
polls.  If they are equal than you could affect the result (but have said 
you do not care).  If not equal then it does not matter how you vote.
> 
> (5)  I haven't seen much evidence that anything more complicated than
> Approval is needed.  For myself, I have yet to participate in an
> election where I would have had difficulty voting an Approval ballot.  I
> CAN think of elections where ranking all the candidates would have been
> difficult, especially when there are many other open seats and ballot
> measures to contend with. 

So we have different views as to which is simpler for voters.
> 
> (6)  It would seem to make sense to implement the simpler, cheaper
> method first.  Then if something more elaborate and expensive is needed,
> implement it only where needed-- probably in larger elections that
> attract many candidates.  Elections with only 3 or 4 candidates should
> probably remain Approval even if Condorcet equipment becomes available.

If you are not buying equipment then you live with whatever you have.

If you are buying equipment then it SHOULD have all the capabilities you 
MIGHT want sometime.  Here I would demand Condorcet-capable.  If the 
equipment is capable then ALL elections for single seats should be 
Condorcet (with two candidates Condorcet is Plurality; with more, 
Condorcet has its own attributes so that, for simplicity, all should be 
Condorcet).

Note that in all this I rank IRV below Condorcet because they are similar 
except for IRV's spoiler problems.
-- 
davek at clarityconnect.com  http://www.clarityconnect.com/webpages3/davek
  Dave Ketchum   108 Halstead Ave, Owego, NY  13827-1708   607-687-5026
            Do to no one what you would not want done to you.
                  If you want peace, work for justice.