[Election-Methods] Borda-elimination, a Condorcet method for public elections?

Fri Dec 21 22:22:50 PST 2007

On Dec 22, 2007, at 1:05 , Ian Fellows wrote:

> Markus,
>
> Thank you for your insight. I certainly agree with you that only  
> the best
> method should be used, but I would pose to you the question: Why is  
> it that
> the best method isn't used?
>
> You and I (though not some others) would agree that the condorcet  
> criterion
> is the correct one when determining the outcome of single winner  
> elections,
> and yet they are not used in any public election anywhere in the  
> world.
> Though the current best methods (Yours, and Ranked Pairs), are  
> relatively
> new, Condorcet methods have been around for quite a long time. So the
> newness of the methodology can't be the reason. The difficulty in  
> changing
> an electoral system once it has been started certainly plays a  
> part, but IRV
> seems to be making significant inroads in this area whereas Condorcet
> methods are not.

Some reasons why Condorcet metods are not as popular as one could  
assume, and reasons why IRV seems to make progress.
- the Condorcet strategy problems exist and can be used against it,  
and it is possible to create dramatical looking examples
- IRV favours large parties (it is natural that if the current large  
party members identify this feature they are likely to like the idea)  
(in some cases IRV can be seen as a clear reform but with minimal  
changes to the existing power balance)
- there are many different Condorcet methods and no clear consensus  
among the experts on which one to recommend for use
- for some, maybe related reason there is no very active Condorcet  
campaign around
- Condorcet tends to require computer calculation (no roots in the  
pre-computer age)
- in many Condorcet methods it is difficult to say who was second,  
why was the first first etc.
- people find the existing methods (e.g. two-round runoff) "not too  
bad" and are complacent
- IRV may get some additional interest due to its interesting ("good  
TV series like") tournament like elimination of competitors one after  
anoher

> I think the answer lies in looking at the organizations that have  
> adopted
> the Schulze method.
> http://en.wikipedia.org/wiki/Schulze_method
>
> 44 organizations, and almost all of them are technically oriented.
>
> The answer seems to me to be clear, complexity. Though beat-path is  
> the best
> methodology, and the one that I would use in any professional  
> organization
> that I was a part of, it violates a principle of democracy. For an  
> election
> method to be "of the people" the people must be able to understand its
> implementation. They must be able to understand why one leader was  
> picked,
> and not another, and further, how their ballot played a part in that
> decision.
>
> This begs that question of whether there is a Condorcet method  
> simple enough
> for everyone to understand, and yet having the greatest number of  
> desirable
> properties.

I have often talked in favour of the very basic minmax(margins)  
Condorcet method. I think it is much better than usually assumed (on  
this list). It directly implements one clear and simple definition/ 
explanation, "number of additional votes required to win all other  
candidates". Among the Condorcet methods it may be the easiest to  
explain, and it also allows very clear comparison of results after  
and during the election (how many votes needed to gain lead). And it  
has a very natural utility function (good and explainable behaviour  
with sincere votes).

> Perhaps one answer might be in Borda-elimination methods. They
> are the only ones to have ever been used in public elections, and  
> have very
> little added complexity when compared to IRV. IRV has had a great  
> deal of
> success in being adopted, so we know that voters can handle  
> something as
> complex as IRV.
>
> Borda-elimination also stacks up favorably when compared to  
> anything but
> ranked pairs and Schulze. The only criteria that it doesn't pass  
> are local
> IIa, monotonicity and independence of clones.
>
> non-monotonicity, while weird, doesn't imply that the candidate  
> chosen is in
> any way inferior to a candidate chosen under a monotonic rule.

Agreed. In general I don't find the strict criterion based approach  
to be the best possible approach in determining which methods are  
best. Especially the _number_ of criteria that one fulfils is not  
very relevant. There are also criteria that look good at first glance  
but that sometimes may also make the results worse, or that are in  
conflict with some other positive criteria. Sometimes it is ok and  
enough to meet some criteria only typically but not necessarily in  
all cases. (Experienced readers may note that the properties of minmax 
(margins) have a connection to these explanations.)

> I would have
> thought that the main reason why you would want a monotonic rule is  
> so that
> people would accept it as valid. This does not appear to be an  
> issue as IRV
> is non-monotonic, and is well liked. There are some possible issues
> regarding additional sussepability to strategy, but I'm not sure  
> how serious
> those would be. Also, like all condorcet methods, Borda-elimination is
> monotonic if there is a Condorcet winner.
>
> local IIa and independence of clones are not passes, and this is an
> inferiority. but at least it passes them when there is a Condorcet  
> winner. I
> seriously doubt that clones would be a big problem outside FPP, where
> vote-splitting is rampant.
>
> So guess I'd ask if the minor theoretical deficiencies are not made  
> up for
> by the additional simplicity in populations that would have difficulty
> understanding beat-path? Why do you think that no Condorcet method  
> has been
> adopted by any government?

I hope I managed to explain my opinions on this question clearly  
enough already above.

Juho

>
>
> Ian
> http://thefell.googlepages.com
>
>
>
> -----Original Message-----
> From: election-methods-bounces at lists.electorama.com
> [mailto:election-methods-bounces at lists.electorama.com]On Behalf Of
> Markus Schulze
> Sent: Friday, December 21, 2007 1:47 PM
> To: election-methods at electorama.com
> Subject: Re: [Election-Methods] Borda-elimination, a Condorcet method
> for public elections?
>
>
> Dear Ian Fellows,
>
> the Nanson method and the Baldwin method violate
> monotonicity and independence of clones. They also
> violate the desideratum that candidates, who are not
> in the Smith set, should not have any impact on the
> result of the elections.
>
> When you try to get a Condorcet method adopted somewhere,
> you will not only be attacked by the FPP supporters and
> the IRV supporters. You will also be attacked by the
> supporters of all kinds of election methods. Therefore,
> it will not be sufficient that you argue that the
> proposed method is better than FPP and IRV; you will
> rather have to argue that the proposed method is the
> best of all methods. Therefore, it is useful to propose
> a Condorcet method that satisfies a large number of
> criteria.
>
> Furthermore, I don't think that it makes much sense to
> try to find a Condorcet method that looks as much as
> possible like IRV or as much as possible like Borda.
> The best method according to IRV's underlying heuristic
> will always be IRV; the best method according to the
> underlying heuristic of the Borda method will always
> be the Borda method. It makes more sense to propose
> a Condorcet method that stands on its own legs.
>
> Markus Schulze
>
>
>
> ----
> Election-Methods mailing list - see http://electorama.com/em for  
> list info
>
>
> ----
> Election-Methods mailing list - see http://electorama.com/em for  
> list info

___________________________________________________________ 
The all-new Yahoo! Mail goes wherever you go - free your email address from your Internet provider. http://uk.docs.yahoo.com/nowyoucan.html