[EM] VoteFair representation ranking recommended for Czech Green Party

[VoteFair at SolutionsCreative.com]

> From Peter Zbornik:
> It seems a bit unusual to keep switching methods.

On the surface, it does seem unusual.

What seldom gets pointed out about STV (single transferable vote) is that it
only produces somewhat reasonable results for electing the first-seat and
second-seat winners.  (This assumes that the underlying unfairness, namely
looking for a majority or quota of currently-top-choice votes, can be
tolerated.)  After that, using STV to fill a third seat (and fourth seat,
etc.) easily produces unrepresentative results (although not unproportional
results, ironically).

With that in mind ....

> I don't understand how proportionality is achieved.

The proportionality can be thought of as a balance between the "majority"
and an "opposition."  This can be regarded as a two-sided balance, such as
between a 60% majority and a 40% minority.  The switching between methods
allows one method (Condorcet-Kemeny) to select majority-based candidates,
and the other method (VoteFair representation ranking, with Condorcet-Kemeny
used to calculate popularity) to select candidates who are in the minority
(or minorities).

Note that very small minorities do not get direct (!) representation using
this approach.  (They do get representation, but not with a council member
who is specifically identified as representing that particular small
minority.)  Here is the reason:

In order to get additional levels of proportionality (i.e. representation
for small minorities), there would need to be a way to categorize the
subgroups within your group (the Czech Green party).  For example, if there
were categories within the Green party -- such as environmentalists,
social-justice members, pacifists, consensus-decision-making members, etc.
-- then there are many ways (including additional aspects of VoteFair
ranking) to ensure that the elected members represent those subgroups
roughly in proportion to their numbers (percentages).  But all those methods
require allowing voters to identify which subgroup they most want to
support.  Without that self-categorization, there is no way to identify, and
then ensure representation for, those small minorities.

There does exist a way to get proportional results for smaller (but not the
smallest) minorities without self-categorization.  I've created a website at
NegotiationTool.com that does that kind of calculation.  However, that
method is complex, and I don't recommend it for your situation.

> I would appreciate if Votefair ranking would have some mathematical
> description and at least well described and discussed in some
> peer-reviewed paper.

If you replace the word "paper" with the word "article," there is a widely
peer-reviewed, published description of the Condorcet-Kemeny method.  (Below
I'll get to the issue of VoteFair representation ranking.)  It's in
Wikipedia under the name "Kemeny-Young method."  The description is
rigorous, which is what I presume you mean by "mathematical."

(I presume you don't mean mathematical in the sense of using mathematical
symbols, because that would have to be translated back into plain language
for your members to understand.)

The "Kemeny-Young method" article includes a list of which voting-method
criteria are satisfied by the Condorcet-Kemeny method, and that list has
been heavily peer-reviewed, so I'll summarize it here:

* The Condorcet-Kemeny meets the criteria named "monotonicity" and
"summable," which someone in another message points out are the most
important voting criteria for your situation.

* There are other criteria that the Condorcet-Kemeny method automatically
meets and fails as a result of meeting the Condorcet criterion (i.e. they
apply to all Condorcet methods), but those are not as important as meeting
the Condorcet criterion.

* The additional criteria of importance that the Condorcet-Kemeny method
meet are: unrestricted domain, pareto efficiency, Smith criterion,
independence of Smith-dominated alternatives, reinforcement, and reversal
symmetry.  Translation: the method is very good.  (Again, this list is from
the peer-reviewed "Kemeny-Young method" article.)

* The significant criteria that the Condorcet-Kemeny method fails are:
independence of clones and invulnerability to push-over.  In your situation
(having about 2000 members), these failed criteria are very unlikely to be
noticeable even after careful analysis of the results.

* Also consider that some of the above failed criteria can only fail in
cases that involve circular ambiguity.  To visualize an example of circular
ambiguity, suppose a group of people at a table are asked for their first
choice and point to the person on their right, and then when asked for their
second choice each person points to the person on their left, and finally
when asked for their third choice everyone points to the person at the head
of the table.  These situations (when there is no Condorcet winner) are not
common (although neither are they rare).

In other words, the Condorcet-Kemeny method has great fairness
characteristics.  Also recall that several people in this forum have
recommended that your Green-party President be elected using a Condorcet
method.  Also note that the reason that the U.S. cities of Aspen (Colorado)
and Burlington (Vermont?) recently stopped using IRV is that recent official
winners were not the Condorcet winners.

As for a peer-reviewed description of VoteFair representation ranking, I
have repeatedly offered (here and elsewhere in the online voting-methods
community) to write a description some place where it can be peer-reviewed,
but there has been no request for such a description.

Ironically, if the Czech Green party adopts VoteFair representation ranking
(along with the Condorcet-Kemeny method), then the method would qualify as
notable, and then I would be allowed to describe it in Wikipedia, where it
would get peer reviewed.  It's a chicken-and-egg problem; each requires the
other.

A fan of VoteFair ranking in Canada encouraged me to propose VoteFair
ranking to a committee in the Canadian province of Ontario when they
solicited recommendations for improved voting methods.  I submitted my
proposal for a subset of VoteFair ranking, and it was reviewed by numerous
people, and I was among those asked to submit a concise summary (because
many other submissions were unnecessarily lengthy and/or complex), and that
adds a bit more credibility.  (No new method was adopted for Ontario
province because the voting-method experts, although supposedly unbiased,
slanted information to support an unfair PR method that was then defeated in
the province-wide vote.)

> According to the description votefair ranking looks like STV.

To answer this question, let's separate my recommendation into three parts:

* The ballots are similar. (However, VoteFair ranking does not have
restrictions that are commonly imposed on STV ballots.)

* The popularity of choices are calculated differently, with STV assuming
that the choice with the fewest votes is the least popular, whereas
Condorcet-Kemeny uses a deeper-looking algorithm.

* Finally, STV uses a quota cut-off and basically only looks at the
top-ranked choices, whereas VoteFair representation ranking identifies
"already-used" ballots in a way that strongly resists strategic voting
attempts.

If these differences are kept in mind, then the part of VoteFair ranking I
recommend for your situation does have similarities with STV, especially
when contrasted with methods such as approval voting and range voting.

> I also have some concerns about the vote-counting.
> We would need to make sure that the vote counting cannot not be
> manipulated
> and that the count is independently verifiable.

This can be done by having three interested people independently (but
probably with assistance from others) enter the ballot information into the
VoteFair.org site as three different elections.  As a further precaution,
alias names (such as alpha, beta, gamma, delta, etc.) can be used instead of
the candidates' real names.  If all three results match one another (after
converting back to real names), the vote-counting has been independently
verified.

In addition, if any voter disputes the results, he or she can look at the
pairwise counts (which appear on every VoteFair results page) to verify that
each winner was preferred over their losing favorite candidate.  Notice that
this simple comparison of numbers is much easier to understand compared to
keeping track of eliminations sequences and vote transfers as used in IRV
and STV.

> Is the vote-counting program possible to install on a computer?
> Is it open source?

I have not yet released the software for use outside the server.  At a later
time I will release it on an open-source basis.  (It's the chicken-and-egg
problem again.)

> Is the count implementable by a reasonably skilled programmer?

In theory, yes, but in reality, no.  It took me several years to figure out
how to do the Condorcet-Kemeny calculations quickly (for all cases).  And it
took me about a year to write the code for VoteFair representation ranking.
In both code segments there are many "edge" effects to handle, such as
dealing with ties, accommodating voters who rank multiple candidates at the
same preference level, and more.  Also, I spent extra time writing code that
does not reject ballots as "spoiled"; most reasonably skilled programmers
would choose (especially initially) the lazier approach of discarding those
ballots as spoiled.  For clarification, I have a degree in Physics (from the
University of California at Davis), and this gives me a mathematical
advantage over average programmers.

When other proposals claim that the software for their method does not yet
exist, but that it would be easy to create, don't count on that happening
quickly.  It always takes much longer than expected to write new code.  And
programmers typically fail to budget time to encounter rare special cases
that the programmer forgot to handle, and then to fix those bugs.

As a reminder, the voting method I recommend for electing your council
members has already been implemented in software and debugged, and it has
been running successfully for more than five years.

I hope this reply answers your questions.  Please ask more questions as they
arise.  I apologize for the delay in answering these questions; I was busy
preparing for a presentation I gave yesterday.

Again, thank you for considering the use of the Condorcet-Kemeny method and
VoteFair representation ranking for electing your council's members.


Richard Fobes