[EM] VoteFair representation ranking recommended for Czech Green Party

[VoteFair at SolutionsCreative.com]

To Peter Zbornik, per your request for a proportional election method for
the Czech Green party:

I recommend that you use VoteFair representation ranking to achieve your
goal of fairness in electing your Green Party's council members.

VoteFair representation ranking has these characteristics:

* It is relatively easy to explain and understand.  (It is explained below.)

* Reliable software to do the calculations (and optionally the balloting) is
available for free at VoteFair.org.

* Drafts of statutes to implement it already exist, and I can modify those
for your situation.

* It has been successfully used in a similarly adversarial election of
directors.

* Most importantly, it produces fair results when a group is split into a
few different sub-groups.

Here is a testimonial from Allan Barber who coordinated the use of VoteFair
representation ranking for electing directors of the San Francisco Bay Area
Curling Club:

"Our club is extremely pleased with multiple aspects of the VoteFair system.
The ability to vote online meant an extremely high voter turnout,
approximately 70-75%!  Equally as important are the concepts underlying the
VoteFair system.  Using a comparison system instead of the more common
method of voting for a single candidate we came out knowing that we had
voted in the candidates our club members preferred to have in the seats.
Not only were there a number of good candidates, which could have split a
conventional vote to the point of electing a non-preferred candidate, but
our club is essentially split between 2 facilities and some candidates were
known better in one or other of the facilities.  VoteFair [ranking] gave us
the ability to balance that out transparently.  Thanks!"

Verbally I was told that everyone in the club -- except the people who did
not get re-elected -- liked the results.

Before explaining the method, please consider that the reason your group's
voters are "dishonest" is that the current voting rules allow a voter to
vote strategically in a way that gives that voter (or that voter's subgroup)
increased (compared to other voters) influence over the results.  A
well-designed voting method does not allow the results to be influenced by
strategic voting.  In other words, widespread strategic voting reveals that
the voting method, not the voters, are flawed.

Regarding strategic voting, range voting is vulnerable to strategic voting
by using an approval-like approach where the approved candidates are given
the highest score and the disapproved candidates are given the lowest score.
(I presume the re-weighted version has the same basic weakness.)  IRV and
(all versions of) STV also are well-known to be vulnerable to strategic
voting.  These reasons alone are enough to disqualify them for use in your
situation.  The fact that they do not necessarily elect a Condorcet winner
is yet another flaw.

As you recognize, the Condorcet criteria is important for electing your
president.  You want to ensure that he/she is pairwise preferred over each
of the other candidates.

To achieve the Condorcet portion (but not yet the proportional portion) of
the outcome, I recommend using the Condorcet-Kemeny method.  For a simple
description of the method, here is the first paragraph of its description in
the "Condorcet method" Wikipedia article 
(http://en.wikipedia.org/wiki/Condorcet_method#Kemeny-Young_method):

"[This] method considers every possible sequence of choices in terms of
which choice might be most popular, which choice might be second-most
popular, and so on down to which choice might be least popular. Each such
sequence is associated with a Kemeny score that is equal to the sum of the
pairwise counts that apply to the specified sequence. The sequence with the
highest score is identified as the overall ranking, from most popular to
least popular."

Of course you would have to add a description of pairwise counting, but
Wikipedia and other sources (indicated below) provide simple and clear
descriptions of pairwise counting.

The second paragraph provides a visual way to think of the Condorcet-Kemeny
method:

"When the pairwise counts are arranged in a matrix in which the choices
appear in sequence from most popular (top and left) to least popular (bottom
and right), the winning Kemeny score equals the sum of the counts in the
upper-right, triangular half of the matrix (shown here in bold on a green
background)."

A disadvantage of the Condorcet-Kemeny method (emphasized by Markus Schulze)
is that it is difficult to write software to calculate the results quickly,
and it is difficult to write the code that handles cases of circular
ambiguity and multiple highest Kemeny scores.  Yet this software-writing
disadvantage disappears by using the software at VoteFair.org; I've already
resolved those software-writing challenges.  Anyone can use that
server-based software for free.  During the last 10 years it has been used
for hundreds of real-life polls and surveys and dozens of (non-government)
elections, so it is fully debugged.

In your situation, the Condorcet-Kemeny method would be used to identify
which candidate is most popular.  That person would be elected president.

Now I'll describe VoteFair representation ranking.  It would be used to fill
the second seat, which in your case would be the vice president.  (There has
been some debate about whether the president and vice president should be
elected separately from the other council members, but I suggest keeping the
process simple; the approach I'm recommending will produce fair results.)

The core of VoteFair representation ranking is to reduce the influence of
the voters who just elected the winner of the first seat.  Those voters, who
clearly constitute a majority (because they elected the president), would
have their collective influence reduced to the degree that they exceed a
majority (50 percent of the voters).  For example, in a simplistic case, if
60 percent of the voters favor the about-to-be-president as their first
choice, then their collective influence would be scaled back to what can be
thought of as 10 percent (the amount beyond 50 percent), so that the
remaining 40 percent of the voters can (to the extent they are in agreement
with one another) elect their first choice as the vice president.

To prevent strategic voting (such as by marking an obscure candidate as the
first choice), there is an adjustment for identifying which voters account
for the winning of the first seat (the presidency in this case).  It is a
two-step process.  First, all the voters who ranked the new president as
their first choice would have their ballots ignored temporarily, and the
most popular candidate (based on the remaining ballots and the remaining
candidates, and using the Condorcet-Kemeny method) would be calculated.
This person we'll call the "alternative winner."  As the second step, the
ballots identified for reduced influence (for filling the second seat) are
the ones in which the new president is ranked higher than the "alternative
winner."  With this approach, a voter cannot strongly favor the new
president and also strongly oppose the likely winner of the second seat.  In
fact, if there is a strategic way to vote under this method, I don't know
what it is.

A recent example of how well VoteFair representation ranking works occurred
in a poll for American Idol contestants.  The results are at this URL:
http://www.votefair.org/results-43200-51085-40733.html
In this poll, the majority of voters were fans of Clay Aiken (which caused
him to be ranked as most popular) and those voters insincerely ranked Adam
Lambert very low, so that Adam Lambert was seventh out of twelve according
to Condorcet-Kemeny calculations.  Yet VoteFair representation ranking
reveals that Adam Lambert is actually second-most popular -- or second-most
"representative."

(Clarification: The word "popular" has two different interpretations, where
one refers to how many people approve of the choice and the other refers to
how strongly people like the choice.  As an example, TV stations basically
only care about how many people watch the show, not how much the viewers
like the show -- beyond what it takes for them to choose to watch it.)

After identifying (via Condorcet-Kemeny) the most popular-and-representative
candidate (your president), and after identifying (via VoteFair
representation ranking) the second-most representative candidate (your vice
president), the third-seat winner is identified using the Condorcet-Kemeny
method among the remaining candidates (and all the ballots).  The fourth
candidate uses VoteFair representation ranking where the just-elected choice
is the the third-most representative choice.  And so on.  This process
(which is executed using a single mouse click) would identify the top five
or seven most-representative candidates as your council members (with the
top-ranked ones also being designated as president and vice president).

I've just explained VoteFair representation ranking.  Do you think your
Green-party members will understand this method?

An even clearer explanation of VoteFair representation ranking is in my book
titled "Ending The Hidden Unfairness In U.S. Elections."  It can be read
online (free) using Google books; just search for "Richard Fobes".  Chapter
15 is the one that describes VoteFair representation ranking.  Chapter 12
clearly describes VoteFair popularity ranking, which is mathematically
equivalent to the Condorcet-Kemeny method (which I was unaware of when I
created the method).  (If there are any access limits on these chapters,
please let me know and I can resolve that.)

Another resource is Wikipedia.  The Condorcet-Kemeny method is described in
the "Kemeny-Young method" article.  Currently I cannot write a Wikipedia
article that explains VoteFair representation ranking because the method
hasn't been published in an academic publication, and because I am its
originator.  However, if the Czech Green party chooses this method and
someone else expands the "VoteFair ranking" article (which now just
redirects to the Condorcet-Kemeny article), I would be happy to refine the
article to include a description of VoteFair representation ranking (and
other components of VoteFair ranking, which also includes party-based
proportionality methods).

Yet another description of what is mathematically equivalent to the
Condorcet-Kemeny method appears in my how-to book on creative problem
solving, which is titled "The Creative Problem Solver's Toolbox."  The book
has been translated into Czech, so you might find that there is a
Czech-language description of that method in the Czech edition.  (I haven't
seen the Czech edition.)

As for ballots, your voters could vote online using the interactive ballots
at VoteFair.org, but duplicates would have to be removed (probably based on
randomly assigned ID numbers).  For this purpose consider that the VoteFair
site can handle high levels of voting traffic, presumably even if all 400
ballots are cast within the same minute.  Or you can collect digital data
from 1-2-3 ballots (also known by the redundant phrase "preferential
ballots") from some other source, and I can write code that converts its
output into VoteFair XML importable code.  I don't know of any open-source
software that reads paper ballots, but if you find such software, we can
similarly import it into the VoteFair XML format.

Perhaps a more viable option would be to digitally photograph all the paper
ballots, distribute those photographs to two or three groups of members, and
each group can have 5 or 10 people manually enter the preference
information, with the groups using separate/independent VoteFair election
IDs.  If the results are different between the groups, the errors
(intentional or not) can be tracked down (especially if two of three groups
get the same result).  When you create the paper ballots, I suggest using
the layout that is used in the interactive ballots; otherwise you will be
tempted to ask voters to write numbers, and those numbers are often
illegible.

The VoteFair site does have a limit of 12 choices (per question).  On my
computer I can handle more choices.  However, I find that voters are
overwhelmed if they have to rank more than 12 choices -- even though they
can rank multiple candidates at the same preference level.  If this
limitation is a problem, there are other alternatives, one of which is to
use an informal approval-voting process to dismiss candidates who do not
have significant support.

I have not yet answered all your questions, but at this point I have a
question.  Is this approach of interest to you?  If so, I would be happy to
assist you in developing a proposal to your group, and then making it
happen.

As the author of a how-to book on creative problem solving that has been
published in nine languages, I'll point out that unfair voting methods are
the cause of many of the world's biggest problems.  Helping your group would
create a path for others to follow as we take democracy to higher levels of
fairness.

Richard Fobes