[EM] New election system in Hungary

[VoteFair]

On 7/4/2017 2:53 AM, Magosányi Árpád wrote:
 > ...
 > If there are flaws in this system, what should be the alternative?

Overall I am impressed by what you have designed!  Yet I suggest the 
following refinement to what you propose:

* For the party-list ballot -- which I assume is different for each 
political party -- I suggest that each candidate's name appear next to 
two ovals, one labeled "Approve" and the other labeled "Disapprove".  If 
neither oval is marked, or if both ovals are marked, the ranking is 
regarded as "Neutral".  (Spoiled ballots would only occur when it is not 
clear whether an oval is marked or unmarked.)  The result is a 
three-level ranking.

* This three-level ranking would be counted using any Condorcet method. 
Very importantly the software must handle ties at any level.  (My 
VoteFair popularity ranking software on GitHub can handle this, and yes, 
it can rank 200 candidates, although only the top-ranked candidates are 
confirmed to match Condorcet-Kemeny results in the extremely unlikely 
case where there is circular ambiguity [also called the 
"rock-paper-scissors" situation].)

* Your description mentions using fractional/decimal vote counts, but 
they are not needed if you are using a Condorcet method that fully ranks 
all the candidates, and which allows multiple candidates to be tied at 
any ranking level.  (My VoteFair ranking software handles this, and any 
good implementation of the Condorcet-Schultze method should do so too.)

* As far as I know, the best voting strategy is for a voter to heavily 
use the Approve and Disapprove categories, and use the Neutral category 
for the remaining candidates.  In other words, a voter is wasting their 
vote if they only mark a few candidates as Approved or Disapproved.

* If multiple candidates are tied and the limit for that party's won 
seats occurs within that tie, the tie would be broken by the party's 
ranking of the candidates.  There are fairer ways to resolve such a tie, 
but this approach gives the impression that the party has some influence 
on the results.  In reality this might affect only a seat or two for 
each party.  Ties at other ranking levels do not need to be resolved 
because each group of tied candidates win seats if they are more popular 
than the threshold level, or they don't win seats if they are less 
popular than the threshold level.

* Most importantly, notice that there is no limit as to how many 
candidates can be marked as Approved, and no limit as to how many 
candidates can be marked as Disapproved.  Voting methods that require 
such constraints are especially vulnerable to strategic voting!!

* This ballot approach is simpler than the one you describe, which makes 
it easier to learn to mark, and virtually eliminates spoiled ballots. 
In contrast, the ballot type you describe would result in very many 
spoiled ballots.

* As for the proportional winning of seats by party, I suggest that you 
use the recommendation from Kristofer Munsterhjelm regarding awarding 
some extra seats to the party that ranks highest according to the 
Condorcet ranking of political-party preferences (where the cumulative 
ballot can be interpreted as a ranking based on the number of ovals marked).

It is possible that I have misunderstood the method you are proposing, 
in which case I might be overlooking something.  If so, please clarify.

I realize that what you are doing is very challenging.  After all, 
combining proportional results with Condorcet counting is a new frontier.

 > But I don't want to base such a proposal on feelings, but rather on
 > mathematical proof.
 > Please advise me on how to work it out: what are the results I can build
 > my proof on?

I think your desire for mathematical proofs will NOT be of interest to 
most voters and most politicians.  The remainder of this message 
contains some more compelling concepts for supporting your electoral reform.

 > ...  My role is to propose an ideal system, for the
 > following tactical communication reasons:
 >  - show how the current system is fscked up compared to an ideal one

The failure of both single-mark ballots (with plurality counting) and 
non-Condorcet methods (such as instant-runoff voting) is conveyed with 
just nine voters in this scenario:

4 voters prefer A > C > B

3 voters prefer B > C > A

2 voters prefer C > B > A

In this example the single-mark ballot (plurality) winner is A, and the 
instant-runoff winner is B, and the Condorcet winner is C.  Notice that 
the Condorcet winner is discarded in the first round of instant-runoff 
voting.  And notice that the plurality winner is actually the least 
popular candidate.  As has been pointed out to me by an election-method 
reform advocate in Canada, this is an example that can be explained 
while doing the counting on the fingers of both hands.

 >  - pressure politicians to agree on something in which they could be
 > successful based on their instinctive behaviour

Here are some motivation-related insights:

* The number of political parties is not as important as who controls 
them -- the voters or the biggest campaign contributors.  The use of a 
Condorcet method increases voter influence, which decreases money influence.

* The biggest unfairness in all major "democracies" is that money is 
used to control political parties in ways that override what the voters 
prefer.  This can occur only because those democracies use vote-counting 
methods that can be exploited to control each political party.  (And 
typically the biggest campaign contributions come from a few people who 
(both directly and indirectly) give money to candidates in MORE THAN ONE 
political party.)  In other words, when flawed voting methods are used, 
contributions of money (not votes) control political parties.

* When politics is controlled by the biggest campaign contributors, the 
result is corruption in the form of customers and clients being 
essentially forced (through legal means) to buy (pharmaceutical 
products, genetically modified foods, insurance, legal assistance, etc., 
often at excessive prices) from the businesses that are owned by those 
biggest campaign contributors.

* The result of corruption is a weak economy.  Better elections will 
lead to less corruption, which will lead to a stronger economy.  This 
was dramatically demonstrated when the United States cut off control 
from Britain; and the current bad economy here in the U.S. occurs 
because the biggest campaign contributors have learned how to control 
primary elections, which unfairly fail to ask voters to provide more 
information than just a single choice in each race.

* Everybody wants economic prosperity.  That overrides all other 
political issues, including most religious issues.

* Storytime: When the inventor of an electric motor in Britain was asked 
by a government official what advantage it provided, the inventor said 
"you can tax it!"  He got the investment money he requested.

* If the economic pie is being divided to give a few insiders bigger 
slices of the economic pie, that leaves less economic pie for everyone 
else, namely the majority of voters.

OK, the last few points may not be relevant to your election-reform 
efforts, yet I think it's useful to keep a clear perspective about 
what's really going on.

Good luck!!

If you have further questions, just ask!

Richard Fobes


On 7/4/2017 2:53 AM, Magosányi Árpád wrote:
 > Hi,
 >
 > Please help, it is really important!
 >
 > We are in the process of designing the new election system of Hungary. A
 > strong movement is emerging for that purpose, and there is a high chance
 > that even if we fail at first, everything we say will be influental for
 > the future of our election system.
 >
 > Our team have came out with a proposal as a basis of discussion between
 > the parties (most of them will be participating). It is designed to be
 > not too shockingly new. My role is to propose an ideal system, for the
 > following tactical communication reasons:
 >  - show how the current system is fscked up compared to an ideal one
 >  - pressure politicians to agree on something in which they could be
 > successful based on their instinctive behaviour
 > Andt he long term communication goal is of course to put good election
 > methods on the political agenda. In case of the ruling party not
 > accepting the compromise proposal of parties (almost certain), most
 > probably sizeable factions of the resistance will nominate the ideal
 > system as the core issue we are fighting for.
 >
 > Our proposal as basis of discussion is a purely party list system, with
 > proportional representation and no entry threshold.
 >
 > I would like to propose something within this framework as the ideal
 > system, with the same results from the game theory standpoint, as
 > preferential Condorcet for a commitee:
 > - The winnig strategy for candidates is collaboration
 > - The winning strategy for voters is honest voting
 > - In the long run there is no two-party system
 >
 > Also, I would like to have easy ballots.
 >
 > What I have came up with, and why:
 >
 > Each voter can nominate one party for the election. Nomination needs
 > active participation from the voter (phisically walking in to a
 > government office), to make strategic nomination hard. The 20 parties
 > with the highest number of nominations will be in the ballot.
 >
 > There is a ballot for parties, and there is a ballot for candidates of
 > each party.
 >
 > The party ballot is a cumulative voting ballot, where six votes can be
 > allocated, and at most 3 can be given to one party.
 >
 > The candidate ballot is also a kind of cumulative one: the voter can
 > indicate at most 10 approvals, and at most 5 disapprovals (for a
 > 200-member list).
 >
 > The results from candidate ballots are computed using shulze method, and
 > ties are broken using the order of names (the preference indicated by
 > the nominating party).
 >
 > The result from party list ballot is computed by first creating a
 > pairwise defeat table, where
 > - the cell in the row of the party will contain the number of wins over
 > the other candidate
 > - in case of tie, both cells receive +0,5
 >
 > The sums of each row are computed, and seats are allocated based on them.
 >
 > Regarding the candidate list, it is a condorcet method, with a bit more
 > constrained ballot, but based on the size of the constituency (10M) and
 > human behaviour, I think that the constraint should not change anything.
 >
 > My understanding is that the party list method is somewhere between
 > range voting and condorcet, with a very simplified ballot. As condorcet
 > comes with the above game theory results, and in range voting majority
 > condorcet is strategically forced, I feel that this method should also
 > have the same game theory results.
 >
 > But I don't want to base such a proposal on feelings, but rather on
 > mathematical proof.
 > Please advise me on how to work it out: what are the results I can build
 > my proof on?
 > If there are flaws in this system, what sould be the alternative?