[EM] New election system in Hungary

[Kristofer Munsterhjelm]

On 07/05/2017 08:47 AM, Magosányi Árpád wrote:
>
>
> 2017-07-04 19:58 GMT+02:00 Jameson Quinn <jameson.quinn at gmail.com
> <mailto:jameson.quinn at gmail.com>>:
>
>>     Before you jump straight into designing mechanisms, it's important
>>     to be clear about what you're looking for: the values you want the
>>     method to fulfill. On that matter, you've said the method should be:
>>
>>      1. not too shockingly new
>>      2. "proportional representation...
>>      3. ...and no entry threshold"
>>      4.
>>         The winning strategy for candidates is collaboration
>>      5.
>>         The winning strategy for voters is honest voting
>>      6.
>>         In the long run there is no two-party system
>>
>>     In the end, you're definitely going to have to compromise to at
>>     least some degree on points 1, 4, and 5.
>
>
> #1 is not a requirement in my work. There are two proposals: one is a
> basis for discussion for parties, which should be conservative, the
> second is an ideal system as the proposal of the movement, which can
> bring in anything new, but if possible should build on the previous one.
> I am talking about the later now.

I think the thing that comes closest to the properties above would be 
Schulze STV, though it's candidate based, not party based.

1. may not be satisfied all that well (Schulze STV is from somewhere 
between 2000 and 2010; I'm not sure of the exact year).

2. It provides proportional representation since it passes Droop 
proportionality - and it is Condorcet in the single-winner case.

3. It's not party based, so there are no thresholds.

4. Is probably not completely satisfied, but since Schulze STV satisfies 
weak invulnerability to Hylland free riding, there shouldn't be much of 
a risk of coordinated vote-management behavior.

5. For the same reason, there shouldn't be much risk of uncoordinated 
vote-management behavior, either.

6. Since Schulze STV reduces to LR-Droop when everybody bullet votes for 
a party, and D'Hondt (which is closest to Droop among the divisor 
methods) don't lead to two-party rule, Schulze STV should also be safe.

Any candidate-based method can be turned into a party list method by 
just replacing a party with every candidate in that party in order. E.g. 
if the parties are A, B, C, and their lists are A1 A2 A3, B1 B2 B3, C1 
C2 C3, then a vote of

A>B>C

becomes

A1>A2>A3>B1>B2>B3>C1>C2>C3.

Schulze STV is very complex, however, which could make it less 
attractive for practical use.

Alternatively, you could use Sainte-Laguë or modified Sainte-Laguë with 
leveling seats or (even better) biproportional apportionment. Modified 
with leveling seats seems to work pretty well in the Scandinavian 
countries, and in my opinion, replacing leveling seats with 
biproportional representation would only improve the outcome.