[EM] Proportional election method needed for the Czech Green party - Council elections

[Juho]

On May 3, 2010, at 3:51 AM, Peter Zbornik wrote:

> Dear all,
>
> if the single-winner president or the proportionally elected VPs
> sometimes are not a member of the set of proportionally elected
> council members (which is likely),

Possible but maybe not very common.

> then I would also like to ask you
> for a proposal on the last "conservative" method, thus it would not be
> optional, as I wrote below.  Thus, in all I ask you for three
> proposals.
>
> Motivation:
> I can imagine that there could be a method, which elected the rest of
> the council members after the the president and some or all vice
> presidents have been elected using proportional ranking (see the
> proposal of Markus Schulze for an example).

Word "after" sounds a bit dangerous because of strategic voting.  
Voters that have gotten their representatives as president and vice  
president may get a second set of representatives if the election of  
the rest of the council is independent of the election of the  
presidents. In order to maintain good proportionality in the full  
council one could reverse the order of the elections (council first)  
and limit the choice of the presidents to the council members, or use  
the same ballots to elect both presidents and the council. (I note  
that later on you seem to propose using the same ballots in both  
elections.)

If term "conservative" means "already widely used and tested in  
politics" then maybe proportional ranking based methods fall outside  
of this category. But if you allow some fresh winds then such  
"locking" methods could be used.

Since the first vice president seems to be a more important position  
than the second and later vice presidents similar locking could be  
used throughout the hierarchical chain of presidents. In my first  
proposal I locked only the president and let the vice presidents be  
equal. Proportional ranking (in methods that aim at electing good  
compromise candidates first) would do the same trick to all vice  
president positions. The other council seats are equal, so  
proportional ranking is not useful there. But since the distorting  
effect of such "compromise oriented proportional ranking" may be  
considered just noise in the last seats it is not impossible to use  
proportional ranking to elect all the presidents and council members  
at one go.

(What I mean by "distorting effect" is that if you have left, centre  
and right, and centre has less first place support than the other two,  
then a good approach may be to elect C if one elects only one  
representative. But if one elects two then one could pick L and R (to  
be proportional). This means that the proportional ranking (or  
locking) approach always makes a mistake, either in the case of one or  
two representatives. But in the case of electing the presidents it may  
be well justified to elect C as president (the most important job,  
expected to represent all sections of the party) and then elect either  
L or R as the first vice president. And the other one as third. Fair  
enough although the "team of two" is not proportional.)

>
> Such an election method would sacrifice a proportionality in the
> council in order to elect the president unambigously and to achieve
> proportionality between the president and some of the vice presidents
> (at least the 1st VP).
>
> The election of the rest of the council members would be done to
> maximize the proportionality of the elected council, maybe by using a
> modified version of STV, where the pre-elected president and vice
> presidents would be considered elected to the council at the start of
> the STV election, using the same ballots as for the  proportional
> ranking election. Less important VPs (like for instance a fourth VP),
> would be elected from the council.

As noted above, and if you want to emphasize simplicity, using the  
serial / proportional ranking approach to elect also the council  
members would not be a big distortion in the proportionality of the  
whole council. Note also that already electing the president outside  
of the ("otherwise to be") council would mean a minor (and not  
probable) distortion to the proportionality of the full council. My  
thinking is thus that if we want to serialize the election of all the P 
+VPs anyway, then one alternative is to use that same basic method all  
the way (since the resulting additional distortion will be smaller  
towards the end of the chain).

This kind of a method provides a complete proportional ranking of the  
elected council members. This is quite unnecessary towards the end of  
the list, but what is interesting at the beginning of the list is that  
there is no need to define the exact number of vice presidents since  
one can just pick as many of them from the chain as needed.

>
> Thus, the rest of the council members would be elected in such a way,
> that the council would be as much proportional as possible given the
> pre-elected P and VPs.
>
> The advantage of this method is, that there would be no ambiguity
> relating to the legitimacy of the elected president and the most
> important VPs.
>
> What methods would you recommend for this scenario?
>
> Would there be a software that could be able to handle this problem
> (possibly after some slight modification)?
>
> I think the scenario above (i.e.first proportional ordering of P and
> some VPs, then balancing election of the rest of the council to
> achieve proportionality in the council and finally possibly electing
> some less important VPs from the elected council) is the variant which
> would come closest to the way the elections are done in our party
> today, while attaining proportionality.
>
> Right now the scenario above seems to be the optimal solution, which I
> would like propose to the party.

Based on my discussion above I propose to consider both approaches,  
locking the presidents one by one and then using some proportional  
method to determine the rest of the council (using the same ballots)  
as you propose, and using one proportional serial method to elect all  
members of the council.

If the requirement of conservative / already used methods (like STV)  
is not strict, then one could well use some Condorcet method as a  
basis in the serialization. This would also remove some the STV  
elimination process related problems. Markus Schulze proposed a  
Schulze method based proportional ranking approach for some uses. Also  
others like Ranked Pairs based are possible. I'm interested also in  
use of minmax(margins) due to its simplicity, explainability and good  
behaviour with sincere votes. But on this question I might be in the  
minority on this list, so also others than minmax(margins) must be  
ok :-).

I ask the members of this list to comment if the benefits of using one  
serial proportional method to elect both P + VPs + council members  
outweigh the problems of using that approach. (This may mean also use  
of and problems of "serial proportional Condorcet" vs. STV vs. "full  
proportional Condorcet" (e.g. CPO-STV) in different parts of the  
method.)



Some further notes.

I note that the criteria (of individual voters) to elect the council  
member vs. criteria to elect the presidents and vice presidents may be  
somewhat different. Maybe this difference is so small that there is no  
big problem to use the same ballots for both purposes. If that is a  
problem, then maybe election of the presidents among the council  
members after electing the council would do.

The serial proportional approach that locks the choices one by one  
(starting from the most important one and then continuing downwards)  
suits very well the need to elect the presidents. I note that the  
requirements are somewhat different if this approach is used in the  
primary elections to elect the candidates for the actual election. I  
understood that in the parliamentary elections the used election  
method is closed list based but has also some open list  
characteristics (http://electionresources.org/cz/). For the closed  
list part the order of the candidates needs to be determined. A serial  
proportional approach (proportional ranking) yields one such order. As  
noted above (in the L-C-R example) the serial approach and  
proportionality are in conflict. But the open list method forces us to  
use one. The list could be optimized so that it gives the ideal result  
with the number of representatives that the party expects to get. One  
would thus ideally not start the locking process from the first  
candidate but by first determining the set of the expected number of  
candidates and then work out the the shorter and longer lists. I  
believe the Czech Green Party is trying to get its first seat, so in  
this case the optimization could start from one representative, but in  
the future things might be different. The problem is that if one  
expects to get two representatives, then the two first candidates on  
the list should maybe be L and R and not C and A. If the numbers of  
representatives is higher these problems will mostly fade away.

I hope the number of candidates is within reasonable limits, so that  
the approach of ranking all (or all relevant) candidates is not too  
tedious to the voters. There are different ballot formats with  
different properties (and different possibilities to read the ballots  
by computers).

You mentioned also the possibility of additional requirements (in the  
P.S. mail). It is possible e.g. to elect certain minimum number of  
males and females. In the serial approach that was discussed above one  
simple approach would be to just eliminate all remaining male or  
female candidates at some appropriate in the serial process when all  
the remaining representatives must be of same sex. That approach would  
maintain proportionality is all dimensions quite well. Note that many  
political elections implement also regional proportionality in  
addition to political proportionality. Also within the party one could  
require that all regions are represented well enough. One could in  
principle follow the same approach as with males and females, but if  
one adds too many such rules then the election of last representatives  
may have some problems like some not having any northern young females  
left. But that is already another topic that we might discuss later.

>
> It wasn't easy to arrive to the model above, but now I think the
> problem is well-defined.
>
> Best regards
> Peter Zbornik

Thanks for sending more information on your preferences (and expected  
preferences of the Czech Green Party). This helps at least me in  
balancing the different requirements. For example the importance of  
the first vice president position and the fact that use of one ballot  
is possible influenced my text above. Now I'm wondering how to balance  
the requirements of simplicity and ease of understanding of the method  
vs. use of conservative methods.

Juho


>
> 2010/4/29, Peter Zbornik <pzbornik at gmail.com>:
>> On Thu, Apr 29, 2010 at 2:45 AM, Juho <juho4880 at yahoo.co.uk> wrote:
>>
>>>  On Apr 28, 2010, at 9:19 PM, Peter Zbornik wrote:
>>>
>>> Hello,
>>>
>>> I have some catching up to do here.
>>> I need to think more about some of the different methods and the
>>> proposals
>>> I have gotten.
>>> Some of the methods are new to me.
>>> As I am a layman it takes time to understand them.
>>>
>>> Condorcet methods have not been used in politics yet, I think.
>>> Are there by any chance other methods to elect centrist presidents?
>>>
>>>
>>> The most common single-winner methods (plurality, top-two runoff,  
>>> IRV)
>>> are
>>> known not to be "centrist oriented". For example in the quite common
>>> set-up
>>> where we have two extremists and one centrist with slightly less  
>>> support
>>> they all elect one of the "extremists".
>>>
>>> Approval and Condorcet are more centrist oriented. They are both  
>>> also old
>>> and well tested methods although they have not been widely used in  
>>> public
>>> political elections.
>>>
>>> One problem with Approval (that was not mentioned yet) is the  
>>> limited
>>> expressive power of the Approval vote and resulting problems in  
>>> choosing
>>> the
>>> right strategy, e.g. when there are three leading candidates and one
>>> should
>>> approve either one or two of them. That is problematic e.g when  
>>> left wing
>>> has 2 candidates and is is bigger than right wing that has 1  
>>> candidate.
>>> (In
>>> Approval voters are generally assumed to vote strategically and  
>>> not just
>>> sincerely list candidates that they consider "approvable".)
>>>
>>> If I have understood the discussion correctly, you are still in  
>>> draft
>>> phase.
>>>
>>>
>>> Many of the presented drafts can be quickly made into concrete and  
>>> exact
>>> proposals. Some of the used components are more mature / well  
>>> tested than
>>> others. There are some details left to decide, e.g. which STV  
>>> variant (/
>>> which proportional method) or which Condorcet variant (/ which
>>> single-winner
>>> method) to use.
>>>
>>> When some of you feel you have a good proposal at hand in this
>>> discussion, feel free to summarize and put forward a draft.
>>>
>>>
>>> I'm trying t listen to you to understand which one of the  
>>> proposals would
>>> be the best for your needs :-). Now my understanding is that you  
>>> still
>>> want
>>> all the listed requirements to be met and you want to use as "well
>>> tested"
>>> (and simple/explainable) methods as possible.
>>>
>> Yes, the requirements are set.
>> People have to understand what they are voting about, if they are to
>> support
>> the method.
>> The people who like the majoritarian system can be expected to  
>> spread FUD
>> (fear uncertainty doubt).
>> It has at least to be a method in use in some organizations.
>> Maybe I shouldn't have excluded Schulze STV right away, since it is  
>> in use
>> at some places.
>>
>>
>>> Are you btw ok with the idea of limiting the choice of P+VPs to the
>>> (already elected of simultaneously elected) council members?
>>>
>> Yes as one variant
>>
>>>
>>>
>>> Using the same ballots with different vote counting techniques  
>>> seems like
>>> quite an elegant and interesting solution.
>>> I never thought of that possibility.
>>> If more than one ballot is to be used, then I prefer to having the  
>>> P and
>>> VP elected before the councilmembers.
>>>
>>>
>>> There is a small problem here. If one elects the P and VPs first,  
>>> then
>>> the
>>> voters may not give the already elected P+VPs any votes in the  
>>> second
>>> (council) round for strategic reasons (or maybe they are not even
>>> candidates
>>> there any more but considered "already elected"), and as a result
>>> the groupings of P and VPs would be over-represented in the  
>>> council. This
>>> is
>>> not ok if you want the council (that includes P+VPs) to be  
>>> proportional.
>>> (For this reason my first draft used the same ballots for all  
>>> elections
>>> and
>>> the second draft elected the council first and P+VPs among the  
>>> council
>>> members only after that.)
>>>
>>>
>>> Just to avoid misunderstandings:
>>> The president is the party leader as in most political parties  
>>> around the
>>> world.
>>> He is the main guy simply, the person appearing on television  
>>> etc., the
>>> one
>>> people know best in the streets.
>>> The president also chairs the meetings of the national council  
>>> (sometimes
>>> I
>>> have used the term "board", the meaning is the same in this  
>>> context).
>>> Think Gordon Brown or Angela Merkel, small scale :o)
>>> We have discussed splitting the external and internal funcions of  
>>> the
>>> president, but it is not politically feasible to do.
>>>
>>>
>>> Two separate jobs then, and someone might be competent for one job  
>>> and
>>> someone else for the other. This would make the election process  
>>> more
>>> complex. The simplest approach would be to elect these two persons  
>>> among
>>> the
>>> council members in two separate elections and forget  
>>> proportionality with
>>> respect to these two jobs. VPs could still be elected  
>>> proportionally (but
>>> they could be close to the two Ps => the set of Ps+VPs is not fully
>>> proportional).
>>>
>>> It is neither feasible to change the role of the president to be a
>>> stricly internal guy or to have him elected by the council.
>>> The president has to be elected by the delegates as is the case  
>>> today, if
>>> the proposal should have a chance to pass.
>>>
>>>
>>> This should be ok. (No limitations on who can vote although in some
>>> scenarios the set of candidates must be limited to the already  
>>> elected
>>> (or
>>> simultaneously elected) council members.)
>>>
>>>
>>> There was some discussion on if it is ok to elect a variable  
>>> number of
>>> VPs
>>> (exact number not known beforehand) or if it is more acceptable to  
>>> modify
>>> the traditional methods a bit (an "innovative" addition).
>>>
>> Well normally, the VPs are variable and are not regulated by the  
>> party
>> stautes, thus not regulated by any method which should be specified  
>> in the
>> statutes.
>>
>>
>>> The option of allowing some deviation from full proportionality in  
>>> the
>>> Ps+VPs set (while the council would still be fully proportional)  
>>> is one
>>> more
>>> alternative to consider (this would keep the methods bit
>>>
>>  more traditional / less "innovative").
>>>
>>> The rule of electing a compromise president may in some cases  
>>> distort the
>> proportionality a bit anyway ->(in all scenarios), but the most  
>> common case
>> (if we allow extra deviation) would probably be one where >one of  
>> the VPs
>> would be close to the P, giving that section double-representation  
>> in the
>> P+VPs set (but >not in the council). You may have to pick one of  
>> these
>> problems :-), a varying number of VPs, some >innovativeness in one  
>> of the
>> methods, some (more) distortion in the P+VPs set proportionality,  
>> or maybe
>>> some other new (slightly problematic) solution.
>>
>> It is indeed appropriate, that at least the first VP is from the  
>> "opposite
>> side".
>> This gives the party peace and quiet in the party and appropriate
>> "balance/division of powers".
>> Maybe one might argue, that an election method without this  
>> "balance of
>> power" is not truly proportional, since not all council members are  
>> equal.
>> I would personally prefer complete proportionality, but I am not  
>> sure how
>> it
>> would be greeted.
>>
>> On the other hand, a representative council would be a vast  
>> improvement
>> even
>> with non-proportionality in the P an VP.
>>
>> Just to comment on the competitiveness in the party. I would say it  
>> is
>> pretty high. That is why we have the infighting going on. I am not  
>> able to
>> quantify hovever :o)
>>
>> So I guess we would have two-three proposals in play:
>> 1. optimal with proportionally elected P and VPs after council  
>> elections,
>> 2. "traditional" (without proportionality in the P. and VPs. set)  
>> elected
>> after council elections.
>> 3. (optional) "conservative" (preserving the order of election today)
>> - elect the P. and the VPs (either one-by one or together or first  
>> P and
>> VPs
>> together), Then elect the rest of the council members - using a  
>> method
>> which
>> is not overly complicated, while possibly sacrificing proportionality
>>
>> I think it is good to have several variants, in order to check them  
>> with
>> some other members before proposing one specific model.
>>
>>> This might already get too complex, so maybe your proposal to  
>>> propose some
>> complete solutions to >the problem (and list their benefits and  
>> problems)
>> after this discussion (that hopefully clarifies the >requirements)  
>> could be
>> the simplest way forward.
>> I aggree, lets not make it overly complex and let us round up the
>> discussion
>> with some specific proposals.
>> A brand new method, however promising, would first need to be  
>> tested and
>> evaluated using some real-life ballots, so I couldn't propose it to  
>> be sent
>> right into the statutes.
>> I guess that is standard when introducing any new "technology", to
>> have alpha and beta tests.
>> Those of you, who have sent me recommendations of methods only in  
>> private,
>> please send your specific proposals publicly.
>>
>> I don't know if there is a "standard public database" of real  
>> ranked-order
>> or range ballots on which the different methods could be tested.
>> Such a database exists in the data-mining field:
>> http://www.kdnuggets.com/datasets/
>> Maybe that could be something to consider for benchmark purposes.
>>
>> Maybe we could test some of the more "innovative" methods in the  
>> party on
>> real ballots.
>> This would allow us to evaluate.
>> Such a test would need to be consulted with the party first and it  
>> would
>> take some time.
>>
>> A motivation why PR systems are better to use in organizations than
>> majority
>> systems, might be of help.
>>
>> I am overwhelmed of the interest in this request.
>> I quite didn't expect such response, so thanks.
>> Please do not have too high hopes on me, it might as well end up  
>> that in
>> the
>> end, it will be decided to leave things as they are.
>>
>>
>>>
>>> Juho
>>>
>>>
>>>
>>> The vice president's are ordered first and second and third. The  
>>> number
>>> of VP chan vary. The VP are the ones who stand in for the  
>>> president or
>>> party
>>> leader (in that order).
>>> The president and the vice presidents are all member of the board,  
>>> which
>>> currently has seven members.
>>>
>>> Best regards
>>> Peter Zborník
>>>
>>> On 4/28/10, Jonathan Lundell <jlundell at pobox.com> wrote:
>>>>
>>>> On Apr 28, 2010, at 7:34 AM, Raph Frank wrote:
>>>>
>>>>> On Wed, Apr 28, 2010 at 4:47 AM, Juho <juho4880 at yahoo.co.uk>  
>>>>> wrote:
>>>>>> You assume that there is only one VP.
>>>>>
>>>>> Well, if more than 1 VP is possible, then the election could be
>>>>>
>>>>> - Elect council with PR-STV
>>>>> - The condorcet winner (only including the councillors) is  
>>>>> President
>>>>> - Elect 2 of the councilors as VPs using PR-STV
>>>>>
>>>>> However, there is still the question of what exactly the VPs and
>>>>> President is supposed to do.
>>>>>
>>>>> If they are to chair council meetings, then it is better to just  
>>>>> elect
>>>> them.
>>>>>
>>>>>> We could have also two and keep track
>>>>>> of which members are elected first, second and third.
>>>>>
>>>>> I still disagree with using order of election in a PR-STV  
>>>>> election.
>>>>> It provides an additional incentive for dishonest rankings.
>>>>
>>>> It also encourages strategic nominations. The whole idea of using
>>>> order-of-election in STV for *anything* should be drowned in a  
>>>> bathtub
>>>> ASAP.
>>>>
>>>>>
>>>>> It would have most of the same problems that plurality has where  
>>>>> you
>>>>> need to vote for one of the top-2.
>>>>
>>>>
>>>> ----
>>>> Election-Methods mailing list - see http://electorama.com/em for  
>>>> list
>>>> info
>>>>
>>>
>>> ----
>>> Election-Methods mailing list - see http://electorama.com/em for  
>>> list
>>> info
>>>
>>>
>>>
>>> ----
>>> Election-Methods mailing list - see http://electorama.com/em for  
>>> list
>>> info
>>>
>>>
>>
>
> -- 
> Odesláno z mobilního zařízení