On 09/02/2016 07:10 AM, Richard Lung wrote:

I just read about an interesting idea in this vein by Brams. First you
pick your legislature (using whatever method), then once a majority of
the parties decide to form a government, the parties sequentially claim
ministries according to an order given by a divisor method.

For instance, suppose the following parties are represented in
legislature, with seat counts:

Party A: 55
Party B: 10
Party C: 7
Party D: 1
Party E: 48
Party F: 29
Party G: 10
Party H: 9

Now suppose that, to make things interesting, parties E, F, and H choose
to form government. Furthermore suppose there are 10 ministries, and the
method this legislature uses for the executive ordering is Sainte-Laguë.
Sainte-Laguë would apportion as follows:

First seat: E
Second: F
Third: E
Fourth: F
Fifth: E
Sixth: H
Seventh: E
Eighth: F
Ninth: E
Tenth: E

The share of seats is: 60% E, 30% F, 10% H, compared to the relative
legislature share of 56% E, 34% F, 10% H when only seats belonging to
those three parties are counted.

Anyway, the idea of Brams was to ask the parties (in order) to choose
ministries. So in the assignment above (EFEFEHEFEE), E would first be
asked to pick a ministry/department (and would probably choose the PM).
Then F gets to pick the next, E gets the next after that one, then F, E,
H, and so on.

The point of the procedure is that it would weaken the need to negotiate
for a share of the government. If party X is part of the majority
coalition that decides to form government, how many seats X gets is not
something the coalition can influence. Furthermore, since Sainte-Laguë
works like a proportional ordering, it is fair at every step of the
process (assuming that going first is always beneficial).

Presumably the same idea could be used for more advanced proportional
orderings, but the system would become more opaque.

Unfortunately, Brams then discovered that sometimes it does pay to not
go first, and he suggested that trading might help with it. One
possibility for trading could be that a party, when it's its turn to get
a ministry, proposes to exchange a free ministry seat with one that's
already taken. If the other party (who holds that other ministry seat)
agrees, they switch. Suppose, for instance, that at some point in the
process, E has claimed the ministry of justice, and the ministry of
education is free when F is to choose. F might then propose to E that F
gets the ministry of justice in exchange for E getting the ministry of
education, even though nominally the ministry of justice has already
been claimed.

I suppose Brams' system could be used for supermajority-support
executives, but then the confidence agreement would have to be supported
by a supermajority. Perhaps the system above would make this easier
since less negotiation would be required beforehand, and thus diminish
the chance of a deadlock or, looking at it another way, enable greater
supermajorities to reach a decision about what government to have for
the same risk of deadlock.

More complex apportionment systems could base the number of executive
seats on support among the voters rather than on legislative support.
However, it may be difficult to use that kind of apportionment, which is
fundamentally party-based, if the legislature is elected using
non-party-list methods like STV.

Hello Kristofer,
Your last paragraph is perhaps the bottom line (in the American idiom).
As you might know from the example, you set me, for my invention of
Binomial STV, and indeed from the 19th century inventors of STV,
elections, pure and simple, are a matter of choice, which is what the
word, elections means. That is to say the (STV) election system does not
presume to do the voters voting for them, by restricting choice to a
(party) group vote or a (monopolistic) single member district.
The first lesson I learned, from reading books on philosophy of science,
was not to presume what one is supposed to be trying to prove (that the
voters are herd partisans) and not to allow ambiguity in ones
experiments or tests, such as allows one to believe anything one wants
to: for example, whether or how much the single member constituencys
party-monopolising candidates have personal appeal.
My free ebook Scientific Method of Elections includes a fuller
discussion, as well as of Binomial STV, and many other matters.
from Richard.

On 14/09/2016 21:06, Kristofer Munsterhjelm wrote:

--
Richard Lung.
http://www.voting.ukscientists.com
Democracy Science series 3 free e-books in pdf:
https://plus.google.com/106191200795605365085
E-books in epub format:
https://www.smashwords.com/profile/view/democracyscience

On 9/14/2016 1:06 PM, Kristofer Munsterhjelm wrote:

Here is yet another approach:

http://www.negotiationtool.com/cabinetministers.html

The MPs (members of parliament) rank their preferences for possible
cabinet members, and then a variation of VoteFair ranking calculations
fills the cabinet positions.

It mostly uses the algorithms explained at VoteFair.org
(http://www.votefair.org/calculation_details.html).  Of course a few
modifications are needed.

Have questions?  Please ask.  But be patient because I'm writing a
time-sensitive election article that I will announce later.

Richard Fobes

On 9/14/2016 1:06 PM, Kristofer Munsterhjelm wrote:

On 09/16/2016 05:52 AM, VoteFair wrote:

I've always had the impression that ranked voting isn't going to work in
parliamentary settings because the MPs will feel obligated to strategize
maximally to get what they want, and there are few enough of them that
it is possible to do so.

In a system with strong party discipline, the party leaders or whips
could very well order their respective MPs to vote in a particular way.

Am I wrong when making that assumption?

Few random observations follow.

• Rankings may be too simplifying when forming a cabinet. I think it is typical that some grouping agrees on a program, and then one forms a (majority, multiparty) government/cabinet based on what was the best identified program and majority supporting that program in the parliament. A mechanistic ranking based approach to form the cabinet might be a too straight forward approach for this. Or maybe the idea was to form totally new kind of cabinets (whose structure could not be negotiated by politicians).

• But instead of electing directly the cabinet members, one could use ranked methods also to elect the best government coalition. The candidates could be e.g. A-B-C, A-B and B-C-D, where A, B, C and D are different parties.

• Also not using ranked methods might lead to ranking method style strategic voting. I mean that if the alternatives are A, B and C, and there is a ABCA preference loop, then it may be up to the chairman which alternative will win. If chairman's own party prefers alternative A, then it would make sense to first decide (using plurality) whether B or C is favoured. B wins. And then B will lose to A in the final plurality voting.

• I tend to think that Condorcet strategies are quite difficult to apply successfully even in environments where the opinions of all the voters are quite well known, and when there is a strong party discipline that can be used to force MPs to vote in some particular way (either against their own will, or to strategically support their own targets). There are strategical vulnerabilities (always, for sure), but I still have not seen any good article on how voters (individual voters alone, or party leaders with the help of party discipline) could successfully apply strategies in Condorcet elections. I'm waiting for someone to write it, or otherwise give general guidance on how the system can be cheated (without too much risk of bad results). The fact that sometimes it is possible to see afterwards that some strategy would have worked is not enough (assuming 100% information on how everyone will vote, and still with ability to change afterwards the votes of oneself and other members of one's own party). I'm assuming that there is no 100% accurate information on the preferences of other voters / parties, and there is no 100% accurate information on what strategies each party is planning to apply. So, where are the practical guidebooks if Condorcet methods are so vulnerable to all he strategies?

Juho

On 9/15/2016 11:52 PM, Kristofer Munsterhjelm wrote:

The existing and (other) proposed PR (proportional representation)
ranked voting methods I've seen are, indeed, vulnerable to strategic
voting, but that doesn't mean that all PR ranked voting methods also
have to have those vulnerabilities.

Strategies only work if the voting method is vulnerable to strategic voting.

I've designed VoteFair ranking and the software negotiation tool to
resist strategies.

This concept is analogous to gerrymandering in the sense that a
well-designed voting system cannot be skewed by gerrymandering tricks.

Strong party discipline is great for taking advantage of strategic
vulnerabilities.  If the voting method does not have strategic
vulnerabilities then party discipline loses its advantage, and MPs
(members of parliament) become free to express their real preferences.

I forgot to mention that in the example of using the software
negotiation tool to elect cabinet ministers, each "proposal" is for a
specific MP to be elected to a specific cabinet position.  The software
takes care of conflicts so that when an MP wins a specific cabinet
position then all other incompatible proposals, such as that MP being
chosen for other cabinet positions, are ignored.

Richard Fobes

On 9/15/2016 11:52 PM, Kristofer Munsterhjelm wrote:

On 9/16/2016 2:47 AM, Juho Laatu wrote:

Good point!

I could be wrong, but it seems to me that when someone points out a
vulnerability in a Condorcet method, it's an edge case.  And edge cases
seldom happen in real elections where there are more than a few dozen
independently acting voters.

I tend to think of election-method vulnerabilities in terms of how often
a specific vulnerability can/could occur -- rather than yes/no
checklists of whether a vulnerability is possible.

If this concept were converted into visual maps of where the edge cases
and vulnerabilities are, the difference between unfair methods
(plurality) and medium-quality methods and the best/fairest methods
would stand out clearly.

Richard Fobes

On 9/16/2016 2:47 AM, Juho Laatu wrote:

emb

One way to do this would be to say:
Each party gets to "bid" portions of their votes for each of the
ministries. The ministries are then allocated such that for each ministry
and party, you calculate the total votes for that party, divided by the
total "bid" that that party made for all the ministries it "wins", times
the "bid" for the ministry in question, and whichever party has that
quantity greatest "wins". (This might not lead to a stable solution, and if
not, you resolve the mess in favor of the largest party.)

Say there were 3 ministries. Party E can guarantee it gets ministry 1 by
bidding 30 for it; then, it can probably get 2 or 3 by bidding the
remaining 18 on one of them. Or, it can guarantee getting at least 2
ministries by bidding at least 15 on each, putting the final 3 on the one
it wants the most. Thus, party F should bid 19 on the ministry it wants the
most, and 10 on the one it thinks E wants least.

Vulnerabilities are often theoretical vulnerabilities in the the sense that if one group of voters could go back in time after they have seen how others voted, and change the way they voted, they would get a result that all the members of this group find better than the original result. Of course it is not possible to go back in time, or to allow only one afterwards identified group to apply a strategy, or not let other voters change the way they vote, and it is difficult to make each member of the identified group to change his vote in the way that the strategy requires.

I agree that plain lists of vulnerabilities of different methods to different theoretical vulnerabilities is not the best possible approach to estimate the quality of different methods in practical elections. Theoretical analysis and exact knowledge of different theoretical vulnerabilities is needed too, but if one wants to know which election methods would be best in different real life environments, it makes more sense to analyse which vulnerabilities might pose threats in such typical practical environments (when used by regular voters).

One approach to studying this kind of practical vulnerabilities is to try to formulate different vulnerabilities as guidance for practical elections ("if condition A then modify your vote (or votes of all similar minded voters) in way B"). That guidance could be intended for individual voters or for leaders that can study the situation and then give strategic guidance to similar minded voters. I have seen many sensible formulations e.g. on how regular voters should vote in Approval (in practical elections). But as already noted, none for Condorcet methods. This means that it is at least quite difficult to find such working practical strategies for real life Condorcet elections. If there are none, then it may be the best strategy to vote sincerely, which would be good news for practical Condorcet elections.

As I already noted, some basic assumptions for practical elections are that there is no complete information on the opinions of other voters and their planned strategies. One can not (in most cases) fully control how all the voters of some particular interest group will vote. And one must give some space also to possible counter strategies (or loss of support of some candidates because of their strategic plots), when other parties / voters learn that some others are planning to vote strategically. I'm still waiting for the first such practical rule for Condorcet methods. I think there are some situations where practical strategies could work, but such strategies tend to work only in some special situations or under some special assumptions (well known preferences, limited number of candidates, strict party discipline etc.). I'm not aware of any simple practical strategic voting rule that could be applied in all Condorcet elections or with only vague information of the preferences of the voters.

Maybe the question is, what if we would have Condorcet presidential elections in country x next month, do you have some strategic voting recommendation available to all or one of the parties or their members? You may assume availability of some rough poll information now (before you give the recommendation). Or even better if the recommendation is so generic that the voters can make their strategic voting decision based on the last (inaccurate, maybe multiple differing) polls that will be published few days before the election.

Juho