better multiwinner method for The Netherlands

[Victor Beffers]

Hi all!

I've been looking at several methods described in this group and elsewhere, hoping to improve the situation of the election of the parliament of the Netherlands but haven't found a satisfying one yet.

Currently:

- every voter has one combined vote for a candidate and for the party that this member belongs to

- there are 150 seats

- last time we had 1114 candidates from 28 party-lists; 13 parties made  in in parliament with: 33, 20, 19, 19, 14, 14, 9, 5, 4, 4, 3, 3 and 2 seats ;

  4 parties with a combined 76 seats make up the goverment (they determine ministers and prime-minister)

problems:

- voters don't know what coalition of parties they make more likely with their vote to become the goverment.

- while 49% of the seats has support from 49% of the people and can talk and vote in parliament but they are usually ignored by the goverment (except for a few small suggestions with almost unanymous support)

- many voters and nonvoters feel disconnected and discontent from the goverment and increasingly from the democratic process itself

- the quality of the public discussions and debates in parliament is very low and more and more polarizing. Oftentimes good ideas from smaller parties are not seriously considered or understood.

- many voters betray their favorite party out of strategic behaviour ( they consider a vote on a likely coalition-party worth more )

Requirements:

- any proposal that (obviously) hurts smaller/bigger/center/extreme parties will not get the needed almost unanimous support from all parties to change the system

- enough buit not more-then-needed changes to make a difference towards more fairness and voter-satisfaction 

maybe reweighted rangevoting?

pro's and con's?

any thoughts/ideas?

thanks!

[parker friedland]

Approval list voting:

Voters get to chose 1 party, and approval vote for any of the candidates withen that party (discribed here: https://www.kialo.com/the-us-should-adopt-a-better-voting-system-for-elected-bodies-5589/5589.0=5589.72)

ALV solves what I believe to be the biggest problem with PLV (party list voting), and that is the closed party lists. Allowing voters to controll how candidates rank on their party list rather then making party oligarchs make those decisions should help the voters feel less disconnected with the goverment.

[Victor Beffers]

The Netherlands already has Approval list voting..

I can assure you this does not even remotely adresses the problems that I (partly) described above.

Party-discipline in the netherlands for voting is above 99%. There are ideas to make splitting of from the party even more difficult. A bigger role for more independant candidates is a long way from happening here. Every now and then there is a candidate that is put unelectable low on the partylist that get's in by preference vote. Maybe 1 or 2 in each parliament. By itself that is a good thing that I like but it is no way a solution for the breakdown of democracy that I see happening. I think we need much more drastic change by subtle means to find a better, fairer, more durable system...

There is one party that has a subgroup that's considering Mixed Member electorla system. It's a dead duck in teh water. No change it'll happen because it would mean less power for the parties so they will stop it.

Thanks for your thoughts though!

[parker friedland]

I checked wikipedia and acording to the wiki, the netherlands uses normal open list voting, which is party list voting + SNTV, not party list voting + approval voting. The difference is that in normal OLV, voters only chose 1 candidate from their favorite party. As a result, there is vote splitting withen the party lists and popular candidates get more votes then they need and are not compinsated by having their faction of the party take up a larger portion of the party list. This means that the centroid of public opinion amoung voters in that party may not match the centroid of public opinion amoung the elected candidates from that party. Factions of the party can also be strategic by encouraging their supporters to evenly split their votes amoung a slate of candidates.

Proportional approval voting has been used in Sweden before Sweden switched to PLV, and voting methods that have been tried in goverment elections already tend to be less scary to voters. But like RRV, PAV highly encourages the strategic voting behavior known as free riding. This is because voter's votes are not as deweighted when they don't vote for popular candidates whom they believe are garenteed to win anyways.

ALV can however be mixed with proportional approval voting, so that voters can approve of what ever candidates they want regardless of party and by doing so they would also be approving of each party they approve a candidate from. Then each party's candidates are ranked by approval rating and PAV is used to determine how many seats each party gets. Thus they will only have influence over the parties in which they approve of. This combination is much better then normal PAV because when approving of a party, you don't need to worry about whether they are gareenteed to win a seat because even if they are, approving them will just help them win more seats. However this combination of ALV and PAV is more complicated then either ALV or PAV, so you might want to just stick with one or the other.

[Victor Beffers]

I kinda understand what you say but it doesn't seem very relevant for The Netherlands in my opinion. Off course I understand your enthousiasm for your suggestion but I think it may only improve things in a US-context.
With 28 partys to choose from, within-party diversity is very limited. Also party-websites and public voting-advice-website (based on f.i. 30 topics with answers from all parties) offer no room to communicate any individual candidate or subfraction opinons. (Only party-positions) The TV-debates are between top-ranking candidates of the major parties. There is very little room for within-party-policy-differences (candidates that are more autonomous would not go up in the years climbing in the rankings within those parties or they start their own new party and make sure that the new partylist is loyal to the top of that party). Any serious campaigning for a candidate is unthinkable here. A very small percentage of voters does know about the subtle differences in opinions (but know that when it comes to voting, there's almost always party-dicipline). More common are gender character or regional preferences but I think those are less important then in a district-system. It realy is about party-differences.

[parker friedland]

> many voters betray their favorite party out of strategic behaviour  ( they consider a vote on a likely coalition-party worth more )

There are two main types of favorite betrayal in party list voting methods:
1. Voters don't want to vote for their favorite party because they are afraid that their favorite party is too radical to be able to form a coalition with more moderate parties or vise versa with a moderate party being to moderate to form a coaltition with a more radical party
2. Voters don't want to vote for their favorite party because their favorite party is too small to surpass the threashold needed to win any seats at all.

I believe you are talking about the first type. Unfortunitly I don't believe any simple voting method can adress this problem without changing more then just how members of parliament are elected. The only solution to this is that I can think of is a complete overhall of the parliamentery procedure where candidates competing for prime minister propose goverments before the election (a list of how every single position is filled) and during the election, either voters directly vote for their favorite prime minister + goverment with a utilitarian voting method or the parliament members elected vote for the prime minister + goverment with a utilitarian voting method. No coalitions needed. If your favorite party is in 100% agreement with you, then there is no reason not to vote for them because your prefered prime minister + goverment has to except votes from your party. However I'm not sure whether such a system would even be classified as a parliamentery system anymore and I don't expect the Netherlands to acept a change to their goverment as radical as this one.

Unlike the first type of favorite betrayal, there is a simple solution to the second type. First award parties a seat for every 151th % of the vote they get and then for each party make the candidate next in line to get elected the holder of that party's remaining votes. Then allow those candidates to trade their votes amoung each other in a process known as asset voting (https://www.kialo.com/the-us-should-adopt-a-better-voting-system-for-elected-bodies-5589/5589.0=5589.58) for a specified period of time. After that time is up, elect the remaining candidates with the most seats. I'm not a huge fan of delegated voting methods but in this case the use of asset voting is very limited and it preserves the simplicity of party list voting.

[Victor Beffers]

Yep nr. 1 is one of my need-to-solve concerns, nr 2 is more of a nice-to-solve addition.

Asssuming that you mean 'radical' in a statistical way. Currently most smaller partys are not extremists but usually they are the first to realistically prioritize certain themes: environment, privacy-rights, tax-policy or they come from a certain religious or cultural background etc.

Interesting suggestion!

I'm going to think about that some more.

Probably a big overhaul is needed like you say (and probably too much of a change for voters and partys) but perhaps an in between solution that would be acceptable is possible.

Op maandag 13 augustus 2018 23:24:51 UTC+2 schreef parker friedland:

[Sara Wolf]

We're working on STAR-PR here in Oregon, USA for multi-winner elections. It's similar to Re-weighted Range Voting but with a runoff added to help make the system more resistant to strategic voting. (If that turns out to even be a concern.) We're digging into the science to try and get better data of how those 2 options compare, and also how variations to the algorithm effect results and voter behavior. But we're confident that the 0-5 ballot is the way to go. This is the multi-winner version of STAR Voting, which is definitely top shelf for single winner elections. 

This is an article from our website on it. We don't have proportional representation here in Oregon yet so this digs into the issue and a few different points of view on the options. 
https://www.starvoting.us/pr

[parker friedland]

> but with a runoff added to help make the system more resistant to strategic voting.

I strongly reccomend that you revise your STAR-PR proposal so that it only uses a runoff after the final round of re-weighting i.e. elect the first N-1 winners with RRV, and then use STAR during the finnal round. Otherwise the runoffs will have an efect oppisite of the one they are intended to have.

If you are going to push for STAR-PR in Oregon, please advocate for the revised STAR-PR algorithem (or any other proportional STAR algorithem that is both monotonic and actually proportional)

[Sara Wolf]

That is the leading algorithm option. As I've said, we are studying the effects of algorithm variations to find the best and most simple option. 

[parker friedland]

> That is the leading algorithm option. As I've said, we are studying the effects of algorithm variations to find the best and most simple option.

Great. However the page you linked to says otherwise:

> Voters would still give each candidate a score from 0-5. The two highest scoring candidates are finalists, and the finalist that was preferred by more voters wins the first seat on the council. The
> ballots are then recounted again with the remaining candidates until each seat is filled.

That discribes the old STAR-PR method in which there was a runoff every single round rather then just in the last round after the final ballot reweighting.

> STAR-PR hasn't been tested in a real world government election. So far it's been tested in some non-governmental elections and in simulations.

I look forward to seeing those simulations.

[Victor Beffers]

I'm still contemplating but wondering whether STAR-PR is worth the added complexcity if the first 149 are like RRV and only the last 150th seat is STAR-voting.
At first glancing over the arguments in other threads is seems there are very big differences between 3-7 candidates for 1-3 seats even within a method with the same name (f.i. STAR or RRV). And having 1100 candidates from 28 parties for 150 seats seems a whole other ballgame.
I very much like the practice of just one round (coming from STAR), an extra round would be much harder to get supported in NL (i know some countries like france are already used to it). I think Dutch voters who are emotionally used to have just one favorite as a vote would have a problem acceppting/using approval of several partylists (and 1 candidate). Then score-voting (RRV) seems much more acceptable: they can give their favorite party a MAX-score. Without wanting to make it like a complete ranking system or a forced point distribution (like songfestival) I would like to prevent a min-max strategy and force some spreading of scores.
I'm contemplating something like this:
1)give at least 5 or maximum 28 of the parties a score 0-99 or X.
2) a specific score can be used only once
(except X)
3) at least 1 score between 60-75
4) at least 1 score between 25-40
5) from the party with your highest score choose 1 favorite candidate (or leave blank)

Thoughts?

[parker friedland]

2) a specific score can be used only once

Well that rule would introduce alot of strategic voting. If it were added to single winner STAR voting, then voters would be encouraged to give a front runner candidate their highest score, and thus they can't give that highest score to their favorite candidate as well. The whole point of the rating system is so that 1 candidate's rating is independent of anouther candidate's rating. If you add a rule that requires voters to only use each score once, then your voting method is effectively one in which voters rank all the candidates when the number of possible ratings equals the number of candidates.

3) at least 1 score between 60-75
4) at least 1 score between 25-40

I would advise against introducing ridiculous rules like this to a voting method without fully understanting what the consiquences of those rules are. Rules designed to prevent dishonesty often result in even more dishonesty just to get around those rules. When designing a voting method, I tend to fallow the logic of occam's razor, which is that archacic rules like that are usally bad.

[Victor Beffers]

I taste from you reasoning that you are used to a completely different culture, system and context (the US one) than the one I know (the Netherlands). This collision/meeting is interesting. I think most Dutch voters would agree that at least 75% of the parties add some value to the whole of the parliament. Off course they do differ in their opinions between prefered proportions between them. So that's what I'm searching for.
a) acceptable changes in a foreseeable future (so no two rounds, no presidential election, no party-less system, no single or few seats from local districts combined into parliament, no Mixed member system)
b) more perceived emotional expressiveness for voters
c) something fair is being done in the translation from votes to parliament-seats
d) not threat to party-system
e) minimally possibility of strategic voting
f) force a cognitive state that is not just black and whole but more nuanced
g) make it possible to be very nuanced about all parties (but not compulsary completeness for all voters)
h) encourage parties to not only have fans but also different degree of lesser support.

remember: 150 seats, 1 district, ca 28 parties, >=4 parties to get >75seats (=government)

thoughts anyone?

[parker friedland]

Well, STAR-PR, RRV, and PAV are all designed to be used in small multi-winner districts. If you tried to use any of those methods in a single 150 seat district, you would need a giant ballot. When your not using plurality voting, for every seat in an election, there are usally at least 3 candidates competing for that seat. So if you were to use a method like STAR-PR, RRV, or PAV in a single 150 seat multi-winner district, you would have to have a ballot containing atleast 450 candidates on it, and if you were using STAR-PR or RRV, you would also have to provide space for voters to rate each of those (probably alot more then) 450 candidates. And calculating the results under any of those methods ... that would be a statistical nightmare. The Netherlands does allow you to directly vote for candidates through their open party system but in order to avoid a ballot with 500 candidates on it, they assign candidates to party lists based off of their preformence in small multi-winner districts. Thus the number of seats each party wins is still determined by their preformence across the entire country but the number of seats a party is assigned is split up amoung various multi-winner districts and in each of those multi-winner districts the top n candidates from s party with the most votes win the first n seats allocated to that party to that district. Party agnostic voting methods like STAR-PR, RRV, or PAV ideally shouldn't be used in districts with more then 10 candidates, so in order to use these methods in the Netherlands, you have to sacrafice a bit of their electoral system's perfect proportionality by dividing the country into multiple multi-winner districts. You can also add an MMP-style top up to preserve some of that proportionality but I would advise against doing so because MMP isn't very frendly to independents and an extra MMP top up is would be anouther added layer of complexity.

[Victor Beffers]

not precisely: yes there are 20 districts but almost all parties have 100% the same candidate list in all 20 districts. A few parties have a few regional candiates lower on their list. However 99% of all 1100+ candidates are the same for the whole of The Netherlands, so it's best to consider it as single district with 28 parties which looks like this: http://www.amstelveenweb.com/afbeeldingen/2017-stembiljet-dignahof-15maart.jpg

[Victor Beffers]

in my contemplated system from above I would not suggest scoring all 1100+ candidate (or not even just a few of them).
I really intend to score-vote the 5-28 parties not the individual candidates. On top of that for the heighest scored party that voter may give just one prefered candidate (which is in dutch context practically irrelevant except maybe a feel-good-voter).

[parker friedland]

Well it's a good thing they use open party list voting because it would be nearly impossible to run PAV, RRV, STAR-PR, STV, etc. in a 1100+ candidate 150 seat district.

[Victor Beffers]

Yes, in theory the voters have some influence over the order in which candidates from a party-list are elected but in practice with parties winning many seats (instead of 1 or max a few seats) and the presented list is made up by the party itself (so only party-loyal candidates), it really functions more as a closed list. While I agree on theoretical grounds that's not a system I would design for a blank sheet situation but in NL it really is a lost battle to try to change this.

[parker friedland]

> in my contemplated system from above I would not suggest scoring all 1100+ candidate (or not even just a few of them).

> I really intend to score-vote the 5-28 parties, not the individual candidates.

That should make things a lot easier however since you are electing 150 seats, there is going to be a lot of reweighting which means that it would be impossible to run this election by hand. With voting methods that use reweighting, there is always a number that you can multiply all the ballot weights by to get rid of all the fractions but I'm sure that in an election with 150 seats, that number would be absurdly large. PAV definitely seems like a better choice for an election of this magnatude.

On Tue, Aug 14, 2018 at 2:38 PM, Victor Beffers <vbef...@gmail.com> wrote:

in my contemplated system from above I would not suggest scoring all 1100+ candidate (or not even just a few of them).
I really intend to score-vote the 5-28 parties not the individual candidates. On top of that for the heighest scored party that voter may give just one prefered candidate (which is in dutch context practically irrelevant except maybe a feel-good-voter).

--
You received this message because you are subscribed to a topic in the Google Groups "The Center for Election Science" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/electionscience/2P39eMRktcA/unsubscribe.
To unsubscribe from this group and all its topics, send an email to electionscience+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Kevin Baas]

I have ran computer simulations. The fairest outcomes happen under strict mathematical observance of "one person one vote".

* the total points each person allots is made the same, by multiplying so that it equals a constant. (equality)

* points that contribute to the election of a candidate are removed from that voters remaining points, in proportion, such that after all candidates are elected, no points remain uncounted or unused. (transference)

[NoIRV]

Just use Asset Voting. No bother with complicated RRV or STARPR rules. Just say "mark 5 candidates" (or however many winners the district allows) and provide enough time for negotiations.

Candidates should be allowed to bring calculators into the negotiation room.

[parker friedland]

> I have ran computer simulations.  The fairest outcomes happen under strict mathematical observance of "one person one vote".

> * the total points each person allots is made the same, by multiplying so that it equals a constant. (equality)

> * points that contribute to the election of a candidate are removed from that voters remaining points, in proportion, such that after all candidates are elected, no points remain uncounted or unused. (transference)

Can I see these simulations?

[Sara Wolf]

Parker, you're right that the description up on the site describes STAR-PR with a runoff for each round. We're confident that would be a good, accurate reform. Still, many/most of us believe that that standard could be maintained with the simplified version as well. Before we move forward to a proposal ready for govt. elections we plan to have solid simulations to back up those theories. Once we have data to support a change we will update our recommendation and the website if needed. 

Simulations are out of my department and our core group is focused on our ballot initiative for single winner STAR through November, but I would welcome people to dig into that kind of science, specifically comparing those 2 variations of STAR-PR as well as RRV and other types of algorithms Jameson is interested in for a 0-5 ballot. 

Aside from determining the perfect algorithm, I wouldn't support these kinds of systems being used on huge multi-member districts. I'm concerned that over 5 seats per district or so would allow for too small of thresholds, and too many candidates for voters to learn about. I also think that geographical representation is important. PR is complex and there are a lot of considerations for a good proposal. 

[parker friedland]

> Still, many/most of us believe that that standard could be maintained with the simplified version as well.

This is not about simplicity. What are you going to say when someone asks you why STAR-PR is proportional? If someone were to ask me why PAV or RRV is proportional, my response would be that any group of voters has the ability to guarantee that their party wins at least the number of seats that it proportionally deserves (±1 seat). This is always true in PAV or RRV. It is sometimes true in STAR-"PR". However if you only used a runoff after the final round of reweighting, this would also always be true of STAR-PR as well, which would make it a lot easier to explain why STAR-PR is proportional. Without this revision, I expect individuals to begin to put STAR-PR's proportionality into question.

And there is an added bonus: When using only one runoff after the final round of reweighting, STAR-PR is monotonic (by CES's definition of monotonicity). This, however, is not the case in the current STAR-PR method.

On Thu, Aug 16, 2018 at 6:14 PM, parker friedland <parkerf...@gmail.com> wrote:

> Still, many/most of us believe that that standard could be maintained with the simplified version as well.

This is not about simplicity. What are you going to say when someone asks you why STAR-PR is proportional? If someone were to ask me why PAV or RRV is proportional, my respoce would be that any group of voters has the ability to garentee that their party wins the number of seats that it proportionally deserves. This is allways true in PAV or RRV. It is sometimes true in STAR-"PR". However if you only used a runoff after the final round of reweighting, this would also allways be true of STAR-PR as well, which would make it alot easier to explain why STAR-PR is proportional. Without this revision, I expect individuals to begin to put STAR-PR's proportionality into question.

And their's an added bonus: When using only one runoff after the final round of reweighting, STAR-PR is monotonic (by CES's definition of monotonicity). This however is not the case in the current STAR-PR method.

--

[Sara Wolf]

That sounds compelling and lines up with my own thought experiment and conclusions, but I'm not a math person and I can't say I fully get all the criteria, how the details effect things, and what you actually mean by proportionality when voters show support for multiple candidates. 

When the proportions of voters and seats line up perfectly (like if you had 5 seats and 20% support for each candidate) and so those 5 are elected then you would say that it's proportional. But what if the numbers don't line up perfectly? What if group A has 19% and group B has 41%. Does A still win a seat or does B win 2 seats? Which version meets the proportionality you are talking about and which is actually better? And what happens if voter blocks overlap in various ways?

I think we have to agree on an ideal outcome to aim for before we can debate what system versions accurately accomplish that. I also think that the kind of proportionality you are talking about here is only one kind of proportionality that is important. Other considerations include geographical proportionality, and minimum acceptable quota/seats per district. 

[parker friedland]

> and what you actually mean by proportionality when voters show support for multiple candidates.

What I mean is that if a group of voters wants to use their colective voting power to elect candidates from just one political party or ideology, that party/ideology should get at least the number of seats they proportionally deserve (±1 seat). For example, if more then 40% of the voters in a 5 winner district are progressive and they concentrate their voting power amoung all 3 progressive candidates (5 stars for progressives, 0 stars for moderates/conservatives/other), then atleast 2 of those 3 liberal candidates should win because 0.4 * 5 = 2.

> But what if the numbers don't line up perfectly? What if group A has 19% and group B has 41%. Does A still win a seat or does B win 2 seats?

Thats why I said that groups have the ability to win the number of seats they deserve ±1 seat. That means that they might win up to 1 seat less then they deserve. This means that depending on the voting method as well as how other people voted, if group A gave max scores to A candidates and min scores to every other candidate, then proportional voting methods might award group A 0 seat because group A deserves atleast 0.95 (0.19 * 5) seats and 0 seats is only 0.95 away from 1 (and proportionality errors of less then 1 seat are ok) when electing 5 winners. Since both PAV and RRV operate under the droop quota, under both PAV and RRV, group A would win atleast 1 seat and group B would win atleast 2 seats because in those methods the quota for winning atleast 1 out of 5 seats is 1/6th of the vote (if those voters are giving max scores to A candidates and min scores to everyone else) and for winning 2 out of 5 seats, that quota is only 1/3rd of the vote. Group A meats the 1/6th quota and group B meats the 1/3rd quota so atleast 1 candidate from party/ideology A will win and atleast 2 candidates from party/ideology B will win (provided that atleast 1 candidate from party/ideology A and 2 candidates from party/ideology B) are running. This is why PAV and RRV are both proportional.

However, since STAR-PR currently does not meat this definition of proportionality, equal vote is going to need to come up with anouther definition of proportionality to justify their claim that STAR-PR is proportional.

Multi-winner district voting methods that fit this definition of proportionality:

• PAV
• RRV
• STV
• CPO-STV
• Schulze STV
• Harmonic voting
• Monroe's method
• Least squares method
• Bucklin Transferable Voting
• MSPAV
• Culmaltive voting (as long as voters evenly divide their support amoung the candidates they want to elect)
• SNTV (as long as voters evenly divide their support amoung the candidates they want to elect)
• ... (I could keep going)

Multi-winner district voting methods that do not fit this definition of proportionality:

• voting methods that never claimed to be proportional
• STAR-PR

[Ciaran Dougherty]

If someone were to ask me why PAV or RRV is proportional, my response would be that any group of voters has the ability to guarantee that their party wins at least the number of seats that it proportionally deserves (±1 seat). This is always true in PAV or RRV.

That is technically true, but misleading.  The only way for a small faction can guarantee they win a seat they deserve when there is a markedly larger in the election, is for the smaller faction to bullet vote.  That's the entire reason I developed Apportioned Range Voting in the first place: I have significant reservations about any method that forces people to behave strategically in order to get the representation that they deserve, and so created a method that didn't.

[Jameson Quinn]

I agree with Ciaran on this. Apportioned STAR, apportioned Range, and PLACE are all PR methods I'd strongly support (either with lexical sorting as I support, or using lowest average as a tiebreaker as Ciaran supports). All of these are apportioned methods, which guarantees proportionality (and in a way that I think is reasonably easy to explain — "each time somebody wins they use up a quota of votes, so each quota of votes has the power to elect a winner"). RRV is certainly better than plurality but still pretty seriously flawed IMO.

2018-08-18 21:50 GMT-04:00 Ciaran Dougherty <lona...@gmail.com>:

If someone were to ask me why PAV or RRV is proportional, my response would be that any group of voters has the ability to guarantee that their party wins at least the number of seats that it proportionally deserves (±1 seat). This is always true in PAV or RRV.

That is technically true, but misleading.  The only way for a small faction can guarantee they win a seat they deserve when there is a markedly larger in the election, is for the smaller faction to bullet vote.  That's the entire reason I developed Apportioned Range Voting in the first place: I have significant reservations about any method that forces people to behave strategically in order to get the representation that they deserve, and so created a method that didn't.

--
You received this message because you are subscribed to the Google Groups "The Center for Election Science" group.
To unsubscribe from this group and stop receiving emails from it, send an email to electionscience+unsubscribe@googlegroups.com.

[parker friedland]

> The only way for a small faction can guarantee they win a seat they deserve when there is a markedly larger in the election, is for the smaller faction to bullet vote.

Not true. If faction A wants to garentee that an A candidate wins a seat, they can all bullet vote for a single A candidate. But giving max scores to all A candidates (and min to everyone else) also works, and it's a much less risky strategy too when different A voters are not sure what A candidates other A voters are voting for.

> I have significant reservations about any method that forces people to behave strategically in order to get the representation that they deserve

Thats why this definition of proportionality is just the bare minimun definition of proprotionality that any voting method should be able to pass to call itself proportional. I consider SNTV and culmultive voting to be simi-proportional voting methods rather then proportional voting methods and even they pass this bare minimum of a proportionality definition. So if you think that PAV and RRV are not always that proportional and that this definition is to laxed that even PAV and RRV pass it, that just means that the fact that STAR-PR does not pass it should be even more significent to you.

[parker friedland]

> Apportioned STAR voting

That's yet anouther method that passes this definition of proportionality and yet anouther candidate for the title of STAR-PR.

To unsubscribe from this group and stop receiving emails from it, send an email to electionscien...@googlegroups.com.

[Ciaran Dougherty]

On Sat, Aug 18, 2018 at 7:39 PM, parker friedland <parkerf...@gmail.com> wrote:

> The only way for a small faction can guarantee they win a seat they deserve when there is a markedly larger in the election, is for the smaller faction to bullet vote.

Not true. If faction A wants to garentee that an A candidate wins a seat, they can all bullet vote for a single A candidate. But giving max scores to all A candidates (and min to everyone else) also works, and it's a much less risky strategy too when different A voters are not sure what A candidates other A voters are voting for.

So "Bullet Voting" in Party List scenarios, and "Min/Max Voting" in Slate scenarios.  Whatever you want to call it, you're agreeing with me that in order to receive the seats they justly deserve, in at least some scenarios, small factions must use strategic voting, where they are compelled to score down options that they do not genuinely believe deserve a minimum score.

 

Thats why this definition of proportionality is just the bare minimun definition of proprotionality that any voting method should be able to pass to call itself proportional. I consider SNTV and culmultive voting to be simi-proportional voting methods rather then proportional voting methods and even they pass this bare minimum of a proportionality definition. So if you think that PAV and RRV are not always that proportional and that this definition is to laxed that even PAV and RRV pass it, that just means that the fact that STAR-PR does not pass it should be even more significent to you.

I'm not saying anything about STAR, multi-seat or otherwise; I try not to, these days, since a certain ...individual called me a troll for bringing up what were, to me, obvious problems with STAR.

All I was doing is pointing out that your assertion regarding the Harmonic Reweighting (PAV & RRV) is not true in certain scenarios (as seen in the real world: try using RRV with CA's electors from the 2016 election) is only true in cases of strategic ballots.

Well, that and plugging the algorithm I prefer for multi-seat cardinal methods.

Incidentally, another reason that I feel my "difference from ballot average" method is useful is that it can be applied without meaningful modification to basically any cardinal ballot.  As such can be used for "Apportioned Approval" and "Apportioned 3-2-1," in addition to Apportioned STAR and Apportioned Range that Jameson mentioned.

[parker friedland]

> pointing out that your assertion regarding the Harmonic Reweighting (PAV & RRV) is not true in certain scenarios (as seen in the real world: try using RRV with CA's electors from the 2016 election) is only true in cases of strategic ballots.

But my assertion is true.

Here was my assertion:

Under PAV, RRV, (and alot of other voting methods) any group of voters has the ability to guarantee that their party wins at least the number of seats that it proportionally deserves (±1 seat).

I never said that a group of voters will aways guarantee that their party wins at least the number of seats that it proportionally deserves (±1 seat). I said that any group of voters has the ability to guarantee that their party wins at least the number of seats that it proportionally deserves (±1 seat). This is true, because any group of voters does have this ability. It is also true of any other multi-winner district voting method that claims to be proportional with the humiliating exception of the current STAR-PR method.

--

[NoIRV]

My reaction to the last 5 comments: "Ugh, too much quibbling over specifics! Just use Asset Voting!"

[Sara Wolf]

any group of voters has the ability to guarantee that their party wins at least the number of seats that it proportionally deserves (±1 seat).

As a novice here forgive me if I'm in left field, but the definition above is troubling. "Plus or minus a seat"? I have a problem with the "Plus". I don't think that candidates who didn't make the minimum threshold should get a seat. Period. I worry about Nazis and other assholes. Better that blocks with support get more seats than people without support get representation they didn't earn.

 

I personally don't think that the threshold should be super low, nor the district super big.

So, reframing the question for a 0-5 PR system and set some parameters: 

• L
  ets limit
  each
   
  election
   to a 10 seat
  max
  district
  . For this example we'll use 10 seats.  
• Divide that into 5 precincts. 1/2+ of those seats are then elected with single winner STAR Voting by local precinct and the other 1/2- are PR. 
• Candidates need to earn 

  at least

  1/5 of the 

  total possible score to earn a PR seat.

  
  
• No district should be more than a 1 hour drive across. 
• Candidates who didn't meet the threshold are eliminated off the top
  . 
• B
  allots wh
  ich
  preferred
  eliminated
   candidates are rescaled to give their remaining favorite a 5 etc
  with scores rounded up as needed
  . 
• The location for public meetings of this council or legislature should rotate from precinct to precinct. 
• The time for public meetings should be evenings or weekends so working people can participate in government. 

Do you have other criteria or parameters that any PR system should meet? If so what are they?

Here are a couple ideas for PR or Quasi PR systems. Maybe they're lame, maybe not:

• F
  or
  a
  way
   too big 10 seat district
  a candidate needs to earn 1/10 of the 
  total possible score to earn a seat. 

• Candidates who didn't meet the threshold are eliminated off the top and ballots who preferred those candidates are rescaled to give their remaining favorite a 5 etc. 

• Total all scores given to all remaining candidates
  .

• Award 1st seat to candidate with the 
  highest
  score. 

• Option A: 

• On all ballots which had not given that candidate a 5, add 1 point to all other candidates supported.
• Total all scores given to all remaining candidates and
  award next seat to candidate with the highest score
• On all ballots 
  w
  h
  ich
   had not given that candidate a 5
  +
  , add 1 point to all other candidates supported.
   
• Repeat until done.

• Option B: 

• Award remaining seats using re-weighting, apportioning, and/or runoffs.
• Study options. Pick option that best meets new 5 Pillars of 0-5 PR: Equity, Honesty, Accuracy, Simplicity, Accountability. 

• Option C:

• Total all scores given to all remaining candidates and make a pie chart of who got what percentage of that total.   
• Award seats to 
  all candidates who got 10%
• For remaining seats: 
• Option A: Remove scores for candidates who were elected or eliminated. Make a pie chart, divide 100% by remaining seats to determine new threshold. Award seats to all who met threshold. Repeat until done.
• Option B: Determine top 2 scoring and elect candidate preferred by more voters. Repeat until done
  .

--

You received this message because you are subscribed to the Google Groups "The Center for Election Science" group.

To unsubscribe from this group and stop receiving emails from it, send an email to electionscience+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--

-Sara Wolk

PDX Equal Vote Coalition

PDX Represent.Us

971-222-9364

sa...@equal.vote

​www​

.

​starv​

o

​ting​

​

.

​

​us​

www.equal.vote

https://www.youtube.com/watch?v=Q1PjRgs2ZPE&t=13s

“The fact is that FPTP, the voting method we use in most of the English-speaking world, is absolutely horrible, and there is reason to believe that reforming it would substantially (though not of course completely) alleviate much political dysfunction and suffering.”

-Jameson Quinn, Election Science expert and PhD candidate in statistics at Harvard University

[Ciaran Dougherty]

On Sun, Aug 19, 2018 at 5:51 PM, Sara Wolf <sa...@equal.vote> wrote:

As a novice here forgive me if I'm in left field, but the definition above is troubling. "Plus or minus a seat"? I have a problem with the "Plus". I don't think that candidates who didn't make the minimum threshold should get a seat. Period.

Allow me to translate what is generally meant by "+/- 1 seat:"  Subject to rounding, based on the requirement to have whole numbers of persons elected.

Also, why do you have a problem with "plus" but not call out a problem with "minus"?

[Ciaran Dougherty]

I'm sorry, but any method that allows Candidates to negotiate among themselves which of them wins, rather than the people determining that, is a complete non-starter for me.

If a voter wanted their support to go to candidate B1 rather than B3, they would have voted for B1 rather than B3. 

While it technically solves the "voters split their support across more candidates than there are seats" problem, Asset Voting strikes me as perilously close to "Backroom Wheeling & Dealing," and likely would to the populace at large, too.

[Sara Wolf]

why do you have a problem with "plus" but not call out a problem with "minus"?

I get that you have to round one way or another. You can't get a magician out on election night and saw the lady in 1/2. But lets say the threshold is 10%. If I got 6%, I deserve no seats. It sounds like a system could round up to give me a seat anyways? That kind of rounding up is especially concerning if you deserve none, but got one. It seems more fair to me to say that if you didn't get the votes you don't get the seat. Obviously if you do it that way at some point you'd likely end up with a seat extra at the end. In that case I'd rather see it go to a more popular candidate then a less popular one. 

Asset Voting strikes me as perilously close to "Backroom Wheeling & Dealing," and likely would to the populace at large, too.

Agreed, though looking at the other options it's starting to look better and better. None of these suggestions are even remotely transparent. At least Asset is simple. 

[Ciaran Dougherty]

On Aug 20, 2018 00:42, "Sara Wolf" <sa...@equal.vote> wrote:

why do you have a problem with "plus" but not call out a problem with "minus"?

I get that you have to round one way or another. You can't get a magician out on election night and saw the lady in 1/2. But lets say the threshold is 10%. If I got 6%, I deserve no seats. It sounds like a system could round up to give me a seat anyways? That kind of rounding up is especially concerning if you deserve none, but got one. It seems more fair to me to say that if you didn't get the votes you don't get the seat. Obviously if you do it that way at some point you'd likely end up with a seat extra at the end. In that case I'd rather see it go to a more popular candidate then a less popular one. 

That's just it, with 10% corresponding to each seat, your 6% means you are the most popular candidate not yet seated, having about 60 of the vote left over for the last seat. 

Unless you're saying that a group with 41% of the vote should get a four seats for their four quotas, plus a 5th seat for their 1% beyond that...

Asset Voting strikes me as perilously close to "Backroom Wheeling & Dealing," and likely would to the populace at large, too.

Agreed, though looking at the other options it's starting to look better and better. None of these suggestions are even remotely transparent. At least Asset is simple. 

Really? I am surprised to hear that. Apportioned Range(/STAR/approval/3-2-1, etc) are pretty transparent, if you report the results for each seat,  I thought...

[parker friedland]

Sarah, look at it this way:

In a two winner district, group A gives max scores to A candidates and min scores to B candidates and vise versa with group B.

50.001% of the voters belong to group A.
49.999% of the voters belong to group B.

A wins the first seat.

Do you think that with just 50.001% of the vote A should win both seats? That result would hardly be proportional because under that logic, when only electing two winners a majority of voters should always be able to dominate all the seats in a two winner multi-winner district just like they can in a single winner district.

Yet due to the weakness of my bare minimum proportionality definition, even a voting method that would award A both seats in this example would still pass that definition of proportionality.

> I don't think that candidates who didn't make the minimum threshold should get a seat. Period.

All voting methods have to round up at some point.

Suppose that instead of being two groups of voters, there were three.

34% of the voters belong to group A.
34% of the voters belong to group B.
32% of the voters belong to group C,

Like it or not, any voting method would in this senario have to award a seat to a candidate with less then 50% of the vote. It's just not possible to always round down. If you eliminate all of the candidates with less then 50% of the vote straight off of the bat, you might not have anybody left to elect.

[Sara Wolf]

Unless you're saying that a group with 41% 

The systems we're talking about are not party list. Am I right? There's no "group". Just individual candidates who have earned votes or haven't. As I understand it.

[Sara Wolf]

Those kinds of examples make sense and sound fair when we talk about blocks just bullet voting, but isn't the whole point of a system like STAR-PR that voters wouldn't just bullet vote? 

[parker friedland]

> The systems we're talking about are not party list. Am I right? There's no "group". Just individual candidates who have earned votes or haven't. As I understand it.

We are talking about when there are groups of voters, A, B, C, etc. and there are also groups of candidates, A, B, C, etc.
A voters give A candidates max scores and everyone elce min scores.
B voters give B candidates max scores and everyone elce min scores.
C voters give C candidates max scores and everyone elce min scores.
...

However, A, B, and C candidates dont have to vote that way. They vote the most effective way possible to elect as many of their candidates as possible, however in most voting methods, that is how they should vote if they want to maximize the number of candidates from their group that they elect.

[parker friedland]

> Those kinds of examples make sense and sound fair when we talk about blocks just bullet voting but isn't the whole point of a system like STAR-PR that voters wouldn't just bullet vote?

Yeah, but if STAR-PR doesn't give voters the ability to guarantee that their party wins at least the number of seats that it proportionally deserves (±1 seat), then what is its claim to proportionality? Why should it be categorized as proportional? It does deweight the ballots of voters who already helped elect some candidates, so it is at least semi-proportional but a lot of voting methods can deweight ballots a bit and also be called semi-proportional. What definition of proportionality does STAR pass that officially makes it a proportional voting method?

[Sara Wolf]

The thing that's problematic is when we get into low thresholds so 50% isn't a great example. So, I'd maybe agree to 10% as a super low bar. But then it turns out you can win with just 6%? What about less than that? If people give low scores across the board you could get in with almost no support? 

I've been going to a ton of different groups meetings lately and there are a ton of toxic, power grabbing, argumentative, and nit-picky individuals thrown in with the altruistic people who are there to make a difference. With super low bars how do you prevent those people from getting elected, and how do you vote them out again if the do make it into office? 

 

They vote the most effective way possible to elect as many of their candidates as possible, however in most voting methods, [bullet voting] is how they should vote if they want to maximize the number of candidates from their group that they elect.

That doesn't sound good. Do we really want to encouraged polarized partisan blocks. It sounds better than the duopoly, but the whole thing about STAR that I like is that it encourages candidates to reach out and make a point of representing people beyond their normal base. Are you saying that would that be a good strategy in STAR-PR as described? I hope not. 

I get that bullet voting makes for much easier examples, but when I try and think beyond to how these systems would behave in a non-duopoly political spectrum or in a local election where you are dealing with individuals who you know, and who may have very different styles, I see issues, or else it's too confusing to tell what the implications would be.

[Sara Wolf]

"..guarantee that their party wins at least the number of seats that it proportionally deserves (±1 seat), then what is its claim to proportionality?

"

People aren't told that. They are told "if you represent 1/4 of the voters you'll win 1/4 of the seats"

I never heard of the + or - until I asked on this forum. I think when you explain the concept of PR people assume it's proportional. You can define "proportionality criteria" as above, or as "if you get the proportion, you get the seat." Whatever is the most fair. If I went to the store and tried to buy a $20 shirt, but I only had $10.01 should I get it? No. But if I was the highest bidder at an auction I would. 

The question about rounding up or down or both has huge implications and it shouldn't be taken for granted that splitting it down the middle is best. I see strong arguments that could be made for both sides.

But I'm really still just thinking all this through. I have initial feelings and opinions, but I want to be clear that I don't feel like I have a well educated opinion at this stage. 

I worry that there was this question posed to you all at some point.. "what is the best algorithm to achieve this PR criterion" and now, many years later it's taken for granted that that's the right question. But is it? If we were to reframe the question with goals about the kind of legislature we hope to elect I wonder if we wouldn't get new ideas that are worth looking at. 

[parker friedland]

> how do you vote them out again if the do make it into office?

It's quite simple. You don't. That's one of the sacrafices proportional voting methods make in order to make it so that a legislative body resembles the populace at large as closely as possible. This is why PR is the closest thing to direct democracy. If 10% of voters are racist, 1% of politicians are also racist, however just like racist policy ballot iniatives would tend to be voted down by the populace at large, racist legislation would also tend be voted down by a proportionally elected congress of which represents the populace at large.

In this way, saying that a NAZIS are not allowed to get elected is like saying that NAZI voters should not be allowed to vote for ballot initiatives. Allowing everyone representation is just one of the consequences of approaching direct democracy.

Ofcourse direct democracy is not allways the most efficient solution to politics (sometimes 2 sheep and 3 wolfs may vote what to have for dinner) and thats where utilitarian voting methods sometimes have an upperhand however this atvantage that utilitarian methods have over proportional methods is canceled out by the distortion the way districts are drawn has on public opinion in the legislative body. But on average, as a rule of thumb, direct democracy weilds pretty good results (legislation) from a utalitarian perspective: https://rangevoting.org/BRmulti.html

[Sara Wolf]

 If 10% of voters are racist, 1% of politicians are also racist

No, what you guys described above is that if 6% of voters are racist, 10% of politicians are also racist. 

[parker friedland]

If 10% of voters are racist, 1% (Correction: I mean 10%) of politicians are also racist

> No, what you guys described above is that if 6% of voters are racist, 10% of politicians are also racist.

Yeah, but the rounding ups are going to cancel with the rounding downs and on average, this number is actually going to be closer to 6% (if not less then 6% because PAV and RRV tend to favor larger groups of voters more often then they favor smaller groups of voters when rounding).

On Mon, Aug 20, 2018 at 2:54 AM, Sara Wolf <sa...@equal.vote> wrote:

 If 10% of voters are racist, 1% of politicians are also racist

No, what you guys described above is that if 6% of voters are racist, 10% of politicians are also racist. 

[parker friedland]

>> No, what you guys described above is that if 6% of voters are racist, 10% of politicians are also racist.

> Yeah, but the rounding ups are going to cancel with the rounding downs and on average, this number is actually going to be closer to 6% (if not less then 6% because PAV and RRV tend to favor larger groups of voters more often then they favor smaller groups of voters when rounding).

What your response might be: But what about when you're just using PAV or RRV with just one multi-winner district? Surely the rounding ups cant cancel with the rounding downs then right?

My reply to that hypothetical response:

PAV and RRV both use the Droop quota, meaning that when voters are just voting on party lines, each party is only guaranteed to win a seat for every 1/(1+seats) share of the vote it receives. This means that in a 9 district election when 5 nazi parties are taking full advantage of the upward rounding, each of the nazi parties would need to win at least 1/10th of the vote in order to be guaranteed a seat that represents 1/9th of the population. If to be able to guarantee a majority and pass Nazi legislation, each Nazi party needs 1/10th of the vote, then all of them combined need at least 5/10ths of the vote or 50% in which case the Nazis would already make up a majority of the populace. However, noticed that I used the word guarantee. Even if some Nazi party parties had less than 1/10th of the vote, all 5 of them could still win seats if the other parties were much smaller in comparison (meaning that in order for the Nazi parties to take advantage of the downward rounding, they have to already be more popular than the other parties). For example, if all the other parties had less then 1/100th of the vote, then PAV and RRV would elect at least 1 candidate from each of the Nazi parties but the irony is under such a situation, the Nazi parties would actually be the larger parties.

[Ciaran Dougherty]

Parker, please stop saying that PAV & RRV use quotas, because they don't.

I have shown, on this list, a scenario where two factions that won more than 3 Hare Quotas between them would get zero votes under RRV.

PAV and RRV have absolutely nothing to do with quotas. Nothing.

You received this message because you are subscribed to the Google Groups "The Center for Election Science" group.

To unsubscribe from this group and stop receiving emails from it, send an email to electionscience+unsubscribe@googlegroups.com.

[parker friedland]

> Parker, please stop saying that PAV & RRV use quotas, because they don't.

> I have shown, on this list, a scenario where two factions that won more than 3 Hare Quotas between them would get zero votes under RRV.

> PAV and RRV have absolutely nothing to do with quotas. Nothing.

They don't use quotas in the traditional sence, but when voters just vote on party lines, they are equivalent to the D'Hondt method, and in the D'Hondt method if a party gets a droop quota worth of the vote they are garenteed to win atleast one seat.

On Monday, August 20, 2018 at 8:29:56 AM UTC-7, Ciaran Dougherty wrote:

Parker, please stop saying that PAV & RRV use quotas, because they don't.

I have shown, on this list, a scenario where two factions that won more than 3 Hare Quotas between them would get zero votes under RRV.

PAV and RRV have absolutely nothing to do with quotas. Nothing.

On Aug 20, 2018 03:27, "parker friedland" <parkerf...@gmail.com> wrote:

>> No, what you guys described above is that if 6% of voters are racist, 10% of politicians are also racist.

> Yeah, but the rounding ups are going to cancel with the rounding downs and on average, this number is actually going to be closer to 6% (if not less then 6% because PAV and RRV tend to favor larger groups of voters more often then they favor smaller groups of voters when rounding).

What your response might be: But what about when you're just using PAV or RRV with just one multi-winner district? Surely the rounding ups cant cancel with the rounding downs then right?

My reply to that hypothetical response:

PAV and RRV both use the Droop quota, meaning that when voters are just voting on party lines, each party is only guaranteed to win a seat for every 1/(1+seats) share of the vote it receives. This means that in a 9 district election when 5 nazi parties are taking full advantage of the upward rounding, each of the nazi parties would need to win at least 1/10th of the vote in order to be guaranteed a seat that represents 1/9th of the population. If to be able to guarantee a majority and pass Nazi legislation, each Nazi party needs 1/10th of the vote, then all of them combined need at least 5/10ths of the vote or 50% in which case the Nazis would already make up a majority of the populace. However, noticed that I used the word guarantee. Even if some Nazi party parties had less than 1/10th of the vote, all 5 of them could still win seats if the other parties were much smaller in comparison (meaning that in order for the Nazi parties to take advantage of the downward rounding, they have to already be more popular than the other parties). For example, if all the other parties had less then 1/100th of the vote, then PAV and RRV would elect at least 1 candidate from each of the Nazi parties but the irony is under such a situation, the Nazi parties would actually be the larger parties.

On Mon, Aug 20, 2018 at 3:01 AM, parker friedland <parkerf...@gmail.com> wrote:

If 10% of voters are racist, 1% (Correction: I mean 10%) of politicians are also racist

> No, what you guys described above is that if 6% of voters are racist, 10% of politicians are also racist.

Yeah, but the rounding ups are going to cancel with the rounding downs and on average, this number is actually going to be closer to 6% (if not less then 6% because PAV and RRV tend to favor larger groups of voters more often then they favor smaller groups of voters when rounding).

On Mon, Aug 20, 2018 at 2:54 AM, Sara Wolf <sa...@equal.vote> wrote:

 If 10% of voters are racist, 1% of politicians are also racist

No, what you guys described above is that if 6% of voters are racist, 10% of politicians are also racist. 

--
You received this message because you are subscribed to a topic in the Google Groups "The Center for Election Science" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/electionscience/2P39eMRktcA/unsubscribe.

To unsubscribe from this group and all its topics, send an email to electionscien...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "The Center for Election Science" group.

To unsubscribe from this group and stop receiving emails from it, send an email to electionscien...@googlegroups.com.

[Ciaran Dougherty]

If they don't use quotas, and they require significant strategy to even approximate quotas, it is incredibly misleading to say that they use quotas.

To unsubscribe from this group and stop receiving emails from it, send an email to electionscience+unsubscribe@googlegroups.com.

[parker friedland]

> If they don't use quotas, and they require significant strategy to even approximate quotas, it is incredibly misleading to say that they use quotas.

How quotas are incorperated in your appointment methods: if a certain quota worth of the voters vote on party lines for a certain party, that party is garenteed to win atleast one seat.

How quotas are incorperated in PAV and RRV: if a certain quota worth of the voters vote on party lines for a certain party, that party is garenteed to win atleast one seat.

In both types of voting methods, a party with less then a certain quota of the electorate voting on party lines for them can still get elected depending on how everyone else votes. The main difference is that in your appointed methods, quotas are directly described in the algorithems, while in PAV and RRV, they are an indirect consequence of the algorithems.

To unsubscribe from this group and stop receiving emails from it, send an email to electionscien...@googlegroups.com.

[Ciaran Dougherty]

And if they don't party-line vote, they aren't, even if they have more than one full quota. As such, those quotas aren't used so much as emergent.

To unsubscribe from this group and stop receiving emails from it, send an email to electionscience+unsubscribe@googlegroups.com.

[NoIRV]

For what it is worth: Asset Voting tends to the droop quota (if I have lots of votes, I would keep a droop quota and give out the rest), although it may be possible to do a hare quota by declaring that all but one winner have more than a hare quota of votes. (All but one, so that two rival factions cannot block the final seat entirely by holding their assets.)