Hot off the presses! SRV-PR!

Depends on how many seats there are.

I don't think anyone's going to propose PR with fewer than 3 seats.

Thank you for posting this. An important criterion for evaluating a single-winner system is how well it serves as a stepping stone to PR, and I was about to ask what SRV-PR would look like.

The formula for ballot influence is given as “1 / (1 + sum/max)”. I assume “max” refers to the maximum permissible score, which I believe has been 9 or 5 in most of your examples. If so, I recommend that you instead use the maximum permissible score that the voter in question actually gave to a candidate. If for example the highest score a particular voter gives to any candidate is 4, then for that voter, “max” would be 4.

The reason I’ve found for doing this (there might be others) is that otherwise a large bloc of voters who all hate an important minority candidate can deliberately underscore their preferred candidates. Depending on various factors like the size of the bloc, this underscoring could let them elect their preferred candidates while still hogging ballot influence for the runoff elections so as to block the hated minority candidate.

One fundamental characteristic of PR is that there’s no effective way to vote against someone you don’t like. (This goes hand in hand with minorities being able to elect candidates.) Instead, you focus on voting in favor of the people* you do like, and then it’s their job to go be effective legislators or whatever, advancing things they're in favor of and blocking things they're not. Presumably, having a two-candidate runoff of any kind undermines this characteristic (and therefore means the system qualifies as semiproportional rather than full PR); but as long as each voter makes use of their maximum score, the “ballot influence” mechanism should compensate somewhat. I haven’t looked at this thoroughly.

*Or for parties you like, in some implementations of PR.

An important criterion for evaluating a single-winner system is how well it serves as a stepping stone to PR

This assumes that PR has been shown by empirical evidence to produce the most utilitarian policy. But that has not been shown.

I recommend that you instead use the maximum permissible score that the voter in question actually gave to a candidate.

You have to be extremely careful proposing rules like this. Historical efforts to counter such strategies have been shown to cause more problems than they solve. For instance, you might kill the proportionality theorem that RRV satisfies.

The history of voting methods is littered with attempts to "fix" existing systems by introducing various semi-random ideas like this, and then not vetting them with math PhD's. Mark kind of got lucky on his Score Runoff Voting idea in that it still manages to perform well. That's usually not the case.

If you want to read up on the state of the art in PR theory, I'd check out this page.

https://www.youtube.com/watch?v=oSNNav2eYwk

This assumes that PR has been shown by empirical evidence to produce the most utilitarian policy.

No it doesn't. I was stating an opinion, and my personal concept of what is valuable in a voting system doesn’t automatically match yours.

The history of voting methods is littered with attempts to "fix" existing systems by introducing various semi-random ideas like this, and then not vetting them with math PhD's.

What I’ve done here is (1) point out a problem and (2) suggest a fix. And sure, I readily believe that in your experience, most of the so-called fixes people suggest are not helpful. So back up a step and look at the problem: SRV-PR can be thought of as having two mechanisms -- the scoring and the runoff -- through which voters exercise clout. I’ve pointed out that a voter can sacrifice their scoring clout to get extra runoff clout, and I’ve implicitly opined that this is bad. (And now I’ll opine it explicitly: This is bad.) SRV-PR advocates should be aware of this and decide whether they care. If they do care, they should do some thinking about possible solutions, including the one I proposed.

Hah! It was luck, buddy. You couldn't have known for sure until Jameson tested it for you.

I blame Rob Richie, since he's the one who really came up with it. I'm just a good listener ;-).

I was stating an opinion, and my personal concept of what is valuable in a voting system doesn’t automatically match yours.

This isn't subjective. Elections have objective effects on human welfare.

Clay: Correct me if I'm wrong, but it sounds like you're saying there's one true method of voting system evaluation, and if that particular method rates system A higher than system B, then everyone is supposed to agree that system A is better. Well forget that. People are individuals, and we have opinionated disagreements on what characterizes a good voting system just like we have opinionated disagreements on almost everything else.

To your other comment: Obviously elections have effects on human welfare, but could you please clarify what you mean by “objective” and how it relates to my holding the opinion that an important criterion for evaluating a single-winner system is how well it serves as a stepping stone to PR?

(If http://www.icasualties.org/ is accurate and I’m interpreting it correctly, 418 US soldiers died in Operation Enduring Freedom/Afghanistan in the year 2011. That’s a big change in human welfare that, yes, was affected by a considerable variety of elections that have taken place over time. Possibly including, to give one example, Richard Nixon’s election to congress in 1946, which may have affected his becoming president in 1969, which may have affected William Rehnquist being nominated to the supreme court in 1971, which may have affected George W. Bush becoming president in 2001, which may have affected the existence of Operation Enduring Freedom and its casualties in 2011, three years after Bush was replaced by the winner of yet another election. Are you speaking of an objectivity that can be discerned, or one that’s based in a hypothetical omniscience?)

Of course, all this is a sidetrack. My main point is that in SRV-PR, a voter can sacrifice their scoring clout to get extra runoff clout. I think SRV-PR advocates ought to care about this; and if they do care, they should do some thinking about possible solutions, including the one I proposed.

RE: "My main point is that in SRV-PR, a voter can sacrifice their scoring clout to get extra runoff clout."

One of the reasons that the SRV-Single Winner system was invented, as I understand it, is that the runoff clout helps incentivize voters to be more honest and less strategic. Specifically, it pays to not bullet vote EDIT: or do tactical minimization or maximization that would change your rankings (sacrifice scoring clout) so that your preferences are clear in the runoff. The two strategies for the two rounds are contradictory so you're best off just being honest.

I can't say I fully understand what you are saying about how scoring clout and runoff clout effect strategy in SRV-PR. For that matter I'm not sure I even fully get the algorithm in the first place (it's a lot simpler than IRV-PR/STV though.) But, if it's a similar effect to the trade off in the single winner version, could this clout conundrum be a balancing effect that helps to incentivize honest voting?

In other words, yes, it may help you in the runoff to have a lower score that your total voting weight is divided by. This would keep your votes weight in later rounds slightly higher, but by doing so you make it so that your favorite candidate is less likely to get elected in the first place. In SRV you can be strategically dishonest if you want, but the odds are that this will backfire and hurt you overall. It seems like the same is true in SRV-PR.

The real question is how dangerous this strategy could be for the electorate at large. It seems like the strategy doesn't give the strategic voter an unfair advantage, (on the contrary it's likely a disadvantage,) but a choice where to throw their weight. Seems fair enough to me and highly unlikely that people would try it anyways.

there's one true method of voting system evaluation

I think you probably knew before posting that Clay would not say anything like "yes, that's exactly what I mean"

I encourage everyone to keep working to use Rapoport's Rules for most effective discussion.

I'm not talking about bullet voting, I'm taking about using exclusively low scores.

Let’s say there’s a 5-winner election for Portland’s city council using SRV-PR with a 0-5 scale. One of the candidates wants to ban bicycles from the city, and I care more about keeping that candidate out than letting anyone else in, so I give them a 0 and each of the other candidates a 1.* So my “ballot influence” remains fairly strong all the way to the last runoff, even if candidates I like happen to be elected.

This is a moderately effective way for me to vote against the ban-the-bikes candidate, relative to the fact that I still have just the one vote amidst hundreds of thousands of other voters. Which means SRV-PR as currently defined isn’t really PR, because in PR there’s no way to cast an effective vote against a particular candidate or party. Everyone votes for the candidates they want, and you get a proportional body that includes minorities.

*Technical nuance: I might score some of the other candidates 0 instead of 1, if I think they’re likely to win without being in a runoff against the ban-the-bikes candidate.

Seems fair enough to me and highly unlikely that people would try it anyways.

People would definitely try it. Under plurality, millions of people deliberately sacrifice their right to vote for a candidate they want to vote against a candidate they don’t want. Lesser of two evils. I don’t know how thoroughly this empirical evidence would transfer from plurality to SRV-PR, but I think a lot would.

As for fairness, well, do we think PR is fairer than winner-take-all? I do, so I’m opposed to letting voters throw their weight against a candidate rather than toward a candidate.

The real question is how dangerous this strategy could be for the electorate at large.

Why? Is there a downside to fixing the problem I’ve pointed out with the current definition of SRV-PR? Would it be hard? Would there be unfortunate consequences?

I said: People would definitely try it. ... I don’t know how thoroughly this empirical evidence would transfer from plurality to SRV-PR, but I think a lot would.

Mmmm, let me walk this back a bit. I do think people would definitely try it, but "a lot"? Not sure.

But like I said, I don't see a downside to fixing this particular loophole so why not fix it.

I'm convinced that PR is good. I'm unconvinced that SRV-PR is better than simple RRV. I'm unconvinced that STV is all that bad relatively (it seems a lot less problematic than single-winner IRV, despite the same overall mechanisms). I'm not sure, but inclined to think that this adapted STV approach looks pretty good (possibly significantly better than STV and without the issues Fillard is worred about with SRV-PR).

Aaron, Clay -- I would welcome the opportunity to talk with each of you on the phone. I don't have your contact information, but you can find mine at https://www.loomio.org/d/eDuxfmGF/welcome-please-introduce-yourself . Thanks.

I'd suggest a group chat sometime via https://meet.jit.si perhaps (works in Firefox or Chrome/Chromium), I'd be happy to chat if we can find a good time that isn't taking too much away from other high-priority time for other things for me.

Aaron, I’ll cut to the chase: My wording was a little strong, but I meant what I said. I was in fact attempting -- consistent with a part of the rules you suggest -- to clearly, vividly, and fairly express Clay’s position. I acknowledged that I might be wrong and invited him to correct me.

If you don’t think my guess as to Clay’s position was reasonable, consider the exchange from https://www.loomio.org/d/wJrTfUR3/private-introduce-yourself where Nathan Hunter said “No abstraction is going to be the right tool for every job”, and Clay responded, “Unless that abstraction is utilitarianism.” Then go back and reread the discussion in this thread between Clay and me with that exchange in mind.

I recognize that the above review would take a little time. If you don’t have that kind of time, I certainly sympathize, but maybe calling other people’s sincerity into question is not the best activity for you to be engaging in right now.

In the event that you’d like to discuss this any further I suggest you call me. (Remember I can’t call or email you directly unless you provide your contact information, which I don’t believe you have.)

I didn't mean to call you out as being rude or anything or aggressive. I only was hoping to keep you and everyone to the highest standards (which I have regularly failed to meet myself, and I apologize for that).

I think Clay has been far more inclined to assert bold, simplistic things as part of a whole pattern on all sides of less-than-ideal discourse. It's not like he himself followed Rapoport's Rules. And I had no intention of saying that you (Fillard) had been at-fault for anything. I think you've done as well as anyone, and better than me, at staying really respectful, and I admire that.

So, let me restate that: I have zero doubt about your sincerity and good faith. I'm completely certain that you are acting in the best of faith and intentions and never meant otherwise. The entire point I was attempting to make was that I thought your particular wording was a fair interpretation of Clay to the extent that his argument could be read that way, but that it was unlikely he actually meant things as simplistically as he came across.

So, I was hoping to push those people like you who seem most amenable, mature, respectful and prepared to actually reach the highest standards to hope that you could be a part of the small number of people who aim to set a higher-bar by modeling the best discourse. I.e. I thought you were totally non-antagonistic and so had a chance and interest in doing even better (which is the sort of push I always appreciate from others for myself).

You've been a wonderful, positive, respectful part of this group, and I sincerely have nothing negative to say about you at all. I think you're the sort of person with the capacity to actually be effective at saying the sort of careful, disarming things that help others learn to improve themselves.

I regret that my brief text failed to carry this meaning well. I won't have much time for any of this going forward, but I care about real democracy and want to see this group succeed. Thanks for everything you're doing.

@fillardspringrhyne RE: "I'm not talking about bullet voting, I'm taking about using exclusively low scores." Sorry, I meant to be a bit more broad than Bullet Voting too. I edited that sentence to now read "Specifically, it pays to not bullet vote EDIT: or do tactical minimization or maximization that would change your rankings (sacrifice scoring clout) so that your preferences are clear in the runoff." I think I did get what you meant there despite the imperfect terminology usage!

RE: "One fundamental characteristic of PR is that there’s no effective way to vote against someone you don’t like." I'm not sure that this is actually ideal or good. It sounds good, and it may well be... My only thought is that in PR I kind of would like there to be a mechanism that helps monitory groups, while at the same not giving any extra advantage to other minority groups that are actively antagonistic to others.

Assuming that this is desirable I'd say that your fix for SRV-PR (changing MAX to mean max score given, not max score possible.) is a simple fix that corrects for the strategy you described.

The strategy you described above is a really smart strategy and I'm impressed you worked it out! I don't doubt that it would work but it's a lot more complicated to get then favorite betrayal in plurality. I think that most people would feel wary trying a strategy like that with obvious risk (major minimization of your favorite) unless they totally get it 100% and I think honestly most people wouldn't get it.

FRAUD AND RECOUNTS
For me, I'm still on the fence for PR in general, even though I love the goal, because I think it's complicated enough to be really easy to hack into unnoticed and hard to catch fraud. This is also a big reason I'm not in support of IRV. In general I'm in favor of doing exit polling for every election (as accurate as possible) and then calling an automatic hand recount for results outside the margin of error (or something like that, double the margin?) I'm concerned that even if recounts are possible there would be a huge cost barrier and they wouldn't get done enough.

STV seems to be the most complicated algorithm. RRV is a lot simpler, but still too complicated for the voters to actually get it. SRV-PR is a tiny bit more complicated than RRV but still in that same ballpark. I wish there was something that was simple enough to be explained without an algorithm and that would be cost effective and simple to count and recount. (Precinct by precinct!) Can you guys talk about this pr issue a bit as it relates to each system?

I heard about another super simple idea for PR from Clay @clayshentrup that I think Warren came up with recently. I couldn't describe it accurately but hopefully Clay can!

The simplest PR system is probably Asset Voting. You vote for one candidate. After the votes are totaled, the candidates have some period of time (an hour, a day, whatever) to redistribute their votes so none are "wasted". Suppose three Greens have between them enough votes for a seat. They can consolidate their votes so that one of them can take that seat. Concessions might be made so that even the ones who don't get a seat can have some influence to help their constituents. You can also modify it so voters can divvy up their vote to multiple candidates, so X gets half a vote and Y gets the other half.

http://scorevoting.net/Asset.html

As for RRV, it's simple enough that it can be done in a spreadsheet. I got the Berkeley city council to unanimously adopt it for prioritizing legislation. Doing the process with the city clerk didn't seem too daunting.

If you really want simple, you can do RRV with Approval Voting. Then it just becomes a simple weighting. Your ballot is divided by 1 + W where W is the number of winners you approved. So if two of the people you approved have been elected, your ballot is 1/3 as powerful as it started out. Super simple.

Warren Smith kind of revolutionized PR a year or so ago with some research that's outlined here.
http://scorevoting.net/CanadaOverview.html

One of the interesting systems there is like an improved MMP. MMP (used in Germany and New Zealand) works like this. You vote for a local representative and for a party. Parties get a number of seats proportional to the votes they got, via party lists (that are typically decided in smoke filled rooms). You subtract out the local seats they won, then "top up" from the party lists to create proportionality.

The problems are: 1) using Plurality Voting to pick the local seats, and 2) using party lists instead of just letting voters vote. Warren's improved system is like MMP, but it uses Score Voting for the local MP, and Asset Voting for the "top up". After deciding the local seats with Score Voting (the candidate with the most points wins each district), we award "ballots" to the candidate(s) with the top score. So if you gave Bob and Alice a 5 on your 0-5 scale, they each get half a ballot which they can redistribute as we described above for Asset Voting. The only caveat is, anyone who was already elected to a local seat loses a seat's worth of ballots, which is equivalent in influence. That local MP is then free to redistribute her remaining ballots.

This is a little weird but based on sound principles, and actually simpler than STV or RRV.

Thanks Clay, that was what I was thinking of! I kind of like the Asset Voting plan. It would allow 2 minor parties to compile their vote and then have 1 rep that represents them both.. or would foster goodwill between parties that give their extra votes and others that get them. Its a bit indirect in that the final sharing is out of the hands of the voters but you've empowered your favorite to act on your behalf so it seems fair. Setting up the candidates working together and sharing on day one could have epic positive psychological side effects and make a big impact on teamwork down the line. What do the rest of you think?

Doing RRV with Approval voting seems like it would have the same centrist favoring properties where you still have to vote for the lesser evil to play it safe so the real favorites can never win. But maybe since there are multiple seats that wouldn't be as much of an issue in PR-Approval?

I'll read up on the rest of the stuff above!

I made this report card for a few of the single winner systems using our 6 criteria and it really helped me lay out the differences and the pros and cons. If anyone wanted to do the same thing but for PR I bet it would help a lot of us! It best if someone knowledgeable on the ins and outs of each does it and that is not me at the moment!
https://dreamtimecompass.wordpress.com/2017/02/24/the-quest-for-election-reform/

I'm not in favor of a system where voters don't elect clear winners and people have to negotiate and horse trade in order to figure out who gets seats. That is basically a recipe for buying and selling seats, never mind buying and selling votes.

I like STV and RRV as well as MMP, not sure what the problem is.

Sure nothing is perfect, but I don't see a compelling reason to try to come up with something better. I don't think any of the above mentioned alternatives to these three are any superior.

@fillardspringrhyne

What about with RRV? I'm curious to hear what your opinion is regarding RRV as a PR system and how it compares to STV.

RE: "My main point is that in SRV-PR, a voter can sacrifice their scoring clout to get extra runoff clout."

I seriously doubt that enough voters could be persuaded en masse to downgrade their votes in order to theoretically block a hated minority candidate from winning the last seat. In order to pull that off you would have to have a large population of voters who care little about who wins and only care about blocking one candidate. Perhaps if there is a literal Nazi candidate. Short of that I don't see enough people being motivated enough.

The downside to reducing the max vote according to each voters actual max vote for any candidate is that it would add a huge amount of complexity to the tabulation.

All that being said I'm currently persuaded that RRV is good as is as a PR system to pair up with SRV, and the new SRV-PR proposal isn't really necessary.

But I reserve the right to be persuaded otherwise given more examples or simulations, etc.