Best Algorithms and Design for Proportional Representation Elections

[sa...@equal.vote]

There's been a lot of talk about different PR algorithms including STV, RRV, STAR-PR variations, and other proposals and variants mentioned in other long threads. Different PR systems may have somewhat different goals, but let's have a focused conversation about the best ways to meet each goal as simply as possible with each type of ballot. I'm nominating Emily Dempsey to dive in here from the Portland Equal Vote chapter for where my math skills will inevitably fall short, but bonus points if you can explain the ideas behind the numbers. 

I'm most interested in STAR-PR, aka finding the best PR system possible with a 0-5 ballot. (Runoffs optional.) I'd love to know what the best perfect mathematical proportionality would be, and also if there is an algorithm that gives an advantage to less polarizing candidates while still allowing good representation for less divisive candidates across the spectrum. (There's another thread where you can debate if that's desirable or not.) Ideally these algorithms or at least the concept would be explainable to a lay person without using numbers or algebra in a paragraph or less i order to be viable in the real world. 

Canada and BC among other places are strongly considering an STV/IRV package deal. I assume that's not most people here's first choice, so I'd like to explore how our proposals compare with that. Also, how do our proposals compare if it was a better single winner/PR ranked ballot option (~Majority Judgement or Ranked Pairs or...) 

I'm sure a number of you will want to talk about PR with an Approval or otherwise less expressive ballot so go for it. 

I'd love to come up with a way to compare and contrast different options/algorithms across the 5 pillars of a just PR voting system and talk about how to study them: Equity, Accuracy, Honesty, Expressiveness, Simplicity. Simulations would be great for looking at VSE and measuring accuracy. Here's a super subjective article comparing voting systems by that metric and it'd be great to have something similar to compare and contrast options using each ballot type and then a version comparing the "best" option for each ballot type: https://www.starvoting.us/report_card

OK go! 

[Jameson Quinn]

"Canada and BC are strongly considering an STV/IRV package deal" : I was in Vancouver as an organizer of the BC Symposium on PR, and that's not really the impression I got. Yes, FairVote Canada favors that kind of idea, but frankly it seemed to me a lot less dogmatically than FairVote US favors RCV; they're willing to be open-minded about things, and in particular seem to support things like FDPR, an IRV/STV/MMP hybrid. (I know, that's still pretty STV/IRV heavy, but it's a lot farther than I could see FairVote USA going.)

----

So, if I were proposing a package deal (single- and multi-winner with a common ballot format), it would be 3-2-1 with PAD. 

(If you really wanted more expressivity in this ballot, you could allow ballots that (for instance) filled in both "good" and "OK" for a certain candidate to count as half a "good" vote when choosing semifinalists and as a full vote for that candidate below a "good" and above an "OK" when choosing a winner. In PAD it would just be a half-and-half vote. Similarly for "OK" and "bad". This would mean that the ballot effectively had 5 levels of support, but wouldn't change the dominant strategies. So aside from requiring some extra instructions on the ballot, it wouldn't sacrifice the main advantages of 3-2-1 and PAD, because naive or lazy votes would still be likely to be strategically optimal.)

I'd also consider the package deal of STAR and P-STAR to be a very good option. I think that the simplicity and honesty of this proposal is slightly below those of 3-2-1 and PAD, but the expressivity is slightly higher.

On a meta-level, if I were drafting a bill/initiative, I'd make it put the decision in the hands of a Citizens' Assembly, the way they did in BC in 2004. It shouldn't be up to self-selected experts like us; or politicians with a self-interested stake in the outcome; or ordinary voters without time to study the issues. A citizens' assembly is exactly what you need to get ordinary voters and then give them time to study the issues. The likely outcome would be that a large supermajority of the CA would agree on an option, giving a very clear signal that I don't think you'll ever get out of an argument on this mailing list, or a lowest-common-denominator initiative campaign in a distracted media environment.

And of course if a citizens' assembly did happen, I'd consider it an unparalleled honor to be able to testify before them. I'd be happy to be sandwiched between Rob Richie and Bill Tieleman (the anti-PR guy from BC) and any number of other people with whom I disagree entirely. I'd trust the voters to get it right in the end. Maybe not perfectly right, but at least a lot righter than Tieleman.

So my initiative would be something like: "Oregon shall create a Citizens' Assembly, consisting of one man and one woman from each state house district, chosen through a random process, to study possible changes to the voting system of Oregon, including proportional representation. This assembly shall have access to expert testimony and administrative assistance, and shall be paid a modest stipend and reimbursed for transportation, childcare, and other necessary support, to meet for at least 40 days over a period of 9 months, preferably on weekends and/or holidays. At the end of that time, any proposals which have support from over 2/3 of the assembly will be put to voters as initiatives in the next regularly-scheduled election, requiring 50% to pass. The Citizens' Assembly shall also be responsible for creating the arguments for and against each such proposal that accompany that election."

--
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.

[Brian Olson]

This is my entry into the PR algorithm space, based on ratings ballots (0-5 should be fine):

https://bolson.org/voting/IRNRP.html

It definitely requires a computer to solve for, but the code is easy to write and easy to verify.

1. Maintain a 'weight' for each choice, staring at 1.0.
2. Sum up the normalized de-weighted preferences from the voters
   For each vote:
   • multiply each voted preference by the current weight for that choice (deweight)
   • divide these by the square root of their summed squares (normalize)
   • add those results to the per-choice sums
3. The total vote is the sum of the vote summed up across choices. The quota is (total vote)/(seats + 1)
4. For each choice with a sum greater than the quota, that choice's weight becomes ((old weight) * (quota / sum)). This will distribute surplus vote to voters' other choices on the next round.
   If negative ratings are allowed changes may cause a choice's sum to fall, in which case if it was above the quota it may fall below the quota. It's weight should be adjusted upwards to be (weight * (quota/sum)), but it's weight shall be limited to 1.0.)
5. If no choices are greater than quota* then the choice with the lowest sum has its weight set to zero. This disqualifies that choice and voters will have their vote distributed to their other choices on the next round.
6. Repeat steps 2-5 until the right number of candidates have a sum greater than or equal to quota.

There's a meta-discussion of what do we value in PR systems?

Algorithmic simplicity?

Ability to run it by hand?

Specific kinds of outcome measures? (Proportionality, Social Utility, Gini Inequality, etc?)

Mathy provable qualities? (Monotonicity, Irrelevance of Equivalent Options, Clone-proof, etc?)

I tend towards utilitarian arguments, but the only thing I can think of to verify in simulation is Proportionality. How do you measure social utility when some people get one of their choices elected and some people get seven of their choices elected but some people should get more because they're part of a big coalition that needs to be proportionally represented? So, the best test I can think of is to just create clear voting blocs of a set size, and see if they get that number of seats to within roundoff.

[Sara Wolf]

I love the Citizens Assembly idea. To the extent possible this is what we've been doing here in PDX. Attempting to have an open, inclusive and well informed, non-biased process leading up to choosing a reform is definitely key to building consensus. Official support and funding would be rad! 

Of course the hard part is finding volunteers and citizens willing to engage to the extent needed in order for them to make a well informed and non-biased decisions. Many people don't really care or want to take the time to learn so that's a thing. Paying them would certainly help! Those that are passionate about this are sometimes not super open minded, understandably, as many have spent significant personal time and energy in pursuit of a specific goal. Being invested in a specific reform isn't the best, especially if the original opinion was based on false or misleading claims or fear based propaganda tactics. That's where I was coming from as an IRV advocate in the past. Even presented with solid info it took a while to change my perspective. Partly because of that personal experience, my focus over the past year has been in a large part to come together around common ways to talk about all of this. When we get to a point where we can skip the "is not/is too" stage and just debate our priorities and the pros and cons we will have a much better shot at educating the public. 

One idea we kicked around on the "Cross Team Dialogue" Loomio was coming up with a certification process for articles, websites, and speakers. This could be essentially a stamp of approval (like peer review) that could go on websites, articles and more. It would state that while we don't agree on proposals and we may not agree on opinions, the source in question doesn't contain false or misleading info. Groups that agreed to be vetted could each get a person on the verification committee. If the committee found problematic claims or wording it would get adjusted to be accurate and not misleading. This group could also come out with a list of problematic and misleading sources and common false claims that would be posted. 

I feel like we've come a long way towards that and there is a good chance FV, Rep.Us, CES, CRV, and EV and more could come together to create that framework or similar. FV WA in particular has made a serious effort and news that FV will take down their 3 most offensive articles (anti-approval/score and star) is a huge step. EV has also taken down stuff or made changes where requested. Once people have access to good info I'd feel a lot better about putting a decision like that in their hands. Getting educated would take a lot less time if we didn't have to figure out who to believe. Thoughts? 

--

-Sara Wolk

Chief Petitioner for "STAR Voting for Multnomah County"

Portland Equal Vote

971-222-9364

sa...@equal.vote

​www​

.

​starv​

o

​ting​

​

.

​

​us​

www.equal.vote

“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, The Center For Election Science

[Jameson Quinn]

Citizens' Assemblies are also sometimes called "Citizens' Juries" or "Deliberative Polls". There are several white papers with best practices out there. For instance:

http://www.activedemocracy.net/articles/cj_handbook.pdf

https://www.ourvoice.scot/719/documents/1024

http://dewa2016.ecodistricts.org/wp-content/uploads/2016/11/11-Susanna-Haas-Lyons.pdf

... There's another document from some organization in Canada, I believe Ontario, that I can't find right now that's even better than the three above. I've a friend who knows to send me that link and I'll share it when I get it.

Obviously, the Oregon Citizens' Initiative Review is a relevant model as well, though not exactly the same thing I'm discussing. That model just reviews existing initiatives; I'm imagining something that would draft one (or more— perhaps one for single-winner and one for multi-winner).

[Clay Shentrup]

Reminds me of a Twitter thread I just saw.

https://twitter.com/davidshor/status/998397696313626624