Approval voting vs instant runoff voting, condorcet, etc

Core Development
1

i saw this discussion in which @gpshead suggested switching from approval voting to instant runoff voting to “better capture voter preferences and provide more meaningful feedback to candidates.”

to be clear, this would have the opposite effect. extensive voter satisfaction efficiency calculations from princeton math phd warren smith (RangeVoting.org - Bayesian Regrets shown graphically) and harvard stats phd jameson quinn (Voter Satisfaction Efficiency (VSE) summary | Voter Satisfaction Efficiency Simulator) show that approval voting is more accurate at capturing voter preference, with virtually any model, including variations of strategic vs honest voting.

this is true not just in the sense of determining a winner, but in terms of surfacing an accurate measure of support for non-winners as well, due to the later-no-harm flaw in IRV (causes it to fail precinct summability).

more sophisticated ranking methods such as condorcet can in some models slightly outperform approval voting and even score voting (approval voting is just score voting on a 0-1 binary scale), but this very tiny potential improvement comes at great cost in terms of complexity, and lack of summability/transparency.

thanks

cc @ncoghlan

3 Likes
2

FYI, Veritasium has just released a video on voting systems: Why Democracy Is Mathematically Impossible (youtube.com)

2 Likes
3

yeah. it’s funny because this is the video i’ve been wanting him to make for years, and then he does it and:

  1. fails to mention that approval voting is used in fargo and st louis.
  2. calls score voting a “form of approval voting”, when in fact approval voting is the binary 0-1 form of score voting.

but whatever, i’ll take it.

1 Like
4

guido mentioned STAR voting here, which is basically the best single-winner voting method, per voter satisfaction efficiency metrics. some other methods can perform a scintilla better in certain strategy models, but STAR voting is radically simpler and more transparent.

full disclosure: i’m the co-inventor of STAR voting.

5

So then let me ask you, why does STAR have 6 values instead of 5, like reviews on Amazon etc.?

(EDIT: Hm, that sounded harsher than I meant. Sorry. It is an honest question I have, and the FAQ at Q: Why is it a 0 through 5 star rating? Not more or less? - STAR Voting doesn’t really seem to address it. Happy to elaborate if my question is still unclear.)

6

i let a princeton math phd who’s arguably the world’s greatest expert on social choice theory answer that one.

https://www.rangevoting.org/RateScaleResearch

tl;dr - having an even number of options seems to have a benefit of tricking people into thinking more carefully rather than just lazily choosing the middle score. and having a 0 as the lowest score is extremely helpful in disambiguating cardinal scores from ordinal ranks, 1-2-3.

as for the number of options, STAR was initially proposed to have 0-9, but then it was later modified to be 0-5 since that seemed closer to familiar scales like yelp’s “star ratings”, and the greater accuracy has decreasing marginal utility.

unrelated, but kind of cool: a paper warren co-authored with ron rivest.
https://www.rangevoting.org/Rivest3B

6 Likes
7

Loved that paper! I guess I can live with 0-5 stars even though on Yelp etc. there’s no such thing as a zero-star review.

8

i think a yelp competitor with zero star reviews could be a hit! funny enough, when i was the 2nd dev at sharethrough, we were in yelp’s old office on 650 mission. it’s a small world.

1 Like
9

I was sort of surprised that blanks are zeros (assuming you sum stars)[1]. I found the relevant FAQ entry:

For the purpose of post-election data analysis, the number of candidates left blank, vs those explicitly bubbled in as a 0 does provide some additional data that could be interesting for campaigns, and could be published as part of the full election data analysis.

Do you know if there are any examples of such data published?

  1. Naively, I thought a different blank score could conceivably reduce the importance of money in a political campaign or help polarisation. It also feels a little pessimistic that an unknown choice is the worst possible choice. ↩︎

10

Okay, going through the case studies on the website, the only one I found with any detailed information is Independent Party of Oregon STAR Voting Primary Spotlight on the Data: | by STAR Voting | Medium

This case study is interesting, because the runoff was very close and it turns out a large number of voters didn’t express a preference between the finalists. The article also speculates (based on affinity to first choice votes) that if voters had put more effort into expressing their preferences, the result would probably have been different.

I’d still be very curious to see blank vote data. But maybe a non-worst blank score is a red herring, and a better solution is that bullet voting leads to delegating your vote to your candidate or something.

11

there are more sophisticated things you can do, like this quorum i proposed.
https://www.rangevoting.org/BetterQuorum

the star team went with simplicity, for the sake of political viability. i doubt it makes much difference in practice.

This case study is interesting, because the runoff was very close and it turns out a large number of voters didn’t express a preference between the finalists.

first, it’s important not to focus too much on any single anecdote. we want to look at statistical aggregates over zillions of randomized elections, which is what our the aforementioned simulation authors did.

second, the right candidate still won as far as we can tell.

**Did the candidate who best represented the electorate win?

Yes. Kim Thatcher-R was the candidate who was the highest scoring overall, she also was the candidate with the biggest block of supporters. In the runoff, Thatcher was preferred by more voters than Smith, the other finalist, and she was also the Condorcet winner, meaning that in head to head match-ups she was preferred over any of her opponents.
3 Likes
12

Just a moderator note. Some of the posts here were temporarily hidden by automation concerned about a new user posting multiple times linking to the same domain unknown to the spam filter. I restored all the posts and marked them as not spam. Sorry for the inconvenience.

IMG_8702
IMG_87021125×2436 226 KB

7 Likes
13

I have created two follow-up topics:

3 Likes
14

The argument that you should use proportional star voting is a complex nuanced topic, and it is not at all established that proportionality is inherently Superior for producing outcomes, which best reflect the will of the people.

that said, if you’re going to do proportional representation, I would recommend just doing this.

https://www.rangevoting.org/RRV

you could certainly apply the top two automatic runoff to every round, but given proportionality already mitigates strategy, it’s questionable whether that’s really worth it.

also there 's are some variations on how to do the proportionality formula for star voting. I can say more if people are interested about why I think The method described in that link is the best balance of accuracy and simplicity.

1 Like
15

NOTE: The article linked to is not independent; it’s by the same person who just linked to it. Please, if you’re going to link to your own article, at least say so.

But more importantly, the article doesn’t establish in any way that proportionality is bad. You’re pushing a particular line of argument without any actual backing beyond that it’s your “personal suspicion”. Unless there’s a very good reason to say otherwise, I would assume that proportional voting IS a superior outcome, since it allows minorities a much better chance of not being completely ignored.

8 Likes
16

i mean…this is an ad hominem fallacy. either the evidence supports the argument or it doesn’t. it doesn’t matter who said it.

it’s not relevant. either the evidence supports the conclusion or it does not.

nor does it even make the argument that proportionality is bad. it just counters the common assumption that proportionality is inherently good. it specially links to a comparative analysis of the pro-PR and anti-PR literature, from some of the most respected experts in history, as summarized by princeton math phd warren smith, arguably the greatest expert on voting methods. (and note: i am not claiming, a la “ad hominem”, that this makes his position correct: just that it’s a basic sanity check; i encourage you to look at the primary data from hain, katz, amy, et al.)

http://scorevoting.net/PropRep

this is deeply incorrect. i demonstrated, using a simple thought experiment, that PR is obviously not inherently superior, and that the question of whether it is superior to block methods is simply an empirical matter. i’m not taking a strong position on what the data says as to that empirical question (which is what you seem to be arguing about), but rather i’m simply saying that it is an empirical question. further, i cited some of the best research that exists on the question. although, admittedly, there is extremely scant research, since no “good” advanced block voting methods have been used, so there’s a paucity of data. that is, existing research compares PR to obviously horrible multi-winner block plurality voting, which is subject to vote splitting. there’s virtually no data on e.g. block score voting, approval voting, etc.

my summary would be that your response is essentially an ad hominem fallacy plus a counter-argument to something i’m not even saying. i’m not saying that PR is known to be worse than block centrist-favoring methods. i’m simply saying that any assumption that we have an answer on this is naive. it’s just gut intuition absent of data.

Unless there’s a very good reason to say otherwise, I would assume that proportional voting IS a superior outcome, since it allows minorities a much better chance of not being completely ignored.

and i just showed why that’s an unfounded assumption. it’s just an intuition. there’s virtually no actual evidence for it.

6 Likes
17

Maybe you could both step back from personal attacks, or I may petition the moderators to dampen the discussion a bit.

4 Likes
18

Uhh, I did not attack you by saying this, I simply added a NOTE, like the word said, to clarify the fact that you were linking to your own article, not to someone else’s supporting evidence. I think that it’s polite to tell people that you’re linking to yourself.

An ad hominem argument would be something like “you don’t use any capital letters therefore your position is wrong”. This is not that. This is a point of courtesy, asking you to be clear about what you’re linking to before we click through.

Except that it doesn’t. It doesn’t actually have any line of argument. You provide a thought experiment and then you state that your “personal suspicion” - that’s not scare quotes, that’s your own words - is that centrist-favouring is better. You haven’t proven anything, you haven’t countered anything.

But, I get the feeling that rational debate on this Discourse isn’t what people are looking for. So I’ll bow out. Have whatever voting scheme you like, there’s little chance of me having any representation regardless.

3 Likes
19

i certainly don’t plan to engage in any personal attacks.

20

i didn’t say anything about “attacking” me. an ad hominem fallacy is when you’re criticizing the person making an argument rather than the argument/evidence itself. “this is not independent, it’s written by the person citing it”, etc.