Proportional Approval IRV

For the serious discussion of weighty matters and worldly issues. No off-topic posts allowed.

Moderators: Azrael, Moderators General, Prelates

User avatar
Qaanol
The Cheshirest Catamount
Posts: 3058
Joined: Sat May 09, 2009 11:55 pm UTC

Proportional Approval IRV

Postby Qaanol » Fri May 16, 2014 12:23 am UTC

I just had a thought for a possible voting system. It achieves proportional representation, voters fill out simple approval-style ballots, and tallying is done in an instant-runoff fashion.

How it works:

Suppose that n seats are being contested. As a voter, you can check yes or no for as many candidates as you like. Say you approve k of them: then each of the candidates you approve gets 1/k points from your ballot.

The points are added up from all ballots, and the candidate with the lowest total is eliminated. Ballots that voted for that candidate are recalculated: your other choices now get 1/(k-1) points from your ballot, because one fewer candidate you approve of remains.

The totals are updated, and again the candidate with the lowest score is eliminated. This continues until only n candidates remain, and they are declared the winners.

Notable properties:

• The outcome is proportional
• Voters have a simple “check yes or no” ballot
• Tallying is complex, and pragmatically must be done by computer
• Each ballot is always worth exactly 1 point as long as any approved candidates remain

So what do you think?

And what should we call it? Personally I’m leaning toward “Field-approval instant-runoff” voting.

A possible variant would be to let voters give a score to each candidate, then weight each ballot so it is worth 1 point in total.
wee free kings

Derek
Posts: 2178
Joined: Wed Aug 18, 2010 4:15 am UTC

Re: Proportional Approval IRV

Postby Derek » Fri May 16, 2014 2:18 am UTC

I believe there can exist a situation where an insincere ballot is strategically optimal. Insincere here means you vote "yes" for some candidate that you prefer less than a candidate for whom you vote "no".

Let's say you like two candidates in the field. Your favorite is very popular, he is a shoo-in. Your second favorite one is borderline, it's very close whether he wins a seat or not. If you vote for both candidates, your second favorite will only receive half a vote and may be eliminated. If you only vote for your second favorite, and insincerely do not vote for your absolute favorite, then your second favorite can win and your favorite is unaffected.

I haven't constructed a concrete example, but it should be possible. Keep in mind that all voting systems have some flaw or another, so just because a scenario like this exists doesn't mean your system is bad (that would require a more detailed analysis).

User avatar
Jplus
Posts: 1711
Joined: Wed Apr 21, 2010 12:29 pm UTC
Location: Netherlands

Re: Proportional Approval IRV

Postby Jplus » Fri May 16, 2014 10:24 am UTC

I think the most important drawback of that system is that the ballot is complicated to handle for the voter (despite having to just check yes/no for each candidate). Do you vote for all candidates that you find acceptable, or only for the ones that you >95% agree with, or something in between? How do you, as a voter, decide what is most likely to make you content with the final result of the election? Also, how do you decide which candidates to include if your scale of acceptability is not one-dimensional? This may lead to confusion, slowdowns and possibly errors when the voter is actually filling out the ballot. I should say I do find it an interesting hybrid of IR and PR voting, though. I haven't calculated it through, but I believe it's neither transitive nor pareto-efficient, but rather a bit of both.

Note that the system you propose is not proportional. All candidates that remain get 1 seat, regardless of the number of votes they received. On top of that, the most popular unchosen candidate might have nearly as many votes as the least popular chosen candidate. The system would be proportional if and only if more votes (for the same candidate) means more seats (for the candidate and whomever they defer to, e.g. a party they are enlisted in).

I've thought about something vaguely related in the past. Instead of letting voters split their vote over as many candidates as they want, each voter gets exactly three votes which are always counted. The voter may give all three votes to the same candidate or split them over up to three candidates. The votes are then counted as in a regular party-list voting system, resulting in a true proportional representation where voters still have some room to express preference. I doubt it would offer real benefits over regular party-list voting systems with one candidate-vote per voter, though.

All of that being said, I think democracies have more pressing concerns than the voting system. In my opinion, the most important is the principle that parties that represent the electorate in parliament/congress also get executive power (deliver ministers/secretaries of state), usually either just the largest party in two-party systems or a majority coalition in multi-party systems. This leads to a host of problems, such as diminished separation of powers, effectively handicapping the (usually significant) proportion of parliament/congress that is not in power, and the fact that it encourages "strategic" voting. The latter tends to promote two-party systems, an effect that is often further amplified by laws that make it harder for small parties to enter the elections. Such laws are usually in place because it is easier to form a coalition if there aren't many small parties, but that problem wouldn't exist in the first place if parties simply didn't form a coalition.

Ministries usually have a lot of expertise on board, including all the means to propose laws and legislations. A parliament/congress that proportionally represents the electorate can check all those proposals and only pass those that a majority agrees with, whether there is a coalition or not. I don't see why on top of that, large parties should be allowed to temporarily place an ideologically motivated leader at the top of a ministry. As far as I can tell it raises many problems while the only problem that it solves (coordination of the plans of the ministries) can also be solved by other means (anything that involves a council of ministers with a chairperson).
"There are only two hard problems in computer science: cache coherence, naming things, and off-by-one errors." (Phil Karlton and Leon Bambrick)

coding and xkcd combined

(Julian/Julian's)

User avatar
Qaanol
The Cheshirest Catamount
Posts: 3058
Joined: Sat May 09, 2009 11:55 pm UTC

Re: Proportional Approval IRV

Postby Qaanol » Fri May 16, 2014 11:00 am UTC

Derek wrote:I believe there can exist a situation where an insincere ballot is strategically optimal. Insincere here means you vote "yes" for some candidate that you prefer less than a candidate for whom you vote "no".

Let's say you like two candidates in the field. Your favorite is very popular, he is a shoo-in. Your second favorite one is borderline, it's very close whether he wins a seat or not. If you vote for both candidates, your second favorite will only receive half a vote and may be eliminated. If you only vote for your second favorite, and insincerely do not vote for your absolute favorite, then your second favorite can win and your favorite is unaffected.

I haven't constructed a concrete example, but it should be possible. Keep in mind that all voting systems have some flaw or another, so just because a scenario like this exists doesn't mean your system is bad (that would require a more detailed analysis).

Ah yes, that is a problem. I suppose some sort of threshold, like the Droop quota, could be incorporated to ameliorate that issue.

So, once a candidate has enough votes to guarantee a seat, all votes they receive are scaled down by the appropriate factor so their total is exactly at the threshold, and the other choices on those ballots are scaled up to make each ballot still have value 1.

It doesn’t completely eliminate the scenario you describe, but it greatly weakens the effect and it simultaneously helps ensure that the “shoo-in” candidate doesn’t accidentally lose on account of everyone using the reasoning in your example.

Jplus wrote:I think the most important drawback of that system is that the ballot is complicated to handle for the voter (despite having to just check yes/no for each candidate). Do you vote for all candidates that you find acceptable, or only for the ones that you >95% agree with, or something in between? How do you, as a voter, decide what is most likely to make you content with the final result of the election? Also, how do you decide which candidates to include if your scale of acceptability is not one-dimensional?

This is a feature, not a bug. The proposed system gives the voter choices, just like plain approval voting does for single-winner races. It is one of the main goals of the system.

Response to the rest of Jplus’s post:

Spoiler:
Jplus wrote:Note that the system you propose is not proportional. All candidates that remain get 1 seat, regardless of the number of votes they received. On top of that, the most popular unchosen candidate might have nearly as many votes as the least popular chosen candidate. The system would be proportional if and only if more votes (for the same candidate) means more seats (for the candidate and whomever they defer to, e.g. a party they are enlisted in).

I think you have a fundamental misunderstanding about what proportional representation means. The proposed system is definitely proportional, just like STV. I designed this as a modification of STV to simplify the ballot and improve the proportionality of outcomes.

Jplus wrote:I've thought about something vaguely related in the past. Instead of letting voters split their vote over as many candidates as they want, each voter gets exactly three votes which are always counted. The voter may give all three votes to the same candidate or split them over up to three candidates. The votes are then counted as in a regular party-list voting system, resulting in a true proportional representation where voters still have some room to express preference. I doubt it would offer real benefits over regular party-list voting systems with one candidate-vote per voter, though.

This is a known system that has serious vote-splitting flaws.

Jplus wrote:All of that being said, I think democracies have more pressing concerns than the voting system. In my opinion, the most important is the principle that parties that represent the electorate in parliament/congress also get executive power (deliver ministers/secretaries of state), usually either just the largest party in two-party systems or a majority coalition in multi-party systems. This leads to a host of problems, such as diminished separation of powers, effectively handicapping the (usually significant) proportion of parliament/congress that is not in power, and the fact that it encourages "strategic" voting. The latter tends to promote two-party systems, an effect that is often further amplified by laws that make it harder for small parties to enter the elections. Such laws are usually in place because it is easier to form a coalition if there aren't many small parties, but that problem wouldn't exist in the first place if parties simply didn't form a coalition.

The voting system is what promotes the two-party system. Plurality voting (aka. first past the post, or “vote for one”) which the USA uses is especially terrible. Instant-runoff voting is slightly better, but still not great. Approval and score voting are much better, and a proportional representation system could be better still.

Australia uses STV for its Senate, but that system tends to get people voting “above the line” for pre-made party lists. In other words, people end up voting for parties rather than candidates. Now, the system I propose could also have an “above the line” party-list option, but then it would be a form of approval voting for parties, with the outcome still proportional.

Jplus wrote:Ministries usually have a lot of expertise on board, including all the means to propose laws and legislations. A parliament/congress that proportionally represents the electorate can check all those proposals and only pass those that a majority agrees with, whether there is a coalition or not. I don't see why on top of that, large parties should be allowed to temporarily place an ideologically motivated leader at the top of a ministry. As far as I can tell it raises many problems while the only problem that it solves (coordination of the plans of the ministries) can also be solved by other means (anything that involves a council of ministers with a chairperson).

I am not quite clear how you intend to select the ministers.
wee free kings

User avatar
Jplus
Posts: 1711
Joined: Wed Apr 21, 2010 12:29 pm UTC
Location: Netherlands

Re: Proportional Approval IRV

Postby Jplus » Fri May 16, 2014 1:05 pm UTC

Qaanol wrote:[...] The proposed system gives the voter choices, just like plain approval voting does for single-winner races. It is one of the main goals of the system.

Well, given that you start from STV, I understand why you might want to do this. Still, more choice isn't always better.

Unquoted because I believe it to be on topic:

Jplus wrote:Note that the system you propose is not proportional. [...]

I think you have a fundamental misunderstanding about what proportional representation means. The proposed system is definitely proportional, just like STV. I designed this as a modification of STV to simplify the ballot and improve the proportionality of outcomes.

I understand perfectly well what proportional representation means, thank you.
I wrote:The system would be proportional if and only if more votes (for the same candidate) means more seats (for the candidate and whomever they defer to, e.g. a party they are enlisted in).

Wikipedia wrote:Proportional representation (PR) is a principle applied to voting systems to elect an assembly or council to ensure that the number of seats won by a party or group of candidates is proportionate to the number of votes received. For example, under a PR voting system, if 30% of voters support a particular party then roughly 30% of seats will be won by that party.

I might, however, have understood your system. Let's make this clear once and for all. Suppose we have 4 seats and 5 candidates, A, B, C, D and E. A gets 99 votes, B gets 101, C gets 200, D gets 400 and E gets 1200. If I understood your system correctly, B, C, D and E will each get one seat. In proportional representation it depends a bit on the rest seat calculations, but a likely outcome is that D gets one seat and E gets the remaining three; in any case, E should get at least two seats and B shouldn't get any. Unless I understood your system incorrectly, it isn't proportional. I should point out that your system doesn't transfer votes of sure winners, unlike STV (if I understood your system correctly).

Spoiler:
Jplus wrote:I've thought about something vaguely related in the past. Instead of letting voters split their vote over as many candidates as they want, each voter gets exactly three votes which are always counted. [...]

This is a known system that has serious vote-splitting flaws.

Well, I didn't believe I was the first to come up with the idea anyway. After all, it's basically the same as a plain vanilla party-list system where each voter does three independent votes. Would you mind telling me what the system is called and what are its flaws?

The voting system is what promotes the two-party system. Plurality voting (aka. first past the post, or “vote for one”) which the USA uses is especially terrible. Instant-runoff voting is slightly better, but still not great. Approval and score voting are much better, and a proportional representation system could be better still.

The USA is the worst example that I know of. However, even the Netherlands, where I live and which has a pretty decent proportional representation system (based on party-list voting, method D'Hondt) with on average 8–10 parties in the parliament, suffers from strategic voting effects and tendencies to a few-party system. Periodically we have proposals resurface to make a greater threshold for parties to enter parliament, exactly because it would make it easier to form a coalition. Obviously, this would badly reduce diversity and choice.

All in all, no matter what voting system is used, if political parties in parliament get to also have executive power it will always work against proportional representation.

Jplus wrote:Ministries usually have a lot of expertise on board, including all the means to propose laws and legislations. A parliament/congress that proportionally represents the electorate can check all those proposals and only pass those that a majority agrees with, whether there is a coalition or not. I don't see why on top of that, large parties should be allowed to temporarily place an ideologically motivated leader at the top of a ministry. As far as I can tell it raises many problems while the only problem that it solves (coordination of the plans of the ministries) can also be solved by other means (anything that involves a council of ministers with a chairperson).

I am not quite clear how you intend to select the ministers.

Internal voting in the ministries.
"There are only two hard problems in computer science: cache coherence, naming things, and off-by-one errors." (Phil Karlton and Leon Bambrick)

coding and xkcd combined

(Julian/Julian's)

User avatar
Qaanol
The Cheshirest Catamount
Posts: 3058
Joined: Sat May 09, 2009 11:55 pm UTC

Re: Proportional Approval IRV

Postby Qaanol » Sat May 17, 2014 1:02 am UTC

Jplus wrote:Suppose we have 4 seats and 5 candidates, A, B, C, D and E. A gets 99 votes, B gets 101, C gets 200, D gets 400 and E gets 1200. If I understood your system correctly, B, C, D and E will each get one seat. In proportional representation it depends a bit on the rest seat calculations, but a likely outcome is that D gets one seat and E gets the remaining three; in any case, E should get at least two seats and B shouldn't get any. Unless I understood your system incorrectly, it isn't proportional.

I suggest you re-read the text you quoted from Wikipedia: it does not say that any individual candidate would hold more than one seat (although there are some delegated-vote systems with that property) but rather that parties or groups of candidates will receive a number of seats proportional to the votes they receive. So if A, B, C, D, and E are candidates for four seats, then yes, four of them will each get a seat. But if instead they are parties, each fielding several candidates, then indeed it is quite likely that 3 candidates from party E would each get a seat, and 1 from party D.

Jplus wrote:I should point out that your system doesn't transfer votes of sure winners, unlike STV (if I understood your system correctly).

The original version of my idea did not, but the revised version as described in my previous post does indeed transfer excess votes from sure winners.

Jplus wrote:
Jplus wrote:I've thought about something vaguely related in the past. Instead of letting voters split their vote over as many candidates as they want, each voter gets exactly three votes which are always counted. [...]

This is a known system that has serious vote-splitting flaws.

Well, I didn't believe I was the first to come up with the idea anyway. After all, it's basically the same as a plain vanilla party-list system where each voter does three independent votes. Would you mind telling me what the system is called and what are its flaws?

Cumulative voting, and its big flaw as I mentioned is vote-splitting. It was used in Illinois for a long time in the specific case where each voter gets 3 votes, and there are 3 seats in each district. The result was that most districts became firmly 2-D’s-and-1-R or 2-R’s-and-1-D. So to an extent it is better than single-winner plurality, but not much.
wee free kings

Derek
Posts: 2178
Joined: Wed Aug 18, 2010 4:15 am UTC

Re: Proportional Approval IRV

Postby Derek » Sat May 17, 2014 3:56 am UTC

The original system is definitely proportional. I'll show a very handwavy proof of this for a simplified model where parties nominate at least as many candidates as there are seats, and voters approve every candidate of their favored party, and no other candidates. This is not a particularly realistic model, but it makes for an easy comparison with party-list proportional voting. Let's say that there are n seats and each party wins a portion p_k of the vote. If they have fielded c_k candidates, then each candidate has p_k/c_k votes. Now candidates will start being eliminated, starting with the lowest p_k/c_k, until there are n candidates left. At this point each remaining candidate must have 1/n = p_k/c'_k votes, where c'_k is the number of winning candidates for party k. Then c'_k = p_k*n, so the number of seats they win is proportional to their portion of the vote.

Additionally this shows that there is no advantage or disadvantage for fielding additional similar candidates, a useful property, and there is no advantage or disadvantage to fielding fewer candidates as long as you field enough to fill the expected number of seats that you will win.

Of course in reality we expect (and hope) that voters will not vote along straight party lines, but this shows the system works at least as well as party-list proportional voting. One problem with doing a more complex analysis of proportionality is that it's not clear what "proportional" means when voters can voter for multiple candidates.

Qaanol wrote:Ah yes, that is a problem. I suppose some sort of threshold, like the Droop quota, could be incorporated to ameliorate that issue.

So, once a candidate has enough votes to guarantee a seat, all votes they receive are scaled down by the appropriate factor so their total is exactly at the threshold, and the other choices on those ballots are scaled up to make each ballot still have value 1.

It doesn’t completely eliminate the scenario you describe, but it greatly weakens the effect and it simultaneously helps ensure that the “shoo-in” candidate doesn’t accidentally lose on account of everyone using the reasoning in your example.

Presumably this Droop quota would apply before dropping additional candidates? Also, how is "enough votes" defined? It's not enough to say 1/n votes, because in certain (probably unlikely) scenarios no candidate can receive more than 1/n votes. I'm sure there is a way to compute this, but it's not obvious to me how. I think you need a system that can always define one "shoo-in", but if that's the case then you would fill seats with a series of "shoo-ins" instead of by eliminating low performing candidates. So I think there are a number of issues to be clarified here.

There are proportional approval voting systems that work by filling in the winners one at a time, and reducing the weight of the voters who have had a successful candidate. Here is one such model, but there's no analysis.

User avatar
Qaanol
The Cheshirest Catamount
Posts: 3058
Joined: Sat May 09, 2009 11:55 pm UTC

Re: Proportional Approval IRV

Postby Qaanol » Sat May 17, 2014 12:31 pm UTC

Derek wrote:Presumably this Droop quota would apply before dropping additional candidates? Also, how is "enough votes" defined? It's not enough to say 1/n votes, because in certain (probably unlikely) scenarios no candidate can receive more than 1/n votes. I'm sure there is a way to compute this, but it's not obvious to me how. I think you need a system that can always define one "shoo-in", but if that's the case then you would fill seats with a series of "shoo-ins" instead of by eliminating low performing candidates. So I think there are a number of issues to be clarified here.

The Droop quota is the lowest threshold that guarantees no more than n candidates will be elected. Numerically, if v votes are casts, it is (v/(n+1)) + 1 rounded down. Note that v/(n+1) is the highest vote total that n+1 candidates could simultaneously achieve, so making it 1 more than that ensures that only n candidates can reach the threshold. Also, v is the number of ballots that approve at least one candidate, which could change as candidates are eliminated.

Concretely, the thresholds are:
For 1 seat, 50% (plus 1 vote)
For 2 seats, 33⅓% (plus 1 vote)
For 3 seats, 25% (plus 1 vote)
For 4 seats, 20% (plus 1 vote)
etc.

Initially, there is no guarantee that anyone reaches the threshold. So the least-popular candidates are eliminated until someone does.

The way I envision the calculation going is:
• How many remaining candidates did you approve? Call it k
• Each of them gets 1/k points from your ballot
• Did a candidate reach the quota? If so scale their total to equal the threshold, and call the scaling factor s
• Your ballot now contributes s/k to that candidate
• Scale your votes for other sub-threshold candidates so your ballot is worth 1 point total
• Did another candidate reach the quota? If so, repeat. If not, eliminate the weakest.

So, as candidates you approve get eliminated, your votes for other candidates get strengthened. And as candidates you approve top the threshold to win a seat, your votes for other candidates get a share of the surplus.

Now, I don’t know how well this will work in practice. For starters, the tallying system is rather complicated just to explain.

Derek wrote:There are proportional approval voting systems that work by filling in the winners one at a time, and reducing the weight of the voters who have had a successful candidate. Here is one such model, but there's no analysis.


As your link mentions, there is another way to make approval voting proportional, as described in this video. The idea is to do a standard approval voting tally, and whichever candidate the most voters approve of wins the first seat. Then if you approved that candidate, your remaining choices count half as much. Again the totals are found, and the leading candidate gets the second seat. Ballots are reweighted again, so that whenever m candidates that you approve of have won, your choices count for 1/(m+1).

I haven’t tried comparing the two, though I’ll note that the version in the video has the advantage that there are no “elimination” steps. On the other hand, that version does not transfer votes from sure winners, which means it is susceptible to the same tactic you mentioned, about not wanting to vote for a sure winner.
wee free kings

User avatar
Jplus
Posts: 1711
Joined: Wed Apr 21, 2010 12:29 pm UTC
Location: Netherlands

Re: Proportional Approval IRV

Postby Jplus » Sun May 18, 2014 3:48 pm UTC

Qaanol wrote:
Jplus wrote:Suppose we have 4 seats and 5 candidates, A, B, C, D and E. A gets 99 votes, B gets 101, C gets 200, D gets 400 and E gets 1200. If I understood your system correctly, B, C, D and E will each get one seat. In proportional representation it depends a bit on the rest seat calculations, but a likely outcome is that D gets one seat and E gets the remaining three; in any case, E should get at least two seats and B shouldn't get any. Unless I understood your system incorrectly, it isn't proportional.

I suggest you re-read the text you quoted from Wikipedia: it does not say that any individual candidate would hold more than one seat (although there are some delegated-vote systems with that property) but rather that parties or groups of candidates will receive a number of seats proportional to the votes they receive. So if A, B, C, D, and E are candidates for four seats, then yes, four of them will each get a seat. But if instead they are parties, each fielding several candidates, then indeed it is quite likely that 3 candidates from party E would each get a seat, and 1 from party D.

That might be likely, but consider the case where candidate A is the sole member of Party1, B and C are the members of Party2 and D and E are the members of Party3. Now explain to me how your system leads to a proportional distribution of the seats over the parties.
"There are only two hard problems in computer science: cache coherence, naming things, and off-by-one errors." (Phil Karlton and Leon Bambrick)

coding and xkcd combined

(Julian/Julian's)

User avatar
Qaanol
The Cheshirest Catamount
Posts: 3058
Joined: Sat May 09, 2009 11:55 pm UTC

Re: Proportional Approval IRV

Postby Qaanol » Mon May 19, 2014 1:09 am UTC

Jplus wrote:That might be likely, but consider the case where candidate A is the sole member of Party1, B and C are the members of Party2 and D and E are the members of Party3. Now explain to me how your system leads to a proportional distribution of the seats over the parties.

If the parties nominate fewer candidates than the number of seats they could have won, that is their own fault. Luckily, the system I described lets every party nominate a full slate of candidates for all available seats, without penalty.

Contrast that with cumulative voting, where a party that nominates more candidates than the number of seats it can actually win, is quite likely to win fewer seats than it could have.
wee free kings

User avatar
Jplus
Posts: 1711
Joined: Wed Apr 21, 2010 12:29 pm UTC
Location: Netherlands

Re: Proportional Approval IRV

Postby Jplus » Mon May 19, 2014 7:10 am UTC

Irrelevant. Let's say that Party3 also has candidates F and G, but neither gets any votes. How does your system lead to proportional representation?
"There are only two hard problems in computer science: cache coherence, naming things, and off-by-one errors." (Phil Karlton and Leon Bambrick)

coding and xkcd combined

(Julian/Julian's)

User avatar
Qaanol
The Cheshirest Catamount
Posts: 3058
Joined: Sat May 09, 2009 11:55 pm UTC

Re: Proportional Approval IRV

Postby Qaanol » Mon May 19, 2014 11:22 pm UTC

Jplus wrote:Irrelevant. Let's say that Party3 also has candidates F and G, but neither gets any votes. How does your system lead to proportional representation?

It is not irrelevant. Are you conjecturing a political party so fragmented that, even under a ballot system that allows approving unlimited candidates, the majority of its supporters would bullet-vote of one or two of its nominees and no one at all would support the rest? That is, frankly, preposterous. Such a party would split in two, or start nominating better candidates.

I am trying to give you the benefit of the doubt here, and answer your questions forthrightly, but it is hard for me to believe you are acting in good faith when you are not even debating the merits of the voting system I proposed, and instead you seem to be arguing about the definition of “proportional representation” even though that is a well-known term in voting theory.

It is as if someone were talking about single transferable vote as a proportional representation system, and you started arguing, “But what if every voter just marked their first choice and didn’t rank the rest of the candidates, and one specific candidate received the overwhelming majority of the votes, and no one else from that same political party got any votes?”

Well, yes, if STV voters bullet-vote, and if people in my proposed system bullet vote, then indeed the outcome reduces to a plurality vote with the top n candidates each getting a seat. But you know what? That is still a proportional representation system. Not a very good one, but the point of alternative voting systems is to provide voters the ability to cast more-informative ballots, so the results become even more representative of the constituency.

So, do you have a reason to suspect that a significant number of people would bullet-vote in my proposed system? Because to me it seems like a much better idea to vote for at least half the number of candidates as there are seats, if not a whole slate or even more.
wee free kings

Derek
Posts: 2178
Joined: Wed Aug 18, 2010 4:15 am UTC

Re: Proportional Approval IRV

Postby Derek » Mon May 19, 2014 11:27 pm UTC

Qaanol wrote:It is not irrelevant. Are you conjecturing a political party so fragmented that, even under a ballot system that allows approving unlimited candidates, the majority of its supporters would bullet-vote of one or two of its nominees and no one at all would support the rest? That is, frankly, preposterous. Such a party would split in two, or start nominating better candidates.

Furthermore, it suggests that the party itself doesn't have that much support, only that particular politician.

However I think there is a broader problem of how to define a "proportional" outcome in any system that allows voters to vote for multiple candidates or to rank candidates, either of which is desirable for allowing more voter expression.

We know that these systems compare favorably to party-list voting (basically the FPTP of proportional voting) when voters just block vote, but when voters have more refined (and realistic) opinions, how can we compare which system gives the more "proportional" outcome?

BattleMoose
Posts: 1993
Joined: Tue Nov 13, 2007 8:42 am UTC

Re: Proportional Approval IRV

Postby BattleMoose » Tue May 20, 2014 6:47 am UTC

I would favour the Australian system with a very minor modification. And its a very similar idea to what you propose.

What I don't like about your system is that my absolute preferred party gets "less" of my vote for every other party that I like, and may hinder it in the first few knock offs. In the Australian system my absolute preferred party gets all of my vote as long as it is in the running, then my whole vote goes to my 2nd preferred.

The issue with the Australian system is that you have to rank all the political parties or your vote is void. And with 130+ parties after your vote it becomes far too much of an effort and functionally very very very few people do this. I would just change the Australian system so that you can rank as many parties as you like and if all your parties get knocked off then your vote gets voided.

User avatar
Qaanol
The Cheshirest Catamount
Posts: 3058
Joined: Sat May 09, 2009 11:55 pm UTC

Re: Proportional Approval IRV

Postby Qaanol » Tue May 20, 2014 1:41 pm UTC

Derek wrote:how can we compare which system gives the more "proportional" outcome?

That really is the question.

BattleMoose wrote:What I don't like about your system is that my absolute preferred party gets "less" of my vote for every other party that I like, and may hinder it in the first few knock offs. In the Australian system my absolute preferred party gets all of my vote as long as it is in the running, then my whole vote goes to my 2nd preferred.

It is really a tradeoff. My version treats all your approvals as equal, and whenever any of them get eliminated your votes for the remaining ones get stronger.

By contrast, STV treats your vote as 100% for your top choice, so it does nothing to help your 2nd or 3rd choices stay in the race. So that is one of the weaknesses I am trying to shore up.

BattleMoose wrote:The issue with the Australian system is that you have to rank all the political parties or your vote is void. And with 130+ parties after your vote it becomes far too much of an effort and functionally very very very few people do this. I would just change the Australian system so that you can rank as many parties as you like and if all your parties get knocked off then your vote gets voided.

Is it really true that in Australia your ballot is void if you don't rank all the parties? I would have thought your version was the default, since it makes so much more sense.
wee free kings

BattleMoose
Posts: 1993
Joined: Tue Nov 13, 2007 8:42 am UTC

Re: Proportional Approval IRV

Postby BattleMoose » Tue May 20, 2014 11:52 pm UTC

Qaanol wrote:
BattleMoose wrote:The issue with the Australian system is that you have to rank all the political parties or your vote is void. And with 130+ parties after your vote it becomes far too much of an effort and functionally very very very few people do this. I would just change the Australian system so that you can rank as many parties as you like and if all your parties get knocked off then your vote gets voided.

Is it really true that in Australia your ballot is void if you don't rank all the parties? I would have thought your version was the default, since it makes so much more sense.


It came in at a time when there weren't many parties so it wasn't a problem. Most people don't vote this way though. At some point it became acceptable to vote for a "party" and basically your vote then gets assigned the predefined preferences of the party that you voted for.

Its resulted in much shenaningans between the parties and trading of preferences for other parties.

Curiously it took this to happen for people to realize that this might be a problem.

Ricky Muir represented AMEP in Victoria at the election and won a 14.3 percent quota after preferences on a record-low primary vote of 0.51 percent or 17,122 first preferences[13][14][15] (previous record held by Family First's Steve Fielding in 2004 on 1.9 percent in Victoria)[16] and will assume his seat from 1 July 2014.

http://en.wikipedia.org/wiki/Australian ... iast_Party

Derek
Posts: 2178
Joined: Wed Aug 18, 2010 4:15 am UTC

Re: Proportional Approval IRV

Postby Derek » Wed May 21, 2014 12:32 am UTC

Yeah, that's a problem with any system that requires you to rank every candidate/party. There will always be too many and voters cannot possibly reasonably rank all of them. So they end up voting for a party's official ranking or they just run down the unimportant candidates from top to bottom.

Therefore any ranking system needs to allow an incomplete ballot to be useful.

User avatar
Qaanol
The Cheshirest Catamount
Posts: 3058
Joined: Sat May 09, 2009 11:55 pm UTC

Re: Proportional Approval IRV

Postby Qaanol » Thu May 22, 2014 1:50 am UTC

BattleMoose wrote:At some point it became acceptable to vote for a "party" and basically your vote then gets assigned the predefined preferences of the party that you voted for.

That is indeed a point of concern. I would much rather that people vote for candidates, not parties. But if we’re thinking about proportional representation, then even in a district with as few as 6 seats and 12 parties, that is already 72 candidates—way too many for voters to know and vote on all of them.

So in proportional representation, there might be no way around voting for parties. And if that is the case, perhaps the most reasonable approach is something like what Sweden does. Namely, you vote for exactly one party, and within that party you have the option to vote for candidates. The seats are allocated to parties proportionally, and within each party the ordering of the candidates is by vote totals.

The simplest option for selecting candidates within a party is, “Vote for one candidate.” A somewhat better method would be “Rank the candidates STV style.” Better yet would be, “Vote for one or more.” And score voting / majority approval would be even better still.
wee free kings

BattleMoose
Posts: 1993
Joined: Tue Nov 13, 2007 8:42 am UTC

Re: Proportional Approval IRV

Postby BattleMoose » Thu May 22, 2014 2:10 am UTC

The issue isn't that you vote for a party and not a candidate.

But that you the majority of voters relinquish their control over their preferences to the party that they voted for. Preferences that then get traded between the parties.

Derek
Posts: 2178
Joined: Wed Aug 18, 2010 4:15 am UTC

Re: Proportional Approval IRV

Postby Derek » Thu May 22, 2014 3:46 am UTC

Candidate overload isn't a problem if the voters don't have to consider all candidates to cast a valid ballot. They can just consider the candidates that they like and/or are likely to win seats. Honestly this is a pretty easy criteria to satisfy, most proportional voting systems, including common STV variants, do just fine here. It sounds like Australia's system is just poorly designed though.

mike-l
Posts: 2758
Joined: Tue Sep 04, 2007 2:16 am UTC

Re: Proportional Approval IRV

Postby mike-l » Fri May 23, 2014 4:58 pm UTC

So EVE online has yearly elections for player representatives, and last year they switched to STV. Now the cool bit is that since this is all online, it's very easy to publish complete voting data, and this leads to crowd sourced analysis. What was found was that some 98% of your voting power went to your top 5 choices.

Now these were 14 seat elections with only about 30 candidates, so you'd probably need more ranks to be fully expressive in a real world election, but it does demonstrate that in STV a full ballot is far from necessary.
addams wrote:This forum has some very well educated people typing away in loops with Sourmilk. He is a lucky Sourmilk.

User avatar
aldonius
Posts: 105
Joined: Sun Jan 17, 2010 2:33 am UTC

Re: Proportional Approval IRV

Postby aldonius » Sun May 25, 2014 12:17 pm UTC

Here in Australia we've recently finished the post-election submissions to our Joint Standing Committee on Electoral Matters; the general consensus is that we should introduce ticket-level optional-preferential voting and completely eliminate the Group Voting Tickets.
The end result of that is a massive disincentive for microparties to run in the Senate compared to the current situation, where the final seat is demonstrably a bit of a lottery.

For the Lower Houses in NSW and Queensland, we've used optional preferential IRV (OPV) for quite a while now. I believe NSW has ticket-level OPV for its Upper House too. Queensland of course is unicameral.

OPV works OK on the ground (as now you don't have to rank candidates you don't care about), but the results aren't particularly proportional. Case in point: the last Queensland state election, where the current government won about half of first preferences, got 60% on two-party preferred vs the incumbents... and walked away with 78 of 89 seats. But I digress.

peregrine_crow
Posts: 180
Joined: Mon Apr 07, 2014 7:20 am UTC

Re: Proportional Approval IRV

Postby peregrine_crow » Mon Jun 02, 2014 1:42 pm UTC

Qaanol wrote:So in proportional representation, there might be no way around voting for parties. And if that is the case, perhaps the most reasonable approach is something like what Sweden does. Namely, you vote for exactly one party, and within that party you have the option to vote for candidates. The seats are allocated to parties proportionally, and within each party the ordering of the candidates is by vote totals.


It might be the most reasonable way, but it can still have some pretty nasty side effects. Specifically, there is no way for voters to indicate a dislike for a certain party, thus promoting more radical viewpoints (that are popular with some, but intensely unpopular with others). I thought about this a bit, but I can't seem to think of a good way to solve it.

The best I could think up is to give each voter, aside from their regular positive vote, one negative vote that they can use to subtract one vote from one party. The downside is that this only means that everyone will aim to be the second most radical (rather than the most radical) party. Also it would be a strategically valid choice for an unpopular party to split between a more radical and less radical party so that the more radical party eats up all the negative votes.

Are there any satisfying party based, proportional representation voting systems that take more than just the top of your preference order into account?

Marginally off topic real world example:
Spoiler:
This is currently a fairly serious issue in the Netherlands where there is a moderately large party (the PVV) that is highly unpopular with the majority of the country but very popular with a reasonably (frighteningly) large minority. Now, it's not that likely that they'll ever be part of any coalition government (especially after a stunt they pulled last time anyone gave them anything resembling governmental power (it's a long story)), but they are clogging up the democracy and they are making us look bad internationally.
Ignorance killed the cat, curiosity was framed.

User avatar
Qaanol
The Cheshirest Catamount
Posts: 3058
Joined: Sat May 09, 2009 11:55 pm UTC

Re: Proportional Approval IRV

Postby Qaanol » Mon Jun 02, 2014 11:55 pm UTC

Yeah, in single-winner elections approval voting is great for that: you need broad support to win, so radical candidates won’t get elected.

Of course, under a proportional system it stands to reason that radical parties should get seats corresponding to their level of support. But then the question becomes, what does “level of support” mean, and how can it be reliably measured on a ballot.

Again, for single-winner elections, approval voting provides the answers: “level of support” is the number of voters who approve of a candidate, and it can be measured by checking yes or no for each contender.

But with proportional representation, I have not yet seen a really good solution.
wee free kings

User avatar
Thesh
Made to Fuck Dinosaurs
Posts: 6166
Joined: Tue Jan 12, 2010 1:55 am UTC
Location: Colorado

Re: Proportional Approval IRV

Postby Thesh » Mon Dec 22, 2014 7:51 am UTC

I was thinking about this today, and I think you might want to go with the opposite approach. Every yes vote is one point for each candidate, regardless of how many you vote for. A quota is calculated, based on the number of ballots cast and seats to fill. The candidate with the most approvals is given all of the ballots that approved of them and they are locked in. If they exceed the quota, then the excess votes are put back into play (ideally, a fraction of every ballot is transferred) and then candidate who has the second highest votes is locked in, and so on and so forth until all seats are filled. No one is eliminated until all seats are filled. It may be necessary to recalculate the quota every round to handle exhausted ballots.

Ideally, every candidate you approve of will meet the quota. It's possible that no candidate meets the quota, which indicates the results may be disproportional. In this case, you might want to eliminate the candidate with the fewest votes and discard the exhausted ballots, recalculating the quota; however, I don't think this is necessary as it can't change who will be locked in when the quota is finally met. I'm not sure you can get truly proportional representation with approval voting.
Summum ius, summa iniuria.

User avatar
Qaanol
The Cheshirest Catamount
Posts: 3058
Joined: Sat May 09, 2009 11:55 pm UTC

Re: Proportional Approval IRV

Postby Qaanol » Sat Dec 27, 2014 3:43 am UTC

Thesh, here is a video from the Center for Election Science showing how proportional representation can be achieved from approval-style ballots. Is that what you had in mind?

https://www.youtube.com/watch?v=_jS7b-0PV9E
wee free kings

User avatar
Thesh
Made to Fuck Dinosaurs
Posts: 6166
Joined: Tue Jan 12, 2010 1:55 am UTC
Location: Colorado

Re: Proportional Approval IRV

Postby Thesh » Sat Dec 27, 2014 3:54 am UTC

Yeah, that looks like it, although I'm not sure how they are dealing with the quotas in that case.
Summum ius, summa iniuria.

Tabish khan
Posts: 3
Joined: Fri Dec 26, 2014 11:11 am UTC

Re: Proportional Approval IRV

Postby Tabish khan » Fri Jan 02, 2015 6:28 am UTC

Approval Voting can be done with the same ballots voters are used to. You just remove the rule that says “vote for only one”.
IRV, on the other hand, requires a new ballot. Because ranking all candidates would take up too much space, the ballots are typically limited to ranking three choices.

User avatar
aldonius
Posts: 105
Joined: Sun Jan 17, 2010 2:33 am UTC

Re: Proportional Approval IRV

Postby aldonius » Sat May 09, 2015 5:12 am UTC

Tabish khan wrote:Approval Voting can be done with the same ballots voters are used to. You just remove the rule that says “vote for only one”.
IRV, on the other hand, requires a new ballot. Because ranking all candidates would take up too much space, the ballots are typically limited to ranking three choices.


Say what? Do the ballots you're used to not consist of a list of all candidates (with a box or equivalent adjacent to each name)? In the case I've just described one simply writes a number rather than making a tick.

I suppose some places do use scantron-style ballots, so that must be what you're talking about in terms of 'too much space'.

User avatar
Thesh
Made to Fuck Dinosaurs
Posts: 6166
Joined: Tue Jan 12, 2010 1:55 am UTC
Location: Colorado

Re: Proportional Approval IRV

Postby Thesh » Sat May 09, 2015 5:31 am UTC

Most ballots in the US are machine-read. I can imagine hand-written ballots with STV would lead to a lot of errors when counting.
Summum ius, summa iniuria.

Derek
Posts: 2178
Joined: Wed Aug 18, 2010 4:15 am UTC

Re: Proportional Approval IRV

Postby Derek » Sat May 09, 2015 5:38 am UTC

That wouldn't really be an issue if the voting was electronic. Instead of writing a number next to each name, you could enter it on a keypad, or names could be dragged into an ordered list. Then the machine could produce a paper ballot that is both human and machine readable.

User avatar
Quercus
Posts: 1731
Joined: Thu Sep 19, 2013 12:22 pm UTC
Location: London, UK
Contact:

Re: Proportional Approval IRV

Postby Quercus » Sat May 09, 2015 7:12 am UTC

Derek wrote:That wouldn't really be an issue if the voting was electronic. Instead of writing a number next to each name, you could enter it on a keypad, or names could be dragged into an ordered list. Then the machine could produce a paper ballot that is both human and machine readable.


Aside on electronic voting:
Spoiler:
As long as the individual voter is able to verify their individual ballot (i.e. the same features that the machine uses to read the ballot are visible to the human eye, and the paper ballot is the one that is counted, rather than merely being used as a verification check) I suppose I don't have a problem with that. The vote counting machines should also be as simple as possible - ideally entirely electromechanical. That way anyone with a stake in the election can verify that the circuit that counts each section of the ballot is exactly the same. As soon as votes pass through an integrated circuit there's a link in the voting chain that's unverifiable and the entire system breaks IMO.

I saw a video which described the UK's hand counting of ballots as "quaint". It's not "quaint", it's a key element of the system of total lack-of-trust that has to operate in order to know you have a fair election. In the UK the ballot boxes are unsealed and every vote is counted in front of every candidate for the seat in question. That makes it very difficult to pull off large scale electoral fraud without it being detected. I've yet to come across and electronic voting system that affords the same degree of protection. Here's a good video which goes into this stuff.

Derek
Posts: 2178
Joined: Wed Aug 18, 2010 4:15 am UTC

Re: Proportional Approval IRV

Postby Derek » Sat May 09, 2015 7:54 am UTC

Yes, I was envisioning a paper ballot that can be personally verified by the voter and is used for the official count either by machine or by hand. The point of my suggestion is to create a paper ballot for ranked voting that can be machine counted, but is not enormously large (filling in bubbles takes O(n^2) space on the paper, O(n log n) is optimal). The details of the electronics aren't even particularly important for this idea. It just needs a printer that can produce a reliably machine-readable ballots and a user-friendly UI.

User avatar
aldonius
Posts: 105
Joined: Sun Jan 17, 2010 2:33 am UTC

Re: Proportional Approval IRV

Postby aldonius » Tue May 12, 2015 3:32 pm UTC

Thesh wrote:Most ballots in the US are machine-read. I can imagine hand-written ballots with STV would lead to a lot of errors when counting.

We hand-count everything here (Australia) for the lower house, at least initially; for the majority of seats a result is clear within 48 hours. Upper house gets hand-counted first-preferences on the night, but later switches to data entry and a computer doing the final preference distribution.

The error rate for both is acceptably low, on the whole.
We also have 'scrutineers' - people from candidates' teams who get to observe the counting process in all its grinding detail. Not sure if that's a thing in the US.

Derek wrote:reliably machine-readable ballots

Just print the numbers in a box next to the candidate names, using OCR-B as the font. There, done. Human and machine readable.

Tyndmyr
Posts: 11213
Joined: Wed Jul 25, 2012 8:38 pm UTC

Re: Proportional Approval IRV

Postby Tyndmyr » Tue May 12, 2015 9:29 pm UTC

aldonius wrote:
Thesh wrote:Most ballots in the US are machine-read. I can imagine hand-written ballots with STV would lead to a lot of errors when counting.

We hand-count everything here (Australia) for the lower house, at least initially; for the majority of seats a result is clear within 48 hours. Upper house gets hand-counted first-preferences on the night, but later switches to data entry and a computer doing the final preference distribution.

The error rate for both is acceptably low, on the whole.
We also have 'scrutineers' - people from candidates' teams who get to observe the counting process in all its grinding detail. Not sure if that's a thing in the US.


Kind of. Election officials are generally supposed to be from both parties and all that. In practice, this may break down in heavily partisan districts, where there simply may not be enough volunteers from one party for representation to be very well done.

User avatar
aldonius
Posts: 105
Joined: Sun Jan 17, 2010 2:33 am UTC

Re: Proportional Approval IRV

Postby aldonius » Wed May 13, 2015 2:42 am UTC

I suggest that under any system that isn't FPTP, these so-called 'heavily partisan districts' will see minor-party or independent candidates coming out of the woodwork within a couple of election cycles, which should partially solve the uneven-representation problem.

Also, to clarify, the scrutineers here get to look but not touch. All the actual work is done by electoral commission staff who are, essentially, under a vow of neutrality.

Tyndmyr
Posts: 11213
Joined: Wed Jul 25, 2012 8:38 pm UTC

Re: Proportional Approval IRV

Postby Tyndmyr » Wed May 13, 2015 7:32 pm UTC

aldonius wrote:I suggest that under any system that isn't FPTP, these so-called 'heavily partisan districts' will see minor-party or independent candidates coming out of the woodwork within a couple of election cycles, which should partially solve the uneven-representation problem.

Also, to clarify, the scrutineers here get to look but not touch. All the actual work is done by electoral commission staff who are, essentially, under a vow of neutrality.


You'd probably see some splitting under non-FPTP, yeah. Sort of the same thing that happens now with disagreements within a party, but happening in a more rapid, visible, and geographically separated way.

I believe in the US, there is also some sort of vow or pledge or whatever, for whatever that's worth. Cross verification seems like a more robust element to rely on. As for who touches what, not entirely sure. There's enough people involved, and enough visibility to *usually* make collusion not a big deal. However, some states do things differently. Different ballots, whatever. Some use electronic machines, even. MD is among those, using the much lambasted diebold ones, I think.

User avatar
aldonius
Posts: 105
Joined: Sun Jan 17, 2010 2:33 am UTC

Re: Proportional Approval IRV

Postby aldonius » Sun Jul 12, 2015 2:59 pm UTC

Thesh wrote:Yeah, that looks like it, although I'm not sure how they are dealing with the quotas in that case.


There aren't any quotas in that CFES system, just "top N are winners". (Single-winner) Approval, range and FPTP all have that problem.

The CFES system also has BattleMoose's problem: minor-candidate voters are disincentivised to approve other, "likely-winner" candidates because their votes will count first for those likely-winners and then get weighted down...

User avatar
Qaanol
The Cheshirest Catamount
Posts: 3058
Joined: Sat May 09, 2009 11:55 pm UTC

Re: Proportional Approval IRV

Postby Qaanol » Fri Aug 28, 2015 2:06 am UTC

A system very similar to this has been put in place for the Hugo award nominations (after a relatively small group coordinated under the old system to control the nominations this year).

The new Hugo system tallies points the same way as in this thread (approval-style ballots weighted as 1/k), but the novelty is that the two candidates with lowest scores are compared, and whichever is approved on fewer ballots is eliminated. More info is here, including a lengthy PDF with lots of graphs showing a worked example.
wee free kings

Derek
Posts: 2178
Joined: Wed Aug 18, 2010 4:15 am UTC

Re: Proportional Approval IRV

Postby Derek » Fri Aug 28, 2015 3:15 pm UTC

Qaanol wrote:(after a relatively small group coordinated under the old system to control the nominations this year).

Not to bring politics into this, but Sad Puppies was (is? I think they're doing it again next year) a reaction to organized voting that was already going on. Their goal was basically to show that the Hugo award selection process was a sham, and they succeeded. The nomination process was always too vulnerable to block voting. Sad Puppies just 1) Made it obvious, and 2) Made Worldcon recognize it as problem.


Return to “Serious Business”

Who is online

Users browsing this forum: No registered users and 6 guests